P. P. Bosshard,* R. Zbinden, S. Abels, B. Böddinghaus, M. Altwegg, and E. C. Böttger

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, Apr. 2006, p Vol. 44, No /06/$ doi: /jcm Copyright 2006, American Society for Microbiology. All Rights Reserved. 16S rrna Gene Sequencing versus the API 20 NE System and the VITEK 2 ID-GNB Card for Identification of Nonfermenting Gram-Negative Bacteria in the Clinical Laboratory P. P. Bosshard,* R. Zbinden, S. Abels, B. Böddinghaus, M. Altwegg, and E. C. Böttger Institut für Medizinische Mikrobiologie, Universität Zürich, 8006 Zürich, Switzerland Received 13 December 2005/Returned for modification 9 January 2006/Accepted 24 January 2006 Over a period of 26 months, we have evaluated in a prospective fashion the use of 16S rrna gene sequencing as a means of identifying clinically relevant isolates of nonfermenting gram-negative bacilli (non-pseudomonas aeruginosa) in the microbiology laboratory. The study was designed to compare phenotypic with molecular identification. Results of molecular analyses were compared with two commercially available identification systems (API 20 NE, VITEK 2 fluorescent card; biomérieux, Marcy l Etoile, France). By 16S rrna gene sequence analyses, 92% of the isolates were assigned to species level and 8% to genus level. Using API 20 NE, 54% of the isolates were assigned to species and 7% to genus level, and 39% of the isolates could not be discriminated at any taxonomic level. The respective numbers for VITEK 2 were 53%, 1%, and 46%, respectively. Fifteen percent and 43% of the isolates corresponded to species not included in the API 20 NE and VITEK 2 databases, respectively. We conclude that 16S rrna gene sequencing is an effective means for the identification of clinically relevant nonfermenting gram-negative bacilli. Based on our experience, we propose an algorithm for proper identification of nonfermenting gram-negative bacilli in the diagnostic laboratory. Gram-negative nonfermenters are primarily opportunistic bacteria ubiquitously present in the environment, causing infections mainly in severely ill and immunocompromised patients. Many of these organisms have become problematic in the hospital in part due to their ability to survive in various habitats, including, e.g., aqueous, moist environments (e.g., Pseudomonas spp.) or dry surfaces (e.g., Acinetobacter spp.) (1). Pseudomonas aeruginosa is a leading cause of nosocomial infections and the most frequent agent of infections due to gram-negative nonfermenters, followed by Acinetobacter spp., Stenotrophomonas maltophilia, and Alcaligenes spp. (23). Variations in drug susceptibility are common among these pathogens (23, 29); in addition, nonfermenting, gram-negative bacilli also differ in terms of pathogenic potential and transmissibility. Identification to species level is thus required for proper clinical management of patients (11). Except for Pseudomonas aeruginosa, accurate identification of gram-negative nonfermenters in the clinical laboratory mainly relies on commercially available phenotypic identification systems. Commercial systems for bacterial identification have been offered for approximately 30 years. Of the original manual systems, only the API 20 NE (biomérieux, Marcy l Etoile, France) has remained available, while fully or partly automated identifications systems, such as VITEK (biomérieux, Marcy l Etoile, France), PHOENIX (BD, Sparks, Md.), and MicroScan (Dade Behring, West Sacramento, Calif.), have now been implemented in many laboratories. These systems have contributed to a more effective management of patients by enabling microbiologists to identify corresponding bacteria * Corresponding author. Mailing address: Institut für Medizinische Mikrobiologie, Universität Zürich, Gloriastrasse 30, CH-8006 Zürich, Switzerland. Phone: Fax: philboss@immv.unizh.ch. Present address: Bio-Analytica AG, 6000 Luzern 6, Switzerland. of clinical relevance more rapidly and accurately (12). However, all phenotypic test systems have potential inherent problems, e.g., (i) not all strains within a given species may exhibit a particular characteristic (6), (ii) the same strain may give different results upon repeated testing (6), and (iii) the corresponding databases are limited (2). Specifically, nonfermenters recovered from cystic fibrosis patients may pose problems due to phenotypic variations and slower growth rates because of the significant antimicrobial pressure that these organisms face in the lungs of these patients (17). Genotypic identification methods are emerging as an alternative or complement to established phenotypic identification procedures. For bacteria, 16S rrna gene sequence analysis is a widely accepted tool for molecular identification (5, 18, 21). Public databases (GenBank, Nucleotide Sequence Database at the European Molecular Biology Laboratory, DNA Data Bank of Japan, RDP II) contain a vast number of bacterial 16S rrna sequences, allowing for rapid analysis and providing phylogenetically meaningful information. Few studies so far have systematically compared molecular and phenotypic identification procedures to determine the usefulness of sequencebased methods for the diagnostic laboratory (2, 3, 8, 10, 11, 16, 22, 24, 25, 27, 28, 30); available studies mainly focused on mycobacteria (8, 16, 22, 25) and gram-positive microorganisms (2, 3, 28). Several limitations are present in those rare studies investigating aerobic gram-negative rods: one study was restricted to isolates which phenotypically gave problematic results (11), one study was restricted to isolates from cystic fibrosis patients (30), and one study was restricted to the commercially available MicroSeq system (27). Given these limitations, and with a view toward developing a diagnostic algorithm for implementation into the microbiology laboratory, we here compared in a prospective fashion phenotypic systems (API 20 NE, VITEK 2) with 16S rrna gene sequencing for 1359

