• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of cmrPermissionsJournals.ASM.orgJournalCMR ArticleJournal InfoAuthorsReviewers
Clin Microbiol Rev. Oct 2004; 17(4): 840–862.
PMCID: PMC523561

Impact of 16S rRNA Gene Sequence Analysis for Identification of Bacteria on Clinical Microbiology and Infectious Diseases


The traditional identification of bacteria on the basis of phenotypic characteristics is generally not as accurate as identification based on genotypic methods. Comparison of the bacterial 16S rRNA gene sequence has emerged as a preferred genetic technique. 16S rRNA gene sequence analysis can better identify poorly described, rarely isolated, or phenotypically aberrant strains, can be routinely used for identification of mycobacteria, and can lead to the recognition of novel pathogens and noncultured bacteria. Problems remain in that the sequences in some databases are not accurate, there is no consensus quantitative definition of genus or species based on 16S rRNA gene sequence data, the proliferation of species names based on minimal genetic and phenotypic differences raises communication difficulties, and microheterogeneity in 16S rRNA gene sequence within a species is common. Despite its accuracy, 16S rRNA gene sequence analysis lacks widespread use beyond the large and reference laboratories because of technical and cost considerations. Thus, a future challenge is to translate information from 16S rRNA gene sequencing into convenient biochemical testing schemes, making the accuracy of the genotypic identification available to the smaller and routine clinical microbiology laboratories.


One area within the practice of clinical microbiology is the craft of putting scientific names to microbial isolates. This is usually done with the intent of giving insight into the etiological agent causing an infectious disease, including pathological associations and possible effective antimicrobial therapy. The historical method for performing this task is dependent on the comparison of an accurate morphologic and phenotypic description of type strains or typical strains with the accurate morphologic and phenotypic description of the isolate to be identified. Microbiologists authoring standard references such as Bergey's Manual of Systematic Bacteriology or the Manual of Clinical Microbiology or compiling results from well-characterized strains such as those found at the Centers for Disease Control and Prevention or the American Type Culture Collection (ATCC) would publish tables summarizing the characteristics of each species of bacteria (35, 54, 60). Clinical microbiologists would try to match the results for their unknown clinical strain with a group in these tables. Not infrequently, there would be no perfect match and a judgment would have to be made about the most probable identification. Although various schema and computer programs were devised to help in these judgements, identification could vary among laboratories (96).

In the 1980s, a new standard for identifying bacteria began to be developed. In the laboratories of Woese and others, it was shown that phylogenetic relationships of bacteria, and, indeed, all life-forms, could be determined by comparing a stable part of the genetic code (111, 113). Candidates for this genetic area in bacteria included the genes that code for the 5S, the 16S (also called the small subunit), and the 23S rRNA and the spaces between these genes. The part of the DNA now most commonly used for taxonomic purposes for bacteria is the 16S rRNA gene (7, 36, 44, 52, 64, 101). The 16S rRNA gene is also designated 16S rDNA, and the terms have been used interchangeably: current ASM policy is that “16S rRNA gene” be used. The 16S rRNA gene can be compared not only among all bacteria but also with the 16S rRNA gene of archeobacteria and the 18S rRNA gene of eucaryotes. Figure Figure11 shows the relationship of major branches of life, the Archaea. Bacteria (procaryotes), and Eucarya, as well as the major branches within the procaryotes based on these gene sequences (62, 64, 111, 113).

FIG. 1.
Universal phylogenetic tree based on the 16S rRNA gene sequence comparisons. Reprinted from reference 62 with permission of the publisher.

The goal of this review is to describe not only the mechanism and limits of bacterial 16S rRNA gene sequence analysis but also the impact and potential contribution that the 16S rRNA gene sequence analysis can make to the understanding of clinical microbiology and infectious diseases. It is hoped that this will promote recognition that the correct identification or taxonomic name assignment can make a difference in our understanding of the pathogenic process and in clinical outcome. A further goal is to help the clinical microbiologist winnow the enormous amount of taxonomic information now being generated in order to promote meaningful and scientifically accurate communications with clinical colleagues.


Choice of the 16S rRNA Gene as the Gene To Sequence

In the 1960s, Dubnau et al. (28) noted conservation in the 16S rRNA gene sequence relationships in Bacillus spp. Widespread use of this gene sequence for bacterial identification and taxonomy followed a body of pioneering work by Woese, who defined important properties. Foremost is the fact that it seems to behave as a molecular chronometer, as pointed out in an excellent review article by Woese (113). The degree of conservation is assumed to result from the importance of the 16S rRNA as a critical component of cell function. This is in contrast to the genes needed to make enzymes. Mutations in these genes can usually be tolerated more frequently since they may affect structures not as unique and essential as rRNA (if a bacterium does not have the gene to make the enzymes needed to utilize lactose, it can use an alternative sugar or protein as an energy source). Thus, few other genes are as highly conserved as the 16S rRNA gene. Although the absolute rate of change in the 16S rRNA gene sequence is not known, it does mark evolutionary distance and relatedness of organisms (44, 49, 62, 100). Problems in assigning a numerical value to this rate of change include the possibility that this rate of change of 16S rRNA gene may not be identical for all organisms (different taxonomic groups could have different rates of change), the rates could vary at times during evolution, and the rates could be different at different sites throughout the 16S rRNA gene. There are so-called “hot spots” which show larger numbers of mutations (101, 104); these areas are not the same for all species. 16S rRNA is also the target for several antimicrobial agents. As such, mutations in the 16S rRNA gene can affect the susceptibility of the organism to these agents and the 16S rRNA gene sequence can distinguish phenotypic resistance to antimicrobial agents (69, 70). However, these characteristics do not obviate or affect the use of 16S rRNA gene sequence for bacterial identification or assignment of close relationships at the genus and species level, as used in clinical microbiology. They can have a greater impact on the assignment of relationships of the deeper (more distantly related) branches (36).

The 16S rRNA gene sequence is about 1,550 bp long and is composed of both variable and conserved regions. The gene is large enough, with sufficient interspecific polymorphisms of 16S rRNA gene, to provide distinguishing and statistically valid measurements. Universal primers are usually chosen as complementary to the conserved regions at the beginning of the gene and at either the 540-bp region or at the end of the whole sequence (about the 1,550-bp region), and the sequence of the variable region in between is used for the comparative taxonomy (11, 75). Although 500 and 1,500 bp are common lengths to sequence and compare, sequences in databases can be of various lengths.

The 16S rRNA gene sequence has been determined for a large number of strains. GenBank, the largest databank of nucleotide sequences, has over 20 million deposited sequences, of which over 90,000 are of 16S rRNA gene. This means that there are many previously deposited sequences against which to compare the sequence of an unknown strain.

Lastly, the 16S rRNA gene is universal in bacteria, and so relationships can be measured among all bacteria (111, 113) (Fig. (Fig.2).2). In general, the comparison of the 16S rRNA gene sequences allows differentiation between organisms at the genus level across all major phyla of bacteria, in addition to classifying strains at multiple levels, including what we now call the species and subspecies level. The occasional exceptions to the usefulness of 16S rRNA gene sequencing usually relate to more than one well-known species having the same or very similar sequences.

FIG. 2.
Dendrogram showing the genetic relationships of many of the major groups of clinically important organisms based on the 500-bp 16S rRNA gene sequence. Most sequences are of type strains from the MicroSeq database. The Leptotrichia buccalis sequence was ...

It is also important to consider whether it is necessary to sequence the whole 1,500-bp length or whether the commonly reported shorter sequences can provide comparable information. Sometimes sequencing the entire 1,500-bp region is necessary to distinguish between particular taxa or strains (84, 85). Sequencing of the entire 1,500-bp sequence is also desirable and usually required when describing a new species. However, for most clinical bacterial isolates the initial 500-bp sequence provides adequate differentiation for identification and in fact can provide a bigger percent difference between strains because the region shows slightly more diversity per kilobase sequenced. Kattar et al. (48) found that 66% of the variability in the 16S rRNA gene sequence among Bordetella species was in the first 500 bp. Evaluations published in the literature, made using the MicroSeq database (Applied Biosystems Inc. [ABI], Foster City, Calif.), are usually based on the 500-bp sequence (42, 66, 97, 98). Other researchers have made identifications using sequences of about 400 bp (6) or even less than 200 bp (109). Data in the figures and calculations in this review refer to the 500-bp length unless otherwise noted. From the MicroSeq databases of over 1,400 organisms for both the 500- and 1,500-bp lengths, we compared the 500- and 1,500-bp sequences for 100 organisms by using each length to generate dendrograms and found the relationships of species to be basically the same with either length. For example, Fig. Fig.33 shows that dendrograms generated using either the 1,500-bp 16S rRNA gene sequence (left side of figure) or the 500-bp 16S rRNA gene sequence (right side of figure) of a group of clinical and type strains of brevibacteria are similar but not identical.

FIG. 3.
A comparison of dendrograms generated using either the 1,500-bp 16S rRNA gene sequence (left) or the 500-bp 16S rRNA gene sequence (right) of a group of clinical and type strains of Brevibacterium.

On a practical note, generating the 500-bp sequence is less expensive and easier since it takes more sequencing reactions to generate the 1,500-bp sequence.

Many other genomic regions have also been used to examine the phylogenetic relationships among bacteria. Whole-genome analysis has been tried, but this is quite difficult because the genomes are of such different sizes and because gene duplication, gene transfer, gene deletion, gene fusion, and gene splitting are common; at present there are less than 100 whole genomes to compare (3, 112). However, it has been observed that the trees based on whole-genomic analysis and the 16S rRNA gene trees are similar (3). Other areas of the rRNA gene have also been used for studying phylogenetic relationships among bacteria. Roth et al. used the 16S-23S rRNA gene internal transcribed spacer sequences to distinguish among Mycobacterium spp., finding it particularly useful for species that were indistinguishable by 16S rRNA gene sequences (82). Others have found the use of 23S rRNA sequences helpful in distinguishing among Streptococcus spp. (72). Although some researchers find that an overall robustness of the method is suggested because the major branching points of the phylogenetic tree were conserved when either the 16S rRNA or 16S-23S rRNA gene sequences were used (82), others find the 16S rRNA gene sequence much more useful for phylogenetic analysis than the 16S-23S rRNA gene region (89). In any case, the method is not widely used, and there are few comparative sequences. For mycobacteria, the gene encoding the 65-kDa heat shock protein is highly conserved and also has been used to define taxonomic relationships (63, 79, 101). Although the gene encoding the 65-kDa heat shock protein sequences are phylogenetically useful, far fewer of them are available in databases. Phylogenetic trees obtained using protein-encoding gene sequence comparisons do not seem to reveal deep-rooted taxonomic and evolutionary relationships as reliably as those obtained using the 16S rRNA gene (36,40, 55).

In general, if one wants to compare strains for epidemiological purposes or to detect a strain having a particular virulence factor, 16S rRNA gene analysis is not usually adequate there is not enough variation, and, obviously, the region does not encode virulence factors. An exception to this is the microheterogeniety in the 16S rRNA gene sequence found by Sacchi et al. (85), which could be used to track Neisseria meningitidis strains. In addition, if one is interested only in differentiating species within a particular genus, a better gene than the 16S rRNA gene might be found to identify species. For example, the citrate synthetase gene in the genera Bartonella and Rickettsia (32, 73) seems to be unique for each species and thus is an excellent tool in experiments to differentiate them. However, no gene has shown as broad applicability over all the taxonomic groups as the 16S rRNA gene. Thus, if the goal is to identify an unknown organism on the basis of no a priori knowledge, the 16S rRNA gene sequence is an excellent and extensively used choice. The new edition of Bergey's Manual of Systematic Bacteriology, the most widely used and authoritative reference on bacterial taxonomy, is organized using 16S rRNA gene sequence analysis as the backbone. Two chapters in the second edition of Bergey's Manual of Systematic Bacteriology are particularly recommended and give an excellent overview (36, 55).

Basics of Sequencing

Nucleic acid sequencing methods have undergone tremendous advances over the past decade. These rapid advances have made it possible for even a small laboratory to determine the sequence of millions of base pairs of DNA per year. The quality of sequence data has improved with the speed and technology available. Thus, one sees that in public databases such as GenBank, the 16S rRNA gene sequences deposited in the early 1990s have many more incorrect and indeterminate bases than do recently generated sequences. We briefly outline the steps required below and in Table Table1.1. Table Table11 also shows the approximate time required for each step of the ABI dye terminator method. Other publications give these steps in more detail, along with diagrams (52, 65, 98). Generally, it is possible to generate a sequence in less than 1.5 working days by using less than a colony or directly from a specimen (9, 22, 98).

