Penetration and Intracellular Growth of Brucella abortus in

INFECTION AND IMMUNITY, JUlY 199, p. 232-2328 Vol. 58, No. 7 19-9567/9/7232-9$2./ Copyright 199, American Society for Microbiology Penetration and Intracellular Growth of Brucella abortus in Nonphagocytic Cells In Vitro PHILIPPE G. DETILLEUX,* BILLY L. DEYOE, AND NORMAN F. CHEVILLE National Animal Disease Center, Agricultural Research Service, U.S. Department of Agriculture, Ames, Iowa 51 Received 1 August 1989/Accepted 16 April 199 In pregnant ruminants, Brucella abortus localizes and replicates within the rough endoplasmic reticulum of trophoblastic epithelial cells. In this study, Vero cells were exposed to B. abortus to investigate its internalization and intracellular growth in nonphagocytic cells. A new double-fluorescence staining procedure to discriminate between extracellular and intracellular bacteria was developed. Studies with the doublefluorescence staining procedure and quantitative bacteriologic culture of disrupted host cells showed that various B. abortus strains replicated within Vero cells, including smooth virulent (strains 238S and 544), smooth attenuated (strain 19), and rough (strains 45/2 and 238R) strains. Rough brucellae were more adherent and entered a greater number of Vero cells. Intracellular replication occurred in a larger percentage of cells with smooth virulent (238S and 544) strains than with smooth attenuated (19) or rough (45/2 and 238R) strains. Differences in adhesiveness and invasiveness were correlated to hydrophobicity of the organism, as measured by hydrocarbon adherence. Ultrastructurally, intracellular smooth (238S) and rough (45/2) brucellae were consistently found within cisternae of the rough endoplasmic reticulum and nuclear envelope. The results suggest that transfer to the rough endoplasmic reticulum is the limiting step in the infection of nonphagocytic cells by B. abortus. Brucella abortus is an intracellular pathogen that causes chronic infections with persistent or recurrent bacteremia in humans and several species of ruminants. The organism has a marked predilection for the gravid uterus of cattle; an important step in the pathogenesis of Brucella placentitis is the invasion of trophoblastic epithelium. Studies in a caprine model (2) indicate that bacteremic B. abortus cells enter and replicate within erythrophagocytic trophoblasts and then colonize the chorioallantoic trophoblastic epithelium. Marked intracellular replication of B. abortus kills trophoblasts, and organisms are shed into surrounding tissues and fluids. The cycle of cell invasion, intracellular replication, and cell death continues, leading to overt placentitis and abortion (2). In trophoblasts, B. abortus replicates within cisternae of the rough endoplasmic reticulum (RER) (1, 2). The same intracellular localization is seen in bovine trophoblasts (26) and in chicken embryo mesenchymal, yolk endodermal, and hepatic cells (7). Replication within RER is a unique mechanism of intracellular bacterial parasitism. Usually, intracellular bacteria inhabit phagosomes, e.g., Salmonella typhimurium (21) and Yersinia enterocolitica (8), or are free in host cell cytoplasm, e.g., Shigella flexneri (35). Most studies on the interactions of B. abortus with host cells deal with monocytes and polymorphonuclear neutrophils. In vitro growth of B. abortus within nonphagocytic cells has been reported, e.g., virulent strains of B. abortus replicate within chicken embryo fibroblasts (16), hamster kidney cells (13), HeLa cells (37), and primary cultures or subcultures of different bovine adult and fetal cells (6, 29, 3). Except for HeLa cells, however, these were all uncharacterized primary cell cultures, precluding meaningful comparisons. In addition, none of these earlier reports provided strong quantitative data nor ultrastructural indication of B. abortus intracellular localization. * Corresponding author. 232 This study was designed to evaluate the internalization and intracellular growth of B. abortus within nonphagocytic cells in vitro. A double-fluorescence staining procedure was used to discriminate between extracellular and intracellular organisms. The effect of various experimental conditions on the ability of B. abortus to establish infection in Vero cells was explored by quantitative means. Replication of B. abortus in Vero cells was shown to occur within the RER. (This research was reported, in part, at the 69th Annual Meeting of the Conference of Research Workers in Animal Disease, Chicago, Ill., 14 and 15 November 1988, abstr. 4.) (This report is part of a dissertation submitted by the senior author in partial fulfillment of the requirements for the Ph.D. degree in veterinary pathology, Iowa State University, Ames.) MATERIALS AND METHODS Bacterial strains. Five strains of B. abortus were used. Strain 238S is a C2-independent virulent smooth strain; strain 544, the international type strain for B. abortus, is a C2-dependent virulent smooth strain; strain 19 is an attenuated smooth strain used worldwide as a live vaccine; strain 45/2 is a rough strain used as a killed-cell vaccine; and strain 238R is a stable rough strain derived from 238S. Cultivation of bacteria. B. abortus was grown on potato infusion agar slants for 48 h at 37 C. Cells were harvested by gentle washing with sterile.85% NaCl and standardized turbidimetrically to a concentration of 1 x 111 CFU/ml. This suspension was then diluted to 1. x 19 CFU/ml in cell culture medium supplemented with 1% fetal calf serum (FCS). In preliminary experiments, we did not find any effect of 1% FCS on the viability of different strains of B. abortus. The brucellicidal effect of bovine serum is complement dependent (5). Because the FCS used in this study was frozen and thawed at least twice before addition to the minimal essential medium (MEM), its complement activity was probably minimal.

