16S rRNA gene sequence analysis is broadly considered the “gold standard” in bacterial identification (
6,
29). In daily clinical diagnostics, accurate bacterial identification is essential in judging whether a bacterial isolate is to be considered the causative agent of an infectious disease or merely a colonizer. In our study, we aimed to characterize the bacterial diversity encountered in a diagnostic laboratory by revealing potentially novel, clinically relevant species, according to the current species definition by the Clinical and Laboratory Standards Institute (
22).
Routine 16S rRNA gene sequencing is implemented in our laboratory and is a fixed part of our diagnostic algorithms for identification of bacterial isolates (
1,
2,
32). We retrospectively reanalyzed 16S rRNA gene sequences collected during 2004 to 2008 to identify potentially novel bacterial taxa of clinical relevance. The Institute of Medical Microbiology (IMM) serves the 850-bed University Hospital of Zurich and surrounding smaller hospitals. Bacterial isolates from blood, cerebrospinal fluid, wounds, joint aspirates, respiratory samples, genitourinary swabs, feces, and urine were recovered by culture on appropriate media according to standard procedures (
19). Isolates that could not be identified by phenotypic methods underwent sequencing. 16S rRNA gene analysis was performed as previously described (
1). Homology analyses were performed using the SmartGene Integrated Database Network System (IDNS) (
24) and NCBI GenBank databases. For the first screening of our large data collection, we selected isolates with sequence homology of <99.0% to members of described taxa, regarding these as potentially novel species; isolates with sequence homology of <95% were regarded as representatives of a novel genus (
2). The boundary for novel families was <87.5% homology and, for novel orders, <78.4% 16S rRNA sequence homology (
30). After the first screening, we used more stringent cutoff values (<97.5% for species) for taxa with significant interspecies 16S rRNA divergence; i.e., members of the
Paenibacillaceae family and the
Clostridiales order (
6,
25).
During the 5-year study period, 1,663 cultured isolates were subjected to 16S rRNA gene sequence analysis (Table ). Of those, 60 isolates (0.4‰; see Table S1 in the supplemental material) had a 16S rRNA gene homology of <99% to members of accepted taxa on the date of the first interpretation. A total of 11 of the 60 sequences with a 16S rRNA homology of <99% in the first-time analysis could be allocated to a species established during the study term as a novel species by others:
Acinetobacter septicus (
16,
20),
Brevibacterium ravenspurgense (
17),
Corynebacterium freiburgense (
12),
Corynebacterium massiliense (
n = 2) (
18),
C. mastitidis (
10,
18),
C. pyruviciproducens (
26),
C. ureicelerivorans (
11,
31),
Neisseria zoodegmatis (
28),
Paenibacillus barengoltzii (
21), and the reclassified
Campylobacter ureolyticus (previously known as
Bacteroides ureolyticus) (
27).
| TABLE 1.Clinical bacterial isolates with 16S rRNA gene homology < 99% (n = 60) |
We calculated dendrograms to assess phylogenetic relationships (Fig. ). Potentially novel streptococcal species clustered with known pathogens. For example, within the streptococci, one isolate (GenBank accession no. GU797873) shared 98.8% sequence homology with S. infantis; another isolate (GU797840) shared 97.8% homology with S. pneumoniae. Within the Bacillales order (Fig. ), eight novel Paenibacillus sequence types were recovered: three of them (GU797882, GU797868, and GU797869) were distantly related (<95% sequence homology) to Paenibacillus chinjuensis-P. validus, and two isolates (GU797838 and GU797854) were related to P. timonensis (97.3% and 94.5% sequence homology). Several bacterial families are represented in the Clostridiales order. The novel sequences recovered in the Clostridiales order all belonged to different families (Fig. ). We found two representatives of the Fusobacteriales order (Fig. ): one isolate (GU797848) was related to Fusobacterium russii (sequence homology 98.8%), and one isolate (GU797890) was most closely related to Leptotrichia buccalis (98.9% sequence homology). Eleven novel sequences belonged to the Pseudomonadales order, and 3 of those (GU797845, GU797842, and GU797892) represented potential novel Acinetobacter spp. (Fig. ). The Actinomycetales order (Fig. ) comprises 25 potentially novel taxa (data for 24 taxa are given in the figure). Six corynebacterial isolates, i.e., Corynebacterium freiburgense (GU797839), Corynebacterium massiliense (GU797864 and GU797833), C. mastitidis (GU797866), C. ureicelerivorans (GU797878), and C. pyruviciproducens (GU797881), clustered with type strain sequences that were established as novel species during the study period. We recovered three Nocardia spp. (GU797846, GU797858, and GU797874); two of them (GU797846 and GU797858) belonged to the Nocardia asteroides complex and one (GU797874) was related to Nocardia flavorosea. Four potentially novel Actinobaculum spp. (GU797861, GU797867, GU797872, and GU797883/GU797308) were attributed to the same taxonomic clade according to 16S rRNA sequence phylogeny results, with Actinobaculum schaalii as the nearest neighboring species (95.7 to 98.5% sequence homology).