Procedures and time to perform 16S rRNA gene sequence analysis for bacterial identification in a routine clinical microbiology laboratorya

Historically, there have been several methods to determine DNA sequence. These methods are detailed in reference 87 (chapters 13 and 14). Bacterial genomic DNA is extracted from whole cells by using a standard method (87) or a commercial system (e.g., PrepMan DNA extraction reagent; ABI). The DNA is used as the template for PCR to amplify a segment of about 500 or 1,500 bp of the 16S rRNA gene sequence. Broad-based or universal primers complementary to conserved regions are used so that the region can be amplified from any bacteria. The PCR products are purified to remove excess primers and nucleotides; several good commercial kits are available (e.g., QiaQick PCR purification kit [Qiagen] and Microcon-100 Microconcentrator columns [Millipore]).

The next step is a process called cycle sequencing. It is similar to PCR in that it uses DNA (purified products of the first PCR cycle) as the template. Both the forward and reverse sequences are used as the template in separate reactions in which only the forward or reverse primer is used. Cycle sequencing also differs from PCR in that no new template is formed (the same template is reused for as many cycles as programmed, usually 25 cycles) and the product is a mixture of DNA of various lengths. This is achieved by adding specially labeled bases called dye terminators (along with unlabeled bases), which, when they are randomly incorporated in this second cycle, terminate the sequence. Thus, fragments of every size are generated. As each of the four added labeled terminator bases has different fluorescent dye, each of which absorbs at a different wavelength, the terminal base of each fragment can be determined by a fluorometer.

The products are purified to remove unincorporated dye terminators, and the length of each is determined using capillary electrophoresis (e.g., ABI PRISM 3100 genetic analyzer with 16 capillaries or ABI PRISM 310 genetic analyzer with 1 capillary) or gel electrophoresis (e.g., the Visible Genetics system). Since we then know the length and terminal base of each fragment, the sequence of the bases can be determined. The two strands of the DNA are sequenced separately, generating both forward and reverse (complementary) sequences. An electropherogram, a tracing of the detection of the separated fragments as they elute from the column (or are separated in the gel) in which each base is represented by a different color, can be manually or automatically edited. It is possible to have the fragments of various lengths so well separated that every base of a 500-bp sequence can be determined. When ambiguities occur, most of them can be resolved by visual reediting of the electropherogram.

We have evaluated two different DNA sequencers for clinical microbiology identifications. The model 3100 ABI sequencer was superior to the model 310 sequencer in the length of reliable sequence (520 and 460 bp, respectively), in addition to the published superiority of ease of use and time of run. In the time and cost analysis, our times and thus our costs are based on data using the ABI PRISM 3100 genetic analyzer.

The generated DNA sequences are usually assembled by aligning the forward and reverse sequences. This consensus sequence is then compared with a database library by using analysis software. Some systems allow comparisons of the single forward or reverse sequences. Well-known databases of 16S rRNA gene sequences that can be consulted via the World Wide Web are GenBank (http://www.ncbi.nlm.nih.gov/), the Ribosomal Database Project (RDP-II) (http://rdp.cme.msu.edu/html/), the Ribosomal Database Project European Molecular Biology Laboratory (http://www.ebi.ac.uk/embl/), Smart Gene IDNS (http://www.smartgene.ch), and Ribosomal Differentiation of Medical Microorganisms (RIDOM) (http://www.ridom.com/). The proprietary MicroSeq 500 bacterial database (version 1.4.2) contains sequences for 1,434 species or subspecies within 235 genera.


Overview of Bacterial Identification and Taxonomic Placement Using 16S rRNA Gene Sequences

In Fig. Fig.2,2, the major branches of important bacteria encountered in clinical practice are shown in dendrogram form, whereas in Fig. Fig.1,1, the major divisions of bacteria are shown with all other life-forms in a star form. For reference, note and compare where Bacillus. Clostridium, and Escherichia are in each figure.

Figure Figure22 was generated by choosing 62 genetically widely disparate strains representing major taxonomic groups or clades of medical interest from the MicroSeq database. Chlamydia trachomatis is the outgroup. The horizontal line at the top (in this case, 16.105%) is used to provide a rough measure of genetic distance. To find the approximate genetic difference between two clades in Fig. Fig.2,2, the two horizontal distances between the species to be assessed are added (the vertical distance does not count) and the total is compared to the top horizontal line. For example, Nocardia asteroides and Corynebacterium diphtheriae differ by approximately 10% whereas Treponema medium and Mycoplasma hominis differ by approximately 28%. The actual numbers generated in this manner are not very accurate, especially for genera that are not closely related. The values also vary somewhat with the computer program used to generate the dendrogram. However, they are useful for estimating relative relatedness. The scale of the dendrogram and the number of organisms that are missing can be better appreciated if one notes that the organism Escherichia coli represents the entire Enterobacteriaceae clade of over 50 species and Vibrio cholerae is the most closely related organism represented in Fig. Fig.2.2. In addition to dendrograms, percent similarity, percent dissimilarity, whole-gene alignment, and concise alignment can be used to compare and evaluate sequences.

Problems in Generating a Sequence

How often are sequences wrong? Although the technology is such that the sequence reading is getting ever more accurate, we estimate that there may be an operator mistake in editing in 1 in 5,000 to 1 in 10,000 bp. Errors at this level do not make a difference in the species name (44). There is also a question whether the DNA sequence must have overlap to be accurate or whether an adequate sequence could be obtained by using just the forward or reverse sequence. Several large and influential laboratories use only the forward sequence (6). In one study of 50 isolates, either forward or reverse sequence could be used to assign a correct species identification, with less than 1% difference between sequences (J. E. Clarridge, Q. Zhang, and S. Heward, Abstr. 101st Gen. Meet. Am Soc. Microbiol. 2001, abstr. C-44, 2001). Other laboratories generate sequences using multiple overlaps, particularly if microheterogeneity is important (84, 85). Since high-fidelity polymerases with proofreading capability became available, PCR-generated errors are very low.

In contrast to the accuracy achievable nowadays with the excellent equipment and reagents available, some of the sequences deposited in public databases such as GenBank, particularly those derived over 10 years ago, are not very accurate (18). This is usually because some sequences were not clearly generated, often because of poor separation of the fragments in the gel electrophoresis steps. When evaluating a sequence generated in another laboratory or another database for which the electropherogram is not available, clues to a poorly read or generated sequence are that they have many ambiguous bases, which are noted as N, R, Y, W, M, S, or K (meaning that the base is unknown, A or G, C or T, A or T, A or C, G or C, G or T, respectively). One can also see the letters B, D, H, and V, representing 3-base ambiguities. Sequences deposited more recently are much more accurate.

It is worthwhile to examine the quality of sequences before using them. During a study of Actinomyces strains (17), we compared sequences from a commercial database (MicroSeq), a public database (GenBank-EMBL), and our own internally generated database. The type strain of Actinomyces turicensis in the GenBank database and the 12 clinical strains that we sequenced showed minimal (none to two) differences and no ambiguous bases, indicating that all were good quality sequences. The sequence of the type strain of A. meyeri from MicroSeq was also of high quality. However, the type strain of A. gravaenitzii in the GenBank database was of poorer quality, with 25 N's in 500 bp sequenced (5%). Thus, it initially appeared that the phenotypically similar A. gravaenitzii strains were genetically more heterogeneous than the A. turicensis strains, although this may not be true.

Although most ambiguities can be resolved by a reediting of the original electropherogram if available (electropherograms are not available from GenBank and do not come with manuscripts for review), there can also be situations in which it is not possible to determine a unique base to a particular position. Usually this is because of some technical problem, e.g., the original specimen was not a pure culture, the yield of labeled product was too low, or the column was malfunctioning. In this case, there are usually many unreadable bases and the whole sequencing procedure must be repeated.

It is also possible that intracellular polymorphisms might cause difficulties in obtaining an easily interpretable sequence; i.e., since there are multiple copies of the 16S rRNA gene within a single-cell genome, there could be several different sequences and thus there could be two different base pairs at a given location. The existence of variant 16S rRNA gene alleles in a single genome has been convincingly demonstrated in several reports (38, 61, 74, 104, 110). Most of the documented intracellular heterogeneities are only one or two polymorphisms per sequence and would not lead to different species identification. Among these reports, the highest rate of polymorphisms found (104) was with stock cultures of Streptomyces spp.; 2.5% of 475 strains showed diversity of more than 1 in 1,500 bp. The greatest diversity of 14 bp, shown by one strain, is still only 1% of the total number of base pairs sequenced. In Streptomyces spp., most of the intracellular polymorphisms are in the hypervariable α-helix region between bp 173 to 195, the same region of the 16S rRNA gene in which the individual Streptomyces species differ the most (86). Not all genera have the same hypervariable region. There is also a report comparing whole-genome sequences that finds widely disparate 16S rRNA gene sequences in a single clone (3), suggesting large-fragment gene transfer. The overall occurrence of polymorphisms may be reevaluated in the future by techniques that could detect a variant allele expressed in minor amounts, but at this point it does not seem to be a problem for accurate identifications in clinical microbiology. During a review of hundreds of sequence analyses from our laboratory, we found that intracellular polymorphism has not been a problem and that only a few Y's or R's have had to be assigned. Even though the intracellular polymorphisms may not be sufficiently numerous to generate a different species name based on each of the sequences, they are to be considered and can be useful when using the 16S rRNA gene sequence for epidemiological strain tracking (38, 85).

One must be certain to distinguish the existence of two (or more) variant 16S rRNA genes in a single genome from a mixture of closely related but distinct strains in the presumed pure culture. In our studies of the “Streptococcus milleri” group, we reported this second possibility, in which there is a mixed culture of two or more separate but highly related strains in the same specimen. For example, two strains of S. anginosus (strains VAMC 417-1 and 417-2, as noted in reference 13) were originally frozen as a single isolate, but with precise work in sorting out the slightly different phenotypes, each substrain could be isolated in pure culture and was shown to be unique in genotype and phenotype. The sequence difference in the isolates was only 4 bp in the 500-bp length, and there were no ambiguous bases. If we had sequenced the mixture, these four positions might have appeared as three R's and a Y, which some could have interpreted as intracellular polymorphism. The situation in which the original specimen is not a pure culture but the contaminating strain is not closely related is easier to resolve since there would typically be over 50 ambiguous bases; under these circumstances, the organism(s) should be reisolated and resequenced.

Generating Dendrograms and Comparing Sequences

Several sequence-comparing software packages are available. We generally use the proprietary software that comes with the MicroSeq method. Other common software packages are PAUP (107), BLAST (1), and Phylip (30, 31; Phylogeny Inference Package, University of Washington). BIBI, a bioinformatics bacterial identification tool, has recently been developed to simplify and automate bacterial identifications using DNA sequence analysis (24); it is available at http://pbil.univ-lyon1.fr/bibi/. An excellent website that compares 194 of the phylogeny software packages and 16 free servers that are used in generating dendrograms as well as addressing other concepts covered in this review is http://evolution.genetics.washington.edu/phylip/software.html. Comparisons are commonly shown as dendrograms and linear alignments (for concise linear alignments, all the identical base pairs are omitted and only the differences are shown).

Methods commonly used for generating dendrograms are the NJ (neighbor-joining) method, the UPGMA (unweighted pair group method with arithmetic averages), and the WPGMA (weighted pair group method with arithmetic averages) (86, 107). The methods are comparable, and the major groupings are preserved if the isolates are closely related. However, when the taxa being compared are less closely related, the dendrogram relationships are more strongly affected by the program used. Figure Figure44 shows the dendrograms generated by the NJ methods for some Streptococcus spp. Figures 4A and B are dendrograms with different outgroups, i.e., the primary sequence against which a sequence is compared. The best choice for the outgroup is a closely related strain just outside the group being studied. When an inappropriate outgroup is chosen, as in Fig. Fig.4A4A (Chlamydia trachomatis is too distantly related), the relationships of the clades can be obscured. S. anginosus may be too closely related but is useful in that the differences are clear (Fig. (Fig.4B).4B). A concise comparison of the same isolates is shown in Fig. Fig.4C.4C. Another method of comparison is what is called a concise alignment: only the differences among the compared sequences are shown (Fig. (Fig.4C).4C). The numbers at the top of the figure are the base position in the sequence and are to be read vertically; for example, strain VAMC5210 differs from strain S7745 at positions 137, 274, and 487.

FIG. 4.
Importance of an appropriate outgroup and a concise comparison of strains. (A and B) Chlamydia trachomata (A) is too distantly related to allow differences easily seen using S. anginosus as outgroup (B). (C) Exact base pair differences between strains ...

