Comparison of Traditional Phenotypic Identification Methods with Partial 5′ 16S rRNA Gene Sequencing for Species-Level Identification of Nonfermenting Gram-Negative Bacilli

Abstract

Correct identification of nonfermenting Gram-negative bacilli (NFB) is crucial for patient management. We compared phenotypic identifications of 96 clinical NFB isolates with identifications obtained by 5′ 16S rRNA gene sequencing. Sequencing identified 88 isolates (91.7%) with >99% similarity to a sequence from the assigned species; 61.5% of sequencing results were concordant with phenotypic results, indicating the usability of sequencing to identify NFB.

Nonfermenting Gram-negative bacilli (NFB) are ubiquitous in the environment and may cause opportunistic infections in immunocompromised patients and individuals with cystic fibrosis (2, 9). Accurate diagnosis and appropriate treatment require species-specific identification of clinically significant NFB isolates. Conventional phenotypic identification may involve a number of methods, including observation of growth and colony morphology on various media, analysis of manual biochemical reactions, and the use of automated and nonautomated commercially available biochemical panels. Unfortunately, commercial phenotypic databases are often outdated and lack current taxonomy (15). Moreover, phenotypic systems often cannot account for the variable characteristics observed among members of the same species, resulting in poor precision upon repeat testing (3).

Identification of NFB by partial 5′ 16S rRNA gene sequencing using the MicroSeq 500 system (Applied BioSystems, Foster City, CA) is more accurate than conventional phenotypic methods and other commercial systems involving fatty acid and carbon utilization profiles (16). As for any identification method, limitations for 16S rRNA gene sequencing exist (11). Bacterial taxonomy and nomenclature continue to change as the genotypic features of organisms are analyzed in greater detail. Furthermore, these analyses have identified unique strains with distinct biochemical and genetic profiles that make definitive identification difficult until the strains are accepted as new species (8, 17). Even with a relatively complete sequence database, 16S rRNA gene sequences from different strains are often identical or closely matched (i.e., >99.5% similar), making expert judgment a requirement for identification (11).

When 16S rRNA gene sequencing is used to identify bacteria, the availability and completeness of databases will affect the accuracy of identification. Previous studies have shown that the MicroSeq 500 16S rRNA gene sequence library is incomplete and outdated for the identification of clinical isolates of Mycobacterium species (4) and Nocardia species (5, 14). As sequencing technology has become more affordable (6), clinical laboratories are now utilizing sequencing methods in conjunction with freely accessible public databases for organism identification. GenBank (http://blast.ncbi.nlm.nih.gov/Blast.cgi) is one such database that has been evaluated and found to contain sequence errors, especially in sequences submitted prior to 1995 (10, 11). The use of commercial databases increases the cost of sequence-based identification, and these databases are often outdated (4, 14). Finally, no criteria for reporting sequence identities exist, likely due to high levels of phylogenetic variation among species (11).

In the present study, implementation of 5′ 16S rRNA gene sequencing using the first 500 to 527 bp for the identification of clinical NFB isolates was assessed. PCR and sequencing methods were performed as described previously (13, 14). Sequence similarity analysis was accomplished using the MicroSeq 500 library (version 500-0125) in conjunction with a new sequence library used previously to assess NFB identification by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) (13). Identifications of 96 clinical isolates by sequencing were compared to identifications established by conventional phenotypic and commercial methods (using the Vitek or API 20NE system [bioMérieux, Durham, NC] or the MicroScan sysem [Siemens Healthcare Diagnostics Inc., Newark, DE]) with the latest version of software available at the time of testing. Isolates were recovered from clinician-requested specimens from patients with suspected infection at four different institutions. Each of the institutions performed phenotypic identifications according to their protocols, which were not necessarily the same among the institutions. While this approach presented a potential limitation for the study, all institutions regularly participated in proficiency surveys and were accredited according to the Clinical Laboratory Improvement Amendments (CLIA). Reporting criteria for identifications based on similarity of sequences were established as follows: excellent species identification, 99.8 to 100% similarity to a database sequence from the assigned species; good species identification, 99.1 to 99.7% similarity; and unlikely species identification, 96.7 to 99.0% similarity. These criteria are consistent with the recommendations reported by Janda and Abbott (11). Of the 96 clinical isolates examined, 64 (66.7%) yielded excellent species identification, 24 (25.0%) yielded good identification, and 8 (8.3%) could be identified confidently only to the genus level by sequencing (Table 1). Fifteen isolates with discrepant identifications by sequencing and phenotypic methods exhibited sequence identity scores between 99.8 and 100%, in support of the sequencing results (Table 1).

