Journal of General Microbiology (199 l), 137, 1431- 1435. Printed in Great Britain
1431
A study of the candidate virulence factors of Bacteroides fragilis FERRY NAMAVAR,'MARIANA. J. J. VERWEIJ-VAN VUGHTand DAVIDM. MACLAREN Department of Medical Microbiology, Research Group for CommensaI Infections, Medical School, Vrije Universiteit, Amsterdam, The Netherlands (Received 4 October 1990; revised 18 January 1991 ;accepted 21 February 1991)
Bactevuides fragifis strains were classified as virulent or avirulent on the basis of their clearance from the subcutaneous tissues of mice. To determine the factors which may contribute to the virulence of B.fragifis strains, we studied encapsulation, hydrophobicity, growth rate, serum sensitivity, agglutination with erythrocytes of different origin, and neuraminidase production. The strains of the virulent group displayed a higher growth rate in broth and a lower sensitivity to the bactericidal activity of serum than the strains of the avirulent group. They also agglutinated different types of erythrocytes more strongly than did the avirulent strains. No significant differences were found between the two groups of strains as regards encapsulation, hydrophobicity and neuraminidase activity.
Introduction Members of the Bacteroides fragilis group are often isolated from various types of mono- and mixed infections (Gorbach & Bartlett, 1974; Duerden, 1980). Such infections usually present as suppurative lesions which may remain localized or may spread to other organs and systems including the bloodstream. Of the fragilis group, B. fragilis itself is the most commonly isolated. In various experimental animal models B. fragilis has been reported to be more virulent than the other species in the B. fragilis group (Onderdonk et al., 1977; Maskell, 1981; Verweij-Van Vught et al., 1986; Brook, 1989). Whether differences in virulence exist between B. ji-agilis strains is not known. Therefore we decided to measure the virulence of B. jiagilis strains by following their clearance from the subcutaneous tissue of mice. Encapsulation, serum sensitivity, haemagglutination activity, growth rate, hydrophobicity and neuraminidase production were studied in relation to survival in a subcutaneous model of infection.
37 "C in an anaerobic chamber (Coy's Manufacturing Co., USA) under an atmosphere of 80% N2, 10% H2 and 10% C 0 2 .Bacterial strains and their sources are listed in Table 1. Virulence studies. The virulence of the B. fragilis strains was measured by following the clearance of the bacteria from the subcutaneous tissue of mice. The details of the model were described by Verweij-Van Vught et al. (1986). Detection of capsules. Capsules were detected by negative staining with Indian ink. A drop of a 24 h broth culture (late exponential phase) was mixed with a drop of 10% (w/v) glucose and a drop of Indian ink on a microscope slide. This was spread over the slide, allowed to air-dry, fixed with ethanol and stained with ammonium crystal violet. Wet preparations were also made by mixing the bacterial suspension, glucose and ink on a slide, placing a coverslip on the mixture and blotting off the excess. The films were examined under light microscopy. Hydrophobicity. The relative surface hydrophobicity of bacteria was measured by their interaction with hexadecane as described by Rosenberg et al. (1980). Hydrophobicity was expressed as the percentage of bacteria recovered in the aqueous phase. Strains were considered relatively hydrophilic when the value obtained was greater than 50% and relatively hydrophobic when less than 50%.
Methods
Growth in broth. The B.fragilis strains were grown in broth for 18 h. A subculture was made by diluting the culture 1 : 100 in fresh medium. Growth was followed by measuring the optical density at 690 nm. The generation time was determined by regression analysis of the optical density values of the exponential phase of growth.
Bacterial strains. The Bacteroides fragilis strains used in this study were collected from various infections in patients at the Academic Hospital of the Vrije Universiteit, Amsterdam, or from faeces of healthy persons. All strains were identified using ATB 32 A (APi) and Minitek system (BBL) kits for the identification of anaerobes. B. fragilis strains were grown in BM broth medium (Shah et al., 1976), supplemented with 5 pg haemin ml-l and 2 pg menadione ml-I at
Serum sensitivity. Serum sensitivity was tested in pooled human serum at concentrations of 10% and 50% (v/v) in phosphate-buffered saline (PBS). Bacteria were grown for 24 h, subcultured (1 :40) in fresh medium and incubated for 4 h in an anaerobic chamber to obtain exponential-phase cells. The cells were washed once in PBS. Serum was inoculated with lo5 c.f.u. as measured by optical density and incubated for 2 h in an anaerobic chamber. Inactivated 10% (v/v) serum (60 min,