2 1360 BOSSHARD ET AL. J. CLIN. MICROBIOL. identification of clinically relevant isolates of aerobic nonfermenting gram-negative bacilli. MATERIALS AND METHODS Clinical isolates. The study was designed to prospectively compare phenotypic with molecular identification for clinically relevant isolates of aerobic nonfermenting gram-negative (non-pseudomonas aeruginosa) rods. The isolates (n 107) investigated were recovered from blood cultures and from relevant clinical material where identification was required. During the study period of 26 months, a total of 2,653 aerobic gram-negative nonfermenters were isolated in our laboratory, including 1,893 strains of Pseudomonas aeruginosa; of the 760 non-p. aeruginosa isolates, 107 strains were found to meet the criterion of clinical relevance and were thus included in the study. Identification using the API 20 NE system. The API 20 NE system covers 61 nonenterobacterial gram-negative taxa. Testing was performed according to the instructions of the manufacturer (biomérieux, Marcy l Etoile, France). Substrate assimilations were read after 24 and 48 h. Interpretation of the results was done after 48 h using the identification software version 6.0. Strains were classified into one of the following 3 groups: (i) identification at species level, (ii) identification at genus level, (iii) no identification (i.e., low discrimination). According to the manufacturer s instructions, strain identification at the species level was divided into 4 subgroups: (i) species identification, percent identification of 99.9% and T value of 0.75; (ii) very good species identification, percent identification of 99.0% and T value of 0.5; (iii) good species identification, percent identification of 90.0% and T value of 0.25; (iv) acceptable species identification, percent identification of 80.0% and T value of 0.0. Identification using the VITEK 2 fluorescent system (ID-GNB card). The VITEK 2 fluorescent system (ID-GNB card) includes 43 nonenterobacterial gram-negative taxa. Testing was performed according to the instructions of the manufacturer. Briefly, strains were cultured on Columbia sheep blood agar (Difco ) or MacConkey agar (BBL ) for 18 to 24 h at 37 C before the isolate was subjected to analysis. Strains which had been stored at 70 C were subcultured twice before analysis. A bacterial suspension was adjusted to a McFarland standard of 0.50 to 0.63 in a solution of 0.45% sodium chloride using the VITEK 2 DensiCheck instrument (biomérieux). The time between preparation of the solution and filling of the card was always less than 1 h. Analysis was done using the identification card for gram-negative bacteria (ID-GNB card) containing 41 fluorescent biochemical tests (13). Cards are automatically read every 15 min. Data were analyzed using the VITEK 2 software version VT2- R03.1. Isolates which were not identified were retested with a fresh subculture. According to the manufacturer s instructions, strain identification at the species level was divided into 4 subgroups: (i) species identification, T value of 0.75; (ii) very good species identification, T value of 0.5 and 0.75; (iii) good species identification, T value of 0.25 and 0.5; (iv) acceptable species identification, T value 0.0 and Sequencing of 16S rrna gene. Sequencing was performed as described previously (2). In brief, a loopful of bacterial cells was digested using lysozyme (Sigma-Aldrich Chemie GmbH, Schnelldorf, Germany) and alkaline lysis. Following nucleic acid purification, the 5 part of the 16S rrna gene (corresponding to Escherichia coli positions 10 to 806) was amplified using primers BAK11w [5 -AGTTTGATC(A/C)TGGCTCAG] and BAK2 [5 -GGACTAC(C/T/A)AGG GTATCTAAT] (4). Cycling parameters included an initial denaturation for 5 min at 95 C; 40 cycles of 1 min at 94 C, 1 min at 48 C, and 1 min at 72 C; and a final extension for 10 min at 72 C. Five microliters of the DNA extract was used for amplification in a total volume of 50 l containing 1.25 U of AmpliTaq DNA polymerase LD (Applied Biosystems, Rotkreuz, Switzerland). Amplicons were purified and sequenced with forward primer BAK11w (4), and the fragments were analyzed using an automatic DNA sequencer (ABI Prism 3100 Genetic Analyzer; Applied Biosystems). Sequence analysis. 16S rrna gene sequences were compared with those available in the GenBank, EMBL, and DDBJ databases using a two-step procedure. A first search was performed with the FASTA algorithm of the Wisconsin GCG program package (9). All positions showing differences to the best-scoring reference sequence were visually inspected in the electropherogram, and the sequence was corrected manually if necessary. Thereafter, a second search was done using BLASTN. Undetermined nucleotides (designated N) in either the determined sequence or the reference sequence were counted as matches. The mean length of the sequences after manual editing was nucleotides (range, 315 to 692 nucleotides) containing undetermined (N) positions (range, 0 to 9 N). For identification, the following criteria were used: (i) when the determined sequence yielded a similarity score of 99% with a reference sequence of a TABLE 1. Molecular versus phenotypic identification for 107 isolates (unresolved data) Identification system No. (%) identified at taxonomic level: Species Genus No identification 16S rrna gene sequencing 98 (92) 9 (8) API 20 NE 58 (54) 7 (7) 42 a (39) VITEK 2 57 (53) 1 (1) 49 b (46) a 11 of the 42 isolates not identified belong to a species not included in the API 20 NE database. b 39 of the 49 isolates not identified belong to a species not included in the VITEK 2 ID-GNB database. classified species, the unknown isolate was assigned to this species; (ii) when the score was 99% and 95%, the unknown isolate was assigned to the corresponding genus; and (iii) when the score was 95%, the unknown isolate was assigned to a family. If the unknown isolate was assigned to a