|Home | About | Journals | Submit | Contact Us | Français|
Anaerobic bacteria can cause a wide variety of infections, and some of these infections can be serious. Conventional identification methods based on biochemical tests are often lengthy and can produce inconclusive results. An oligonucleotide array based on the 16S-23S rRNA intergenic spacer (ITS) sequences was developed to identify 28 species of anaerobic bacteria and Veillonella. The method consisted of PCR amplification of the ITS regions with universal primers, followed by hybridization of the digoxigenin-labeled PCR products to a panel of 35 oligonucleotide probes (17- to 30-mers) immobilized on a nylon membrane. The performance of the array was determined by testing 310 target strains (strains which we aimed to identify), including 122 reference strains and 188 clinical isolates. In addition, 98 nontarget strains were used for specificity testing. The sensitivity and the specificity of the array for the identification of pure cultures were 99.7 and 97.1%, respectively. The array was further assessed for its ability to detect anaerobic bacteria in 49 clinical specimens. Two species (Finegoldia magna and Bacteroides vulgatus) were detected in two specimens by the array, and the results were in accordance with those obtained by culture. The whole procedure of array hybridization took about 8 h, starting with the isolated colonies. The array can be used as an accurate alternative to conventional methods for the identification of clinically important anaerobes.
Anaerobic bacteria are important human pathogens, and infections caused by these bacteria can be serious and life-threatening (6). A recent report from the Mayo Clinic (Rochester, MN) revealed an overall increase in the incidence of anaerobic bacteremias of 74% from 2001 to 2004 compared to that from 1993 to 1996 (20), although the same trend was not found in community hospitals or in an European countries (2, 11). The commonly isolated anaerobic bacteria are the members of the Bacteroides fragilis group and Peptostreptococcus, Clostridium, and Fusobacterium species (3, 6, 20).
Most clinical laboratories use differential biochemical tests for the identification of anaerobic microorganisms (35). However, Simmon et al. (31) found that 24% of the isolates of anaerobic bacteria recovered from blood cultures were misidentified and that 10% isolates were not identified to the species level by phenotypic characteristics. A rapid commercial kit, the Rapid ID 32A kit (bioMérieux, Marcy l'Etoile, France), was evaluated for its ability to identify strains in the Bacteroides fragilis group. The results showed that only 78.4% of the strains were correctly identified to the species level without supplemental tests (15). The success of the Rapid ID 32A system for species identification varied with different taxa (10), and a low identification rate (50%) was observed for fusobacteria (16). Veillonella isolates are relatively easily identified to the genus level, but the differentiation of Veillonella isolates at the species level remains difficult and inconclusive due to the lack of discriminatory tests (14). In recent years, increasing antimicrobial resistance for some anaerobic bacteria (1, 13, 33) were noted, especially for species in the B. fragilis group (40). The rapid identification of anaerobic bacteria and the administration of appropriate antimicrobials play crucial roles in preventing mortality and morbidity in patients (6).
Molecular methods have emerged as accurate alternatives for the identification of anaerobic bacteria (21, 22, 34, 36). Approximately 9% isolates of bacteremic anaerobes could not be identified to the species level by 16S rRNA gene sequencing, although all isolates were correctly assigned to the genus level (31). Other molecular identification methods targeting the rRNA operon include PCR (32), real-time PCR (26), PCR-restriction fragment length polymorphism analysis (39), and matrix-assisted laser desorption ionization-time-of-flight mass spectrometry (37).
The intergenic spacer (ITS) region separating the 16S and 23S rRNA genes has been suggested to be a good candidate for use for the identification of aerobic and anaerobic bacteria (8, 19, 42). Moreover, the DNA array technology has been applied to the identification of a variety of microorganisms (12, 17, 41). The aim of the study described here was to develop an oligonucleotide array based on the ITS sequences to identify 28 clinically important species of anaerobes and Veillonella.