The length of sequence analyzed and the alignment tool used can also affect the comparison of sequences. In Fig. Fig.3,3, even though the relationships shown between the brevibacteria are similar, the dendrogram reflects the fact that the three strains in the Brevibacterium mcbrellneri group show no difference in the first 500-bp sequence but there are five or six differences in the last 1,000-bp sequence. This is a less common occurrence than the case where there is more heterogeneity in the initial segment of the gene (48). However, if the clinical strain VAMC3643 and B. epidermidis are compared using two alignment tools (BLAST and the Needleman Wunsch algorithm), four nonidentical sets of dissimilarity percentages are generated. For both sets of comparisons, the 500-bp sequences show more dissimilarity. More important is that the two programs give somewhat different results; BLAST is faster but less accurate than the Needleman Wunsch algorithm. For example, the percent dissimilarity in the whole 1,500-bp sequence between clinical strain VAMC3643 and B. epidermidis was 0.8% using the Needleman Wunsch algorithm and 1.5% using the BLAST comparison. The same calculations based on the first 500 bp are 1.8 and 2.8%, respectively. The order of relatedness was about the same for all four sets of calculations for the closely related strains. However, this is not true for the more distantly related species shown in Fig. Fig.3.3. More caution must be exercised in using sequence data to assign exact taxonomic relationships between the higher taxa (36, 55).

The dendrogram can also be used to assess an unknown sequence. To demonstrate how one can quickly assess an unknown sequence from any database, I imported a sequence, Leptotrichia buccalis L37788, which is not in the MicroSeq database, from the GenBank database and incorporated it into the dendrogram generated with type strain sequences from the MicroSeq database (Fig. (Fig.2).2). It is appropriately placed in the fusobacteria-streptobacilli lineage. In addition, a strain sequenced in our laboratory, Unknownsp01M1398, was incorporated into the dendrogram to demonstrate the relatedness of the organism to other major groups, even though there is not a closely related sequence with which to compare it in the database. It represents a novel, deeply divergent genetic line that is about 20% divergent from any bacteria in the MicroSeq database and matches only an “uncultureable” strain in the GenBank database (these databases are discussed below). Both strains are seen about halfway down the dendrogram.

There are certain aspects of relatedness that are better shown by dendrogram or by concise alignment comparison than by percent dissimilarity. In Fig. Fig.4B4B the dendrogram shows that strains S5366, S7377, and VAMC6703 are not in the same genogroup and the concise alignment shows the exact base differences, yet the percent dissimilarity from strain 5366 of both S7377 and VAMC6703 is 1.1%. Thus, for example, one cannot assume that because two strains are 1.1% dissimilar to a particular strain, they have a closer relationship to each other (in this case they are 1.2% dissimilar to each other).

Sequence Databases

For phenotypic identifications of microorganisms, we are dependent on a database with an accurate morphologic and biochemical description of type strains or typical strains and standard methods to determine these characteristics for the isolate to be identified. Similarly, for accurate 16S rRNA gene sequence identification of organisms, we are dependent on accurate sequences in databases, appropriate names associated with those sequences, and an accurate sequence for the isolate to be identified.

There are several reasons why sequence databases can vary and may not accurately link a name with a sequence and, further, with a correct relative placement of the sequence among other bacterial sequences. It could be that the type strain or strains in certified collections such as ATCC were incorrectly named or classified by biochemical means or that the descriptions are just wrong. A familiar example of an organism being placed in the wrong genus is “Corynebacterium aquaticum.” Although “C. aquaticum” has, superficially, the same morphology as the corynebacteria, it is genetically distant from the genus Corynebacterium and is more closely related to organisms in the genus Microbacterium. Another example is that when examined by genotypic methods, the ATCC strains previously deposited as Corynebacterium xerosis were found to be diverse and incorrectly identified (23). There are numerous other straightforward errors in taxonomy that are addressed in the new edition of Bergey's Manual of Systematic Bacteriology (36).

There is also the misplacement of well-defined species presumably within a single genus but actually found in many taxonomic groups. Species of the genus Enterobacter are found associated with five different genera. Enterobacter is a good example of what is called a polyphyletic genus, as is Citrobacter. In an attempt to correct this, some of the historic Enterobacter spp. have been given the new genus name Pantoea (for example, P. agglomerans and P. aerogenes). Proteus is a monophyletic genus with all species in a single clade.

A third reason for error applies primarily to unverified databases such as GenBank, which accept any linked name and sequence that is sent to them. In the 1995 investigation of GenBank sequences, a comparison of sequences for the presumed same species showed that almost 20% had more than 2% variability, indicating that many genetically different strains were being deposited under the same species name (18). On some occasions, the name associated with a given sequence may not be correct due to poor phenotypic descriptions or faulty testing. For example, the deposited strains of the genera Alcaligenes and Achromobacter are intermixed, with the descriptions and sequence both overlapping. On other occasions, the sequences deposited in GenBank are not complete or as accurate as they could be. For example, we compared the sequence of the type strain of Mycobacterium szulgai generated in our laboratory to the sequences in the GenBank database and the MicroSeq database (115). The sequence from the MicroSeq database and our sequence agreed, while the GenBank sequence had an unacceptable number of bases that were listed as N or undetermined.

Wilck et al. also have noted that databases such as GenBank, which are very broad (i.e., contain both pathogenic and nonpathogenic strains of human, animal, and environmental origin) and not peer reviewed, tend to contain errors. They suggest that the use of peer-reviewed databases for defining the 16S rRNA gene sequences of bacteria found in humans would improve the validity of this method of organism identification in the clinical microbiology laboratory (109). It was this sort of consideration that propelled the development of the proprietary MicroSeq database of verified type strains and the RIDOM database.

There are few papers on the evaluation of the GenBank, RDP-II (57), and RIDOM databases and Internet-based programs. To demonstrate the quality and accuracy of results provided from available databases, Turenne et al. submitted 79 mycobacterial type strain sequences determined in their laboratory for analysis using the GenBank database, the RDP-II, and, most recently, the RIDOM database (103). They found that all of the type strain sequences they generated had an identical matching sequence when analyzed by RIDOM whereas only 23% of species had a perfect match with sequences from GenBank databases as determined by BLAST and 25% of species had a perfect match as determined by RDP-II. This means that the quality and/or the number of the mycobacterial sequences in RIDOM is higher. However, although the RIDOM database has a particularly good collection of mycobacterial sequences, there is a narror range of other organisms: in 2002 the database had only 237 sequences, of which 158 were mycobacterial. Turenne et al. found that there were species not present in other public databases that were present in the RIDOM database (103). The RIDOM database is broadening and expanding their database with quality sequences.

The value of comparing unknown sequences to several different databases is addressed in Table Table2.2. The sequence of a recent blood isolate that we thought was a Streptococcus sp. (but with sequencing was shown to be Gemella bergiae) was identified using MicroSeq, GenBank, our own database, and the RDP-II and RIDOM databases (Table (Table2).2). In the three major databases, the degree of relatedness is expressed differently. GenBank's main measure is percent identity, MicroSeq's measure is percent dissimilarity and RDP-II's measure is a relatedness value somewhat close to (but lower than) the GenBank percent identity. It is enough to note, without comparing statistical merit, that all are valid. A dendrogram gives additional information by showing that Gemella haemolysans and G. morbillorum are easily distinguishable from each other although both are equidistant from the unknown isolate (Gemella bergiae).

Comparison of databases for the identification of a abscess isolate that seemed to be a Streptococcus sp. by routine phenotypic methods but was identified by 16S rRNA gene sequence homology as Gemella bergena

The problem of intraspecies variability, i.e., that all strains within a species do not have identical 16S rRNA gene sequences, requires the deposition of more than one sequence for each species (18). By examining multiple sequences available through GenBank for the same species and not counting the 20% with variation of over 2% that were misclassified, Clayton found there was still considerable variation in sequence within a species (18). As more clinical specimens are sequenced, interspecies genetic variability may become more obvious. On the other hand, strains with minor variability (less than 1%) are sometimes given separate species designation. The clinical significance of minor variation is becoming clearer and is addressed below (see “microheterogeneity” in the 16S rRNA gene sequence is common”).


Common Definitions of Genus or Species Derived by 16S rRNA Gene Sequence Analysis

At times, and particularly when compared with all the advances in molecular biology, taxonomy has been thought to be a relatively musty and obscure branch of microbiology. Thus, the taxon assignment for an isolate might be considered trivial: “a rose by any other name…” However, within the practice of clinical microbiology, the laboratory's decision on a name often precipitates decisions about clinical treatment and therapy. In addition, it is difficult to correctly elucidate the pathology of organisms or the disease process if a complex of organisms is referred to by a single name or if many names are used for a single species.

At this time, there exists no clear-cut consensus definition of bacterial genus or species by 16S rRNA gene sequence comparisons. This is a critical taxonomic concern, and a thorough discussion is beyond the scope of this review (26, 33, 55, 56, 94, 106, 108).

However, I will outline some of the practices and problems as they affect clinical microbiology. Although a proposal for defining species and genus using DNA-DNA hybridization as the “gold standard” has been published (108), DNA-DNA hybridization is a difficult technique, performed in only a few laboratories, and does not always correlate with other definitions of species (32). The 16S rRNA gene sequence is much easier to determine and thus has become the new gold standard (32, 44). Even though there is a consensus neither on the exact degree of genetic difference that defines a species nor on the mathematical algorithm used to generate the data (Fig. (Fig.3),3), in practice a range of about a 0.5 to 1% difference (99 to 99.5% similarity) is often used (91). Bosshard et al. (6) used ≥99% similarity to define a species and ≥95% to ≥99% to define a genus. Fox et al. proposed that there be a difference of at least 5 to 15 bp in the whole 16S rRNA gene sequence to define a species (33). Turenne et al. (103) designated the reportable range for a species as <0.8 to 2.0% and suggested that a sequence could obviously be called unique, i.e., representing a organism whose sequence has not yet been deposited and thus might be a novel species, if there were at least 20 to 38 bp difference in sequence. This gives a corresponding score of 0.961 to 0.916 in RDP-II (19, 103). Hall et al. adopted a distance score of 0.00% to less than 1.00% as the criterion for species identity (42). However, Tortoli found that Mycobacterium species could be validly named with differences of 4 bp or fewer (101). Tang et al. (97, 98) suggested a 0.5% difference as the limit for species designation. Often, a strain with a small genotypic difference (less than 0.5%) has been considered a subspecies (10). When there is a clear phenotypic uniqueness, genogroups with less than 1.0% differences in sequence have in fact been named as new species (48, 83, 101). For rickettsial isolates, Fournier et al. propose that members of the same species and genus have ≥99.8% and ≥98.1% 16S rRNA gene homology, respectively (32).

In addition, the total amount of intraspecies variability to be allowed is also not clear. For example, using the guideline that an unknown strain should be less than 1% different from the type strain, it might mean that hypothetical strain 1 and strain 2 of the same species are 2% different from each other. In general, there is agreement that all sequences of strains within the same species should be close (no more than 1 to 1.5% difference in base pair sequence) (33, 91). At the next higher level, one should think of establishing a new genus if genogroups are more than 5 to 7% divergent. The next section discusses how well these guidelines pertain to our present nomenclature. Major challenges will be in the nomenclature for organisms named before their correct taxonomy was revealed by 16S rRNA gene sequence comparisons and in dealing with microheterogeneity, i.e., a difference of only a few base pairs in sequence.

To summarize, it is not possible to give a definite similarity or dissimilarity value to define genus and species. This is in part because different values are generated by analyzing separate databases and using different methods (Table (Table2;2; Fig. Fig.3).3). The percent difference can vary if it is calculated using only the first 500 bp or all 1,500 bp and can also vary with the program used for the calculations. It also is probable that a single value for the definition of a genus or species on the basis of the 16S rRNA gene sequence is not appropriate for all genera (32, 44, 101).

Problems with the Present Nomenclature

In addition to the fundamental problems with both the phenotypic definition of species, which has been the standard until this time, and the genotypic definition of species, which is the present gold standard, there are special problems for some organisms in resolving relationships between genotype and phenotype. This is often the case with well-known genera and species that were named based only phenotypic criteria prior to the availability of the 16S rRNA gene sequence.

Examples of problems in assigning taxa and in finding a meaningful correspondence between genotype and phenotype are shown in Table Table3.3. The first two categories in Table Table33 (“Same genotype but different phenotypes” and “Similar genotype but different phenotypes”) show a critical problem for clinical practice and a drawback or deficiency in the 16S rRNA gene sequence identification method: that for some of the species, a sequence can be ambiguous since it does not distinguish between two closely related but distinct and sometimes clinically important species or phenotypes. Another way to say this is that there is more than one phenotype for a given genotype. Several important examples of those with the same genotype (or with only a few base pairs that are different) but different phenotypes are M. tuberculosis and M. bovis. M. avium and M. paratuberculosis, and M. kansasii and M. gastri (42, 103). Of the pathogenic bordetellae, B. pertussis. B. parapertussis, and B. bronchiseptica are not well distinguished by 16S rRNA gene sequence (48). Other such organisms are listed in the review by Fredericks and Relman (34). Roth et al. showed that on occasion, where 16S rRNA gene sequences were indistinguishable, such as for M. kansasii and M. gastri, or highly similar, such as for M. malmoense and M. szulgai, the 16S-23S rRNA gene internal transcribed spacer sequences were a helpful supplement for the differentiation of closely related species (82). Cases of unresolved identification by 16S rRNA gene sequence analysis can also be distinguished by the addition of a few carefully selected biochemical tests (42).