species and the second classified species in the scoring list showed less than 0.5% additional sequence divergence, this was categorized as a species with low demarcation to next species. Discrepant analysis. According to Stackebrandt and Goebel (26), 16S rrna gene similarities of less than 97% indicate that strains belong to different species. If the isolate was assigned to a different species by phenotypic and molecular criteria, the procedure was as follows: (i) if the isolate s sequence showed less than 97% similarity to the sequence of the species assigned phenotypically, it was concluded that the isolate does not belong to the species identified by phenotypic means and the sequencing result was considered correct; (ii) if the isolate s sequence showed 97% to 99% similarity to the sequence of the species assigned by phenotypic means, the isolate was categorized as unresolved. RESULTS Identification by ribosomal RNA gene sequencing. During the study period, a total of 107 clinically relevant isolates of aerobic nonfermenting gram-negative rods (non-pseudomonas aeruginosa) were recovered. The isolates are representatives of 13 different genera. Using the defined criteria, 16S rrna gene sequencing resulted in the assignment of 98 isolates to the species level and 9 isolates to the genus level (Table 1). In 12 of the 98 isolates identified at the species level, sequence comparison with public databases resulted in retrieval of two sequences of different species which exhibited identical similarity scores; thus, the isolate was not assigned to a single taxon but was reported as belonging to either of the two species. As an example, the sequence of an isolate showed 100.0% identity to sequences of Pseudomonas fluorescens and Pseudomonas jessenii. Twenty-five of the 98 isolates identified at the species level were identified as a species with low demarcation to the next species, i.e., less than 0.5% additional sequence difference to another sequence entry. As an example of low demarcation, the sequence of an isolate showed 99.8% similarity to a sequence of Achromobacter xylosoxidans and 99.4% similarity to an Achromobacter piechaudii sequence. Thus, in total, 37 isolates could not be assigned unambiguously to a single species. Fourteen of these isolates were members of the Burkholderia cepacia complex, 9 isolates belonged to the Acinetobacter baumannii-acinetobacter calcoaceticus complex (14) and close relatives, 8 isolates belonged to the genus Pseudomonas, 4 isolates were Achromobacter xylosoxidans, and 2 isolates were strains of the genus Ralstonia (for further details, see Tables 2 and 3). Identification using API 20 NE. API 20 NE identified 58 of the 107 isolates to the species level, yielding, very good, good, and acceptable species identifications in 20, 18, 15,

3 VOL. 44, 2006 IDENTIFICATION OF GRAM-NEGATIVE BACTERIA BY SEQUENCING 1361 TABLE 2. Eighty-eight isolates for which molecular and phenotypic identification were concordant Identification by 16S rrna gene sequencing (n) No. of isolates API 20 NE a result VITEK 2 a, b result Identical species assignment by phenotypic and molecular identification (33) Acinetobacter baumannii 1 Acinetobacter baumannii or A. calcoaceticus; Acinetobacter baumannii or A. calcoaceticus subsp. anitratus 3 Acinetobacter baumannii or A. calcoaceticus; (2), very good (1) genosp. 3; genosp. 3; (3) Acinetobacter calcoaceticus 6 Acinetobacter baumannii or A. calcoaceticus; (1), very good (3), good (2) genosp. 3; (5), very good (1) Acinetobacter calcoaceticus or A. haemolyticus 2 Acinetobacter baumannii or A. calcoaceticus; (1), very good (1) genosp. 3; (2) Burkholderia cepacia complex 8 c Burkholderia cepacia; (5), very good (2), good (1) Burkholderia cepacia; (3), very good (2), good (1), acceptable (2) Pseudomonas aeruginosa 1 c Pseudomonas aeruginosa; very good (1), Pseudomonas aeruginosa; (2) acceptable (1) Pseudomonas aeruginosa 1 Pseudomonas aeruginosa; very good (1), Pseudomonas aeruginosa; (2) acceptable (1) Ralstonia pickettii 1 c Ralstonia pickettii; very good Ralstonia pickettii; good Stenotrophomonas maltophilia 10 Stenotrophomonas maltophilia; (6), very Stenotrophomonas maltophilia; (8), very good (2) good (4) Identical species/genus assignment by API 20 NE and molecular identification, molecular identification more discriminative than VITEK 2 (13) Achromobacter xylosoxidans 2 c Alcaligenes xylosoxidans; good (3), acceptable (1) No identification Achromobacter xylosoxidans 2 Alcaligenes xylosoxidans; good (3), acceptable (1) No identification Agrobacterium tumefaciens 1 Agrobacterium radiobacter; good Rhizobium sp. d ( A. radiobacter) Alcaligenes faecalis 1 Alcaligenes faecalis; very good No identification Pseudomonas fluorescens or 1 Pseudomonas fluorescens; No identification P. jessenii Pseudomonas putida or 1 Pseudomonas putida; good Low discrimination (Shigella group or NF e ) P. parafulva Pseudomonas stutzeri 1 Pseudomonas stutzeri; very good Low discrimination (Pseudomonas aeruginosa or NF) Ralstonia pickettii 1 Ralstonia pickettii; good Low discrimination (Ralstonia pickettii or NF) Stenotrophomonas maltophilia 1 Stenotrophomonas maltophilia; Low discrimination (Stenotrophomonas maltophilia or NF) Acinetobacter sp. 1 Acinetobacter sp. No identification Pseudomonas sp. 1 Pseudomonas sp. Low discrimination (Shigella group or NF) Identical species assignment by VITEK 2 and molecular identification, molecular identification more discriminative than API 20 NE (13) Acinetobacter calcoaceticus subsp. anitratus genosp. 3; genosp. 3; genosp. 