0001-6551 0 1991 SGM
1432
F. Namauar and others
56°C) was also included as control. Viable counts were made on Wilkins-Chalgren agar (Oxoid) after incubation for 48 h. Haemagglutination. Human 0,horse, sheep, guinea pig and chicken blood were collected each week and stored at 10% (v/v) concentration in Alsever's solution (Oyston & Handley, 1990) at 4 "C. Erythrocytes were washed three times with PBS and suspended in PBS to give a concentration of 2% (v/v). Bacterial cultures were grown for 24 h then washed three times in PBS and suspended in PBS to a concentration of about 2.5 x lo9 bacteria ml-l, as estimated by optical density. Haemagglutination was tested qualitatively by mixing 50 p1 bacterial suspension with 5Opl erythrocytes on a ceramic ring slide (Clay Adams, USA) for 5 min. When this screening test gave positive results, a quantitative test was performed; serial twofold dilutions of the bacterial suspension in '50 pl PBS were made in round-bottomed microtitre plates (Flow Laboratories). A volume of 50 pl erythrocyte suspension was added to each dilution, then the plate was gently shaken and kept overnight at 4 "C. Haemagglutination titres were expressed as the reciprocal of the highest bacterial dilution that showed haemagglutinat ion activity . Neuraminidase. Bacterial strains were grown for 48 h then washed in PBS and suspended in PBS to a concentration of 5 x lo8 bacteria ml-l, as estimated by optical density. The washed cells were disrupted for 3 min with an MSE 150 W ultrasonic disintegrator. Fetuin (Sigma), at a final concentration of 5 mg ml-l, was used as the substrate. Equal volumes (0.2 ml) of cell extract and substrate solution were mixed and incubated at 37 "C for 18 h. The reaction was stopped by addition of 0.1 ml 0.2 M-sodium metaperiodate in 9 M-orthophosphoric acid to 0.1 ml cell extract/substrate. Enzymically released N-acetylneuraminic acid was estimated by the method of Warren (1959). The A549of the organic phase was measured in a Gilford spectrophotometer. Clostridium perfringens neuraminidase (Sigma) was used for the calibration plot of N-acetylneuraminic acid. Fraser & Brown (1981) have reported that Bacteroides unformis does not produce neuraminidase. Therefore, we also included the B. uniformis BE 40 in this study.
lo9r
10'-
.-m lo6-
L
a
.-L lo5+J cd
D
% lo46 s
z3!
lo3lo2BE 67
10' -
AP 27 BE 17 AP 14 I
I
2 Days after injection of bacteria 1
I
3
Fig. 1. Clearance of virulent (-) and avirulent (---) strains of B. fragilis from the skin of mice at 1,2 and 3 d after injection of lo8 c.f.u.
Statistical analysis. This was performed with Student's t-test.
Results Virulence of B. fragilis strains and capsule staining
were relatively the most hydrophilic strains, with values of 97.5% and 98.0% respectively (Table 1). No significant difference was found in cell surface hydrophobicity between virulent and avirulent strains. No differences were found between faecal strains and those from infections.
Twenty-five B. fragilis strains were tested for virulence. With 20 strains the bacterial numbers in the mouse fell by a factor of only 10'-lo2 in 3 d. These were judged virulent. With five strains the numbers fell by a factor of lo6-10' in 3 d : these were judged avirulent. Of the latter strains one gave variable results in the ATB 32 A and Minitek system and was excluded. We chose five of the virulent strains, selected at random and the four avirulent strains for further study. The clearance rates of the strains chosen are shown in Fig. 1. All strains tested were capsulated : within a strain, noncapsulated cells were always seen (k 10%). The largest capsules were seen on strains BE 12, BE 13 and BE 67. The smallest ones were seen on strain BE 17.
Measurement of generation time in broth revealed differences between the virulent and avirulent strains (Table 1). All the virulent strains showed a significantly (P< 0.05) shorter generation time than the avirulent ones. The mean generation time of the virulent and avirulent groups was 54.4 & 9.2 and 77.0 & 8.3 min, respectively. Strains BE 17 and AP 14 had longer generation times than the other strains of B.fragilis. The two faecal isolates displayed longer generation times than all other strains.
Cell surface hydrophobicity
Serum sensitivity
The virulent strain BE 25 and the avirulent strain AP 14
Since mouse serum was not available in sufficient
Growth rates in broth
Virulence of B. fragilis
Table 1. Source, generation time and surface hydrophobicity of the B. fragilis strains
B. fragilis strain Virulent BE 2 BE 12 BE 13 BE 25 BE 64 Avirulent AP 14 AP 27 BE 67 BE 17
Source
Generation time (min)*
Hydrophobicity (% in water phase)*?