Insights into the relationships of genotype and phenotypea

By some guidelines, isolates with a small genotypic difference (ca. 0.4 to 0.9%) but a definite phenotypic difference have been considered either separate species or subspecies. In some cases, the isolates are given different names (e.g., S. pneumoniae and S. mitis), and in others they are given the same species name (e.g., some members of the S. anginosus subgroups [13]). However, there is great variability in practice (Table (Table3),3), since a comparison of sequences for several subspecies shows a differences of from 1 to 14 bp.

There is also the category “Similar phenotypes but different genotype.” Another way to say this is that there is more than one genotype for a given phenotype or that the phenotype is polyphyletic, a vexing but not uncommon problem that 16S rRNA gene sequencing is well positioned to solve. Recently we sequenced all our stored clinical Nocardia strains and were surprised to discover that essentially all the strains that, in times past, we had considered to be N. asteroides on the basis of phenotypic characteristics were indeed mostly N. farcinica and a few N. nova. We also found that our S. bovis collection consisted of two separate genogroups, although the clinical significance of this has not been determined (15, 45). Often, on further study, distinguishable phenotypes are found.

Table Table33 also contains the categories “Too distant to be the same species or genus” and “Too close to have different names,” and “Too close to be three different genera.” It is helpful to look at the number of base pair difference and Table Table44 to have an understanding of how variable interspecies and intergenus distances are in present-day practice. The obvious reason why there is great variation in what we call a genus and species, as the rules have not been uniformly applied, is that these were named before 16S rRNA gene sequencing was available. The consideration of subspecies is problematic since the difference between some subspecies is as great as that between some genera; the distance between the subspecies Streptococcus dysgalactiae subsp. dysgalactiae and Streptococcus dysgalactiae subsp. equisimilis is greater than between many genera. There are several additional categories. When the genotype and phenotype overlap, as with Afipia strains, it may be that strains overlap because there are so few isolates that the description may not be sufficiently accurate. Sometimes, if neither genotype nor phenotype is distinct but strains have been given different names, it could be because of historic reasons, simultaneous publication, or carelessness on the part of the authors and the editors.

Comparison of the genetic heterogeneity within generaa

Table Table44 is a way to present data showing the genetic variability within different genera. The data are modified from those presented by Montgomery et al. (S. Montgomery, S. Anderson, M. Waddington, J. Bartell, G. Num, and P. Foxall, Abstr. IXth Int. Congr. Bacterial Appl. Microbiol., abstr. 54, 1999). The interspecies variability for each of the genera listed in column has been calculated. The number of known species is given in column 2, and the percentage of known species that are in the MicroSeq database are shown in column 3. The genera are sorted on the basis of increasing average interspecies variability, which ranges from 0.5 to 17.3%. The last two columns can be used to evaluate genera: they indicate that there may be problems if the minimum distance between species is less than about 0.4 to 0.6% and the maximum distance is above about 5%. Thus, we see that Arthrobacter. Bacillus. Lactobacillus, and Clostridium, to name some of the most egregious examples, have at least two named species that have the same 16S rRNA gene sequence and at least one species that is highly unrelated to others in the same genus. Most of the genera in the first half of the table err only, if at all, in having two names for the same species or too little separation between species.

Table Table44 demonstrates inequalities in broadness between taxa. A well-known example is that the genetic difference between some genera that comprise the Enterobacteriaceae is smaller than the difference between some subspecies of Streptococcus (Table (Table3).3). Further, this difference is far smaller than in the single genus Clostridium. Therefore, by the Enterobacteriaceae standards for designating genera, many species of Clostridium would be given separate genus status. On the other hand, by Clostridium standards, all the members of the Enterobacteriaceae plus all the Vibrio. Aeromonas. Haemophilus, and Pasteurella groups would be considered to belong to only one genus (Fig. (Fig.22).

In general, one sees more variability reported in the 16S rRNA gene of species that are less well described and of lower pathogenicity than in the well-known pathogenic species. This is simply because organisms of low pathogenicity are usually not as well described as pathogens. Thus, in our justifiable lack of knowledge, we may be grouping genotypically and phenotypically heterogeneous isolates within one taxon. As a consequence, commensal or environmental isolates with a given name tend to have more variation in their 16S rRNA gene sequence than do well described highly pathogenic strains. An example of this is that strains identified on phenotypic grounds as S. anginosus or M. flavescens would tend to have more variation in their 16S rRNA gene sequence than strains identified on phenotypic grounds as S. aureus or M. tuberculosis.

Microheterogeneity in the 16S rRNA Gene Sequence Is Common

Sequevars, intraspecies variation, and variant subspecies genotypes are terms expressing the concept of microheterogeneity within a species. Usually this denotes differences of less than 0.5% or only a few base pair per 16S rRNA gene sequence. There is not a consensus about how to name organism groups showing microheterogeneity. The significance of microheterogeneity to clinical microbiologists is that it seems to allow the possibility of distinguishing important phenotype, pathogenicity, and niche differences between strains (10, 38, 84, 85). Microheterogeneity has also been exploited for strain tracking and epidemiological studies (38, 85).

16S rRNA gene sequence microheterogeneity has been found in many taxa. In Nocardia, substitutions of as little as 1 or 2 bp correlated with a unique phenotype (83). However, major differences in drug susceptibility patterns, important for clinical practice, were not found at this level but were found at the 1% difference range. Similar microheterogeneity is seen within the genus Mycobacterium (101, 103). Over 40 new species have been detected since 1990, most of which were grown from clinical samples and are potentially pathogenic. Many of them differ from another by only a few base pairs, but even small changes in the sequence seem to be correlated with unique phenotypic characteristics, clinical significance, and niche (101). Sometimes they are called sequevars (101). Because new sequences are found in almost all studies of clinical mycobacterial strains, the prospect is that many more sequevars will be detected, swelling the numbers of potential subspecies clades.

For the four groups that we have studied, S. anginosus. S. constellatus. H. influenzae, and H. parainfluenzae, the subspecies clade has correlated with a combination of biochemical profile and niche. As an example, 100 clinical strains of S. anginosus that we have sequenced cluster into about 13 groups or clades. Most of the clades differ by only 2 to 4 bp per 500-bp sequence from the next nearest clade, and many of the clades differ from each other by 6 bp (Fig. (Fig.4C).4C). We examined the phenotypic characteristics and niche for each clade. We found that the minute differences in sequence are reflected in phenotypic and niche differences. Some of this information is summarized in Table Table5,5, where methods of reporting and the significance of the eight possible differences at positions 71, 97, 137, 186, 274, 288, 463, and 487 are addressed. Similarly, Chen et al. found that genetic subspecies groupings of Pasteurella multocida might have clinical significance (10).

Possible ways to name organisms to capture all the information available by 16S rRNA gene sequence analysisa

As these microheterogeneities are increasingly shown to be correlated with important clinical or phenotypic characteristics, there is a need to recognize them. However, it is not clear that assigning a new name would be the most practical or informative practice. If species names are given to recognize every microheterogeneity, we may be confronted with a plethora of new names that might add confusion, particularly for the medical practioner. A practical interim solution might be using quantifying modifiers (i.e., the percent difference from the type strain) and pathologic descriptions (the most common site of isolation) attached to the primary species name (Table (Table5).5). A number of possible naming schemes are shown. The advantage of some of the possible nomenclatures is that they allow an immediately apparent relationship with better-known species to be understood without having to do a literature search with each clinical microbiology patient's report.


16S rRNA Gene Sequences Can Better Identify Poorly Described, Rarely Isolated, or Phenotypically Aberrant Strains

Because 16S rRNA gene sequence analysis can discriminate far more finely among strains of bacteria than is possible with phenotypic methods, it can allow a more precise identification of poorly described, rarely isolated, or phenotypically aberrant strains. This is an area in which 16S rRNA gene sequence identification might have an immediate impact on patient care. For example, some species of “viridans” streptococci are far more likely than others to cause endocarditis (2, 8). However, identification of these organisms by phenotypic methods is difficult and subject to error. Thus, judging the importance of isolates of “viridans” streptococci isolated from the blood has been a problem for the infectious-disease physician. However, 16S rRNA gene sequence analysis provides accurate identification at the species level and can clarify their clinical importance (14, 34; S. M. Attorri, M. Waddington, and J. E. Clarridge, Abstr. 100th Gen. Meet. Am. Soc. Microbiol. 2000, abstr. C-348, 2000).

Mistakes in identifying poorly described, rarely isolated, or phenotypically aberrant strains are probably quite common in the routine clinical laboratory. Even in large and complex research and reference laboratories, which have more time and phenotypic tests available for identification of bacteria, unusual bacteria often cannot be identified. Drancourt et al. published a list of 177 such difficult organisms (27). It is interesting that these organisms had been referred to the bioMérieux, Inc. (Marcy l'Etoile, France), laboratory by routine laboratories and intensive prior attempts to identify the organisms by biochemical methods had been made without success. For about 80% of the isolates, there was a close match with a described species. Another 10% represented a new species within a described genus, and about 10% of the organisms represented novel taxa. (27). Tang et al. compared a variety of identification systems including cellular fatty acid profiles, carbon source utilization, and conventional biochemical identification with the 16S rRNA gene sequence to evaluate both unusual aerobic gram-negative bacilli and coyneform organisms isolated from clinical specimens (97, 98). They found that 16S rRNA gene sequence provided more rapid, unambiguous identification of the difficult bacterial isolates than did convential methods and that this identification could translate to improved clinical outcomes (98). Bosshard et al. (6) found that only a minority of the clinical laboratory isolates of aerobic gram-positive rods could be correctly identified by phenotypic methods whereas rRNA gene sequencing is an excellent method for identifying these organisms, which are difficult to identify by convential methods. Woo et al. found that the MicroSeq 500 16S rRNA gene-based bacterial identification system for most clinically important bacterial strains with ambiguous biochemical profiles was limited only by the degree of completeness of the database (114).

However, it may be difficult for the primary laboratory to tell if an organism is difficult to identify or that it is incorrectly identified. For example, twice we isolated from blood an organism that was consistently identified by widely used phenotypic methods as Neisseria meningitidis or Actinobacillus actinomycetemcomitans but was unambiguously identified by sequencing as Francisella tularensis subsp. novicida (16). There was no history compatible with tularemia. Since two strains have been isolated from our hospital patients but none reported from the rest of the world, it is probable that other strains may be isolated but not correctly identified, since they were not thought to be “difficult” based on routine tests and thus were not sequenced. Similarly, we reported the isolation of Actinomyces israelii from a cervix and did not think it was a difficult organism. Later results of sequencing performed as part of a survey of our Actinomyces spp. showed it to be Bifidobacterium boum, yet it had a colony and biochemical profile similar to A. israelii and grew in air supplemented with CO2, despite Bifidobacterium boum being described as a strict anaerobe in the literature.

Some species of Nocardia are difficult to distinguish phenotypically but are readily identified by 16S rRNA gene sequence (83, 105). Sequence analysis has led to reevaluation of the strains that had been called N. asteroides and to reassignment of clinical significance (83). In our laboratory, we found that all of the presumed N. asteroides isolates associated with brain abscess (three) were actually N. farcinica when analyzed by 16S rRNA gene sequencing.

16S rRNA Gene Sequences Can Be Routinely Used for Identification of Mycobacteria

Mycobacteria are in general slow-growing and/or difficult to identify. Thus, they were an important group of organisms in early important studies establishing the usefulness of 16S rRNA gene sequencing for clinical microbiology (4, 5, 7, 29, 51, 80, 81, 99). More recently, there have been several additional studies comparing the identification of mycobacteria by 16S rRNA gene sequence and phenotypic methods (19, 22, 42, 50, 66, 92, 103). In all of these studies, the accuracy of 16S rRNA gene sequencing in the identification to the species level was judged to be superior overall to phenotypic methods. Overall, by providing for the accurate identification of species in the database and the taxonomic placement if not complete identification of novel species, 16S rRNA gene sequence analysis of mycobacteria seems to be the most accurate method available.