3; 1 c Low discrimination (Acinetobacter baumannii or A. calcoaceticus) Acinetobacter calcoaceticus or 1 Low discrimination (Acinetobacter baumannii or A. haemolyticus A. calcoaceticus) Acinetobacter calcoaceticus or 1 Unacceptable profile (CDCgr.IVC-2 A. haemolyticus Wautersia paucula) Burkholderia cepacia complex 1 c Low discrimination (Burkholderia cepacia) Burkholderia cepacia; acceptable Burkholderia cepacia complex 1 c Unacceptable profile (Ochrobacter anthropi) Burkholderia cepacia; acceptable Burkholderia cepacia complex 1 c Low discrimination (Pasteurella spp.) Burkholderia cepacia; Burkholderia cepacia complex 1 c Doubtful profile (Pasteurella haemolytica) Burkholderia cepacia; Pseudomonas aeruginosa 1 Low discrimination (Pseudomonas aeruginosa) Pseudomonas aeruginosa; Stenotrophomonas maltophilia 1 Low discrimination (Burkholderia cepacia) Stenotrophomonas maltophilia; Stenotrophomonas maltophilia 2 Unacceptable profile (Burkholderia cepacia or S. maltophilia) Stenotrophomonas maltophilia 2 Unacceptable profile (Brevundimona vesicularis or Shewanella putrefaciens) Molecular identification more discriminative than API 20 NE/VITEK 2 (28) Achromobacter xylosoxidans 1 c Low discrimination (Alcaligenes faecalis) No identification Achromobacter xylosoxidans 1 c Low discrimination (Alcaligenes xylosoxidans) No identification Achromobacter xylosoxidans 1 Low discrimination (Comamonas testosterone or No identification Pseudomonas alcaligenes) Achromobacter xylosoxidans 1 Low discrimination (Moraxella sp.) No identification Achromobacter xylosoxidans 2 Low discrimination (Pasteurella sp.) No identification Achromobacter sp. 1 Low discrimination (Alcaligenes faecalis) No identification Acidovorax temperans 2 Low discrimination (Comamonas testosterone or Pseudomonas alcaligenes) Stenotrophomonas maltophilia; (1), good (1) Stenotrophomonas maltophilia; (1), very good (1) No identification Continued on following page

4 1362 BOSSHARD ET AL. J. CLIN. MICROBIOL. TABLE 2 Continued Identification by 16S rrna gene sequencing (n) No. of isolates API 20 NE a result VITEK 2 a, b result Acidovorax temperans 1 No identification Low discrimination (Chryseobacterium indologenes or NF) Acinetobacter johnsonii 2 Acinetobacter sp. No identification Acinetobacter junii 1 Acinetobacter sp. No identification Acinetobacter lwoffii 1 Acinetobacter sp. No identification Acinetobacter ursingii 1 Acinetobacter sp. No identification Agrobacterium larrymoorei 1 Low discrimination (Ochrobactrum anthropii) No identification Bordetella sp. 1 Low discrimination (Comamonas testosterone No identification or Pseudomonas alcaligenes) Oligella urethralis 1 Brevundimonas diminuta or Oligella ureolytica No identification Oligella urethralis 1 Low discrimination (Comamonas testosterone No identification or Pseudomonas alcaligenes) Pseudomonas aeruginosa 1 Low discrimination (Pseudomonas No identification aeruginosa) Pseudomonas aeruginosa 1 No identification Low discrimination (Bordetella sp. or NF) Pseudomonas mendocina 1 c No identification Low discrimination (Ralstonia pickettii or NF) Pseudomonas pseudoalcaligenes 1 Low discrimination (Comamonas testosterone No identification or P. oleovorans or Pseudomonas alcaligenes) Pseudomonas stutzeri 1 c Low discrimination (Pseudomonas stutzeri) Low discrimination (Pseudomonas aeruginosa or NF) Pseudomonas sp. 1 Low discrimination (Pseudomonas stutzeri) Low discrimination (Ralstonia pickettii or Burkholderia cepacia) Pseudoxanthomonas mexicana 1 Low discrimination (Sphingomonas paucimobilis) Low discrimination (Brevundimonas vesicularis or Stenotrophomonas maltophilia) Wautersia paucula 1 Low discrimination (CDCgr.IVC-2 No identification Wautersia paucula) Wautersia sp. 1 Doubtful profile (CDCgr.IVC-2 Wautersia No identification paucula) API 20 NE more discriminative than molecular identification/vitek 2 (1) Wautersia sp. 1 CDCgr.IVC-2 Wautersia paucula; good No identification a Quality of phenotypic identification; number in parentheses corresponds to number of isolates. In cases of no identification with API 20 NE/VITEK 2, the species given in parentheses corresponds to the systems suggestion. b genosp., genospecies. c These isolates were categorized as species with low demarcation to the next species, i.e., less than 0.5% additional sequence difference from the next best matching sequence entry. d Agrobacterium tumefaciens is reclassified as Rhizobium radiobacter (synonym A. radiobacter) (31). e NF, nonfermentative gram-negative bacteria. Downloaded from and 5 cases, respectively. In 7 cases, identification at the genus level was achieved; 42 isolates could not be identified (Table 1). In 14 isolates identified to the species level, not a single species but a mixed taxon was reported: in 12 cases, Acinetobacter baumannii or A. calcoaceticus (both belonging to the A. baumannii-a. calcoaceticus complex), in 1 case, Acinetobacter junii or Acinetobacter johnsonii, and in 1 case, the Pseudomonas putida group. Identification using VITEK 2. VITEK 2 identified 57 of the 107 isolates to the species level, yielding, very good, good, and acceptable species identifications in 37, 12, 4, and 4 cases, respectively. In 1 case, identification at the genus level was achieved; 49 isolates could not be identified (Table 1). In 12 cases identified at the species level, a mixed taxon was reported, with all of these representing Acinetobacter baumannii or A. calcoaceticus or Acinetobacter genomic species 3 (belonging to the A. baumannii-a. calcoaceticus complex). Comparing 16S rrna gene sequencing with API 20 NE. For 45 of the 58 strains identified to the species level by API 20 NE (for raw data, see Tables 2 and 3; for resolved data, see Table 4), molecular identification assigned the isolate to the same taxon (18/20, 15/18, 10/15, and 2/5 cases with, very good, good, and acceptable species identification by API 20 NE, respectively). In 1/58 isolates, sequencing gave genus level assignment (Wautersia sp.) and the API system resulted in species level identification (Wautersia paucula). Discrepant results were found in 12 of the 58 isolates identified by API 20 NE to species level (Table 3). In 10/12 discrepant cases, 16S rrna gene sequencing assigned the isolate to a different species than API 20 NE; in 2/12 discrepant cases, the isolates were