A collection of 310 target strains (122 reference strains and 188 clinical isolates) representing 28 species of anaerobic bacteria and Veillonella spp. was analyzed (Table (Table1).1). Reference strains were obtained from the American Type Culture Collection (ATCC; Manassas, VA); the Belgian Coordinated Collections (BCCM/LMG; Ghent, Belgium); the Bioresources Collection and Research Center (BCRC; Hsinchu, Taiwan); the Culture Collection, University of Göteborg (CCUG; Göteborg, Sweden); and the Japan Collection of Microorganisms, Riken BioResource Center (JCM; Saitama, Japan). Clinical isolates, identified by use of the Rapid ID 32A system, were obtained from the National Cheng Kung University Hospital (Tainan, Taiwan) and the National Taiwan University Hospital (Taipei, Taiwan). In addition, 98 nontarget strains (51 species) were used for specificity testing of the oligonucleotide array (see Table S1 in the supplemental material). All anaerobic bacteria were cultured on CDC anaerobe 5% sheep blood agar (BBL, Becton Dickinson Microbiology Systems, Cockeysville, MD) and incubated in an anaerobic chamber at 35°C, while aerobic and facultative anaerobic bacteria were cultured on blood agar or chocolate agar plates and incubated in ambient air at 35°C.
The boiling method was used to extract DNA from the bacteria (24). The ITS sequences of some anaerobes were determined in this study and were submitted to GenBank (Table (Table2).2). The bacterium-specific universal primers 2F (5′-TTGTACACACCGCCCGTC-3′) and 10R (5′-TTCGCCTTTCCCTCACGGTA-3′) were used to amplify the ITS regions, as described previously (41). The TOPO TA cloning kit (Invitrogen, Carlsbad, CA) was used for cloning of the ITS region for species that possessed multiple ITS fragments with different lengths and sequences, according to the manufacturer's instructions. The ITS fragments of positive clones were amplified by PCR and sequenced (41).
Thirty-five oligonucleotide probes (18- to 30-mers) (Table (Table2)2) were designed to identify the anaerobic bacteria listed in Table Table1.1. These probes included 33 species- and group-specific probes and 2 positive control probes (designed from the 3′ ends of the 16S rRNA genes). Each probe except the positive control was spotted on the array as a single dot; the positive control dot contained a mixture of two probes at equal concentrations (Table (Table2).2). Ten or 15 additional bases of thymine were added to the 3′ end of each probe to increase the hybridization signal (7). An irrelevant probe (code M) (5′-digoxigenin-GCATATCAATAAGCGGAGGA-3′) labeled with digoxigenin at the 5′ end was used as a position marker on the array (Fig. (Fig.1).1). The oligonucleotide probes were diluted with a tracking dye, drawn into wells of a 96-well microtiter plate, and spotted onto positively charged nylon membranes (Roche, Mannheim, Germany), as described previously (41). The arrays (0.7 by 0.7 cm, 7 by 7 dots) were fabricated with an automatic arrayer (model SR-A300; EZlife Technology Co., Taipei, Taiwan) by use of a solid pin (diameter, 400 μm). The layout of the different probes on the array is shown in Fig. Fig.11.
The ITS region of the test bacterium was amplified by PCR with primer pair 2F and 10R, with each primer being labeled with a digoxigenin molecule at the 5′ end. The reagents and procedures used for prehybridization, hybridization (50°C for 90 min), and color development with enzyme-conjugated antidigoxigenin antibodies were described previously (41). The hybridized spots (diameter, 400 μm) could be read by the naked eye. A strain was identified as one of the species listed in Table Table11 when both the positive control probe and the species-specific probe (or at least one of the two probes designed to be specific for a species) were hybridized (Table (Table2).2). Identification was determined to the species level; subspecies-level identification was not considered.
In cases in which the result of array identification did not correspond with the original species name of a strain, the test with the Rapid ID 32A system was repeated to check the species name of the strain. If the result of one of the two tests with the Rapid ID 32A system agreed with that provided by the array, a concordant identification was considered for the strain. If the discrepancy continued to exist, the identity of the strain was determined by sequencing nearly the complete length of the 16S rRNA gene (27). The sequences determined were used for a BLAST search of the sequences in public databases. The following criteria were used for identification: (i) when the comparison of the sequence determined with a best-scoring reference sequence of a classified species yielded an identity of ≥99%, the isolate was assigned to that species; and (ii) when the identity was <99% and ≥95%, the isolate was assigned to the corresponding genus (4). When discrepant identification occurred, the result of 16S rRNA gene sequencing was considered the final identification.