Exceptions are the known instances in which 16S rRNA gene sequencing could not differentiate among a limited number of species, e.g., M. avium and M. paratuberculosis. M. chelonae and M. abscessus, and M. tuberculosis and others in the M. tuberculosis group (19, 42, 66, 80). In a large reference laboratory, Hall et al. found that 16S rRNA gene sequence could identify 243 of 328 clinical isolates with a distance score of <1% (42). In this group, the agreement with phenotypic identification was 90.1%, with the discrepant results being those in the groups which sequencing cannot distinguish, as given above. The remaining 85 isolates had distance scores above 1% but below 4% and thus were determined to be within the genus Mycobacterium: in this case, either novel species or species that exhibited significant genotypic divergence from an organism in the database with the closest match. The power of sequence analysis is demonstrated since the 85 organisms were “known to be unknown” by sequence analysis whereas the phenotypic testing identified most of these incorrectly as known species. The authors recommended integration of nucleic acid sequencing into the routine mycobacteriology laboratory after the use of genetic probes for the most common species. The use of the MicroSeq 500 microbial identification system and internal (Mayo Clinic) databases containing additional sequences significantly reduced the number of organisms that could not be identified by phenotypic methods. Cloud et al. also used the MicroSeq 500 system and found that another database (RIDOM) was needed to provide sequences for additional species (19). Most of the recent studies agree that a few known organisms such as M. lentiflavum are missing from the MicroSeq 500 database and that about 7 to 12% of the isolates were novel strains (18, 42). Patel et al. found that 12 of 113 strains had over 0.9% difference from any known strains (66). Not surprisingly, older studies found a larger percentage of unidentified strains because of the less extensive databases available at the time (92). A European consortium, using phenotypic testing, high-performance liquid chromatography, and 16S rRNA gene sequencing, also found that about 7% of mycobacterial strains represented novel species (102). A few examples of the many good papers in which novel mycobacterial species are described using both biochemical and genotypic descriptions and more than one strain are included in the reference list (90, 91, 101).

16S rRNA Gene Sequence Analysis Can Lead to the Discovery and Description of Novel Pathogens

In clinical microbiology practice, novel organisms are generally first recognized by an aberrant phenotype or niche. If these observations are followed by 16S rRNA gene sequencing, the sequence often indicates that the organism is only an unusual phenotype of a known taxon (Table (Table3).3). However, at this time, perhaps 10 to 20% of the isolates might not match any other described organism and thus might be a novel organism and an even higher percentage might be previously described organisms but not strains usually encountered in clinical practice (25, 76, 102). A medium to large clinical microbiology laboratory can be expected to isolate a few novel organisms per month.

Where should one look for new species? We look for novel organisms among the groups that are poorly described or difficult to grow. We have found that there is little potential for finding novel genogroups within the well-studied major pathogenic species, since these taxonomic groups have been so well and accurately described that there is little confusion or variability in their 16S rRNA gene sequence. Species with this genetically homogeneous characteristic that we have tested include M. tuberculosis. H. ducreyi. H. influenzae group b, Streptococcus dysgalactiae. S. pneumoniae (blood isolates), and Staphylococcus aureus.

If we plotted all known strains in a dendrogram such as in Fig. Fig.2,2, we would see that some taxa are very, very crowded. For example, the family Enterobacteriaceae has poor potential for harboring novel types, since so many phenotypes have already been described. However, other genetic groups are not as well studied, and few have had as many distinct species grown, isolated, and sequenced. At this time, some other taxa such as the Leptotrichia-Streptobacillus-Fusobacterium group remain quite sparse. It is probable that many future isolates in this group may represent novel species. Other potentially fruitful taxa which are beginning to be more fully described include the Actinomyces-Actinobacillus genera and the less common catalase-negative gram-positive cocci (6, 17, 20, 21, 43, 89). One can also examine Table Table4;4; where the interspecies difference is great, it is probable that there are undiscovered bacteria in the group. In a recent review of taxonomic groups of mycobacteria, Tortoli (101) presented the essential contribution made by 16S rRNA gene sequencing not only to distinguishing 42 new species but also to realigning of classically known species of slow and rapid growers into new groupings. Large numbers of anaerobes remain to be described (90, 91). Another fruitful area is that of organisms associated with animals; a seemingly large percentage remain to be well described (21).

In the last 15 years, there have been thousands of publications utilizing 16S rRNA gene sequence as part of a species description. Some representative journals reflect both the numbers of novel organisms being described and the uses of 16S sequencing. In 2001 in Applied and Environmental Microbiology, there were 116 reports of such studies, many of which used 16S rRNA gene sequencing to assess the proportions of species in populations of bacteria in a particular environment (see, e.g., reference 88; see also reference 46). In 2003, there were 181 such reports. In contrast, in the Journal of Clinical Microbiology, many of the 94 citations in 2001 and 91 citations in 2003 concerned the correct description of a novel potential pathogen (e.g., Helicobacter winghamensis) (58). In 2003 in the International Journal of Systematic and Evolutionary Microbiology, about 300 novel species were reported with the 16S rRNA gene sequence as part of the description.

16S rRNA Gene Sequence Analysis Can Identify Noncultured Bacteria

Although the 16S rRNA gene sequence is an essential part of the description of a novel organism, for many noncultured bacteria it may be the only taxonomic description (34, 76, 78). Since all organisms are obviously able to grow under the proper conditions, the terms “noncultured” or “not easily cultured” are preferable to “nonculturable,” which is sometimes used. Indeed, we recently grew and described a strain (Fig. (Fig.2,2, strain 01-1398) whose sequence in GenBank was identified as “nonculturable.”

The basic method to derive a sequence for a noncultured bacterium is to use universal primers against the 16S rRNA gene region in a PCR step to increase the amount of DNA and then to sequence the amplicon (7, 11, 39, 72, 75, 77). This will work well if there is only one organism to detect. A different strategy is imposed if there is a situation in which a mixed culture is likely, such as a clinical specimen from a nonsterile site or an environmental sample.

There are valid reasons why bacterial strains that we may encounter in clinical practice are not recovered. For example, prior antibiotic treatment may render them nonviable. There are several interesting publications in which the etiological agents of “culture-negative” endocarditis (8, 9, 37, 39, 47, 109) were identified by molecular analysis using broad-range PCR primers complementary to the 16S rRNA gene, sequencing, and database searches with different software. The organisms usually can be grown (e.g., Streptococcus salivarius [109] and Capnocytophaga canimorsus [our laboratory]), but in these cases the specimens were submitted once the patient was being given antibiotics.

A second reason is that the organisms may be genuinely hard to grow. Bartonella henselae, the causative agent of bacillary angiomatosis, was primarily identified and associated with the disease in this way (77). Many of the reports of Whipple's disease are based solely on PCR amplification and subsequent sequencing of the 16S rRNA genes (9, 78), because the etiologic agent, Tropheryma whipplei, has demanding growth requirements beyond the capabilities of most clinical laboratories. It is interesting that Celard et al. found T. whipplei as the causative agent in two cases of endocarditis without there being previous evidence of related disease (9). Brouqui and Raoult (8), using broad-based PCR amplification of the 16S rRNA gene, found that the most common etiologic agents associated with culture-negative endocarditis were Bartonella quintana and Coxiella burnetii, both of which require special conditions to grow. We routinely freeze at −70°C a portion of the removed valves if at that time we are still uncertain about the cause of a case of endocarditis. If the pathogen is not subsequently identified by culture, we subject the blood culture bottles and valve to PCR with universal primers and perform sequence analysis on the amplified product. In most cases, the process is successful only if the organisms are present in sufficient numbers to be seen by electron or light microscopy (37, 109). In an excellent presentation of 29 cases of histologically confirmed infective endocarditis, direct amplification and sequence analysis was confirmatory in 21 cases and essential to the diagnosis in 5 cases (37). Because many Mycobacterium spp. are slow-growing and costly to identify, it has been useful to identify them by direct sequencing of amplified DNA (7, 80).

A third situation in which these methods are useful is when there are mats of adherent, diverse, and unknown organisms. Periodontal disease and biofilms have been studied in this way (25). The diversity of organisms present in the subgingival pockets of patients with periodontitis and acute necrotizing ulcerative gingivitis was examined by amplification of the 16S rRNA gene using PCR with a universal forward primer and a spirochete-selective reverse primer. The amplified DNA was cloned into Escherichia coli. The DNA segment was then sequenced, and the sequences were compared as described above. Novel genotypes are commonly found, such as those representing Atopobium species and a new genus, Olsenella gen. nov. (25).

Similar studies of the environment are also yielding a wealth of novel organisms. The medically important order Chlamydiales has long been considered to contain a few closely related bacteria which occur exclusively in animals and humans. However, employing techniques similar to those described above, Horn and Wagner (46) found at least four novel evolutionary lineages of Chlamydiales in environmental sludge. These findings suggest that some wastewater treatment plants represent reservoirs for a diverse assemblage of environmental chlamydiae and suggest that the environment may be a source of novel organisms that might have public health consequences.

Automatic instruments used to detect growth in blood culture bottles sometimes flag positive in the absence of apparent growth. A unique use of direct 16S rRNA gene sequence analysis with universal primers was to search for noncultured bacteria as a source of the presumably false-positive blood cultures (71). The investigators did not find any previously undetected bacterium as the cause of the false-positive blood cultures.


At this time, for routine identifications, 16S rRNA gene sequence analysis is more expensive than most traditional identification methods. However for difficult organisms, multiple identification methods often must be used, which increases the cost. Thus, Hall et al. (42) found that for the identification of unusual mycobacteria, sequence analysis was less than $2 more expensive than standard methods. Patel et al. estimated the costs to be $144 for a single isolate but only about half of that when multiple specimens were analyzed at the same time (65, 66). Wilck et al. published a similar figure, finding that the material costs, excluding labor, amounted to $59 per specimen (this included the deparaffinization and DNA extraction from the surgical specimen, materials for PCR, PCR cleanup, sequencing, and the processing of a control) (109). Cook et al. (22) found even lower costs for identification of nontuberculous mycobacteria: without estimating capital equipment charges, the cost for labor and materials were $48 Canadian (about $32 U.S.).

We examined the costs for microbial identification based on 16S rRNA gene sequences with a sequence analyzer based in the clinical microbiology laboratory. We used the MicroSeq software and two different DNA sequences. ABI model 3100 was superior to ABI model 310 in the length of reliable sequence (520 and 460 bp, respectively), in addition to the published superiority of ease of use and time of run. This decreased the editing time and even allowed a correct identification with only the forward sequence (Clarridge et al., Abstr. 101st Gen. Meet. Am. Soc. Microbiol. 2001). Bottgers group routinely uses just one sequence. We found, not including instrument purchase, that the cost per isolate is about $44 (material costs of $20 and labor costs of $24; the labor costs were up from the 2001 estimate in the abstract mentioned above). Our costs assume some batch preparation of specimens. Our calculations were made on the basis of using the reagent and disposables for the ABI PRISM 3100 genetic analyzer (with 16 capillaries), but they are comparable to the cost for the ABI PRISM 3100-Avant Genetic analyzer (with 4 capillaries). We do not process extra specimens as controls since the success of the process can be judged by the final product. If the instrument is on a reagent-purchase agreement for other tests such as human immunodeficiency virus genotyping, instrument costs can be lower. Cook et al. (22) calculated that for the identification of nontuberculous mycobacteria, one had to perform 100 analyses per month to have the cost of identification, including the instrument, be lower than cost of using conventional identification methods. In addition, especially if a sequence analyzer is available, being shared, for example, with human immunodeficiency virus genotyping tests or research, the analysis can be more cost-effective and timely.

If one does not have a sequence analyzer in-house, sequence analysis is available at many university core laboratories. The costs can be about $30 to $50 for 500 bp if the isolate is sent and $12 to $18 if the DNA is provided in a condition ready to be put on the analyzer. These costs include both the forward and reverse sequences. The quality of sequences from such institutions can vary considerably. To assemble the sequence and make a comparison with databases with an evaluation, there is usually an additional $50 to $80 charge. Laboratories offer the complete service of identification of bacteria by sequence analysis for about $100 to $150.

All reports cited above agree that these costs for the definitive identification provided by 16S sequence analysis are reasonable when the costs of multiple investigations done in the attempt to establish an identification are considered. In one study, the turnaround time for a mycobacterial identification was shortened to 24 h and the results were reported much earlier (42). The additional costs of possible patient-related sequelae due to improper or lack of identification of an isolate, such as broad-spectrum intravenous antibiotics or incorrect diagnosis, were not estimated.


This section summarizes some standards that reviewers and editors of clinical microbiology and infectious-disease journals and clinical laboratory personnel who may not be actively engaged in sequence analysis themselves need to be aware of to evaluate sequence-based data for laboratory use or publication.

(i) Novel organisms. The 16S rRNA gene sequence is an essential part of the description of a novel organism. The authors should be required to include 16S rRNA gene sequence data in every publication describing a new strain, preferably the complete gene sequence (94). It is helpful to reviewers if this information is accessible in electronic form. A summary of the differences from type strains could be published, but the whole sequence need not be published since it wastes space (68). The sequence should contain fewer than 1% undetermined bases without an explanation.

(ii) Poorly described organisms. The 16S rRNA gene sequence identification should be performed (even if not reported) on isolates for which unusual disease associations are being asserted if the isolate is not extremely well known and easy to identify. For example, it is permissible to refer to M. tuberculosis or to S. pyogenes without sequence data but not to refer to the rapidly growing environmental mycobacteria or to the viridans streptococci as a whole or most of the species that comprise the viridans streptococci or Actinomyces spp., because in these groups the phenotype is often polyphyletic and descriptions may be ambiguous. Examples of other ambiguous identifications are given in Table Table3.3. The large changes in the names and descriptions of gram-positive rods and anaerobes make sequence identification necessary for any publication in which an association is claimed between an unusual microbe and a disease.