Appendicitis Perianal abscess Appendicitis Sinusitis Pus
56-5 (16.2) 57.5 (1.0) 38.0 (8.8) 60.0 (15.0) 60.0 (8-4)
60.4 (5.2) 93.5 (3.5) 88.9 (4.4) 97.5 (1.2) 57.8 (7.3)
Faeces Faeces Pus Peritoneal cavity
84.5 (6.3) 82.5 (3.5) 66-0 (3.0) 75.0 (8.0)
98.0 (5.1) 75.5 (2.8) 74.5 (9-1) 61.4 (9.9)
* All values shown are the mean of three experiments: SD is given in parentheses. t The higher the value, the less hydrophobic the strain. amounts, pooled human serum was used to measure serum sensitivity of the strains. The mean percentage of viable bacteria after 2 h incubation in 10% serum as compared with the original count was 115 & 16 for strains belonging to the virulent group and 21 & 29 for strains belonging to the avirulent group. When 50% serum was used, the mean percentage of viable bacteria was 105 & 19 and 9 & 16 for virulent and avirulent strains, respectively (Table 2). The avirulent strains were significantly more sensitive to serum than the virulent ones ( P < 0.01). When 10% inactivated serum was used, the mean percentage of viable bacteria after 2 h was 102.8 & 9.6 for the virulent group and 106.0 & 6.2 for the
1433
avirulent group. These results point to the role of complement in the serum sensitivity of the avirulent group. Haemagglutina tion
In the qualitative test, the strains of the virulent group strongly agglutinated all five blood types tested (Table 3). Agglutination of chicken erythrocytes was most marked and gave the most clear-cut results. The strains of the avirulent group did not agglutinate the five blood types or did so very weakly. Chicken erythrocytes were used to measure the agglutination quantitatively. Strains BE 12, BE 13 and BE 64 were the most strongly positive, agglutinating chicken erythrocytes at a concentration of 7.6 x lo7 bacteria ml-l. The strains of the avirulent group agglutinated erythrocytes only at a concentration of 2.5 x lo9 bacteria ml-l. Neuraminidase
Table 2 shows the amounts of neuraminidase produced by virulent and avirulent strains of B. fragilis. The bacteria were grown for 24 h in broth and the cell extract was tested for neuraminidase activity. The enzyme was in all cases cell-bound, with only small amounts detectable in culture supernatants. All the B. fragilis strains produced neuraminidase, in varying amounts. Mean neuraminidase activity for the virulent and avirulent groups was 35-7 & 8.8 and 31.1 8.9 nmol min-l per lo8 c.f.u., respectively. The difference between the two groups is not significant. B. uniformis BE 40 produced neuraminidase, but at a very low level (10 nmol min-l per lo8 c.f.u.).
Table 2. Neuraminidase production and serum sensitivity of B . fragilis strains
B. fragilis strain Virulent BE 2 BE 12 BE 13 BE 25 BE 64 Avirulent AP 14 AP 27 BE 67 BE 17
Neuraminidase (nmol min-l per lo8 c.f.u.)*
Discussion
Serum sensitivity (% of inoculum after 2 h)* in: 10% serum
50% serum
38.0 (5.2) 36.3 (5.6) 41.5 (1.0) 42.3 (1 - 5 ) 20.5 (2.1)
140.0 (17.0) 101.0 (8-0) 124.0 (15.0) 113.0 (20.0) 101.0 (16-0)
133.0 (13.0) 85.0 (2.0) 116.0 (17.0) 92.0 (18.0) 102.0 (10.0)
23.3 (2.3) 23.5 (2.1) 39.0 (1*4) 38.6 (5.5)
65.0 (23-0) 9-0 (10.0) 10.0 (7.0) 0.0
34-0 (20-0) 2.0 (1 -0) 1.0 (2.0) 0.0
* All values calculated are the mean of three experiments: SD is given in parentheses.
In this study we examined a number of B .fragifis strains, some from infections, some from faeces, and classified them as virulent or avirulent on the basis of their rate of clearance from the subcutaneous tissue of the mouse. This is a very artificial-model, but-economical i n + k use of experimental animals, and the progress of the infection can be followed quantitatively. It has proved its value in confirming the clinical observation that B. vulgatus is less virulent than B. jiagifis (Verweij-Van Vught et a f . , 1986). Differences in virulence between strains of B. fragilis were shown in this model. We therefore studied several virulence factors which might explain these differences. All B.fragifisstrains examined possessed a capsule, the capsule size varying between strains ; both capsulated
1434
F . Namavar and others Table 3 . Haemagglutinatwn of erythrocytes of diferent origin by strains of B. fragilis
0, No agglutination with 2.5 x lo9 bacteria m1-I; 1 , 2.5 x lo9 bacteria ml-l; 2, 1.2 x lo9 bacteria ml-I; 3, 6.1 x lo8 bacteria ml-I.