To assess the clinical relevance of the microbiological findings, we selected 30 isolates for which sufficient clinical data were available and performed reviews of patient charts (Table ). We reviewed the patient charts for clinical signs and symptoms of infection, inflammation parameters such as fever, leukocyte count, C-reactive protein (CRP) and procalcitonin levels, radiological and laboratory findings (serology and bacterial culture results), previous infections or bacterial isolates, antibiotic treatment, and clinical diagnosis (see Table S2 in the supplemental material). We established a clinical score incorporating the different parameters mentioned above to determine the likelihood (codified as “yes,” “likely,” “unlikely,” “no”) of an infectious disease in each case and to assess the association of the bacterial isolate found with disease.
| TABLE 2.Subset analysis of microbial and clinical data of 30 patientsa |
Clinical relevance was attributed to 18 isolates. In 10 cases, patient history, laboratory findings, and clinical course following antibiotic therapy guided by the isolate's drug susceptibility testing results were compatible with a pathogenic role for the isolated microorganism. In eight cases, we concluded that the bacterial isolate was likely to have been the cause of an infection. The putative novel disease-associated species mostly belonged to the
Actinomycetales order of Gram-positive rods. The recently described
Acinetobacter septicus (GU797892) (
16), two
Actinobaculum spp. (GU797883 and GU797872), and a
Gardnerella sp. (GU797857) were each identified as present in samples from patients with urinary tract infections. In one case, we recovered a potentially novel
Actinobaculum sequence type (GU797872) from several blood cultures of a patient suffering from urosepsis, underlining the pathogenic potential of species of the
Actinobaculum genus in ascending urinary tract infections. A potentially novel
Acinetobacter sp. (GU797845), most closely related to
Acinetobacter calcoaceticus, was identified in the peritoneal dialysate of a 22-year-old patient with kidney failure. He exhibited infective monobacterial peritonitis as a complication of continuous ambulatory peritoneal dialysis (CAPD). A previously unknown
Actinomyces sp. (GU797891) was isolated from a 49-year old female patient suffering from severe, acute, suppurative parotitis.
Neisseria zoodegmatis (previously
Neisseria CDC EF-4 group [
28]) was found in an isolate from a patient with a wound and a history of a cat bite (GU797849). A novel sequence type of
Paenibacillus barengoltzii (
21) was cultured from a central venous catheter in the jugular vein of a patient with a burn injury (GU797838). A potentially novel
Streptococcus sp. (GU797859), belonging to the
Streptococcus anginosus group, was found in an isolate from a septic patient suffering from cholangitis. A potentially novel
Streptococcus sp. related to
Streptococcus oralis (GU797840) was cultured from a hip joint aspirate of a patient with hip prosthesis infection. Bacterial isolates that were found to be irrelevant to a patient's clinical disease entity (12/30) were considered to represent skin flora (e.g.,
Paenibacillus spp. or
Ruminococcus spp.) or to belong to environmental bacteria (e.g.,
Aquabacterium spp. or
Kineosphaera spp.).