VOL. -58, 199 INFECTION OF NONPHAGOCYTIC CELLS BY B. ABORTUS 2321 JISsu culture cells. African green monkey kidney fibroblasts (Vero), Madin-Darby bovine kidney (MDBK), Madin- Darby canine kidney (MDCK), and porcine kidney (PK-15) epithelal cells were obtained from the National Veterinary Service Laboratories (Ames, Iowa). Human choriocarcinoma (Jeg-3, ATCC-HTB36) cells were purchased from the American Type Culture Collection (Rockville, Md.). Rat osteosarcoma (ROS) cells were a gift from T. Reinhard (NatibnalhA-imal Disease Center, Ames, Iowa). Cells were routll'ngrown in Eagle MEM (GIBCO Laboratories, Grand_L4and, N.Y.) supplemented with L-glutamine (2 mm), FCS (1%), and penicillin-streptomycin. For monolayer inoculation, 35-mm plastic tissue culture dishes (Becton Dickinson Labware, Oxnard, Calif.) each containing a glass cover-,slip. (22 by 22 mm) were seeded with 3 ml of a suspension-of 15, to 2, cells per ml. After 5 h of incubation, subconfluent monolayers were washed with MEM at 37TC and were further incubated in 1 ml of MEM supplemented 'with.25% FCS and 2 mm glutamine but without antibiotics. Monolayer inoculation. After overnight incubation, medium was aspirated from culture dishes and 2 ml of bacterial suspension was added. The culture dishes were centrifuged for 2 min at 55 x g at room temperature and placed in a humidified incubator with an atmosphere of 5% CO2 at 37 C. At intervals, cover slips were removed from the petri dishes, washed in two changes of phosphate-buffered saline (PBS) at 37 C, placed in six-well tissue culture plates (Costar, Cambridge, Mass.), and further incubated in MEM supplemented with.25% FCS and gentamicin (5,ug/ml) (Gentocin; Sterling Corp., Kenilworth, N.J.) at bactericidal level to kill extracellular brucellae. In a preliminary experiment, we established that a 1-h exposure of 19 CFU of B. abortus 238S per ml of MEM to 5 p,g of gentamicin per ml reduce their numbers to 14 CFU/ml (data not shown). Medium was replaced at 24 h. The inoculation period was defined as the period between the exposure of the monolayers to B. abortus and the introduction of gentamicin. Experiments were done in triplicate or duplicate and were repeated at least twice for each experimental treatment. Double-fluorescence staining. At intervals, cover slips were washed in PBS, fixed for 5 min in methanol at 4 C, dehydrated for 5 min in acetone at 4 C, and air dried. Brucellae were labeled by indirect immunofluorescence with goat or rabbit anti-b. abortus 238S antiserum (smooth organisms) and rabbit anti-b. canis antiserum (rough organisms) as primary antibody. Secondary antibody was either goat antirabbit or rabbit anti-goat fluorescein isothiocyanate-conjugated antiserum (National Veterinary Service Laboratory). After being washed in PBS containing 3% bovine serum albumin, cover slips were incubated for 1 h in a 1-to-5 dilution of the primary antiserum. After washes in PBSbovine serum albumin, cover slips were incubated for 1 h in a 1-to-5 dilution of the appropriate secondary antiserum. After immunolabeling, cover slips were washed in PBS and treated for 5 min with propidium iodide (Sigma Chemical Co., St. Louis, Mo.) (25,ug/ml in PBS) to stain DNA. Propidium iodide fluoresces bright red when intercalated in double-stranded nucleic acid and exposed to UV light. After staining, cover slips were washed in PBS, rinsed in distilled water, and mounted on glass slides with a 9% solution of glycerin in PBS (ph 8.5). Specimens were examined by epifluorescence microscopy with either a blue (excitation at 49 nm and emission at 515 nm) or green (excitation at 545 nm and emission at 59 nm) filter. For discrimination between intracellular and extracellular bacteria, the primary antibody was applied onto the cover slips for 3 min at 4 C before fixation. Antibodies do not penetrate through the plasma membrane of unfixed cells (11, 19); consequently, only adhering extracellular bacteria reacted with the primary antibody. After being washed, monolayers were fixed and stained as described above, starting after the primary antibody labeling. Since propidium iodide was applied after solubilization of the cells by methanolacetone fixation, both bacterial and host cell DNAs were stained. Enumeration of brucellae. The number of intracellular viable B. abortus was determined at 4, 8, 16, 24, 36, and 48 h postinoculation (p.i.). Except for the 4-h samples, gentamicin was introduced in all samples 8 h after inoculation. The 4-h and 8-h samples were treated with gentamicin for 1 h before bacterial counts. After exposure to gentamicin, inoculated monolayers on cover slips were washed in PBS and incubated for 1 to 15 min in 2 ml of a.1% deoxycholate in distilled water. This procedure disrupted the host cells without affecting the viability of brucellae. Samples of the lysate were serially diluted in PBS for quantitation of CFU of B. abortus on tryptose agar plates containing 5% heatinactivated bovine serum. Brucella colonies were identified by colony morphology and growth characteristics (4). Enumeration of infected cells. For each of three cover slips, the number of cells per square millimeter was estimated at a total magnification of x4 by counting cell nuclei in five fields located along a diagonal across the cover slip (15 to 25 cell nuclei per field). The surface of the field was defined by an eyepiece reticle. The number of infected cells per square millimeter was then estimated at a total magnification of x 1 by counting infected cells in 25 fields (1 to 2 infected cells per field). The results were expressed as percentage of infected cells. In selected experiments, the percentage of Vero cells containing, 1, 2, 3 to 5, 6 to 1, and >1 intracellular brucellae was estimated by examining, at a magnification of x 1,, a total of 2 randomly selected