B. fragilis strain
Chicken
Sheep
Human
Horse
Guineapig
Virulent BE 2 BE 12 BE 13 BE 25 BE 64 A virulent AP 14 AP 27 BE 67 BE 17
and noncapsulated cells were present in populations of all strains. The same observation has been reported by other investigators (Patrick et al., 1986; Oyston & Handley, 1990). Since the avirulent strains also possessed a capsule, the documented importance of a capsule for the virulence of B. fragilis (Kasper, 1976) was not confirmed in this study. There was no correlation between virulence and hydrophobicity. Growth rate studies in vitro, however, showed that avirulent strains grew more slowly than virulent ones. If these differences in growth rate are mirrored in the in vivo situation, where growth conditions are more difficult for bacteria, this might well explain enhanced virulence. It is known that antibiotic-resistant mutants of staphylococci are less virulent in vivo than their parents and this has been correlated with a slower in vivo growth rate (Gorrill, 1958). The virulent strains were much less serum sensitive than the avirulent ones; the fact that strains from both groups survived equally well in inactivated serum indicates that complement is crucial for the killing of avirulent strains. Another clear distinction between virulent and avirulent strains lay in the ability to cause haemagglutination; virulent strains were clearly more active. Vel et al. (1986) and Oyston & Handley (1990) detected different patterns of haemagglutination among their B. fragilis strains. Such differences were not obvious in this study, but we only studied five virulent strains. The avirulent strains produced no (or very weak) haemagglutination. There was no correlation between capsulation (as seen with Indian ink) and haemagglutination. We examined our strains for neuraminidase because it
Chicken (titration) 1 :32 1 :64 1 :64 1 :16 1 :64 1 0 1 0
has been proposed as a virulence enzyme in view of its activity on cell and serum glycoproteins. No differences in neuraminidase activity were found between avirulent and virulent strains. We included B. uniformis as a negative control, although it did seem to produce low levels of neuraminidase. Fraser & Brown (1981) and Riley (1984) have reported the absence of neuraminidase in B. unqormis. The virulence of B. fragilis strains for mice seemed to be associated with faster growth in broth, with resistance to serum killing and with the capacity for haemagglutination. The growth rate and resistance to serum killing may be factors relevant for the mouse model used. The ability of B. fragilis to agglutinate different erythrocytes may be relevant in the in vivo situation of intraabdominal infection, where adherence of bacteria to peritoneal cells may prevent clearance of bacteria from the peritoneal cavity. The intimate association of B. fragilis with the gut mucosa as described previously (Namavar et al., 1989) would support the concept of a great capacity of this bacterium to adhere to the epithelial lining of the large intestine. Further studies are planned to throw more light on the virulence factor(s) of B. fragilis strains.
References BROOK, I. (1989). Pathogenicity of the Bacteroides fragilis group. Annals of Clinical and Laboratory Science 19, 360-375. DUERDEN, B. I. (1980). The isolation and identification of Bacteroides spp. from the normal human faecal flora. Journal of Medical Microbiology 13, 69-78. FRASER,A. G . & BROWN,R. (1981). Neuraminidase production by Bacteroidaceae. Journal of Medical Microbiology 14, 63-76.
Virulence of B. fragilis GORBACH, S. L. & BARTLETT,3. G . (1974). Anaerobic infections. New England Journal of Medicine 290, 1177-1 184, 1237-1245, 1289-1294. GORRILL, R. H . (1958). The establishment of staphylococcal abscesses in the mouse kidney. British Journal of Experimental Pathology 39, 203-2 12.
KASPER,D. L. (1976). The polysaccharide capsule of Bacteroidesfragilis subspecies fragilis: immunochemical and morphologic detection. Journal of Infectious Diseases 133, 19-87. MASKELL, J. P. (1981). The pathogenicity of Bacteroides fragilis and related species estimated by intra-cutaneous infection in the guineapig. Journal of Medical Microbiology 14, 131-140. NAMAVAR, F., THEUNISSEN, E. B. M., VERWEIJ-VANVUGHT, A. M. J . J., PEERBOOMS, P. G. H., BAL, M., HOITSMA,H . F. W . & MACLAREN, D. M. (1989). Epidemiology of Bacteroides fragilis group in the colonic flora in 10 patients with colonic cancer. Journal of Medical Microbiology 29, 171- 176. ONDERDONK, A. B., KASPER,D. L., CISNEROS, R. L. & BARTLETT, J. G. (1977). The capsular polysaccharide of Bacteroides fragilis as a virulence factor: comparison of the pathogenic potential of encapsulated and unencapsulated strains. Journal of Infectious Diseases 136, 82-89. OYSTON,P. C. F. & HANDLEY,P. S. (1990). Surface structures, haemagglutination and cell surface hydrophobicity of Bacteroides fragilis strains. Journal of General Microbiology 136, 94 1-948.
1435
PATRICK,S., REID,J. H. & COFFGY,A. (1986). Capsulation of in vivo and in vitro grown Bacteroides species. Journal of General Microbiology 132, 1099-1 109. RILEY,T. V. (1984). Neuraminidase production by Bacteroides species. FEMS Microbiology Letters 25, 225-227. ROSENBERG, M., GUTNICK,D. & ROSENBERG, E. (1980). Adherence of bacteria to hydrocarbons: a simple method for measuring cellsurface hydrophobicity. FEMS Microbiology Letters 25, 229-232. SHAH,H. N., WILLIAMS,R. A. D., BOWDEN,G . H. & HARDIE,J. M. (1976). Comparison of the biochemical properties of Bacteroides melaninogenicus from human dental plaque and other sites. Journal of Applied Bacteriology 41, 473-492. VEL, W . A. C., NAMAVAR,F., VERWEIJ-VAN VUGHT,A. M. J . J., PUBBEN,A. B. P. & MACLAREN, D. M. (1986). Haemagglutination by the Bacteroides fragilis group. Journal of Medical Microbiology 21, 105-107.
VERWEIJ-VAN VUGHT, A. M. J. J., NAMAVAR,F., VEL, W. A. C., SPARIUS,M. & MACLAREN,D. M. (1986). Pathogenic synergy between Escherichia coli and Bacteroides fragilis or B. vulgatus in experimental infections : a nonspecific phenomenon. Journal of Medical Microbiology 21, 43-47. WARREN,L. (1959). The thiobarbituric acid assay of sialic acids. Journal of Biological Chemistry 234, 1971-1975.