Broad-range 16S rRNA gene amplification readily allows the detection of members of as-yet-unknown bacterial taxa (
4,
5,
9). Major hypervariable regions are present in the first 500 bp of the roughly 1,600 bp comprising the 16S rRNA gene downstream of the conserved primer target sites (
5,
9,
29). Thus, analysis of this part of the gene sequence allows the recognition of potentially novel taxa based on previously established cutoff values of <99% homology for new species and <95% homology for new genera (
1-
3,
7). While such a general cutoff is appropriate for overall first analysis of large data sets, we note that the boundaries for species definition by 16S rRNA sequence homology may be different for different phyla (
13,
25). A less stringent cutoff value (i.e., <99.6% homology) could have been used to delimit different species in bacterial groups such as the
Streptococcus mitis group or nonfermenters (
13). Conversely, for species belonging to the
Paenibacillaceae family and the
Clostridiales order, a more stringent cutoff value (i.e., 97.5% homology) is more appropriate (
6) and was therefore applied after the first selection performed with the 99% cutoff value.
In 2008, the fraction of bacterial isolates submitted for molecular identification was 0.8%. Previous investigations reported rates of 0.5% to 1% for a similar study setup (
7) and a rate of 14% for isolates restricted to aerobic Gram-positive rods (
2). Gram-positive rods and Gram-negative cocci are overrepresented in the group of sequenced isolates in our comparative analysis of phenotypic and 16S rRNA-based identification methods. Of the 1,663 (3.7%) sequences determined during the study period, 60 were judged to be representatives of potentially novel species or novel genera. A recent review (
29), summarizing 16S rRNA gene-based studies published from 2001 to 2007, calculated that 215 unique sequences recovered during this period from human specimens represented potentially novel species. Of the 215, 29 belonged to novel genera. During our study, the number of 16S rRNA sequences deposited in the NCBI nucleotide database increased by a factor of 15. The SmartGene IDNS 16S rRNA database, which is a curated database derived from the NCBI repository, increased in size by a factor of 4. Despite this increase in the number of sequences deposited, the recovery of sequences with <99% homology to members of established taxa in our data set during 2004 to 2008 remained relatively constant at between 2.4% and 5.1%. This finding may reflect the fact that many of the sequences deposited in public databases were the outcome of large-scale ecological or environmental (metagenomic) sequencing projects and did not include sequences of clinical laboratory isolates.
Bacterial taxonomic classification has advanced differently in various taxonomic groups (
25): Phenotypic methods readily allow species determination below the resolution of 16S rRNA-based sequence analysis in studies of enteric Gram-negative bacteria (
14,
15). In contrast, the
Actinomycetales order is a rich but poorly investigated group (
2). For example, within the
Corynebacterium genus, 18 novel species were validly described from 2004 to 2009. A total of 5 of these, namely,
Corynebacterium freiburgense (
12),
Corynebacterium pyruviciproducens (
26),
C. mastitidis (
10,
18),
C. massiliense (
18), and
C. ureicelerivorans (
11,
31), were also identified in our study.
When we calculated phylogenetic trees based on partial 16S rRNA sequences (Fig. ), we found that differentiation was numerically strong (as measured by nucleotide substitutions of base pairs) in the
Actinomycetales,
Clostridiales,
Fusobacteriales, and
Pseudomonales orders whereas it was less profound in the
Lactobacillales order and, more specifically, in the
Streptococcaceae family. Regarding potentially novel
Streptococcus spp., further molecular analysis of additional loci (by, e.g.,
sodA,
rpoB, and
recA sequence homology) would be required to determine exact phylogenetic relationships (
8,
23).
In summary, out of 1,663 bacterial isolates subjected to 16S rRNA sequencing during a 5-year period, we recovered 60 clinical bacterial isolates that were indicative of the presence of putative novel bacterial species. Of these 60 isolates, 9 were established as novel pathogens in the literature during the period of the study. A total of 18 (60%) isolates showed clinical relevance in a subset analysis of 30 of the 60 isolates. Isolates with clinical implications are mostly representatives of genera that comprise known pathogens (i.e., Streptococcus spp., Actinobaculum spp., Actinomyces spp., and Neisseria spp.). Our findings underline the importance of 16S rRNA gene sequencing in routine identification algorithms designed to recognize novel pathogens in the diagnostic laboratory.