cells in each of two to three cover slips. Hydrocarbon adherence assay. Cell surface hydrophobicity of B. abortus (strains 238S, 19, 238R, and 45/2) was measured by determining the degree to which they associate with hydrocarbons (p-xylene, n-heptane, isooctane) by using a modification of the hydrocarbon adherence method of Rosenberg et al. (33). Standardized bacterial suspensions in.85% saline were adjusted to an optical density of.2 at 6 nm (approximately 2 x 19 CFU/ml). Each bacterial suspension was divided into 36 4-ml samples to which various volumes (5, 1, 15, or 2,u) of hydrocarbon were added. Three tubes without hydrocarbon were controls. After agitation for 3 s, tubes were allowed to stand for 15 min to permit phase separation and optical density of the aqueous phase was determined spectrophotometrically at 6 nm. Results were expressed as percentage of the optical density of control tubes. Transmission electron microscopy. To validate the model, we needed to ascertain that intracellular replication of B. abortus occurs within the RER of Vero cells. For transmission electron microscopy, Vero cells were grown on microporous membranes in 12-mm Millicell-HA inserts (Millipore Corp., Bedford, Mass.) placed in a 24-well tissue culture plate (Costar). After overnight incubation in antibiotic-free MEM, 1 ml of B. abortus inoculum was centrifuged onto inserts placed in 35-mm petri dishes as described above for the cover slips. After 8 h of incubation, the inserts were rinsed in PBS, placed in a 24-well plate, and incubated for 4

2322 DETILLEUX ET AL. INFECT. IMMUN. I FIG. 1. Light micrographs of Vero cells infected with smooth (238S) and rough (238R) B. abortus. (A) 238S, 8 h p.i.; Vero cells with few intracellular brucellae. (B) 238S, 24 h p.i.; slight increase in the number of intracellular brucellae. (C) 238S, 48 h p.i.; Vero cells with cytoplasm full of B. abortus. Note the presence of intracellular brucellae in mitotic cells (arrows). (D) 238S, 72 h p.i.; Vero cell rupture due to excessive intracellular growth of B. abortus. (E) 238R, 4 h p.i.; greater adhesiveness of rough organisms (compare with panel G). (F) 238R, 4 h p.i.; clumps of intemalized rough B. abortus (arrow). (G) 238S, 4 h p.i.; note brucellae adhering to areas of the cover slip not covered by Vero cells (panels F and G, arrowheads). (A to E and G) Bar = 2,um. (F) Bar = 1,.m. h in MEM containing 5,ug of gentamicin per ml. Inserts were then washed in cold PBS and fixed for 1 h by immersion in 2.5% glutaraldehyde in.1 M sodium cacodylate buffer (ph 7.4) at 4 C. After fixation, membranes were removed from the inserts and stored in.1 M sodium cacodylate buffer at 4 C. Membranes were postfixed in osmium tetroxide, infiltrated and embedded in epoxy resin, sectioned at 7 to 9 nm, and examined with a Philips 41 electron microscope. Reproducibility of the model. Variation in the percentage of infected cells on different cover slips in one experiment was + 15%. While variation between different experiments was much greater, the relative effect of different treatments was constant. Variation between different experiments was probably related to variations in the condition of Vero cells. RESULTS Microscopic observations. B. abortus 238S entered and replicated within all cell lines tested (Vero, MDBK, Jeg-3, ROS, and PK-15) except MDCK cells (data not shown). The appearance of infected cells was similar with all cell lines. In our hands, the Vero cell line was the easiest to work with as we could easily modulate its growth and obtain uniform confluent monolayers. Uniform confluent monolayers allowed accurate and reproducible calculation of the percentage of infected cells. Because of these results, all further work was done in Vero cells. I I We developed a procedure to discriminate between intracellular and extracellular bacteria based on color differences in the same microscopic field under UV light. Using the blue filter with this technique, brucellae had an intense greenyellow fluorescence, while eucaryotic cell nuclei were red (Fig. 1A to D). When the primary antibody was applied before fixation, yellow-green fluorescence was restricted to extracellular organisms, while intracellular bacteria stained red (Fig. 1E to G). With a green filter, both bacteria and cell nuclei were stained with an intense red fluorescence. Intracellular localization of B. abortus 238S in Vero cells was evident 2 h p.i. At 4 h p.i., 3 to 4% of the cells contained one or more bacteria, and at the end of the inoculation period, 8 h p.i., 4 to 5% of Vero cells contained intracellular brucellae (Fig. 1A). Intracellular brucellae were isolated rods throughout the cytoplasm (Fig. 1A and B). The number of cells containing at least one organism did not increase significantly after 8 h, but in a small proportion of these cells, the number of intracellular bacteria increased sharply between 24 and 48 h (Fig. 1B and C). At 48 h p.i.,.5 to 1% of the Vero cells were filled with bacteria (Fig. 1C) (referred to hereafter as infected cells). Most infected cells were in clusters of two to six cells. Brucellar replication occurred throughout the cytoplasm, except in the nuclear area. Infected cells in different phases of the mitotic cycle were frequent (Fig. 1C). At 72 h, some heavily infected cells