1431
A study of the candidate virulence factors of Bacteroides fragilis FERRY NAMAVAR,'MARIANA. J. J. VERWEIJ-VAN VUGHTand DAVIDM. MACLAREN Department of Medical Microbiology, Research Group for CommensaI Infections, Medical School, Vrije Universiteit, Amsterdam, The Netherlands (Received 4 October 1990; revised 18 January 1991 ;accepted 21 February 1991)
Bactevuides fragifis strains were classified as virulent or avirulent on the basis of their clearance from the subcutaneous tissues of mice. To determine the factors which may contribute to the virulence of B.fragifis strains, we studied encapsulation, hydrophobicity, growth rate, serum sensitivity, agglutination with erythrocytes of different origin, and neuraminidase production. The strains of the virulent group displayed a higher growth rate in broth and a lower sensitivity to the bactericidal activity of serum than the strains of the avirulent group. They also agglutinated different types of erythrocytes more strongly than did the avirulent strains. No significant differences were found between the two groups of strains as regards encapsulation, hydrophobicity and neuraminidase activity.
Introduction Members of the Bacteroides fragilis group are often isolated from various types of mono- and mixed infections (Gorbach & Bartlett, 1974; Duerden, 1980). Such infections usually present as suppurative lesions which may remain localized or may spread to other organs and systems including the bloodstream. Of the fragilis group, B. fragilis itself is the most commonly isolated. In various experimental animal models B. fragilis has been reported to be more virulent than the other species in the B. fragilis group (Onderdonk et al., 1977; Maskell, 1981; Verweij-Van Vught et al., 1986; Brook, 1989). Whether differences in virulence exist between B. ji-agilis strains is not known. Therefore we decided to measure the virulence of B. jiagilis strains by following their clearance from the subcutaneous tissue of mice. Encapsulation, serum sensitivity, haemagglutination activity, growth rate, hydrophobicity and neuraminidase production were studied in relation to survival in a subcutaneous model of infection.
37 "C in an anaerobic chamber (Coy's Manufacturing Co., USA) under an atmosphere of 80% N2, 10% H2 and 10% C 0 2 .Bacterial strains and their sources are listed in Table 1. Virulence studies. The virulence of the B. fragilis strains was measured by following the clearance of the bacteria from the subcutaneous tissue of mice. The details of the model were described by Verweij-Van Vught et al. (1986). Detection of capsules. Capsules were detected by negative staining with Indian ink. A drop of a 24 h broth culture (late exponential phase) was mixed with a drop of 10% (w/v) glucose and a drop of Indian ink on a microscope slide. This was spread over the slide, allowed to air-dry, fixed with ethanol and stained with ammonium crystal violet. Wet preparations were also made by mixing the bacterial suspension, glucose and ink on a slide, placing a coverslip on the mixture and blotting off the excess. The films were examined under light microscopy. Hydrophobicity. The relative surface hydrophobicity of bacteria was measured by their interaction with hexadecane as described by Rosenberg et al. (1980). Hydrophobicity was expressed as the percentage of bacteria recovered in the aqueous phase. Strains were considered relatively hydrophilic when the value obtained was greater than 50% and relatively hydrophobic when less than 50%.
Methods
Growth in broth. The B.fragilis strains were grown in broth for 18 h. A subculture was made by diluting the culture 1 : 100 in fresh medium. Growth was followed by measuring the optical density at 690 nm. The generation time was determined by regression analysis of the optical density values of the exponential phase of growth.
Bacterial strains. The Bacteroides fragilis strains used in this study were collected from various infections in patients at the Academic Hospital of the Vrije Universiteit, Amsterdam, or from faeces of healthy persons. All strains were identified using ATB 32 A (APi) and Minitek system (BBL) kits for the identification of anaerobes. B. fragilis strains were grown in BM broth medium (Shah et al., 1976), supplemented with 5 pg haemin ml-l and 2 pg menadione ml-I at
Serum sensitivity. Serum sensitivity was tested in pooled human serum at concentrations of 10% and 50% (v/v) in phosphate-buffered saline (PBS). Bacteria were grown for 24 h, subcultured (1 :40) in fresh medium and incubated for 4 h in an anaerobic chamber to obtain exponential-phase cells. The cells were washed once in PBS. Serum was inoculated with lo5 c.f.u. as measured by optical density and incubated for 2 h in an anaerobic chamber. Inactivated 10% (v/v) serum (60 min,
0001-6551 0 1991 SGM
1432
F. Namauar and others
56°C) was also included as control. Viable counts were made on Wilkins-Chalgren agar (Oxoid) after incubation for 48 h. Haemagglutination. Human 0,horse, sheep, guinea pig and chicken blood were collected each week and stored at 10% (v/v) concentration in Alsever's solution (Oyston & Handley, 1990) at 4 "C. Erythrocytes were washed three times with PBS and suspended in PBS to give a concentration of 2% (v/v). Bacterial cultures were grown for 24 h then washed three times in PBS and suspended in PBS to a concentration of about 2.5 x lo9 bacteria ml-l, as estimated by optical density. Haemagglutination was tested qualitatively by mixing 50 p1 bacterial suspension with 5Opl erythrocytes on a ceramic ring slide (Clay Adams, USA) for 5 min. When this screening test gave positive results, a quantitative test was performed; serial twofold dilutions of the bacterial suspension in '50 pl PBS were made in round-bottomed microtitre plates (Flow Laboratories). A volume of 50 pl erythrocyte suspension was added to each dilution, then the plate was gently shaken and kept overnight at 4 "C. Haemagglutination titres were expressed as the reciprocal of the highest bacterial dilution that showed haemagglutinat ion activity . Neuraminidase. Bacterial strains were grown for 48 h then washed in PBS and suspended in PBS to a concentration of 5 x lo8 bacteria ml-l, as estimated by optical density. The washed cells were disrupted for 3 min with an MSE 150 W ultrasonic disintegrator. Fetuin (Sigma), at a final concentration of 5 mg ml-l, was used as the substrate. Equal volumes (0.2 ml) of cell extract and substrate solution were mixed and incubated at 37 "C for 18 h. The reaction was stopped by addition of 0.1 ml 0.2 M-sodium metaperiodate in 9 M-orthophosphoric acid to 0.1 ml cell extract/substrate. Enzymically released N-acetylneuraminic acid was estimated by the method of Warren (1959). The A549of the organic phase was measured in a Gilford spectrophotometer. Clostridium perfringens neuraminidase (Sigma) was used for the calibration plot of N-acetylneuraminic acid. Fraser & Brown (1981) have reported that Bacteroides unformis does not produce neuraminidase. Therefore, we also included the B. uniformis BE 40 in this study.
lo9r
10'-
.-m lo6-
L
a
.-L lo5+J cd
D
% lo46 s
z3!
lo3lo2BE 67
10' -
AP 27 BE 17 AP 14 I
I
2 Days after injection of bacteria 1
I
3
Fig. 1. Clearance of virulent (-) and avirulent (---) strains of B. fragilis from the skin of mice at 1,2 and 3 d after injection of lo8 c.f.u.
Statistical analysis. This was performed with Student's t-test.
Results Virulence of B. fragilis strains and capsule staining
were relatively the most hydrophilic strains, with values of 97.5% and 98.0% respectively (Table 1). No significant difference was found in cell surface hydrophobicity between virulent and avirulent strains. No differences were found between faecal strains and those from infections.
Twenty-five B. fragilis strains were tested for virulence. With 20 strains the bacterial numbers in the mouse fell by a factor of only 10'-lo2 in 3 d. These were judged virulent. With five strains the numbers fell by a factor of lo6-10' in 3 d : these were judged avirulent. Of the latter strains one gave variable results in the ATB 32 A and Minitek system and was excluded. We chose five of the virulent strains, selected at random and the four avirulent strains for further study. The clearance rates of the strains chosen are shown in Fig. 1. All strains tested were capsulated : within a strain, noncapsulated cells were always seen (k 10%). The largest capsules were seen on strains BE 12, BE 13 and BE 67. The smallest ones were seen on strain BE 17.
Measurement of generation time in broth revealed differences between the virulent and avirulent strains (Table 1). All the virulent strains showed a significantly (P< 0.05) shorter generation time than the avirulent ones. The mean generation time of the virulent and avirulent groups was 54.4 & 9.2 and 77.0 & 8.3 min, respectively. Strains BE 17 and AP 14 had longer generation times than the other strains of B.fragilis. The two faecal isolates displayed longer generation times than all other strains.
Cell surface hydrophobicity
Serum sensitivity
The virulent strain BE 25 and the avirulent strain AP 14
Since mouse serum was not available in sufficient
Growth rates in broth
Virulence of B. fragilis
Table 1. Source, generation time and surface hydrophobicity of the B. fragilis strains
B. fragilis strain Virulent BE 2 BE 12 BE 13 BE 25 BE 64 Avirulent AP 14 AP 27 BE 67 BE 17
Source
Generation time (min)*
Hydrophobicity (% in water phase)*?
Appendicitis Perianal abscess Appendicitis Sinusitis Pus
56-5 (16.2) 57.5 (1.0) 38.0 (8.8) 60.0 (15.0) 60.0 (8-4)
60.4 (5.2) 93.5 (3.5) 88.9 (4.4) 97.5 (1.2) 57.8 (7.3)
Faeces Faeces Pus Peritoneal cavity
84.5 (6.3) 82.5 (3.5) 66-0 (3.0) 75.0 (8.0)
98.0 (5.1) 75.5 (2.8) 74.5 (9-1) 61.4 (9.9)
* All values shown are the mean of three experiments: SD is given in parentheses. t The higher the value, the less hydrophobic the strain. amounts, pooled human serum was used to measure serum sensitivity of the strains. The mean percentage of viable bacteria after 2 h incubation in 10% serum as compared with the original count was 115 & 16 for strains belonging to the virulent group and 21 & 29 for strains belonging to the avirulent group. When 50% serum was used, the mean percentage of viable bacteria was 105 & 19 and 9 & 16 for virulent and avirulent strains, respectively (Table 2). The avirulent strains were significantly more sensitive to serum than the virulent ones ( P < 0.01). When 10% inactivated serum was used, the mean percentage of viable bacteria after 2 h was 102.8 & 9.6 for the virulent group and 106.0 & 6.2 for the
1433
avirulent group. These results point to the role of complement in the serum sensitivity of the avirulent group. Haemagglutina tion
In the qualitative test, the strains of the virulent group strongly agglutinated all five blood types tested (Table 3). Agglutination of chicken erythrocytes was most marked and gave the most clear-cut results. The strains of the avirulent group did not agglutinate the five blood types or did so very weakly. Chicken erythrocytes were used to measure the agglutination quantitatively. Strains BE 12, BE 13 and BE 64 were the most strongly positive, agglutinating chicken erythrocytes at a concentration of 7.6 x lo7 bacteria ml-l. The strains of the avirulent group agglutinated erythrocytes only at a concentration of 2.5 x lo9 bacteria ml-l. Neuraminidase
Table 2 shows the amounts of neuraminidase produced by virulent and avirulent strains of B. fragilis. The bacteria were grown for 24 h in broth and the cell extract was tested for neuraminidase activity. The enzyme was in all cases cell-bound, with only small amounts detectable in culture supernatants. All the B. fragilis strains produced neuraminidase, in varying amounts. Mean neuraminidase activity for the virulent and avirulent groups was 35-7 & 8.8 and 31.1 8.9 nmol min-l per lo8 c.f.u., respectively. The difference between the two groups is not significant. B. uniformis BE 40 produced neuraminidase, but at a very low level (10 nmol min-l per lo8 c.f.u.).
Table 2. Neuraminidase production and serum sensitivity of B . fragilis strains
B. fragilis strain Virulent BE 2 BE 12 BE 13 BE 25 BE 64 Avirulent AP 14 AP 27 BE 67 BE 17
Neuraminidase (nmol min-l per lo8 c.f.u.)*
Discussion
Serum sensitivity (% of inoculum after 2 h)* in: 10% serum
50% serum
38.0 (5.2) 36.3 (5.6) 41.5 (1.0) 42.3 (1 - 5 ) 20.5 (2.1)
140.0 (17.0) 101.0 (8-0) 124.0 (15.0) 113.0 (20.0) 101.0 (16-0)
133.0 (13.0) 85.0 (2.0) 116.0 (17.0) 92.0 (18.0) 102.0 (10.0)
23.3 (2.3) 23.5 (2.1) 39.0 (1*4) 38.6 (5.5)
65.0 (23-0) 9-0 (10.0) 10.0 (7.0) 0.0
34-0 (20-0) 2.0 (1 -0) 1.0 (2.0) 0.0
* All values calculated are the mean of three experiments: SD is given in parentheses.
In this study we examined a number of B .fragifis strains, some from infections, some from faeces, and classified them as virulent or avirulent on the basis of their rate of clearance from the subcutaneous tissue of the mouse. This is a very artificial-model, but-economical i n + k use of experimental animals, and the progress of the infection can be followed quantitatively. It has proved its value in confirming the clinical observation that B. vulgatus is less virulent than B. jiagifis (Verweij-Van Vught et a f . , 1986). Differences in virulence between strains of B. fragilis were shown in this model. We therefore studied several virulence factors which might explain these differences. All B.fragifisstrains examined possessed a capsule, the capsule size varying between strains ; both capsulated
1434
F . Namavar and others Table 3 . Haemagglutinatwn of erythrocytes of diferent origin by strains of B. fragilis
0, No agglutination with 2.5 x lo9 bacteria m1-I; 1 , 2.5 x lo9 bacteria ml-l; 2, 1.2 x lo9 bacteria ml-I; 3, 6.1 x lo8 bacteria ml-I.