(iii) Number of strains. It is preferable that there be three or more similar isolates for any publication of a new species, strain, or subspecies; some recommend that there be five strains (12, 18). However, single-strain reports of rare isolates continue to be published. Because the phenotype continues to be extremely important to clinical microbiologists, phenotypic tests should be associated with the sequence. The same phenotypic tests for the new isolate should be performed in the same laboratory on several known strains; i.e., it is not right to compare tests performed in the investigator's laboratory only on the strain being described with literature values for the type or known strains with which the strains will be compared: the literature values can be wrong, or methods may not be identically performed.

(iv) At least 0.5% and possibly 1% difference is needed to warrant a new species name. As a rule of thumb until firmer rules are propagated by taxonomists, unless there are overwhelming and well-documented phenotypic reasons, there should be a difference of at least 1 bp per 100 bases sequenced to warrant assigning a new species name. If there are clear phenotypic differences, as little as 0.5% may warrant a new species name. Recall that calculations using mathematically different search algorithms such as BLAST and Needleman Wunsch may not give the same results. When differences between isolates are larger than 5%, consideration should be given to generating a new genus name.

(v) Sequence trumps phenotype. For an unknown strain, when relatedness by sequence and biochemical or morphological characteristics do not agree, in general the sequence trumps the phenotype. This does not mean that the genotype cannot still be ambiguous, as detailed in the first examples of closely related taxa in Table Table3.3. If the genotype-phenotype discrepancy is an unexpected result, it is worth rechecking that the sequence and phenotypic studies have been performed on the same exact strain.

(vi) What the sequencing laboratory needs to tell the client. The report from the laboratory performing the sequence analysis should include the name of the bacterium and the percent difference from the nearest type strain in the database used. If this is a novel isolate or one that does not match closely with a type strain, it would be helpful if information on other similar isolates and an alignment of the sequence with nearest known strain, showing just the differences and the sequence in a text file, were also included. Thus, for example, the client might receive the following report Capnocytophaga canimorsis (2.46%) and a text file of the actual sequence. The client can then examine the quality of the sequence. If not satisfied (using criteria outlined above), the client could ask for a resequencing. If the quality is good but the client considers the 2.46% difference is too large, the sequence in the text file can be compared to all sequences in GenBank or another database by using the free program, BLAST, to see if there is a closer match. We would perform a BLAST search using a public database on a sequence with greater than 1% difference from a known type strain. If there was no closer match to a type strain, the organism could be reported as C. canimorsus (most closely resembling) or C. canimorsus, novel genotype. Table Table55 shows different ways of reporting.


A colleague once remarked that he thought it was astounding that the simple stain developed by Gram would have profound structural and taxonomic significance. I find it remarkable and equally serendipitous that 16S rRNA gene sequence would have almost the exact amount of variability to define a species or at least provide a clinically useful distinction among bacterial strains.

We have seen that identifying bacteria isolated in the clinical laboratory by sequence instead of phenotype can improve clinical microbiology by better identifying poorly described, rarely isolated, or biochemically aberrant strains. 16S rRNA gene sequences allow bacterial identification that is more robust, reproducible, and accurate than that obtained by phenotypic testing. The test results are less subjective. 16S rRNA gene sequence analysis can lead to the discovery of novel pathogens. 16S rRNA gene sequence analysis can identify noncultured bacteria, allowing independence from growth conditions.

The correct designation of organisms is important. For example, when we refer to a complex of organisms by a single name and these organisms have different pathogenic potential, the disease process is obscured. As it is recognized that the correct taxonomy or name assignment can make a difference in clinical outcome, there should be a demand for more widespread use of the accurate identifications that 16S rRNA gene sequence analysis can provide.

A downside to the better discrimination provided by 16S rRNA gene sequence analysis is that it introduces a communication difficulty, since there are many more distinct sequences than names or phenotypic descriptions. Without a one-to-one correspondence, there can be a problem in assigning names to sequences in a meaningful way. The additional data may also be difficult to fully communicate to clinical colleagues. An attempt to address this is presented in Table Table5.5. Further, the information that 16S rRNA gene sequencing has made available so far (and even more so in the future) confronts the clinical microbiologist with having to change some familiar concepts of species identification.

16S rRNA gene sequencing traditionally played a limited role in the identification of microorganisms in clinical microbiology laboratories, mainly due to high costs, requirements for great technical skill, and the lack of user-friendly comparative sequencing analysis software and validated databases. However, the availability of improved DNA sequencing techniques, vastly increased databases and more readily available kits and software, makes this technology a competitive alternative to routine microbial identification techniques for some groups of organisms, such as mycobacteria. The costs can be also comparable to traditional identification methods for other slow-growing and difficult-to-identify organisms, particularly if a sequencer is available for multitasking in other sections of the laboratory.

An additional important function for 16S rRNA gene sequencing is to provide accurately grouped organisms for further study. Despite its accuracy, 16S rRNA gene sequence analysis lacks widespread use beyond the large and reference laboratories because of technical and cost considerations. Thus, a future challenge for the large clinical, reference, and research laboratories is to translate information from 16S rRNA gene sequencing into convenient biochemical testing schemes, making the accuracy of the genotypic identification available to the smaller and routine clinical microbiology laboratories.


I am grateful to the following people for their knowledge, helpfulness, and willingness to share: Silvia Attorri, John Bartell, Sharon Heward, Kristina Hulten, S. Montgomery, Liane Tsai, R. Visanas, Michael Waddington, Qing Zhang, and the people in the Microbiology Laboratories at the VA Medical Center, Houston and Seattle.