identified to mere genus level by sequencing (the sequences showed less than 99.0% similarity to the bestscoring reference sequence). In 7 of the 12 discrepant cases, the isolates 16S rrna gene sequences exhibited 97% similarity to the 16S rrna gene sequence of the species assigned by API 20 NE. According to Stackebrandt and Goebel (26), 16S rrna gene similarities of less than 97% indicate that strains belong to different species. Although only partial sequences were used here, it was thus concluded that these isolates do not belong to the species identified by the API system. For example, one strain was identified as Burkholderia cepacia by API 20 NE; the strain s 16S rrna gene sequence showed 100.0% similarity with Herbaspirillum huttiense (formerly Pseudomonas huttiensis) but less than 95% similarity with B. cepacia. This isolate clearly does not represent B. cepacia but is H. huttiense. In 4 of these 7 cases, the isolate belongs to a species not covered by the API 20 NE database. In 5/12 discrepant cases, the 16S rrna gene sequence of the isolate exhibited between 97% and 99% sequence similarity with the 16S rrna gene sequence of the taxon assigned by API 20 NE (for further details, see Table 3). We can thus not exclude the possibility that the isolates are representatives of the species identified by the API system (unresolved cases). In on April 25, 2018 by guest

5 VOL. 44, 2006 IDENTIFICATION OF GRAM-NEGATIVE BACTERIA BY SEQUENCING 1363 TABLE 3. Molecular and phenotypic identifications: discrepant results (n 19) 16S rrna gene sequencing Identification (n) % Difference from reference sequence Reference sequence a No. of isolates API 20 NE b, c VITEK 2 b, c Result of discrepant analysis Discrepant results between molecular identification and API 20 NE (8) Achromobacter xylosoxidans 0.2 A. xylosoxidans (Ralstonia pickettii 10.0%) 1 Ralstonia pickettii; very good No identification Achromobacter xylosoxidans Acinetobacter baumannii 0.8 A. baumannii (A. junii 2.3% A. johnsonii 3.5%) 1 A. junii or A. johnsonii; acceptable No identification Unresolved Acinetobacter johnsonii 0.5 A. johnsonii (A. lwoffii 1.0%) 1 Acinetobacter lwoffii; good Low discrimination (Ralstonia pickettii, NF) Acinetobacter venetianus d 0.8 A. venetianus (A. baumannii 2.1%) 1 Acinetobacter baumannii or A. calcoaceticus; good Bordetella petrii 0.4 B. petrii (Achromobacter denitrificans 3.0%) 1 Alcaligenes denitrificans; acceptable Unresolved No identification Unresolved No identification Bordetella petrii Herbaspirillum huttiense 0.0 H. huttiense (Burkholderia cepacia 5.0%) 1 Burkholderia cepacia; No identification Herbaspirillum huttiense Pseudomonas aeruginosa 0.0 P. aeruginosa (Ralstonia pickettii 5.0%) 1 Ralstonia pickettii; very good Low discrimination (Brucella Pseudomonas aeruginosa sp., NF) Ralstonia pickettii or R. detusculanense 0.0 R. pickettii or R. detusculanense (Pseudomonas fluorescens 10.0%) 1 Pseudomonas fluorescens; good Low discrimination (Ralstonia pickettii, NF) Discrepant results between molecular identification and VITEK 2 (7) Acinetobacter lwoffii 0.2 A. lwoffii (A. baumannii 2.8%) 1 Low discrimination (Ochrobactrum anthropi or Sphingomonas paucimobilis) Burkholderia cepacia complex d 0.0 B. cepacia (Pseudomonas aeruginosa 10.0%) 1 Low discrimination (Pseudomonas fluorescens) Burkholderia cepacia complex d 0.0 B. cepacia (B. pseudomallei 2.0%) 1 Low discrimination (Pasteurella sp.) Pseudomonas fluorescens or P. veronii or P. marginalis 0.0 P. fluorescens or P. veronii or P. marginalis (P. aeruginosa 5.0%) Pseudomonas monteilii 0.0 P. monteilii (Ralstonia sp. 5.0%) 1 Low discrimination (Pseudomonas putida) Acinetobacter baumannii or A. calcoaceticus or Acinetobacter genosp. 3; Pseudomonas aeruginosa; very good Burkholderia pseudomallei; 1 Pseudomonas fluorescens; good Pseudomonas aeruginosa; Pseudomonas sp. 1.1 P. putida (Ralstonia pickettii 4.0%) 1 Unacceptable profile (Flavimonas oryzihabitans) Pseudoxanthomonas mexicana 0.0 P. mexicana (Brevundimonas vesicularis 10.0%) 1 Low discrimination (Stenotrophomonas maltophilia) Discrepant results between molecular identification and phenotypic identification (4) Alcaligenes sp. 0.2 Alcaligenes sp. (Ralstonia pickettii 5.0%, Pseudomonas aeruginosa 5.0%) Ralstonia sp. Unresolved Burkholderia cepacia complex Burkholderia cepacia complex P. fluorescens Ralstonia pickettii; very good P. monteilii Ralstonia pickettii; very good Pseudomonas sp. Brevundimonas vesicularis; good 1 Ralstonia pickettii; very good Pseudomonas aeruginosa; Burkholderia gladioli 0.2 B. gladioli (B. cepacia 2.2%) 1 Burkholderia cepacia; very good Burkholderia cepacia; very good P. mexicana Alcaligenes sp. Unresolved Chryseobacterium sp. 2.9 Chryseobacterium proteolyticum (C. indologenes 3.2%) 1 Chryseobacterium indologenes; Chryseobacterium Chryseobacterium sp. very good indologenes; very good Pseudomonas gingeri d 0.7 P. gingeri (P. putida 1.3%, Ralstonia sp. 5.0%) 1 Pseudomonas putida group Ralstonia pickettii; very good Pseudomonas sp. e a The best matching reference sequence is given. For comparison, the sequence homology to the species identified by API 20 NE/VITEK 2 is given in parentheses. b Identification and quality of phenotypic identification. c In cases of no identification with API 20 NE/VITEK 2, the species given in parentheses corresponds to the system s suggestion. d These isolates were categorized as species with low demarcation to the next species, i.e., less than 0.5% additional sequence difference the next best matching sequence entry. e This case is unresolved when 16S rrna gene sequencing and API 20 NE are compared; the 16S rrna gene sequence of the isolate exhibited 97% sequence similarity with the 16S rrna gene sequence of the taxon assigned by API 20 NE. When sequencing is compared to VITEK 2, this case is resolved; the sequence of the isolate exhibited 97% similarity with the 16S rrna gene sequence of the taxon assigned by VITEK 2.