VOL. 58, 199 INFECTION OF NONPHAGOCYTIC CELLS BY B. ABORTUS 2323 TABLE 1. Relative infectivity (determined 48 h p.i.) of different B. abortus strains in comparison with strain 238S in Vero cells Strain Type No. of determinations Relative infectivity (mean [SEMI)a 238S Smooth 1 544 Smooth 2 9 (33) 19 Smooth 13 15 (1.9) 238R Rough 2 11(2) 45/2 Rough 2 19 (1) a Results are expressed as percentage of strain 238S (mean of two or three samples). had ruptured and organisms were scattered over the field (Fig. 1D). The most accurate determination of the percentage of infected cells was obtained with cover slips sampled 4 h after the end of the 8-h inoculation period. At that sampling time, the number of infected cells was easily quantitated by scanning the cover slips at low magnification. At later sampling times, because increasing numbers of infected cells became disrupted, the estimation of the percentage of these cells was less accurate. Therefore, in all subsequent experiments in which the number of infected cells was determined, cover slips were sampled 48 h p.i. Interaction of four B. abortus strains (544, 19, 238R, and 45/2) with Vero cells was compared with that of strain 238S. All smooth and rough strains entered and grew in Vero cells, and infected cells at 48 h p.i. had similar morphology. There was no difference between strains 238S and 544, two virulent smooth strains (Table 1). When compared with strain 238S, larger numbers of B. abortus strain 19 adhered to Vero cells but entry of strain 19 was less efficient and, at 48 hrs p.i., resulted in a lower percentage of productively infected cells than with strain 238S (Tables 1 and 2). Rough brucellae were markedly more adherent and invasive than smooth brucellae (Table 2). At 4 h p.i., 95% of Vero cells had associated (extracellular and intracellular) rough brucellae, while with smooth organisms, <65% of Vero cells had associated bacteria (Table 2). Some Vero cells inoculated with rough brucellae were covered by extracellular bacteria (Fig. 1E). In 4 to 24 h p.i. samples, the percentage of Vero cells containing intracellular brucellae was higher with rough organisms (Table 2). Intracellular rough brucellae were frequently seen as clumps of five or more bacteria (Fig. 1F); this was not seen with smooth organisms. Despite the higher adherence and invasiveness of rough strains, the percentage of infected cells at 48 h p.i. was <2% of that with smooth 238S (Table 1). With all strains, extracellular adherent brucellae were more abundant at the periphery of the cell (Fig. 1E). In addition, especially with rough strains, numerous brucellae adhered to areas of the cover slips not covered by Vero cells (Fig. 1E and G). Effect of inoculum concentration and of centrifugation. In experiments in which dilutions of cultures of strains 19 and 238S were incubated with Vero cells, the percentage of infected cells increased with increasing bacterial concentration (Fig. 2). In the experiment represented in Fig. 2, although the percentages of infected cells were low (less than 1%), differences between various inoculum concentrations were significant. Infected cells could not be found when the inoculum contained <2 x 17 CFU of strain 19 and <2 x 15 CFU of strain 238S. As an average, there were approximately 1.7 x 15 cells per cover slip at the time of inoculation, and the minimal multiplicity of infection (ratio of number of bacteria to number of Vero cells at time of inoculation) was 12 for strain 19 and 1.2 for 238S. For all further work, an inoculum of 1 x 19 CFU/ml was used, with a multiplicity of infection of 1 x 14 to 1.2 x 14 brucellae per Vero cell. With strains 238S and 19, centrifugation of the inoculum onto monolayers doubled the number of infected cells (data not shown). Kinetics of B. abortus intracellular growth. The number of viable brucellae in Vero cell lysates was determined at various times after inoculation (Fig. 3). After an initial decrease (after 8 h p.i.), the number of viable brucellae in the cellular lysate increased rapidly (after 16 h p.i.), reflecting intracellular replication. For all strains, the maximal rate of intracellular replication was observed between 24 and 36 h p.i. The initial period of decrease may correspond to the delayed killing of extracellular cell-associated bacteria by gentamicin or may represent intracellular destruction of B. abortus in Vero cells that do not become productively infected. Strains 19 and 238S had similar intracellular growth patterns, but at all sampling times the number of viable brucellae in cellular lysates was about 1-fold lower with strain 19. There was no difference in growth pattern between the two rough strains (238R and 45/2), which differed slightly from the growth pattern of smooth organisms. With rough strains, the initial number of viable brucellae was >1-fold higher than that of 238S and the following decrease lasted longer. In a separate experiment, strain 544 had the same growth pattern as smooth 238S (data not shown). Effect of inoculation period length. To establish the optimal length of the inoculation period, cover slips were inoculated for 2, 4, 6, 8, 1, or 12 h before introduction of gentamicin and were sampled 48 h p.i. The percentage of infected cells doubled when the length of the inoculation period was increased from 2 to 8 h, at which time it reached a plateau. An inoculation period of 24 h did not result in a significantly larger percentage of infected cells (data not shown). Hydrophobicity of various B. abortus strains. Marked differences in adhesiveness were correlated to differences in hydrophobic characteristics. The relative hydrophobicity of two smooth (238S and 19) and two rough (238R and 45/2) B. abortus strains was determined by assessing their affin- TABLE 2. Associationa with and internalization by Vero cells of different strains of B. abortus (determined 4 h p.i.) Mean % of Vero cells with Mean no. of associated Mean % of Vero cells with Mean no. of intracellular Strain associated brucellae brucellae/vero cell intracellular brucellae brucellae/vero cell (SEM) (SEM) (SEM) (SEM) 238S 51.17 (5.53).87 (.16) 35.33 (3.38).6 (.14) 19 65. (3.37) 1.2 (.16) 26.67 (3.13).39 (.1) 238R 95.67 (1.44) 4.63 (.42) 47. (3.53) 1.84 (.3) 45/2 97. (1.21) 4.83 (.42) 48.17 (3.53) 1.99 (.31) a Extracellular and intracellular brucellae.