1. Altschul, S. F., W. Gish, W. Miller, E. W. Myers, and D. J. Lipman. 1990. Basic Local Alignment Search Tool. J. Mol. Biol. 215:729-731. [PubMed]
2. Banks, J., S. Poole, S. P. Nair, J. Lewthwaite, P. Tabona, R. McNab, M. Wilson, A. Paul, and B. Henderson. 2002. Streptococcus sanguis secretes CD14-binding proteins that stimulate cytokine synthesis: a clue to the pathogenesis of infective (bacterial) endocarditis. Microb. Pathog. 32:105-116. [PubMed]
3. Bansal, A. K., and T. E. Meyer. 2002. Evolutionary analysis by whole- genome comparisons. J. Bacteriol. 184:2260-2272. [PMC free article] [PubMed]
4. Boddinghaus, B., J. Wolters, W. Heikens, and E. C. Bottger. 1990. Phylogenetic analysis and identification of different serovars of Mycobacterium intracellulare at the molecular level. FEMS Microbiol. Lett. 58:197-203. [PubMed]
5. Boddinghaus, B., T. Rogall, T. Flohr, H. Blocker, and E. C. Bottger. 1990. Detection and identification of mycobacteria by amplification of rRNA. J. Clin. Microbiol. 28:1751-1759. [PMC free article] [PubMed]
6. Bosshard, P. P., S. Abels, R. Zbinden, E. C. Bottger, and M. Altwegg. 2003. Ribosomal DNA sequencing for identification of aerobic gram-positive rods in the clinical laboratory (an 18-month evaluation). J. Clin. Microbiol. 41:4134-4140. [PMC free article] [PubMed]
7. Bottger, E. C. 1989. Rapid determination of bacterial ribosomal RNA sequences by direct sequencing of enzymatically amplified DNA. FEMS Microbiol. Lett. 65:171-176. [PubMed]
8. Brouqui, P., and D. Raoult. 2001. Endocarditis due to rare and fastidious bacteria. Clin. Microbiol. Rev. 14:177-207. [PMC free article] [PubMed]
9. Celard, M., G. Gevigney, S. Mosnier, P. Buttard, Y. Benito, J. Etienne, and F. Vandenesch. 1999. Polymerase chain reaction analysis for diagnosis of Tropheryma whippelii infective endocarditis in two patients with no previous evidence of Whipple's disease. Clin. Infect. Dis. 29:1348-1349. [PubMed]
10. Chen, H., K. Hulten, and J. E. Clarridge III. 2002. Taxonomic subgroups of Pasteurella multocida correlate with clinical presentation. J. Clin. Microbiol. 40:3438-3441. [PMC free article] [PubMed]
11. Chen, K., H. Neimark, P. Rumore, and C. R. Steinman. 1989. Broad-range DNA probes for detecting and amplifying eubacterial nucleic acids. FEMS Microbiol. Lett. 57:19-24. [PubMed]
12. Christensen, H., M. Bisgaard, W. Frederiksen, R. Mutters, P. Kuhnert, and J. E. Olsen. 2001. Is characterization of a single isolate sufficient for valid publication of a new genus or species? Proposal to modify recommendation 30b of the Bacteriological Code (1990 revision). Int. J. Syst. Bacteriol. 51:2221-2222. [PubMed]
13. Clarridge, J. E., III, C. Osting, M. Jalali, J. Osborne, and M. Waddington. 1999. Genotypic and phenotypic characterization of “Streptococcus milleri” group isolates from a Veterans Administration hospital population. J. Clin. Microbiol. 37:3681-3687. [PMC free article] [PubMed]
14. Clarridge, J. E., III, S. Attorri, D. M. Musher, J. Hebert, and S. Dunbar. 2001. Streptococcus intermedius, Streptococcus constellatus, and Streptococcus anginosus (“Streptococcus milleri group”) are of different clinical importance and are not equally associated with abscess. Clin. Infect. Dis. 32:1511-1515. [PubMed]
15. Clarridge, J. E., III, S. M. Attorri, Q. Zhang, and J. Bartell. 2001. 16S ribosomal DNA sequence analysis distinguishes biotypes of Streptococcus bovis: Streptococcus bovis biotype II/2 is a separate genospecies and the predominant clinical isolate in adult males. J. Clin. Microbiol. 39:1549-1552. [PMC free article] [PubMed]
16. Clarridge, J. E., III, T. J. Raich, A. Sjosted, G. Sandstrom, R. O. Darouiche, R. M. Shawar, P. R. Georghiou, C. Osting, and L. Vo. 1996. Characterization of two unusual clinically significant Francisella strains. J. Clin. Microbiol. 34:1995-2000. [PMC free article] [PubMed]
17. Clarridge, J. E., III, and Q. Zhang. 2002. Genotypic diversity of clinical Actinomyces species: phenotype, source, and disease correlation among genospecies. J. Clin. Microbiol. 40:3442-3448. [PMC free article] [PubMed]
18. Clayton, R. A., G. Sutton, P. S. Hinkle, Jr., C. Bult, and C. Fields. 1995. Intraspecific variation in small-subunit rRNA sequences in GenBank: why single sequences may not adequately represent prokaryotic taxa. Int. J. Syst. Bacteriol. 45:595-599. [PubMed]
19. Cloud, J. L., H. Neal, R. Rosenberry, C. Y. Turenne, M. Jama, D. R. Hillyard, and K. C. Carroll. 2002. Identification of Mycobacterium spp. by using a commercial 16S ribosomal DNA sequencing kit and additional sequencing libraries. J. Clin. Microbiol. 40:400-406. [PMC free article] [PubMed]
20. Collins, M. D., R. A. Hutson, N. Nikolaitchouk, A. Nyberg, and E. Falsen. 2003. Luteococcus sanguinis sp. nov., isolated from human blood. Int. J. Syst. Evol. Microbiol. 53:1889-1891. [PubMed]
21. Collins, M. D., R. Higgins, S. Messier, M. Fortin, R. A. Hutson, P. A. Lawson, and E. Falsen. 2003. Allofustis seminis gen. nov., sp. nov., a novel gram-positive, catalase-negative, rod-shaped bacterium from pig semen. Int. J. Syst. Evol. Microbiol. 53:811-814. [PubMed]
22. Cook, V. J., C. Y. Turenne, J. Wolfe, R. Pauls, and A. Kabani. 2003. Conventional methods versus 16S ribosomal DNA sequencing for identification of nontuberculous mycobacteria: cost analysis. J. Clin. Microbiol. 41:1010-1015. [PMC free article] [PubMed]
23. Coyle, M. B., R. B. Leonard, D. J. Nowowiejski, A. Malekniazi, and D. J. Finn. 1993. Evidence of multiple taxa within commercially available reference strains of Corynebacterium xerosis. J. Clin. Microbiol. 31:1788-1793. [PMC free article] [PubMed]
24. Devulder, G., G. Perrière, F. Baty, and J. P. Flandrois. 2003. BIBI, a bioinformatics bacterial identification tool. J. Clin. Microbiol. 41:1785-1787. [PMC free article] [PubMed]
25. Dewhirst, F. E., B. J. Paster, N. Tzellas, B. Coleman, J. Downes, D. A. Spratt, and W. G. Wade. 2001. Characterization of novel human oral isolates and cloned 16S rDNA sequences that fall in the family Coriobacteriaceae: description of Olsenella gen., reclassification of Lactobacillus uli as Olsenella uli comb. Description of Olsenella profusa sp. Int. J. Syst. Bacteriol. 51:1797-1804. [PubMed]
26. Dixon, B. 2001. What's in a name? Lancet Infect. Dis. 1:212. [PubMed]
27. Drancourt, M., C. Bollet, A. Carlioz, R. Martelin, J. P. Gayral, and D. Raoult. 2000. 16S ribosomal DNA sequence analysis of a large collection of environmental and clinical unidentifiable bacterial isolates. J. Clin. Microbiol. 38:3623-3630. [PMC free article] [PubMed]
28. Dubnau, D., I. Smith, P. Morell, and J. Marmur. 1965. Gene conservation in Bacillus species. I. Conserved genetic and nucleic acid base sequence homologies. Proc. Natl. Acad. Sci. USA 54:491-498. [PMC free article] [PubMed]
29. Edwards, U., T. Rogall, H. Blocker, M. Emde, and E. C. Bottger. 1989. Isolation and direct complete nucleotide determination of entire genes. Characterization of a gene coding for 16S ribosomal RNA. Nucleic Acids Res. 17:7843-7853. [PMC free article] [PubMed]
30. Felsenstein, J. 1981. Evolutionary trees from DNA sequences: a maximum likelihood approach. Mol. Evol. 17:368-376. [PubMed]
31. Felsenstein, J. 1989. PHYLIP—phylogeny inference package (version 3.2). Cladistics 5:164-166.
32. Fournier, P. E., J. S. Dumler, G. Greub, J. Zhang, Y. Wu, and D. Raoult. 2003. Gene sequence-based criteria for identification of new Rickettsia isolates and description of Rickettsia heilongjiangensis sp. nov. J. Clin. Microbiol. 41:5456-5465. [PMC free article] [PubMed]
33. Fox, G. E., J. D. Wisotzkey, and P. Jurtshuk, Jr. 1992. How close is close: 16S rRNA sequence identity may not be sufficient to guarantee species identity. Int. J. Syst. Bacteriol. 42:166-170. [PubMed]
34. Fredricks, D. N., and D. A. Relman. 1996. Sequence-based identification of microbial pathogens: a reconsideration of Koch's postulates. Clin. Microbiol. Rev. 9:18-33. [PMC free article] [PubMed]
35. Funke, G., A. V. Graevenitz, J. E. Clarridge III, and K. Bernard. 1997. Clinical microbiology of coryneform organisms. Clin. Microbiol. Rev. 10:125-159. [PMC free article] [PubMed]
36. Garrity, G. M., and J. G. Holt. 2001. The road map to the manual, p. 119-166. In G. M. Garrity (ed), Bergey's manual of systematic bacteriology. Springer-Verlag, New York, N.Y.
37. Gauduchon, V., L. Chalabreysse, J. Etienne, M. Célard, Y. Benito, H. T. Lepidi, F. Thivolet-Béjui, and F. Vandenesch. 2003. Molecular diagnosis of infective endocarditis by PCR amplification and direct sequencing of DNA from valve tissue. J. Clin. Microbiol. 41:763-766. [PMC free article] [PubMed]
38. Gee, J. E., C. T. Sacchi, M. B. Glass, B. K. De, R. S. Weyant, P. N. Levett, A. M. Whitney, A. R. Hoffmaster, and T. Popovic. 2003. Use of 16S rRNA gene sequencing for rapid identification and differentiation of Burkholderia pseudomallei and B. mallei. J. Clin. Microbiol. 41:4647-4654. [PMC free article] [PubMed]
39. Goldenberger, D., A. Kunzli, P. Vogt, R. Zbinden, and M. Altwegg. 1997. Molecular diagnosis of bacterial endocarditis by broad-range, PCR amplification and direct sequencing. J. Clin. Microbiol. 35:2733-2739. [PMC free article] [PubMed]
40. Gupta, S., and M. C. Maiden. 2001. Exploring the evolution of diversity in pathogen populations. Trends Microbiol. 9:181-185. [PubMed]
41. Haas, W. H., W. R. Butler, P. Kirschner, B. B. Plikaytis, M. B. Coyle, B. Amthor, A. G. Steigerwalt, D. J. Brenner, M. Salfinger, J. T. Crawford, E. C. Bottger, and H. J. Bremer. 1997. A new agent of mycobacterial lymphadenitis in children. Mycobacterium heidelbergense sp. nov. J. Clin. Microbiol. 35:3203-3209. [PMC free article] [PubMed]
42. Hall, L., K. A. Doerr, L. S. Wohlfiel, and G. D. Roberts. 2003. Evaluation of the MicroSeq system for Identification of mycobacteria by 16S ribosomal DNA sequencing and its integration into a routine clinical mycobacteriology laboratory. J. Clin. Microbiol. 41:1447-1453. [PMC free article] [PubMed]
43. Hall, V., M. D. Collins, R. A. Hutson, E. Falsen, E. Inganas, and B. I. Duerden. 2003. Actinobaculum urinale sp. nov., from human urine. Int. J. Syst. Evol. Microbiol. 53:679-682. [PubMed]
44. Harmsen, D., and H. Karch. 2004. 16S rDNA for diagnosing pathogens: a living tree. ASM News 70:19-24.
45. Herrero, I. A., M. S. Rouse, K. E. Piper, S. A. Alyaseen, J. M. Steckelberg, and R. Patel. 2002. Reevaluation of Streptococcus bovis endocarditis cases from 1975 to 1985 by 16S ribosomal DNA sequence analysis. J. Clin. Microbiol. 40:3848-3850. [PMC free article] [PubMed]
46. Horn, M., and M. Wagner. 2001. Evidence for additional genus-level diversity of Chlamydiales in the environment. FEMS Microbiol. Lett. 204:71-74. [PubMed]
47. Jalava, J., P. Kotilainen, S. Nikkari, M. Skurnik, E. Vänttinen, O. P. Lehtonen, E. Eerola, and P. Toivanen. 1995. Use of the polymerase chain reaction and DNA sequencing for detection of Bartonella quintana in the aortic valve of a patient with culture-negative infective endocarditis. Clin. Infect. Dis. 21:891-896. [PubMed]
48. Kattar, M. M., J. F. Chavez, A. P. Limaye, S. L. Rassoulian-Barrett, S. L. Yarfitz, L. C. Carlson, Y. Houze, S. Swanzy, B. L. Wood, and B. T. Cookson. 2001. Application of 16S rRNA gene sequencing to identify Bordetella hinzii as the causative agent of fatal septicemia. J. Clin. Microbiol. 38:789-794. [PMC free article] [PubMed]
49. Kimura, M. 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. Mol. Evol. 16:111-120. [PubMed]
50. Kirschner, P. and E. C. Bottger. 1998. Species identification of mycobacteria using rDNA sequencing. Methods Mol. Biol. 101:349-361. [PubMed]
51. Kirschner, P., B. Springer, U. Vogel, A. Meier, A. Wrede, M. Kiekenbeck, F. C. Bange, and E. C. Bottger. 1993. Genotypic identification of mycobacteria by nucleic acid sequence determination: report of a 2-year experience in a clinical laboratory. J. Clin. Microbiol. 31:2882-2889. [PMC free article] [PubMed]
52. Kolbert, C. P. and D. H. Persing. 1999. Ribosomal DNA sequencing as a tool for identification of bacterial pathogens. Curr. Opin. Microbiol. 2:299-305. [PubMed]
53. Koukila-Kahkola, P., B. Springer, E. C. Bottger, L. Paulin, E. Jantzen, and M. L. Katila. 1995. Mycobacterium branderi sp. nov., a new potential human pathogen. Int. J. Syst. Bacteriol. 45:549-553. [PubMed]
54. Krieg, N. R. and J. G. Holt (ed.). 1984. Bergey's manual of systematic bacteriology. The Williams & Wilkins Co., Baltimore, Md.
55. Krieg, N. R., and G. M. Garrity. 2001. On using the manual, p. 15-19. In G. M. Garrity (ed.), Bergey's manual of systematic bacteriology, 2nd ed. Bergey's Manual Trust. Springer-Verlag, New York, N.Y.
56. Lan, R., and P. R. Reeves. 2001. When does a clone deserve a name? A perspective on bacterial species based on population genetics. Trends Microbiol. 9:419-423. [PubMed]
57. Maidak, B. L., J. R. Cole, C. T. Parker, Jr., G. M. Garrity, N. Larsen, B. Li, T. G. Lilburn, M. J. McCaughey, G. J. Olsen, R. Overbeek, S. Pramanik, T. M. Schmidt, J. M. Tiedje, and C. R. Woese. 1999. A new version of the RDP (Ribosomal Database Project). Nucleic Acids Res. 27:171-173. [PMC free article] [PubMed]
58. Melito, P. L., C. Munro, P. R. Chipman, D. L. Woodward, T. F. Booth, and F. G. Rodgers. 2001. Helicobacter winghamensis sp. nov., a novel Helicobacter sp. isolated from patients with gastroenteritis. J. Clin. Microbiol. 39:2412-2417. [PMC free article] [PubMed]
59. Reference deleted.
60. Murray, P. R., E. J. Baron, M. A. Pfaller, F. C. Tenover and R. H. Yolken (ed.). 1999. Manual of clinical microbiology, 7th ed. ASM Press, Washington, D.C.