B. fragilis strain
Chicken
Sheep
Human
Horse
Guineapig
Virulent BE 2 BE 12 BE 13 BE 25 BE 64 A virulent AP 14 AP 27 BE 67 BE 17
and noncapsulated cells were present in populations of all strains. The same observation has been reported by other investigators (Patrick et al., 1986; Oyston & Handley, 1990). Since the avirulent strains also possessed a capsule, the documented importance of a capsule for the virulence of B. fragilis (Kasper, 1976) was not confirmed in this study. There was no correlation between virulence and hydrophobicity. Growth rate studies in vitro, however, showed that avirulent strains grew more slowly than virulent ones. If these differences in growth rate are mirrored in the in vivo situation, where growth conditions are more difficult for bacteria, this might well explain enhanced virulence. It is known that antibiotic-resistant mutants of staphylococci are less virulent in vivo than their parents and this has been correlated with a slower in vivo growth rate (Gorrill, 1958). The virulent strains were much less serum sensitive than the avirulent ones; the fact that strains from both groups survived equally well in inactivated serum indicates that complement is crucial for the killing of avirulent strains. Another clear distinction between virulent and avirulent strains lay in the ability to cause haemagglutination; virulent strains were clearly more active. Vel et al. (1986) and Oyston & Handley (1990) detected different patterns of haemagglutination among their B. fragilis strains. Such differences were not obvious in this study, but we only studied five virulent strains. The avirulent strains produced no (or very weak) haemagglutination. There was no correlation between capsulation (as seen with Indian ink) and haemagglutination. We examined our strains for neuraminidase because it
Chicken (titration) 1 :32 1 :64 1 :64 1 :16 1 :64 1 0 1 0
has been proposed as a virulence enzyme in view of its activity on cell and serum glycoproteins. No differences in neuraminidase activity were found between avirulent and virulent strains. We included B. uniformis as a negative control, although it did seem to produce low levels of neuraminidase. Fraser & Brown (1981) and Riley (1984) have reported the absence of neuraminidase in B. unqormis. The virulence of B. fragilis strains for mice seemed to be associated with faster growth in broth, with resistance to serum killing and with the capacity for haemagglutination. The growth rate and resistance to serum killing may be factors relevant for the mouse model used. The ability of B. fragilis to agglutinate different erythrocytes may be relevant in the in vivo situation of intraabdominal infection, where adherence of bacteria to peritoneal cells may prevent clearance of bacteria from the peritoneal cavity. The intimate association of B. fragilis with the gut mucosa as described previously (Namavar et al., 1989) would support the concept of a great capacity of this bacterium to adhere to the epithelial lining of the large intestine. Further studies are planned to throw more light on the virulence factor(s) of B. fragilis strains.
References BROOK, I. (1989). Pathogenicity of the Bacteroides fragilis group. Annals of Clinical and Laboratory Science 19, 360-375. DUERDEN, B. I. (1980). The isolation and identification of Bacteroides spp. from the normal human faecal flora. Journal of Medical Microbiology 13, 69-78. FRASER,A. G . & BROWN,R. (1981). Neuraminidase production by Bacteroidaceae. Journal of Medical Microbiology 14, 63-76.
Virulence of B. fragilis GORBACH, S. L. & BARTLETT,3. G . (1974). Anaerobic infections. New England Journal of Medicine 290, 1177-1 184, 1237-1245, 1289-1294. GORRILL, R. H . (1958). The establishment of staphylococcal abscesses in the mouse kidney. British Journal of Experimental Pathology 39, 203-2 12.
KASPER,D. L. (1976). The polysaccharide capsule of Bacteroidesfragilis subspecies fragilis: immunochemical and morphologic detection. Journal of Infectious Diseases 133, 19-87. MASKELL, J. P. (1981). The pathogenicity of Bacteroides fragilis and related species estimated by intra-cutaneous infection in the guineapig. Journal of Medical Microbiology 14, 131-140. NAMAVAR, F., THEUNISSEN, E. B. M., VERWEIJ-VANVUGHT, A. M. J . J., PEERBOOMS, P. G. H., BAL, M., HOITSMA,H . F. W . & MACLAREN, D. M. (1989). Epidemiology of Bacteroides fragilis group in the colonic flora in 10 patients with colonic cancer. Journal of Medical Microbiology 29, 171- 176. ONDERDONK, A. B., KASPER,D. L., CISNEROS, R. L. & BARTLETT, J. G. (1977). The capsular polysaccharide of Bacteroides fragilis as a virulence factor: comparison of the pathogenic potential of encapsulated and unencapsulated strains. Journal of Infectious Diseases 136, 82-89. OYSTON,P. C. F. & HANDLEY,P. S. (1990). Surface structures, haemagglutination and cell surface hydrophobicity of Bacteroides fragilis strains. Journal of General Microbiology 136, 94 1-948.
1435
PATRICK,S., REID,J. H. & COFFGY,A. (1986). Capsulation of in vivo and in vitro grown Bacteroides species. Journal of General Microbiology 132, 1099-1 109. RILEY,T. V. (1984). Neuraminidase production by Bacteroides species. FEMS Microbiology Letters 25, 225-227. ROSENBERG, M., GUTNICK,D. & ROSENBERG, E. (1980). Adherence of bacteria to hydrocarbons: a simple method for measuring cellsurface hydrophobicity. FEMS Microbiology Letters 25, 229-232. SHAH,H. N., WILLIAMS,R. A. D., BOWDEN,G . H. & HARDIE,J. M. (1976). Comparison of the biochemical properties of Bacteroides melaninogenicus from human dental plaque and other sites. Journal of Applied Bacteriology 41, 473-492. VEL, W . A. C., NAMAVAR,F., VERWEIJ-VAN VUGHT,A. M. J . J., PUBBEN,A. B. P. & MACLAREN, D. M. (1986). Haemagglutination by the Bacteroides fragilis group. Journal of Medical Microbiology 21, 105-107.
VERWEIJ-VAN VUGHT, A. M. J. J., NAMAVAR,F., VEL, W. A. C., SPARIUS,M. & MACLAREN,D. M. (1986). Pathogenic synergy between Escherichia coli and Bacteroides fragilis or B. vulgatus in experimental infections : a nonspecific phenomenon. Journal of Medical Microbiology 21, 43-47. WARREN,L. (1959). The thiobarbituric acid assay of sialic acids. Journal of Biological Chemistry 234, 1971-1975.