2324 DETILLEUX ET AL. INFECT. IMMUN. GJ a) la '4- '4.4.2. 1 -. - 17 1o 1u8; 19 1 l) 1; Inoculum Concentration (CFU/ml) FIG. 2. Infection of Vero cells by B. abortus 238S (-) and 19 (---) as a function of the inoculum concentration. Results are expressed as the percentage of infected Vero cells estimated microscopically 48 h p.i. (mean of three determinations + standard error [bars]). ities for three hydrocarbons in aqueous-hydrocarbon biphasic systems (33). More rough than smooth brucellae partitioned in the hydrocarbon phase, indicating that rough strains have a higher hydrophobicity (Fig. 4). While among rough strains, 45/2 exhibited the highest hydrophobicity, there was no differences between the two smooth strains. Electron microscopy. Vero cell monolayers grown on v Cu co O._ as Cu '._ ad Cu to I- o 16. 14. 1 3. 12. I I 1 4 8 12 16 2 24 28 32 36 4 44 48 Time After Inoculation FIG. 3. Infection of Vero cells with 2 x 19 CFU of B. abortus 238S (), 19 (A), 238R (), and 45/2 (A). Results are expressed as the mean number of viable B. abortus per milliliter (mean of two or three determinations standard error [bars]). 1 11 microporous membranes were inoculated for 8 h with smooth (238S) or rough (45/2) B. abortus. Two days after inoculation, most infected cells contained large numbers of bacteria (Fig. 5A). Both smooth and rough strains of B. abortus were within cistemae of the RER (Fig. 5B and C). The limiting membranes of brucella-filled cisternae were discontinuously lined by ribosomes. In the most heavily infected cells, brucellae were also in perinuclear spaces (Fig. 5D). Evidence of cellular degeneration (cell swelling and vacuolation) was minimal, although some of the most heavily infected cells were necrotic as indicated by the increase in electron density of their cytoplasm and the loss of structural details. Clusters of two to four brucellae were also found within phagolysosomes, especially in less infected cells that had no organisms in the RER. While half of these brucellae were morphologically intact, the other half were degraded. Myelin figures were frequent in association with intact and degraded brucellae in phagolysosomes. DISCUSSION Using a double-fluorescence staining procedure and quantitative bacteriologic cultures, we were able to show that B. abortus enters and replicates in nonphagocytic Vero cells. As in trophoblasts of infected ruminants (1, 2, 26), the cytoplasm of infected Vero cells contained large numbers of brucellae which surrounded the nuclei, and this massive intracellular bacterial replication resulted in minimal cytopathic effect. The double-fluorescence labeling procedure provides a useful tool to discriminate microscopically between intracellular and extracellular organisms. Differentiation between the two locations of bacteria is the most difficult problem in microscopic assessment of cell culture monolayer penetration. Different methods have been proposed to overcome this problem, e.g., optical sectioning by light microscopy (27), fluorescence quenching technique (14), combined differential interference (Nomarski) and UV light microscopy (4), combined immunofluorescence and bright-field light

VOL. 58,199 INFECTION OF NONPHAGOCYTIC CELLS BY B. ABORTUS 2325._.C" _E I.-.5.1.15.2.5.1.15.2.5.1.15.2 p-xylene n-heptane iso-octane FIG. 4. Relative affinity of various B. abortus strains toward hydrocarbons expressed as percentage of the initial A6. of the aqueous suspension as a function of hydrocarbon volume (mean of three independent determinations + standard error [bars]). Symbols:, 238S; /, 19; *, 238R; and A, 45/2. microscopy (11, 19), and double-immunofluorescence microscopy (15). These methods are cumbersome and slow, as for each determination they require the observation either of the same field under two different conditions (4, 14, 15, 27) or of two differently processed samples (11, 19). In addition, some of these methods (14) necessitate the observation of unfixed material, a hazardous situation when working with human pathogens like B. abortus. Our procedure allows identification of both intracellular and extracellular organisms within the same microscopic field. By reducing the number of samples, it provides an easier, safer, faster, and more accurate procedure. Electron microscopic analysis showed that both smooth and rough B. abortus replicated within RER of Vero cells. As in trophoblasts of experimentally infected goats (1, 2) and cows (26), intracellular brucellae were within cisternae which (i) were lined by ribosomes on their cytoplasmic faces, (ii) were continuous to normal RER, (iii) had outer membranes continuous with outer nuclear membranes, and (iv) had lumens continuous with the perinuclear envelope, which also contained brucellae. Among intracellular bacteria, only Legionella pneumophila has been described in a similar intracellular location; it inhabits ribosome-lined cisternae in phagocytic (1, 17) and nonphagocytic (28) cells. However, continuity between these cisternae and RER or perinuclear envelope has never been reported. In addition, in vitro studies with blood monocytes (1) indicated that internalized legionellae are enclosed in vacuoles which are later surrounded by ribosomes, suggesting that L. pneumophila inhabits ribosome-studded phagosomes rather than RER cisternae. Transfer to the RER may be required for unrestricted growth of B. abortus within nonphagocytic cells. Although at 8 h p.i., 3 to 5% of the Vero cells contained intracellular brucellae, at 48 h p.i. the maximal percentage of infected cells was 1%. Ultrastructurally, in cells with abundant