Sensitivity was defined as the number of target strains correctly identified (true positives) by the array divided by the total number of target strains tested. Specificity was defined as the number of nontarget strains producing negative hybridization reactions (true negatives) divided by the total number of nontarget strains tested (23). The detection limit was the smallest amount of bacterial DNA that could be detected by the array. Serial 10-fold dilutions of DNAs of Peptostreptococcus anaerobius CCUG 7835, Fusobacterium nucleatum subsp. fusiforme JCM 11024, and Bacteroides fragilis CCUG 4856 were used to determine the detection limit.
A total of 49 clinical specimens were analyzed by use of the array, and the results were compared to those obtained by the conventional methods. The specimens included cerebrospinal fluid (7 samples), pleural effusion (10 samples), ascitic fluid (10 samples), synovial fluid (7 samples), bile (9 samples), pus (5 samples), and tissue (1 samples). All specimens were obtained from National Cheng Kung University Hospital. DNA was extracted from the clinical specimens with DNeasy blood and tissue kit (Qiagen, Hilden, Germany), and the ITS regions were amplified by seminested PCR. The first amplification was conducted with primers 11F (5′-GTTTGATCCTGGCTCAG-3′) and 10R (5′-TTCGCCTTTCCCTCACGGTA-3′). PCR was performed in a reaction volume of 25 μl consisting of 10 mM Tris-HCl (pH 8.8), 50 mM KCl, 1.5 mM MgCl2, 0.6 U GoTaq HotStart polymerase (Promega, Madison, WI), 0.8 mM deoxyribonucleoside triphosphates (0.2 mM each), and 1 μM (each) primer. The thermocycling conditions were as follows: initial denaturation at 95°C for 3 min; 25 cycles of denaturation (94°C, 1 min), annealing (55°C, 1 min), and extension (72°C, 1.5 min); and a final extension step at 72°C for 7 min. One microliter from the first reaction was then used for the second run of the PCR by use of the same thermocycling conditions, except that primer pair 2F (5′-TTCTACACACCGCCCGTC-3′) and 10R was used, each primer was labeled with a digoxigenin molecule at the 5′ end, and 35 cycles of amplification were performed.
The ITS sequences of some anaerobes determined in this study were submitted to GenBank, and the accession numbers are reported in Table Table22.
A total of 35 probes with high degrees of sensitivity and specificity were spotted onto the array (Table (Table2).2). For most species, a single probe was designed to identify an individual species, but two probes were used to identify each of the following species due to intraspecies ITS sequence variations in these species: Anaerococcus prevotii Anaerococcus tetradius, Bacteroides fragilis, Finegoldia magna, Peptoniphilus asaccharolyticus, and Propionibacterium acnes (Table (Table2).2). It should be noted that 1- or 2-base mismatches were intentionally incorporated into some probes to eliminate the nonspecific hybridization caused by some nontarget bacteria (Table (Table22).
Anaerococcus prevotii and A. tetradius had almost identical ITS sequences (data not shown), and two probes (codes Apre4 and Apre5) were used to identify the two species as a group (Table (Table22 and Fig. Fig.2).2). Fusobacterium varium produced weak cross-hybridization with the probe (code Fnec1-2) targeting Fusobacterium necrophorum (Fig. (Fig.2,2, chip 23). Therefore, a strain was identified as F. varium if the species-specific probe (code Fvar3) was hybridized, regardless of whether the F. necrophorum probe (code Fnec1-2) was hybridized. Different Veillonella species had very high ITS sequence similarities (data not shown), and hence, only a genus-specific probe (code Vpar1R) was designed to identify the genus. All five species of Veillonella (Veillonella atypica, V. dentocariosi, V. dispar, V. montpellierensis, and V. parvula) hybridized to the genus-specific probe (Fig. (Fig.2).2). Since subspecies-level identification was not considered, all subspecies of Fusobacterium nucleatum (F. nucleatum subsp. animalis, F. nucleatum subsp. fusiforme, F. nucleatum subsp. nucleatum, F. nucleatum subsp. polymorphum, and F. nucleatum subsp. vincentii) hybridized to a single probe (code Fnuc6R) (Fig. (Fig.22).