TABLE 1.

Phenotypic identifications of NFB isolated from clinical samples compared with identifications obtained using partial 16S rRNA gene sequencing

Phenotypic identification (no. of isolates, no. of discrepancies)	Species-level 5′ 16S rRNA gene-sequencing identification (no. of isolates, no. of discrepancies) scored as:
	Excellent (99.8-100% similarity)	Good (99.1-99.7% similarity)	Unlikely (96.7-99.0% similarity [genus level only])
Species-specific identifications
Acinetobacter baumannii (8, 3)	Acinetobacter genomospecies 3 (1, 1)		Acinetobacter calcoaceticus (1, 1)
	A. baumannii (5, 0)	Pseudomonas beteli (1, 1)
Acinetobacter lwoffii (2, 2)	Acinetobacter grimontii/A. junii (1, 1)		Acinetobacter haemolyticus (1, 1)
Acinetobacter junii (1, 1)			A. haemolyticus (1, 1)
Alcaligenes xylosoxidans subsp.	A. xylosoxidans subsp. xylosoxidans (3, 0)
xylosoxidans (5, 1)	A. xylosoxidans subsp. xylosoxidans/Alcaligenes ruhlandii (1, 0)
	Stenotrophomonas maltophilia (1, 1)
Brevundimonas diminuta (1, 0)	B. diminuta (1, 0)
Chryseobacterium meningosepticum (1, 0)	C. meningosepticum (1, 0)
Delftia acidovorans (1, 0)	D. acidovorans (1, 0)
Flavimonas oryzihabitans (1, 0)	Pseudomonas psychrotolerans/Pseudomonas oryzihabitans (1, 0)
Pseudomonas aeruginosa (15, 0)	P. aeruginosa (15, 0)
Pseudomonas alcaligenes (1, 1)		Achromobacter spanius (1, 1)
Pseudomonas fluorescensa (9, 7)	Pseudomonas plecoglossicida/P. putida/Pseudomonas monteilii (2, 2)	P. putida (1, 1)
	Pseudomonas koreensis (1, 1)	P. putida/P. monteilii (1, 1)
	P. plecoglossicida (1, 1)
	Pseudomonas synxantha/Pseudomonas mucidolens/Pseudomonas libanensis/Pseudomonas gessardii (2, 0)
	P. putida (1, 1)
Pseudomonas putidab (10, 1)	P. plecoglossicida (2, 0)	P. monteilii/P. putida (4, 0)	P. putida (1, 0)
		P. plecoglossicida/P. putida/P. monteilii (2, 0)
		Pseudomonas fulva (1, 0)
		Pseudomonas citronellolis (1, 1)
Pseudomonas stutzeri (6, 3)	Pseudomonas oleovorans (1, 1)	P. beteli (1, 1)
	P. stutzeri (3, 0)
	P. aeruginosa (1, 1)
Ralstonia pickettii (1, 1)	Ralstonia insidiosa (1, 1)
Sphingomonas paucimobilis (1, 1)		S. sanguinis (1, 1)
Stenotrophomonas maltophilia (19, 11)	S. maltophilia (7, 0)	P. beteli (4, 4)	P. beteli (1, 1)
		Pseudomonas hibiscicola (4, 4)	P. hibiscicola/Pseudomonas geniculata/Stenotrophomonas africana (2, 2)
Species group identifications
A. baumannii/A. haemolyticus (2, 2)	Acinetobacter genomospecies 13 (1, 1)	Acinetobacter tjernbergiae (1, 1)
P. fluorescens/P. putida (4, 1)	S. maltophilia (1, 1)		P. putida (1, 0)
	P. plecoglossicida/P. putida/P. monteilii (2, 0)
A. baumannii/A. calcoaceticus (1, 1)	Acinetobacter genomospecies 3 (1, 1)
Genus-only identifications
Acinetobacter species (2, 0)	Acinetobacter genomospecies 3 (2, 0)
Alcaligenes species (3, 0)	Alcaligenes faecalis subsp. faecalis (2, 0)	A. faecalis subsp. faecalis (1, 0)
Chryseobacterium species (1, 0)	Chryseobacterium indologenes (1, 0)
Unable to identify (1, 1)	Brevundimonas nasdae/Brevundimonas intermedia/Brevundimonas vesicularis (1, 1)