6 1364 BOSSHARD ET AL. J. CLIN. MICROBIOL. TABLE 4. Molecular identification versus identification by API 20 NE (resolved data) Taxonomic level API 20 NE 16S rrna gene sequencing (no. [%]) No. of isolates investigated Identical results Results more discriminative Discrepant results Species (76) 13 a (24) Genus 7 2 (29) 5 (71) No identification 42 b 42 c (100) a In 7 cases, sequencing yielded the correct result; in 1 case, API 20 NE resulted in species assignment and sequencing resulted in genus assignment; five cases remained unresolved. b 11 of 42 isolates not identified belong to a species not included in the API 20 NE database. c By 16S rrna gene sequencing, 37 isolates were assigned to species level and 5 isolates were assigned to genus level. one of these 5 cases, sequence analysis assigned the isolate to a species not included in the API 20 NE database. Using API 20 NE, 7/107 strains investigated were identified at the genus level and 42/107 isolates could not be identified at any taxonomic level (Table 1). 16S rrna gene sequencing allowed species assignment in 5 of the former and 37 of the latter cases (for raw data, see Tables 2 and 3; for resolved data, see Table 4). The remaining 7 strains were identified at the genus level by molecular means. Comparing 16S rrna gene sequencing with VITEK 2. Molecular identification assigned the isolate to the same taxon in 46 of 57 strains identified to the species level by VITEK 2 (33/37, 6/12, 3/4, and 4/4 cases with, very good, good, and acceptable species identification by the VITEK 2, respectively) (for raw data, see Tables 2 and 3; for resolved data, see Table 5). Eleven of the 57 results were discrepant (Table 3). In 3/11 discrepant results, the isolates were identified merely to genus level by sequencing (the sequences showed less than 99.0% similarity to the best-scoring reference sequence); in 8/11 discrepancies, sequencing assigned the strains to a different species compared to VITEK 2. The isolates 16S rrna sequences had less than 97% similarity compared to the 16S sequences of the species assigned by VITEK 2 in 8 of the 11 discrepant results; we concluded that these isolates (n 8) were misidentified by VITEK 2 (in 5 of these 8 cases, the isolates belong to a species not covered by the VITEK 2 database). In 3 of the 11 isolates, the determined sequence had a similarity between 97% and 99% with the sequence of the species assigned by VITEK 2 (for further details, see Table 3). In one of these three cases, the result of VITEK 2 was considered wrong (Burkholderia cepacia complex instead of Burkholderia pseudomallei), and 2 cases remained unresolved (in these 2 cases, sequence analysis assigned the isolates to a species not covered by the VITEK 2 database). Using VITEK 2, one strain was identified at the genus level; 16S rrna gene sequencing allowed species identification of this isolate. Forty-nine of 107 isolates could not be identified at any taxonomic level using VITEK 2; sequencing allowed species assignment in 42 and genus assignment in 7 of these 49 isolates (for raw data, see Tables 2 and 3; for resolved data, see Table 5). TABLE 5. Molecular identification versus identification by VITEK 2 (resolved data) Taxonomic level VITEK 2 16S rrna gene sequencing (no. [%]) No. of isolates investigated Identical results DISCUSSION Results more discriminative Discrepant results Species (81) 11 a (19) Genus 1 1 (100) No identification 49 b 49 c (100) a In 8 cases, sequencing yielded the correct result; three cases remained unresolved. b 39 of 49 isolates not identified belong to a species not included in the VITEK 2 ID-GNB database. c By 16S rrna gene sequencing, 42 isolates were assigned to species level and 7 isolates were assigned to genus level. In this prospective study, we have examined the suitability of partial 16S rrna gene sequencing for the identification of aerobic nonfermenting gram-negative bacilli in the diagnostic laboratory. The study was designed to compare phenotypic with molecular identification for nonfermenting, gram-negative rod-shaped isolates of clinical relevance (isolates of P. aeruginosa were excluded from the study). Results of phenotypic identification (API 20 NE and VITEK 2 fluorescent system) were compared to the results of sequencing. 16S rrna gene sequencing is more accurate for the identification of gram-negative nonfermenters than API 20 NE and VITEK 2: 92% of isolates were assigned at the species level by sequencing, compared to 54% and 53% by API and VITEK 2, respectively. In 12 of 45 (API 20 NE) and 11 of 57 (VITEK 2) isolates with identification to the species level by phenotypic procedures, sequence analysis yielded discrepant results. For API 20 NE, we concluded that sequencing was correct for 7/12 strains, and 5/12 strains remained unresolved. For VITEK 2, it was concluded that the molecular approach delivered the correct result for 9/11 strains, and 2 strains remained unresolved. For the API 20 NE, a result seems reliable when a species assignment is reported as an or very good identification; 18 of 20 and 15 of 18 cases in these categories were correctly identified when compared to the results of sequencing. In the categories good species identification and acceptable species identification, 10 of 15 cases and 2 of 5 cases were correct, respectively. Thus, if identification is of lower quality, the result is less reliable and the isolate should be subjected to sequence analysis if accurate identification is of concern. For the VITEK 2 fluorescent system, species identifications are reliable; results of 33 of 37 cases in this category were consistent with molecular analysis. For the other categories, the numbers of isolates were too small to allow for firm conclusions. One major problem of phenotypic test systems is that the available databases are limited. Of the 107 strains analyzed in the present study, 16 and 46 of the isolates corresponded to species not included in the API 20 NE and VITEK 2 fluorescent card databases, respectively. In other words, while the API 20 NE covers the majority of nonfermenters isolated, the VITEK 2 database needs to be expanded. More recently, efforts have been undertaken to enlarge the VITEK 2 database. The revised VITEK 2 database is associated with a new col-