The hybridization patterns of 28 anaerobic bacterial species and Veillonella species are shown alphabetically in Fig. Fig.2.2. Of 122 reference strains analyzed, 118 hybridized to the respective probes and were correctly identified, but 4 strains produced discrepant identifications by use of the array (Table (Table3).3). The array identified Bacteroides ovatus CCUG 35192 as Bacteroides thetaiotaomicron, Propionibacterium acnes CCUG 4945 as Propionibacterium granulosum, and Propionibacterium propionicus LMG 16717 as Propionibacterium acnes (Table (Table3).3). Sequence analyses of the 16S rRNA genes clearly confirmed the accurate identifications made by the array. Another discrepant reference strain (Peptostreptococcus anaerobius CCUG 49327) was not identified by the array; however, 16S rRNA gene sequencing revealed that the strain was Peptostreptococcus stomatis, a nontarget species in this study. In brief, 121 of the 122 reference strains were correctly identified to the species level by the array, with the remaining strain (CCUG 49327) being a nontarget species (P. stomatis). Therefore, the sensitivity of the array for the identification of reference strains was 100% (121/121).
Of 188 target clinical isolates, the array yielded 20 discrepant identifications. Repeat testing of these discordant strains with the Rapid ID 32A system reduced the number of discordant strains to 11 (Table (Table3).3). Of these 11 strains, four (Anaerococcus prevotii A776-1, Anaerococcus prevotii C270, Fusobacterium nucleatum 365, and Fusobacterium nucleatum 5084N2) were correctly identified by the array, as evidenced by the results of 16S rRNA gene sequencing. One target isolate (Finegoldia magna D752-3) was not identified. The remaining six strains (Anaerococcus prevotii C400-3, Fusobacterium nucleatum K789-2, Peptoniphilus asaccharolyticus B619-2, Peptoniphilus asaccharolyticus C302-2, Peptoniphilus asaccharolyticus 3480N, and Parvimonas micra I503) were actually nontarget isolates. Among these six strains, strains C400-3 and K789-2 were misidentified as Finegoldia magna and Fusobacterium necrophorum, respectively, and the remaining four strains were not identified by the array. In addition, one nontarget isolate (Bacteroides caccae 483) was identified as B. fragilis by the array, and the identification was confirmed by 16S rRNA gene sequencing (Table (Table3).3). In summary, the total number of target clinical isolates was 183 (188 − 6 + 1) and 182 isolates were correctly identified, resulting in a test sensitivity of 99.5% (182/183). If reference strains and clinical isolates were taken together, the sensitivity of the array for the identification of anaerobic bacteria was 99.7% [(182 + 121)/(183 + 121)].
A collection of 98 nontarget strains (51 species), including anaerobic and aerobic bacteria, were used for specificity testing (see Table S1 in the supplemental material). Bacteroides caccae L117 was misidentified as B. vulgatus by the array; however, the strain was determined to be B. dorei by its 16S rRNA gene sequence (Table (Table3).3). In addition to the 98 clinical nontarget isolates, another 7 strains (Peptostreptococcus anaerobius CCUG 49327; Anaerococcus prevotii C400-3; Fusobacterium nucleatum K789-2; Peptoniphilus asaccharolyticus B619-2, C302-2, and 3480N; and Parvimonas micra I503), initially included as target strains, were found to be nontarget microorganisms through discrepant analysis (Table (Table3).3). Among the seven nontarget strains, two (C400-3 and K789-2) were misidentified as target species. On the contrary, one nontarget isolate (Bacteroides caccae 483) was found to be B. fragilis, a target isolate. Therefore, a total of 104 (98 + 7 − 1) nontarget strains were analyzed by the array and three strains (C400-3, K789-2, and L117) were misidentified, resulting in an identification specificity of 97.1% (101/104). The detection limits of the array ranged from 10 fg (Peptostreptococcus anaerobius CCUG 7835 and Fusobacterium nucleatum subsp. fusiforme JCM 11024) to 100 fg (Bacteroides fragilis CCUG 4856) per assay. If a bacterial cell has about 4 fg of DNA (18), the detection limits of the array were from 3 to 30 cells per assay.
A total of 49 clinical specimens were analyzed by the array. The array detected two species (Finegoldia magna and Bacteroides vulgatus) in two pus samples, and the results corresponded to those obtained by culture.
In this study, an oligonucleotide array was developed to identify 28 species of clinically important anaerobic bacteria and Veillonella spp. The sensitivity and the specificity of the array were 99.7 and 97.1%, respectively.
Four strains (Bacteroides ovatus CCUG 35192, Peptostreptococcus anaerobius CCUG 49327, Propionibacterium acnes CCUG 4945, and Propionibacterium propionicus LMG 16717) from two strain collection centers were given the wrong species names, as evidenced by the results obtained with the array and by 16S rRNA gene sequencing (Table (Table3).3). Two of the four strains belonged to the genus Propionibacterium, which confirms that the phenotypic identification of the propionibacteria is still problematic and that alternative identification techniques are required for this genus (28). Three Propionibacterium species, i.e., P. acnes, P. granulosum, and P. propionicus, were included in this study and were well differentiated from one another by the array (Fig. (Fig.22).
Bacteria in the Bacteroides fragilis group are important pathogens in polymicrobial infections. The group includes B. fragilis, B. thetaiotaomicron, B. vulgatus, B. ovatus, B. distasonis, B. uniformis, and other species, with B. thetaiotaomicron being much more resistant to many antimicrobials (5, 9). The group accounted for as many as 61% of the anaerobic isolates recovered from blood cultures (3). The members of the group are phenotypically very similar and are frequently misidentified by biochemical tests. Since different species in the B. fragilis group vary in their resistance to antimicrobial agents (40), it is important to differentiate species in the group, especially for severe infections (44). The taxonomy of Bacteroides has undergone major revisions in the last few decades, with more than 20 species now being included in the genus (29, 43). In this study, oligonucleotide probes were successfully applied to differentiate the five important members (B. fragilis, B. ovatus, B. thetaiotaomicron, B. uniformis, and B. vulgatus) of the B. fragilis group (Fig. (Fig.22).
Gram-positive anaerobic cocci are a heterogeneous group of organisms, with the different species displaying major differences in antimicrobial susceptibility patterns (25). In this study, three isolates of Anaerococcus prevotii (isolates A776-1, C270, and C400-3) were found to be misidentifications of Peptoniphilus asaccharolyticus or Finegoldia species (Table (Table3).3). In addition, three clinical isolates (isolates B619-2, C302-2, and 3480N) of Peptoniphilus asaccharolyticus were not identified to the species level by the array or by 16S rRNA gene sequencing, suggesting that the three isolates were misidentified by conventional biochemical tests (Table (Table3).3). Since the three Peptoniphilus isolates were recovered from blood cultures, their significance and real identities warrant further investigation.
Fusobacterium nucleatum and F. necrophorum can cause manifold infections, such as periodontitis, organ abscesses, and bacteremia (30). Two clinical F. nucleatum isolates (isolates 365 and 5084N2) were misidentifications of F. necrophorum, as revealed by the array and 16S rRNA gene sequencing (Table (Table3).3). The identification of clinically relevant Fusobacterium spp. is hampered by their slow growth and the low levels of reliability of biochemical tests (30, 38). In the present study, F. necrophorum was well differentiated from F. nucleatum, including several subspecies (Table (Table11 and Fig. Fig.22).
The array was also assessed for its ability to directly detect anaerobic bacteria in 49 clinical specimens. Two species (Finegoldia magna and Bacteroides vulgatus) were detected in two specimens by the array, and the results were in accordance with those obtained by culture. However, instead of PCR, nested PCR was required to produce enough amplicons for hybridization. These results indicate that the array may have the potential to detect anaerobic bacteria in clinical specimens. However, the number and types of clinical specimens tested in this study were limited, and further comprehensive evaluation is needed to validate this potential. The low rate of detection (4.1%) of anaerobic bacteria in these samples might be due to the use of a high percentage of sterile body specimens.
In conclusion, identification of the species of clinically relevant anaerobes by use of the array described here is highly reliable. The method could be used as an accurate alternative to the conventional methods if adequate species identification is of concern. The whole procedure of array hybridization took about 8 h, starting with the isolated colonies.
This study was supported by grants from the National Science Council (grant 96-2320-B-006-024-MY3), the Center for Frontier Materials and Micro/Nano Science and Technology (grant D97-2720), National Cheng Kung University, and the Department of Health (grant DOH-99-TD-B-111-002), Taiwan, Republic of China.
Published ahead of print on 3 February 2010.
†Supplemental material for this article may be found at http://jcm.asm.org/.