^aThe P. fluorescens group includes P. fluorescens, P. synxantha, P. mucidolens, P. libanensis, and P. gessardii (1).

^bThe P. putida group includes P. putida, P. plecoglossicida, P. monteilii, and P. fulva (1).

Compared with conventional phenotypic identification methods, sequencing featured increased reliability and reproducibility; however, limitations with database accuracy and species discrimination needed to be considered. Phenotypic identification utilizes a less precise scoring system than sequencing, is affected by intraspecies phenotypic variation, and exhibits low-level reproducibility (3), making it difficult to assess accuracy. This study has revealed that 77.8% (7 of 9) of the clinical isolates identified as Pseudomonas fluorescens by traditional phenotypic methods may have been misidentified, as indicated by sequence analysis (Table 1). Some phenotypic databases are likely to be outdated since seven of the nine different 5′ 16S rRNA gene sequences showed 99.8 or 100% similarity to a database sequence (Table 1). Therefore, when P. fluorescens is identified by phenotypic methods, reflex testing using sequencing or another genetic method should be considered if a more accurate identification is indicated.

Nineteen isolates were identified phenotypically as Stenotrophomonas maltophilia, with 11 (57.9%) having a discrepant identification by 5′ 16S rRNA gene sequencing. Of the 11 isolates with discrepant results, 8 had good (99.1 to 99.7%) similarity to S. maltophilia and 3 had unlikely (≤99.0%) similarity. Of the discrepantly identified isolates with good similarity, four were identified as Pseudomonas beteli and four were identified as Pseudomonas hibiscicola by 5′ 16S rRNA gene sequencing. These results suggest that data for P. beteli and P. hibiscicola were either not included in the phenotypic databases or that these species had biochemical profiles indistinguishable from that of S. maltophilia. Genetic comparison shows P. beteli and P. hibiscicola to be similar to S. maltophilia; however, further epidemiologic and genotypic studies are required for definitive taxonomic placement (1).

In conclusion, this study showed that 5′ 16S rRNA gene sequencing could improve the accuracy of species-level identification of NFB. With current taxonomy and nomenclature, however, there is great difficulty in knowing whether a single species should be recognized or multiple genomospecies should be used in classification (18). Comparing conventional phenotypic identifications by multiple laboratories to identifications obtained by 5′ 16S rRNA gene sequencing showed that no single method was reliable and that all methods were limited by incomplete and outdated databases. Additional studies using genotypic, phenotypic, and proteomics analyses are needed to establish assays with consistent and reproducible results for NFB identification. Recently, MALDI-TOF MS was shown to be a powerful technique with good interlaboratory reproducibility (7, 12). With any method, accuracy for the identification of NFB will depend on databases that are updated with the most current taxonomy.