intracellular organisms, B. abortus was always located within RER cisternae, while in cells containing few isolated brucellae, these were within phagosomes or phagolysosomes. This suggests that transfer to the RER, not internalization, is the limiting step for replication of B. abortus in Vero cells and that brucellae that fail to enter the RER are eventually destroyed or eliminated by Vero cells. Differences in infectivity between smooth and rough strains may be correlated to differences in their ability to gain access to RER. A vesicular pathway may be involved in the transfer of B. abortus to the cisternae of the RER. In this as in other studies (1, 7; P. G. Detilleux, B. L. Deyoe, and N. F. Cheville, Vet. Pathol., in press), intracytoplasmic brucellae were always within membrane-bound structures and were never seen free in the cytoplasm. Since intracellular localization within cisternae of the RER is found with various strains and species (Detilleux et al., in press) of Brucella as well as within various types of eucaryotic cells, we suggest that transfer to the RER is Brucella-induced. Cisternae of the RER are the site of synthesis of luminal and membrane proteins. These proteins are then transported via vesicles from the RER to the Golgi apparatus where they are sorted into numerous pathways leading to different final destinations. Retrograde movement of Golgi proteins into the endoplasmic reticulum via tubulovesicular processes has recently been demonstrated (23, 24). Similar mechanisms may be involved in the transfer of B. abortus to the RER, and consequently this organism has the potential of becoming an important tool in cellular biology. Failure of rough B. abortus strains to replicate within some Vero cells may also result from a higher susceptibility of rough strains to intracellular killing. Although rough attenuated brucellae entered a higher proportion of Vero cells, the percentage of infected cells 2 days p.i. was lower than with smooth virulent bacteria. Lack of O-polysaccharide side chain in rough B. abortus is correlated to their higher susceptibility to killing by bovine serum (5). Similarly, components of cell wall lipopolysaccharides are important for the intracellular survival and replication of brucellae in phagocytic cells (12, 31, 32). In vitro, cell walls of smooth brucellae are more resistant to digestion by hydrolytic enzymes than those of rough strains (22). However, studies with extracts of neutrophil and macrophage granules indicate that both smooth and rough brucellae are resistant to lysosomal enzymes (32). Greater susceptibility of rough strains to intracellular killing by phagocytic cells, seems to depend on the myeloperoxidase-h22-halide system (32), which is probably not active in nonphagocytic cells. Despite a high multiplicity of infection (up to 1.1 x 15 brucellae per cell), B. abortus infected only a small percentage of cells in a monolayer. A similar observation was

2326 DETILLEUX ET AL. - --..'.r. 4.J> -, *4wb ;<Ri,d:'>.rtve*zFEpsZ9if5.tr;s!-' INFECT. IMMUN. X8t >7,, ti i *'' ' t' 'F > g t+ w * *..l,-@ ^ w *._ g w. X-. - '-, Aw\; BZ.'t,.s4 v ' w> -- FIG. 5. Electron micrographs of Vero cells infected with B. abortus 238S (A, B, and D) or 45/2 (C) at 48 h p.i. Vero cells containing numerous B. abortus cells are attached to the microporous membrane by cytoplasmic processes (A, arrows). Brucellae are located within cisternae of the RER (B and C) and within perinuclear envelope (D). B. abortus-containing cistemae are discontinuously lined by ribosomes (B and C, open arrows) and are continuous with normal RER (C, arrow). Bar = 1,um. A

VOL. 58, 199 INFECTION OF NONPHAGOCYTIC CELLS BY B. ABORTUS 2327 reported for B. melitensis infection of BHK21 cells; the percentage of infected cells never increased above 2%, even with a multiplicity of infection of 1, bacteria per cell and an inoculation period of 18 h (34). This suggests that monolayers are heterogeneous in terms of susceptibility and/or that B. abortus cells are heterogeneous in terms of invasiveness. In favor of the first alternative is the observation that when incubated with Vero monolayers, S. typhimurium and S. flexneri enter <1% of the cells (3), while with HeLa (9) and Henle 47 (11) monolayers, >5% of the cells are entered. In our study, however, a low percentage of infected cells was observed with five different cell lines. This may indicate that only a fraction of the Brucella inoculum expresses the invasive phenotype. Differences in adherence and invasiveness between smooth and rough strains of B. abortus are correlated to differences in physicochemical properties between the two types of brucellae. Adherence to the cell surface is required for internalization by phagocytic cells, and this adhesion is affected by electrostatic charge and hydrophobicity of both host cells and bacteria (18, 2, 36). Correlation between bacterial hydrophobicity and the degree to which they associate with cells in culture has been demonstrated for S. typhimurium (18) and Y. enterocolitica (36). Loss of polysaccharide side chains increases the hydrophobicity of rough bacteria; this was reported for organisms such as S. typhimurium (25) and Escherichia coli (33) and is confirmed here for B. abortus by the hydrocarbon adherence method (33). Centrifugation increased the percentage of infected Vero cells, indicating that B. abortus attachment is also, at least in part, charge dependent. Centrifugation enhances bacterial attachment by reducing electrostatic repulsion between negatively charged bacteria and host cells (39). Invasion of nonphagocytic cells by smooth B. abortus may be correlated to their virulence in vivo. The attenuated vaccine strain 19 infected a smaller percentage of Vero cells than the virulent strains 238S and 544. Unlike rough B. abortus, this difference was probably not the result of a lower ability to get access to RER or of decreased intracellular survival, because with strain 19 fewer organisms entered Vero cells. Decreased virulence of strain 19 has been attributed to its inability to metabolize erythritol (38). Although we have not tested our Vero cell cultures for the production of erythritol, we think it unlikely that differences in erythritol metabolism are responsible for the observed difference in invasiveness. In summary, we believe that Vero cells provide a useful in vitro model to study B. abortus interaction with nonphagocytic cells. As in ruminant trophoblasts, intracellular replication of B. abortus in Vero cells occurs within cisternae of the RER. This unique mechanism of intracellular parasitism has not been described for any other bacterial pathogen. To gain access to the cisternae of the RER, bacteria of the genus Brucella must have developed specific virulence attributes. Further experimentation with respect to the mechanisnms of transfer of internalized brucellae to the RER are clearly indicated. The procedure presented in this report should help to achieve that goal. ACKNOWLEDGMENTS We acknowledge the technical assistance of Helen Persons, Judi Stasko, Katie Meredith, and Robert Kappmeyer. LITERATURE CITED 1. Anderson, T. D., and N. F. Cheville. 1986. Ultrastructural morphometric analysis of Brucella abortus-infected trophoblasts in experimental placentitis. Am. J. Pathol. 124:226-237. 2. Anderson, T. D., N. F. Cheville, and V. P. Meador. 1986. Pathogenesis of placentitis in the goat inoculated with Brucella abortus. II. Ultrastructural studies. Vet. Pathol. 23:227-239. 3. Barrow, P. A., and M. A. Lovell. 1989. Invasion of Vero cells by Salmonella species. J. Med. Microbiol. 28:59-67. 4. Bukholm, G., B. V. Johansen, E. Namork, and Lassen. 1982. Bacterial adhesiveness and invasiveness in cell culture monolayer. I. A new combined light optical method evaluated by scanning electron microscopy. Acta Pathol. Microbiol. Immunol. Scand. Sect. B 9:43-48. 5. Corbeil, L. B., K. Blau, T. J. Inzana, K. H. Nielsen, R. H. Jacobson, R. R. Corbeil, and A. J. Winter. 1988. Killing of Brucella abortus by bovine serum. Infect. Immun. 56:3251-3261. 6. Corbel, M. J. 198. The effect of synthetic polynucleotides on the intracellular growth of Brucella abortus in vitro and in vivo. Br. Vet. J. 136:175-181. 7. Detilleux, P. G., N. F. Cheville, and B. L. Deyoe. 1988. Pathogenesis of Brucella abortus in chicken embryos. Vet. Pathol. 25:138-146. 8. Finlay, B. B., and S. Falkow. 1988. A comparison of microbial invasion strategies of Salmonella, Shigella, and Yersinia species, p. 227-243. In M. A. Horwitz (ed.), Bacterial-host cell interaction. Alan R. Liss, Inc., New York. 9. Giannella, R. A.,. Washington, P. Gemski, and S. B. Formal. 1973. Invasion of HeLa cells by Salmonella typhimurium: a model for study of invasiveness of Salmonella. J. Infect. Dis. 128:69-75. 1. Glavin, F. L., W. C. Winn, and J. E. Craighead. 1979. Ultrastructure of lung in Legionnaires' disease. Observations of three biopsies done during the Vermont epidemic. Ann. Intern. Med. 9:555-559. 11. Hale, T. L., and P. F. Bonventre. 1979. Shigella infection of Henle intestinal epithelial cells: role of the bacterium. Infect. Immun. 24:879-886. 12. Harmon, B. G., L. G. Adams, and M. Frey. 1988. Survival of rough and smooth strains of Brucella abortus in bovine mammary gland macrophages. Am. J. Vet. Res. 49:192-197. 13. Hatten, B. A., and S. E. Sulkin. 1966. Intracellular production of Brucella L forms. I. Recovery of L forms from tissue culture cells infected with Brucella abortus. J. Bacteriol. 91:285-296. 14. Hed, J. 1977. The extinction of fluorescence by crystal violet and its use to differentiate between attached and ingested microorganisms in phagocytosis. FEMS Microbiol. Lett. 1: 357-361. 15. Heesemann, J., and R. Laufs. 1985. Double immunofluorescence microscopic technique for accurate differentiation of extracellularly and intracellularly located bacteria in cell culture. J. Clin. Microbiol. 22:168-175. 16. Holland, J. J., and M. J. Pickett. 1956. Intracellular behavior of Brucella variants in chick embryo cells in tissue culture. Proc. Soc. Exp. Biol. Med. 93:476-479. 17. Horwitz, M. A. 1983. Formation of a novel phagosome by the Legionnaire's disease bacterium (Legionella pneumophila) in human monocytes. J. Exp. Med. 158:1319-1331. 18. Jones, G. W., L. A. Richardson, and D. Uhlman. 1981. The invasion of HeLa cells by Salmonella typhimurium: reversible and irreversible attachment and the role of bacterial motility. J. Gen. Microbiol. 127:351-36. 19. Kihlstrom, E. 1977. Infection of HeLa cells with Salmonella typhimurium 395 MS and MR 1 bacteria. Infect. Immun. 17:29-295. 2. Kihlstrom, E., and L. Edebo. 1976. Association of viable and inactivated Salmonella typhimurium 395 MS and MR 1 with HeLa cells. Infect. Immun. 14:851-857. 21. Kihlstrom, E., and S. Latkovic. 1978. Ultrastructural studies on the interaction between Salmonella typhimurium 395 M and HeLa cells. Infect. Immun. 22:84-89. 22. Kreutzer, D. L., and D. C. Robertson. 1979. Surface macromolecules and virulence in intracellular parasitism: comparison of cell envelope components of smooth and rough strains of Brucella abortus. Infect. Immun. 23:819-828. 23. Lippincott-Schwartz, J., J. G. Donaldson, A. Schweizer, E. G.

2328 DETILLEUX ET AL. Berger, H. P. Hauri, L. C. Yuan, and R. D. Klausner. 199. Microtubule-dependent retrograde transport of proteins into the ER in the presence of brefeldin A suggests an ER recycling pathway. Cell 6:821-836. 24. Lippincott-Schwartz, J., L. C. Yuan, J. S. Bonifacino, and R. D. Klausner. 1989. Rapid redistribution of Golgi proteins into the ER in cells treated with brefeldin A: evidence for membrane cycling from Golgi to ER. Cell 56:81-813. 25. Magnusson, K.-E., and G. Johansson. 1977. Probing the surface of Salmonella typhimurium and Salmonella minnesota SR and R bacteria by aqueous biphasic partitioning in systems containing hydrophobic and charged polymers. FEMS Microbiol. Lett. 2:225-228. 26. Meador, V. P., and B. L. Deyoe. 1989. Intracellular localization of Brucella abortus in bovine placenta. Vet. Pathol. 26:513-515. 27. Mehiman, I. J., E. L. Eide, A. C. Sanders, M. Fishbein, and C. C. G. Aulisio. 1977. Methodology for recognition of invasive potential of Escherichia coli. J. Assoc. Off. Anal. Chem. 6: 546-562. 28. Oldham, L. J., and F. G. Rodgers. 1985. Adhesion, penetration and intracellular replication of Legionella pneumophila: an in vitro model of pathogenesis. J. Gen. Microbiol. 131:697-76. 29. Richardson, M. 1959. Parasitization in vitro of bovine cells by Brucella abortus. J. Bacteriol. 78:769-777. 3. Richardson, M., and J. N. Holt. 1964. Multiplication of Brucella in cultured lymphoid and nonlymphoid cells. J. Bacteriol. 88:1163-1168. 31. Riley, L. K., and D. C. Robertson. 1984. Ingestion and intracellular survival of Brucella abortus in human and bovine polymorphonuclear leukocytes. Infect. Immun. 46:224-23. INFECT. IMMUN. 32. Riley, L. K., and D. C. Robertson. 1984. Brucellacidal activity of human and bovine polymorphonuclear leukocyte granule extracts against smooth and rough strains of Brucella abortus. Infect. Immun. 46:231-236. 33. Rosenberg, M., D. Gutnick, and E. Rosenberg. 198. Adherence of bacteria to hydrocarbons: a simple method for measuring cell-surface hydrophobicity. FEMS Microbiol. Lett. 9:29-33. 34. Roux, J., and C. Arnaud-Bosq. 1975. La multiplication intracellulaire de Brucella (theorie pour un modele expdrimental). Ann. Microbiol. (Paris) 126A:461-481. 35. Sansonetti, P. J., A. Ryter, P. Clerc, A. T. Maurelli, and J. Mounier. 1986. Multiplication of Shigella flexneri within HeLa cells: lysis of the phagocytic vacuole and plasmid-mediated contact hemolysis. Infect. Immun. 51:461-469. 36. Schiemann, D. A., M. R. Crane, and P. J. Swanz. 1987. Surface properties of Yersinia species and epithelial cell interactions in vitro by a method measuring total associated, attached and intracellular bacteria. J. Med. Microbiol. 24:25-218. 37. Shepard, C. C. 1959. Nonacid-fast bacteria and HeLa cells: their uptake and subsequent intracellular growth. J. Bacteriol. 77:71-714. 38. Sperry, J. F., and D. C. Robertson. 1975. Inhibition of growth by erythritol catabolism in Brucella abortus. J. Bacteriol. 124: 391-397. 39. Vesikari, T., J. Bromirska, and M. Maki. 1982. Enhancement of invasiveness of Yersinia enterocolitica and Escherichia coli in HEp-2 cells by centrifugation. Infect. Immun. 36:834-836. 4. White, P. G., and J. B. Wilson. 1951. Differentiation of smooth and nonsmooth colonies of brucellae. J. Bacteriol. 61:239-24. Downloaded from http://iai.asm.org/ on January 14, 219 by guest