61. Ninet, B., M. Monod, S. Emler, J. Pawlowski, C. Metral, P. Rohner, R. Auckenthaler, and B. Hirschel. 1996. Two different 16S rRNA genes in a mycobacterial strain. J. Clin. Microbiol. 34:2531-2536. [PMC free article] [PubMed]
62. Pace, N. 1997. A molecular view of microbial diversity and the biosphere. Science 276:734-740. [PubMed]
63. Pai, S., N. Esen, X. Pan, and J. M. Musser. 1997. Routine rapid Mycobacterium species assignment based on species-specific allelic variation in the 65-kilodalton heat shock protein gene (hsp65). Arch. Pathol. Lab. Med. 121:859-864. [PubMed]
64. Palys, T., L. K. Nakamura, and F. M. Cohan. 1997. Discovery and classification of ecological diversity in the bacterial world: the role of DNA sequence data. Int. J. Syst. Bacteriol. 47:1145-1156. [PubMed]
65. Patel, J. B. 2001. 16S rRNA gene sequencing for bacterial pathogen identification in the clinical laboratory. Mol. Diagn. 6:313-321. [PubMed]
66. Patel, J. B., D. G. Leonard, X. Pan, J. M. Musser, R. F. Berman, and I. Nachamkin. 2000. Sequence-based identification of Mycobacterium species using the Microseq 500 16S rDNA bacterial identification system. J. Clin. Microbiol. 38:246-251. [PMC free article] [PubMed]
67. Patel, J. B., J. E. Clarridge III, M. S. Schuster, J. Osborne, M. Waddington, and I. Nachamkin. 1999. Leptotrichia species bacteremia in a neutropenic patient. J. Clin. Microbiol. 37:2064-2067. [PMC free article] [PubMed]
68. Patel, R., K. E. Piper, M. S. Rouse, J. M. Steckelberg, J. R. Uhl, P. Kohner, M. K. Hopkins, F. R. Cockerill III, and B. C. Kline. 1998. Determination of 16S rRNA sequences of enterococci and application to species identification of nonmotile Enterococcus gallinarum isolates. J. Clin. Microbiol. 36:3399-3407. [PMC free article] [PubMed]
69. Pfister, P., M. Risch, D. E. Brodersen, and E. C. Bottger. 2003. Role of 16S rRNA helix 44 in ribosomal resistance to hygromycin B. Antimicrob. Agents Chemother. 47:1496-1502. [PMC free article] [PubMed]
70. Pfister, P., S. Hobbie, Q. Vicens, E. C. Bottger, and E. Westhof. 2003. The molecular basis for A-site mutations conferring aminoglycoside resistance: relationship between ribosomal susceptibility and X-ray crystal structures. Chembiochem 4:1078-1088. [PubMed]
71. Qian, Q., Y. Tang, C. P. Kolbert, C. A. Torgerson, J. G. Hughes, E. A. Vetter, W. Harmsen, S. O. Montgomery, F. R. Cockerill, III, and D. H. Persing. 2001. Direct identification of bacteria from positive blood cultures by amplification and sequencing of the 16S rRNA gene: evaluation of BACTEC 9240 instrument true-positive and false-positive results. J. Clin. Microbiol. 39:3578-3582. [PMC free article] [PubMed]
72. Rantakokko-Jalava, K., S. Nikkari, J. Jalava, E. Eerola, M. Skurnik, O. Meurman, O. Ruuskanen, A. Alanen, E. Kotilainen, P. Toivanen, and P. Kotilainen. 2000. Direct amplification of rRNA genes in diagnosis of bacterial infections. J. Clin. Microbiol. 38:32-39. [PMC free article] [PubMed]
73. Regnery, R. L., B. E. Anderson, J. E. Clarridge III, M. C. Rodriguez- Barradas, D. C. Jones, and J. H. Carr. 1992. Characterization of a novel Rochalimaea species, R. henselae sp. nov., isolated from blood of a febrile, human immunodeficiency virus-positive patient. J. Clin. Microbiol. 30:265-274. [PMC free article] [PubMed]
74. Reischl, U., K. Feldmann, L. Naumann, B. J. Gaugler, B. Ninet, B. Hirschel, and S. Emler. 1998. 16S rRNA sequence diversity in Mycobacterium celatum strains caused by presence of two different copies of 16S rRNA gene. J. Clin. Microbiol. 36:1761-1764. [PMC free article] [PubMed]
75. Relman, D. A. 1999. The search for unrecognized pathogens. Science 284:1308-1310. [PubMed]
76. Relman, D. A. 1993. Universal bacterial 16S rDNA amplification and sequencing, p. 489-495. In D. H. Persing, T. F. Smith, F. C. Tenover, and T. J. White (ed.), Diagnostic molecular microbiology: principles and applications. ASM Press, Washington, D.C.
77. Relman, D. A., J. S. Loutit, T. M. Schmidt, S. Falkow, and L. S. Tompkins. 1990. The agent of bacillary angiomatosis. An approach to the identification of uncultured pathogens. N. Engl. J. Med. 323:1573-1580. [PubMed]
78. Relman, D. A., T. M. Schmidt, R. P. MacDermott, and S. Falkow. 1992. Identification of the uncultured bacillus of Whipple's disease. N. Engl. J. Med. 327:293-301. [PubMed]
79. Ringuet, H., C. Akoua-Koffi, S. Honore, A. Varnerot, V. Vincent, P. Berche, J. L. Gaillard, and C. Pierre-Audigier. 1999. hsp65 sequencing for identification of rapidly growing mycobacteria. J. Clin. Microbiol. 37:852-857. [PMC free article] [PubMed]
80. Rogall, T., T. Flohr, and E. C. Bottger. 1990. Differentiation of Mycobacterium species by direct sequencing of amplified DNA. J. Gen. Microbiol. 136:1915-1920. [PubMed]
81. Rogall, T., J. Wolters, T. Flohr, and E. C. Bottger. 1990. Towards a phylogeny and definition of species at the molecular level within the genus Mycobacterium. Int. J. Syst. Bacteriol. 40:323-330. [PubMed]
82. Roth, A., M. Fischer, M. E. Hamid, S. Michalke, W. Ludwig, and H. Mauch. 1998. Differentiation of phylogenetically related slowly growing mycobacteria based on 16S-23S rRNA gene internal transcribed spacer sequences. J. Clin. Microbiol. 36:139-147. [PMC free article] [PubMed]
83. Roth, A., S. Andrees, R. M. Kroppenstedt, D. Harmsen, and H. Mauch. 2003. Phylogeny of the genus Nocardia based on reassessed 16S rRNA gene sequences reveals underspeciation and division of strains classified as Nocardia asteroides into three established species and two unnamed taxons. J. Clin. Microbiol. 41:851-856. [PMC free article] [PubMed]
84. Sacchi, C. T., A. M. Whitney, L. W. Mayer, R. Morey, A. Steigerwalt, A. Boras, R. S. Weyant, and T. Popovic. 2002. Sequencing of 16S rRNA gene: a rapid tool for identification of Bacillus anthracis. Emerg. Infect. Dis. 8:1117-1123. [PMC free article] [PubMed]
85. Sacchi, C. T., A. M. Whitney, M. W. Reeves, L. W. Mayer, and T. Popovic. 2002. Sequence diversity of Neisseria meningitidis 16S rRNA genes and use of 16S rRNA gene sequencing as a molecular subtyping tool. J. Clin. Microbiol. 40:4520-4527. [PMC free article] [PubMed]
86. Saitou, N., and M. Nei. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4:406-425. [PubMed]
87. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual, 2nd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
88. Schmalenberger, A., F. Schwieger, and C. C. Tebbe. 2001. Effect of primers hybridizing to different evolutionarily conserved regions of the small- subunit rRNA gene in PCR-based microbial community analyses and genetic profiling. Appl. Environ. Microbiol. 67:3557-3563. [PMC free article] [PubMed]
89. Song, J., S. C. Lee, J. W. Kang, H. J. Baek, and J. W. Suh. 2004. Phylogenetic analysis of Streptomyces spp. isolated from potato scab lesions in Korea on the basis of 16S rRNA gene and 16S-23S rDNA internally transcribed spacer sequences. Int. J. Syst. Evol. Microbiol. 54:203-209. [PubMed]
90. Song, Y., C. Liu, M. McTeague, and S. M. Finegold. 2003. Ribosomal DNA sequence-based analysis of clinically significant gram-positive anaerobic cocci. J. Clin. Microbiol. 41:1363-1369. [PMC free article] [PubMed]
91. Song, Y., C. Liu, D. R. Molitoris, T. J. Tomzynski, P. A. Lawson, M. D. Collins, and S. M. Finegold. 2003. Clostridium bolteae sp. nov., isolated from human sources. Syst. Appl. Microbiol. 26:84-89. [PubMed]
92. Springer, B., E. C. Bottger, P. Kirschner, and R. J. Wallace, Jr. 1995. Phylogeny of the Mycobacterium chelonae-like organism based on partial sequencing of the 16S rRNA gene and proposal of Mycobacterium mucogenicum sp. nov. Int. J. Syst. Bacteriol. 45:262-267. [PubMed]
93. Springer, B., L. Stockman, K. Teschner, G. D. Roberts, and E. C. Bottger. 1996. Two-laboratory collaborative study on identification of mycobacteria: molecular versus phenotypic methods. J. Clin. Microbiol. 34:296-303. [PMC free article] [PubMed]
94. Stackebrandt, E., W. Frederiksen, G. M. Garrity, P. A. Grimont, P. Kampfer, M. C. Maiden, X. Nesme, R. Rossello-Mora, J. Swings, H. G. Trüper, L. Vauterin, A. C. Ward, and W. B. Whitman. 2002. Report of the ad hoc committee for the re-evaluation of the species definition in bacteriology. Int. J. Syst. Evol. Microbiol. 52:1043-1047. [PubMed]
95. Sultana, F., Y. Kawamura, X. G. Hou, S. E. Shu, and T. Ezaki. 1998. Determination of 23S rRNA sequences from members of the genus Streptococcus and characterization of genetically distinct organisms previously identified as members of the Streptococcus anginosus group. FEMS Microbiol. Lett. 158:223-230. [PubMed]
96. Sussman, J. I., E. J. Baron, M. J. Tenenbaum, M. H. Kaplan, J. Greenspan, R. R. Facklam, M. B. Tyburski, M. A. Goldman, B. F. Kanzer, and R. A. Pizzarello. 1986. Viridans streptococcal endocarditis: clinical, microbiological, and echocardiographic correlations. J. Infect. Dis. 154:597-603. [PubMed]
97. 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. 2000. Identification of coryneform bacterial isolates by ribosomal DNA sequence analysis. J. Clin. Microbiol. 38:1676-1678. [PMC free article] [PubMed]
98. Tang, Y. W., N. M. Ellis, M. K. Hopkins, D. H. Smith, D. E. Dodge, and D. H. Persing. 1998. Comparison of phenotypic and genotypic technique for identification of unusual aerobic pathogenic gram-negative bacilli. J. Clin. Microbiol. 36:3674-3679. [PMC free article] [PubMed]
99. Teske, A., J. Wolters, and E. C. Bottger. 1991. The 16S rRNA nucleotide sequence of Mycobacterium leprae: phylogenetic position and development of DNA probes. FEMS Microbiol. Lett. 64:231-237. [PubMed]
100. Thorne, J. L., H. Kishino, and I. S. Painter. 1998. Estimating the rate of evolution of the rate of molecular evolution. Mol. Biol. Evol. 15:1647-1657. [PubMed]
101. Tortoli, E. 2003. Impact of genotypic studies on mycobacterial taxonomy: the new mycobacteria of the 1990s. Clin. Microbiol. Rev. 16:319-354. [PMC free article] [PubMed]
102. Tortoli, E., A. Bartoloni, E. C. Bottger, S. Emler, C. Garzelli, E. Magliano, A. Mantella, N. Rastogi, L. Rindi, C. Scarparo, and P. Urbano. 2001. Burden of unidentifiable mycobacteria in a reference laboratory. J. Clin. Microbiol. 39:4058-4065. [PMC free article] [PubMed]
103. Turenne, C. Y., L. Tschetter, J. Wolfe, and A. Kabani. 2001. Necessity of quality- controlled 16S rRNA gene sequence databases: identifying nontuberculous Mycobacterium species. J. Clin. Microbiol. 39:3637-3648. [PMC free article] [PubMed]
104. Ueda, K., T. Seki, T. Kudo, T. Yoshida, and M. Kataoka. Two distinct mechanisms cause heterogeneity of 16S rRNA. J. Bacteriol. 181:78-82. [PMC free article] [PubMed]
105. Wallace, R. J., Jr., M. Tsukamura, B. A. Brown, J. Brown, V. A. Steingrube, Y. Zhang, and D. R. Nash. 1990. Cefotaxime-resistant Nocardia asteroides strains are isolates of the controversial species Nocardia farcinica. J. Clin. Microbiol. 28:2726-2732. [PMC free article] [PubMed]
106. Ward, B. B. 2002. How many species of prokaryotes are there? Proc. Natl. Acad. Sci. USA 99:10234-10236. [PMC free article] [PubMed]
107. Waterman, M. S. 1995. Introduction to computational biology, p. 201-202. Chapman & Hall, New York, N.Y.
108. Wayne, L. G., R. C. Good, E. C. Bottger, R. Butler, M. Dorsch, T. Ezaki, W. Gross, V. Jonas, J. Kilburn, P. Kirschner, M. I. Krichevsky, M. Ridell, T. M. Shinnick, B. Springer, E. Stackebrandt, I. Tarnok, Z. Tarnok, H. Tasaka, V. Vincent, N. G. Warren, C. A. Knott, and R. Johnson. 1996. Semantide- and chemotaxonomy-based analyses of some problematic phenotypic clusters of slowly growing mycobacteria, a cooperative study of the International Working Group on Mycobacterial Taxonomy. Int. J. Syst. Bacteriol. 46:280-297. [PubMed]
109. Wilck, M. B., Y. Wu, J. G. Howe, J. Y. Crouch, and S. C. Edberg. 2001. Endocarditis caused by culture-negative organisms visible by Brown and Brenn staining: utility of PCR and DNA sequencing for diagnosis. J. Clin. Microbiol. 39:2025-2027. [PMC free article] [PubMed]
110. Wilson, R. W., V. A. Steingrube, E. C. Bottger, B. Springer, B. A. Brown- Elliott, V. Vincent, K. C. Jost, Jr., Y. Zhang, M. J. Garcia, S. H. Chiu, G. O. Onyi, H. Rossmoore, D. R. Nash, and R. J. Wallace, Jr. 2001. Mycobacterium immunogenum sp. nov., a novel species related to Mycobacterium abscessus and associated with clinical disease, pseudo- outbreaks and contaminated metalworking fluids: an international cooperative study on mycobacterial taxonomy. Int. J. Syst. Evol. Microbiol. 51:1751-1764. [PubMed]
111. Woese, C. R., E. Stackebrandt, T. J. Macke, and G. E. Fox. 1985. A phylogenetic definition of the major eubacterial taxa. Syst. Appl. Microbiol. 6:143-151. [PubMed]
112. Woese, C. R., G. J. Olsen, M. Ibba, and D. Soll. 2000. Comparisons of complete genome sequences allow the most objective and comprehensive descriptions possible of a lineage's evolution. Microbiol. Mol. Biol. Rev. 64:202-236. [PMC free article] [PubMed]
113. Woese, C. R. 1987. Bacterial evolution. Microbiol. Rev. 51:221-271. [PMC free article] [PubMed]
114. Woo, P. C., K. H. Ng, S. K. Lau, K. T. Yip, A. M. Fung, K. W. Leung, D. M. Tam, T. L. Que, and K. Y. Yuen. 2003. Usefulness of the MicroSeq 500 16S ribosomal DNA-based bacterial identification system for identification of clinically significant bacterial isolates with ambiguous biochemical profiles. J. Clin. Microbiol. 41:1996-2001. [PMC free article] [PubMed]
115. Zhang, Q., R. Kennon, M. A. Koza, K. Hulten, and J. E. Clarridge III. 2002. Pseudoepidemic due to a unique strain of Mycobacterium szulqai: genotypic, phenotypic, and epidemiological analysis. J. Clin. Microbiol. 40:1134-1139. [PMC free article] [PubMed]

Articles from Clinical Microbiology Reviews are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...