7 VOL. 44, 2006 IDENTIFICATION OF GRAM-NEGATIVE BACTERIA BY SEQUENCING 1365 FIG. 1. Algorithm for the identification of nonfermenting gramnegative bacilli. orimetric detection card and covers 159 taxa (versus 101 for the ID-GNB card). Twenty of the 46 strains not identified in our study due to noninclusion of the corresponding species in the database are now included in the expanded VITEK 2 database. Evaluation with a large collection of 655 gram-negative strains (including 144 nonfermenters) gave encouraging results (12). Further studies, testing the system under routine laboratory conditions, and comparing the results with genotypic methods are required before firm conclusions can be drawn on the quality of the enlarged VITEK 2 database. 16S rrna gene sequencing has distinct benefits compared to phenotypic identification procedures: (i) it is not restricted to a specific group of bacteria as public databases such as GenBank cover the whole spectrum of phylogenetic diversity; (ii) novel, not yet described species can be assigned to a group of related bacteria; (iii) results are in general unambiguous and not dependent on strain variation or individual interpretation. We have recently demonstrated the potential of partial 16S rrna gene sequencing for identification of gram-positive microorganisms and we have proposed algorithms for integration of molecular identification procedures into the diagnostic work flow (2, 3). In the present study, we have extended our efforts to include 16S rrna gene sequence analysis for accurate identification of gram-negative nonfermenting bacilli. 16S rrna gene sequencing has some limitations. It is in part compromised by a low phylogenetic power at the species or subspecies level (26). In the present study, 35% of isolates could not be unambiguously assigned to a single species. This was particularly true for members of the Burkholderia cepacia complex, the Acinetobacter baumannii-a. calcoaceticus complex and close relatives, some members of the genus Pseudomonas, the genus Achromobacter, and the genus Ralstonia. For the Burkholderia cepacia complex, the reca gene seems more appropriate for recognizing the different genomovars (19). For Acinetobacter, complete 16S rrna gene sequencing has been proposed for identification (15). More recently, the 16S-23S intergenic spacer region has been shown to differentiate closely related bacteria within the Acinetobacter baumannii-a. calcoaceticus complex (7). The quality of public databases, such as GenBank, is critical. Sequences are deposited independently of their quality, e.g., regardless of the correct assignment, the length of the sequence, or the number of ambiguous nucleotides. Of particular concern is that sequences in public databases may be assigned to a naming which possibly no longer is valid due to taxonomic changes or which never has been validly published before. As an example, several sequences are annotated under Acinetobacter calcoaceticus subsp. anitratus in GenBank; however, this is an old denomination which phenotypically corresponds to members of the Acinetobacter baumannii-a. calcoaceticus complex (1). Correct interpretation of sequence data also requires some familiarity with taxonomy and recent taxonomic changes. As an example, a sequence of an isolate with 100% homology to Burkholderia multivorans but 12 mismatches to Burkholderia cepacia should not be reported as B. multivorans per se. Rather, the result reported should reflect that B. multivorans is regarded as a member of the B. cepacia complex; B. multivorans is the second most common species of the B. cepacia complex in cystic fibrosis infections (20). Despite all these minor limitations, 16S rrna gene sequencing does not lead to false identification, and with some knowledge about taxonomy, a sequence can unambiguously be assigned. In this study, partial 16S rrna gene sequencing has been compared with two commercially available systems (API 20 NE, VITEK 2) for identification of nonfermenting gramnegative rods. The majority of strains could not be accurately identified by phenotypic profiling, as species assignment was found to be reliable only when an (or very good) species identification according to the system s criteria was achieved; this was the case in 35% of the isolates. In practical terms, molecular identification is more laborious than phenotypic identification; results of sequencing are usually available within one to two working days. Given these considerations, we have developed an algorithm for the effective and proper identification of gram-negative nonfermenters in the diagnostic laboratory (Fig. 1). Thus, when API 20 NE or VITEK 2 do not yield an (or very good) species identification, nonfermenters should be subjected to 16S rrna gene sequencing if adequate species assignment is of concern. ACKNOWLEDGMENTS We thank the technicians of the Institute of Medical Microbiology for technical assistance. This study was supported by the University of Zürich. REFERENCES 1. Bergogne-Berezin, E., and K. J. Towner Acinetobacter spp. as nosocomial pathogens: microbiological, clinical, and epidemiological features. Clin. Microbiol. Rev. 9:

8 1366 BOSSHARD ET AL. J. CLIN. MICROBIOL. 2. Bosshard, P. P., S. Abels, M. Altwegg, E. C. Böttger, and R. Zbinden Comparison of conventional and molecular methods for identification of aerobic catalase-negative gram-positive cocci in the clinical laboratory. J. Clin. Microbiol. 42: Bosshard, P. P., S. Abels, R. Zbinden, E. C. Böttger, and M. Altwegg Ribosomal DNA sequencing for identification of aerobic gram-positive rods in the clinical laboratory (an 18-month evaluation). J. Clin. Microbiol. 41: Bosshard, P. P., A. Kronenberg, R. Zbinden, C. Ruef, E. C. Böttger, and M. Altwegg Etiologic diagnosis of infective endocarditis by broad-range PCR: a 3-year experience. Clin. Infect. Dis. 37: Böttger, E. C Approaches for identification of microorganisms. ASM News 62: Brisse, S., S. Stefani, J. Verhoef, A. Van Belkum, P. Vandamme, and W. Goessens Comparative evaluation of the BD Phoenix and VITEK 2 automated instruments for identification of isolates of the Burkholderia cepacia complex. J. Clin. Microbiol. 40: Chang, H. C., Y. F. Wei, L. Dijkshoorn, M. Vaneechoutte, C. T. Tang, and T. C. Chang Species-level identification of isolates of the Acinetobacter calcoaceticus-acinetobacter baumannii complex by sequence analysis of the 16S 23S rrna gene spacer region. J. Clin. Microbiol. 43: Cloud, J. L., H. Neal, R. Rosenberry, C. Y. Turenne, M. Jama, D. R. Hillyard, and K. C. Carroll Identification of Mycobacterium spp. by using a commercial 16S ribosomal DNA sequencing kit and additional sequencing libraries. J. Clin. Microbiol. 40: Devereux, J., P. Haeberli, and O. Smithies A comprehensive set of sequence analysis programs for the VAX. Nucleic Acids Res. 12: Drancourt, M., C. Bollet, A. Carlioz, R. Martelin, J. P. Gayral, and D. Raoult S ribosomal DNA sequence analysis of a large collection of environmental and clinical unidentifiable bacterial isolates. J. Clin. Microbiol. 38: Ferroni, A., I. Sermet-Gaudelus, E. Abachin, G. Quesne, G. Lenoir, P. Berche, and J. L. Gaillard Use of 16S rrna gene sequencing for identification of nonfermenting gram-negative bacilli recovered from patients attending a single cystic fibrosis center. J. Clin. Microbiol. 40: Funke, G., and P. Funke-Kissling Evaluation of the new VITEK 2 card for identification of clinically relevant gram-negative rods. J. Clin. Microbiol. 42: Funke, G., D. Monnet, C. debernardis, A. von Graevenitz, and J. Freney Evaluation of the VITEK 2 system for rapid identification of medically relevant gram-negative rods. J. Clin. Microbiol. 36: Gerner-Smidt, P., I. Tjernberg, and J. Ursing Reliability of phenotypic tests for identification of Acinetobacter species. J. Clin. Microbiol. 29: Ibrahim, A., P. Gerner-Smidt, and W. Liesack Phylogenetic relationship of the twenty-one DNA groups of the genus Acinetobacter as revealed by 16S ribosomal DNA sequence analysis. Int. J. Syst. Bacteriol. 47: Kirschner, P., B. Springer, U. Vogel, A. Meier, A. Wrede, M. Kiekenbeck, F. C. Bange, and E. C. Böttger Genotypic identification of mycobacteria by nucleic acid sequence determination: report of a 2-year experience in a clinical laboratory. J. Clin. Microbiol. 31: Kiska, D. L., A. Kerr, M. C. Jones, J. A. Caracciolo, B. Eskridge, M. Jordan, S. Miller, D. Hughes, N. King, and P. H. Gilligan Accuracy of four commercial systems for identification of Burkholderia cepacia and other gram-negative nonfermenting bacilli recovered from patients with cystic fibrosis. J. Clin. Microbiol. 34: Kolbert, C. P., and D. H. Persing Ribosomal DNA sequencing as a tool for identification of bacterial pathogens. Curr. Opin. Microbiol. 2: Mahenthiralingam, E., J. Bischof, S. K. Byrne, C. Radomski, J. E. Davies, Y. Av-Gay, and P. Vandamme DNA-based diagnostic approaches for identification of Burkholderia cepacia complex, Burkholderia vietnamiensis, Burkholderia multivorans, Burkholderia stabilis, and Burkholderia cepacia genomovars I and III. J. Clin. Microbiol. 38: Mahenthiralingam, E., T. A. Urban, and J. B. Goldberg The multifarious, multireplicon Burkholderia cepacia complex. Nat. Rev. Microbiol. 3: Patel, J. B S rrna gene sequencing for bacterial pathogen identification in the clinical laboratory. Mol. Diagn. 6: Patel, J. B., D. G. Leonard, X. Pan, J. M. Musser, R. E. Berman, and I. Nachamkin Sequence-based identification of Mycobacterium species using the MicroSeq S rdna bacterial identification system. J. Clin. Microbiol. 38: Sader, H. S., and R. N. Jones Antimicrobial susceptibility of uncommonly isolated non-enteric gram-negative bacilli. Int. J. Antimicrob. Agents 25: Song, Y., C. Liu, M. Bolanos, J. Lee, M. McTeague, and S. M. Finegold Evaluation of 16S rrna sequencing and reevaluation of a short biochemical scheme for identification of clinically significant Bacteroides species. J. Clin. Microbiol. 43: Springer, B., L. Stockman, K. Teschner, G. D. Roberts, and E. C. Böttger Two-laboratory collaborative study on identification of mycobacteria: molecular versus phenotypic methods. J. Clin. Microbiol. 34: Stackebrandt, E., and B. M. Goebel A place for DNA-DNA reassociation and 16S rrna sequence analysis in the present species definition in bacteriology. Int. J. Syst. Bacteriol. 44: Tang, Y. W., N. M. Ellis, M. K. Hopkins, D. H. Smith, D. E. Dodge, and D. H. Persing Comparison of phenotypic and genotypic techniques for identification of unusual aerobic pathogenic gram-negative bacilli. J. Clin. Microbiol. 36: Tang, Y. W., A. Von Graevenitz, M. G. Waddington, M. K. Hopkins, D. H. Smith, H. Li, C. P. Kolbert, S. O. Montgomery, and D. H. Persing Identification of coryneform bacterial isolates by ribosomal DNA sequence analysis. J. Clin. Microbiol. 38: Van Looveren, M., and H. Goossens Antimicrobial resistance of Acinetobacter spp. in Europe. Clin. Microbiol. Infect. 10: Wellinghausen, N., J. Kothe, B. Wirths, A. Sigge, and S. Poppert Superiority of molecular techniques for identification of gram-negative, oxidase-positive rods, including morphologically nontypical Pseudomonas aeruginosa, from patients with cystic fibrosis. J. Clin. Microbiol. 43: Young, J. M., L. D. Kuykendall, E. Martinez-Romero, A. Kerr, and H. Sawada A revision of Rhizobium Frank 1889, with an emended description of the genus, and the inclusion of all species of Agrobacterium Conn 1942 and Allorhizobium undicola de Lajudie et al as new combinations: Rhizobium radiobacter, R. rhizogenes, R. rubi, R. undicola and R. vitis. Int. J. Syst. Evol. Microbiol. 51: