Utility of 16S rDNA Sequencing for Identification of Rare Pathogenic Bacteria

Abstract

Background

Phenotypic identification systems are established methods for laboratory identification of bacteria causing human infections. Here, the utility of phenotypic identification systems was compared against 16S rDNA identification method on clinical isolates obtained during a 5‐year study period, with special emphasis on isolates that gave unsatisfactory identification.

Methods

One hundred and eighty‐seven clinical bacteria isolates were tested with commercial phenotypic identification systems and 16S rDNA sequencing. Isolate identities determined using phenotypic identification systems and 16S rDNA sequencing were compared for similarity at genus and species level, with 16S rDNA sequencing as the reference method.

Results

Phenotypic identification systems identified ~46% (86/187) of the isolates with identity similar to that identified using 16S rDNA sequencing. Approximately 39% (73/187) and ~15% (28/187) of the isolates showed different genus identity and could not be identified using the phenotypic identification systems, respectively. Both methods succeeded in determining the species identities of 55 isolates; however, only ~69% (38/55) of the isolates matched at species level. 16S rDNA sequencing could not determine the species of ~20% (37/187) of the isolates.

Conclusion

The 16S rDNA sequencing is a useful method over the phenotypic identification systems for the identification of rare and difficult to identify bacteria species. The 16S rDNA sequencing method, however, does have limitation for species‐level identification of some bacteria highlighting the need for better bacterial pathogen identification tools.

Introduction

Accurate identification of clinically isolated bacteria is critical for the determination of the causative agent causing an infection. It is also needed for instituting the most appropriate antimicrobial therapy and infection management strategies 1, 2, 3, 4. Bacteria identification in a routine microbiology laboratory is often done by first culturing the bacteria in a suitable growth medium 1, 3, 5. Pure bacteria culture is obtained and following that, bacteria identification is done by various means. The most common method used largely relies on commercially available phenotypic identification systems such as the API (bioMérieux, Marcy l'Etoile, France) 1, 2, 3, 5, 6. The test is usually easy to perform, and in most cases, results can be rapidly obtained 4, 5. The phenotypic identification method, however, has relatively lower accuracy, reproducibility, and discriminatory power often resulting in inconclusive identification 5, 6, 7. The introduction of matrix‐assisted laser desorption ionization time of flight mass spectrometry (MALDI‐TOF MS) improves the accuracy of bacteria species identification; however, limitations related to cost of equipment and database coverage restricted its broad application 5, 8. 16S rDNA sequencing, on the other hand, has emerged as a widely accepted method of choice for the accurate identification of bacteria 2, 3, 4, 6, 9, 10, 11, 12, and it is also the most well‐represented gene in the public databases (GenBank, European Nucleotide Archive, DNA Data Bank of Japan and The Ribosomal Database Project) 4, 11. This method capitalizes on the presence of unique variable regions in the gene that are genus or species specific to all bacteria 1, 2, 3, 5, 7, 8, 9, 13, and comparing the sequences with those available in the public databases 12. This method is especially useful and practical in healthcare facility settings where a variety of pathogenic bacteria are encountered from patients. The method can be applied especially for identification of fastidious, rare, slow‐growing bacteria and those which possess atypical phenotypic profiles 1, 2, 3, 7. Whenever the identification of bacteria isolates is not possible or doubtful by phenotypic identification systems, 16S rDNA sequencing is likely to be able to identify at least to the genus level mainly due to the availability of public databases that cover the entire extent of known taxa diversity 4. This study aimed at comparing the utility of phenotypic and 16S rDNA identification methods on clinical isolates obtained during a 5‐year study period, with special emphasis on isolates that gave unsatisfactory identity using phenotypic identification systems.

Materials and Methods

This retrospective study included 187 bacteria isolates obtained from patients admitted to University Malaya Medical Center (UMMC) from 2007 to 2011. This study received approval from the University Malaya Medical Center Medical Ethics Committee (MECID. No. 20149‐575). Patient specimens were from various clinics of the UMMC, and samples were cultured on various media. All isolates were initially identified using routine Gram staining, microscopy, and biochemical tests. These tests were complimented using API kits (bioMérieux). The following API kits, API 20A, API 20E, API 20NE, API 20Strep, API 50CHB, API Coryne, API NH, Rapid 20E and Rapid ID 32A, were collectively referred to as “phenotypic identification systems” in this study. When results from the phenotypic identification systems gave weak (<95% probability) or ambiguous genus identification, 16S rDNA PCR and sequencing were performed. In cases when phenotypic identification systems yielded more than one potential species identity, the species with the highest probability was chosen and unknown identity was recorded when the system could not identify the isolate.

Bacteria DNA was extracted using NucleoSpin Tissue kit (Macherey‐Nagel, Düren, Germany) according to the manufacturer's instructions, and the extracted DNA was stored in −20°C until needed. 16S rDNA PCR was performed according to previously published protocols 9, and the amplicons were purified using NucleoSpin Gel Clean‐up kit (Macherey‐Nagel). The purified DNA fragments were subjected to DNA sequencing, and the resulting sequences were assembled using Sequencher 4.10.1 (Gene Codes, Ann Arbor, MI) generating approximately 1,500 nucleotides. Nucleotide sequences were compared against those available in the GenBank using BLAST 12. The entry with the highest sequence similarity was chosen for each isolate. Identities of every isolate determined using the 16S rDNA sequences and phenotypic identification systems were compared for similarity at genus and species level, with 16S rDNA sequencing as the reference method. Identification to the species level was defined as ≥ 99% 16S rDNA sequence similarity to the closest GenBank entry. The isolate was assigned to the corresponding genus when its 16S rDNA sequence similarity was <99% and ≥95% 7.

Results

A total of 187 isolates were obtained over a 5‐year period (2007–2011). 16S rDNA sequence analyses enabled the identification of all 187 (100%) isolates to the genus level and 150/187 (~80%) isolates to the species level. The 187 bacteria isolates belonged to the following genera: Neisseria (n = 21), Streptococcus (n = 20), Enterococcus (n = 15), Moraxella (n = 12), Bacillus (n = 8), Campylobacter (n = 6), Burkholderia (n = 5), Corynebacterium (n = 5), Vibrio (n = 5), Clostridium (n = 4), Globicatella (n = 4), Paenibacillus (n = 4), Ralstonia (n = 4), Abiotrophia (n = 3), Brevibacterium (n = 3), Brevundimonas (n = 3), Gordonia (n = 3), Lactobacillus (n = 3), Nocardia (n = 3), Staphylococcus (n = 3), Aerococcus (n = 2), Aeromonas (n = 2), Eikenella (n = 2), Granulicatella (n = 2), Haematobacter (n = 2), Pantoea (n = 2), Pseudomonas (n = 2), Roseomonas (n = 2), Rothia (n = 2), Serratia (n = 2), and Stenotrophomonas (n = 2). Each of the following genera had only one isolate (n = 1), respectively: Achromobacter, Actinobacillus, Actinobaculum, Aggregatibacter, Anaerobiospirillum, Arcanobacterium, Bacteroides, Brevibacillus, Brucella, Cardiobacterium, Citrobacter, Dialister, Escherichia, Haemophilus, Kingella, Microbacterium, Micrococcus, Mycobacterium, Mycoplasma, Naxibacter, Novosphingobium, Ochrobactrum, Oligella, Paracoccus, Peptoniphilus, Propionibacterium, Rhodococcus, Shewanella, Streptobacillus, Tsukamurella, and Weissella.

The conventional phenotypic identification systems managed to identify ~46% (86/187) of the isolates with genus identity similar to that identified using the 16S rDNA sequencing (Table 1). However, ~39% (73/187) of the isolates showed different genus identity, and ~15% (28/187) of the isolates could not be identified using the phenotypic identification system (Table 1). Both genotypic (16S rDNA sequencing) and phenotypic (API kits) methods managed to determine the species identities of 55 isolates (Table 1). However, only ~69% (38/55) of the isolates showed matching bacteria species identities using both, phenotypic identification systems and 16S rDNA sequencing (Table 1). The 16S rDNA sequencing could not determine the species of ~20% (37/187) of the isolates (Table 1), and those genus for which species identity could not be determined include Moraxella spp. (accession no. LN871842), Neisseria spp. (accession no. LN871843), and Paenibacillus spp. (accession no. LN871844).

Table 1.

Performance of API Phenotypic Identification Systems in Comparison With 16S rDNA Sequencing

	n	%
Bacteria genus determined using 16S rDNA sequencing	187
Similar genus	86	45.99
Not similar genus	73	39.04
Unknown identity	28	14.97
Bacteria species determined using 16S rDNA sequencing	150	80.21
Bacteria species determined using phenotypic identification systems and 16S rDNA sequencing	55
Similar species	38	69.09
Not similar species	17	30.91

Discussion

In this study, 187 clinical bacteria isolates collected over a 5‐year period were examined. Bacteria species identification was determined using the API phenotypic identification systems and 16S rDNA sequencing. The 16S rDNA sequencing identified 100% and ~80% of the 187 isolates to the genus and species levels, respectively. Of the 55 isolates whose species could be determined using 16S rDNA sequencing and phenotypic identification systems, only ~69% species identities matched using both methods. This suggests that at least 30% of the bacteria isolated from patient samples were not accurately identified if only the phenotypic identification systems were used. This represents a substantial percentage of the total isolates, highlighting the possibility that the patients from which the isolates were obtained, could perhaps not optimally treated. The API phenotypic identification systems designated ~46% of the total isolates into similar genus as determined by 16S rDNA sequencing with ~39% of the isolates designated into different genus. Hence, highlighting potentially a huge discrepancy in bacteria identification when API phenotypic identification system is used alone 4, 10, 14.

The API phenotypic identification system, which identifies bacteria based on the fermentation of sugars, assimilation of certain carbon sources, and production of unique metabolites and enzymes 5, 10, had difficulties identifying bacteria with similar biochemical profiles 4, 10, 13. It is worth noting that ~15% of the clinical bacteria isolates could not be identified using the API kits, emphasizing the limitation of using the API kits. In our study, we found that the API phenotypic identification systems could not resolve Abiotrophia defectiva, Clostridium sporogenes, Moraxella osloensis, and Pseudomonas oryzihabitans, even though their identities are present in the API and ID 32 identification databases. This could be attributed to the altered biochemical characteristics of bacteria isolated from patients who may have undergone antimicrobial therapy, unusual biochemical profiles of old cultures, phenotypic variability between different strains of the same species 10, and different results obtained from the same strain upon repeated testing 4.

The phenotypic identification systems could not accurately identify a number of these species, Aggregatibacter aphrophilus, Cardiobacterium hominis, and Globicatella sulfidifaciens (Table 2), despite them being members of the clinically significant classes, Gammaproteobacteria and Bacilli, which are well represented in the API identification database 6. Our findings also demonstrated compelling evidence of the utility of 16S rDNA sequencing for the identification of rare and difficult to isolate bacteria, which were not found in the API and ID 32 identification databases; these include Brevibacterium paucivorans, Corynebacterium sundsvallense, Gordonia terrae, Nocardia cyriacigeorgica, Novosphingobium panipatense, Paracoccus yeei, Peptoniphilus harei, and Tsukamurella tyrosinosolvens (Table 2). This finding is congruent with the earlier report that members of the classes Actinobacteria and Clostridia are the least represented in the API database 6. In addition, our study also found two Alphaproteobacteria members (N. panipatense and P. yeei) (Table 2), which could not be identified using phenotypic identification systems. In the case of Rothia mucilaginosa, however, the API kits we employed were not suitable 15; hence, an unknown identification was scored. The latter exemplified the additional cost that would be necessary to procure additional API kits 3 for the identification of rare and uncommon bacteria species.

Table 2.

Identification of Rare Clinical Bacteria Isolates by API Phenotypic Identification Systems Compared With 16S rDNA Sequencing

Phenotypic identification systems	16S rDNA sequencing	GenBank accession no.	Reference
Identity	Identity
Haemophilus paraphrophilus/Haemophilus aphrophilus	Aggregatibacter aphrophilus	–	9
Rhodococcus spp./Nocardia spp./Gordona spp./Mycobacterium spp.	Tsukamurella tyrosinosolvens	AY254699	10
Unknown	Novosphingobium panipatense	LN871837	This study
Chryseomonas luteola	Pseudomonas oryzihabitans	LN871840	This study
Corynebacterium propinquum	Brevibacterium paucivorans	LN871829	This study
Alloiococcus otitis	Abiotrophia defectiva	LN871828	This study
Eubacterium lentum	Clostridium sporogenes	LN871831	This study
Streptococcus spp.	Corynebacterium sundsvallense	LN871832	This study
Peptostreptococcus spp.	Peptoniphilus harei	LN871839	This study
Rhodococcus spp.	Gordonia terrae	LN871834	This study
Pasteurella multocida	Cardiobacterium hominis	LN871830	This study
Unknown	Rothia mucilaginosa	LN871841	This study
Aerococcus viridans	Globicatella sulfidifaciens	LN871833	This study
Haemophilus parainfluenzae	Paracoccus yeei	LN871838	This study
Sphingomonas paucimobilis	Moraxella osloensis	LN871835	This study
Propionibacterium avidum	Nocardia cyriacigeorgica	LN871836	This study

The 16S rDNA sequencing method, however, is not without its own disadvantages. As shown here, the 16S rDNA sequencing could not provide adequate discrimination for the identification to species level for some bacteria 1, 4, 5, 6, 8 and these include Moraxella, Neisseria, and Paenibacillus. One plausible explanation is that the 16S rDNA gene evolved slowly in these groups of bacteria; hence, it lacks the required resolution to differentiate between the different species 16. Moraxella and Neisseria, for example, once belonged to the same family, Neisseriaceae before a new family, and Moraxellaceae was proposed in 1991 17, suggesting insufficient sequence variations for characterization to the species level. It was also reported that the presence of multiple copies of the 16S rDNA gene restricted Neisseria 13 and Paenibacillus 18 species identification. Another limitation of the 16S rDNA sequencing method is that it depends on the quality of the public databases, as sequences are deposited regardless of the correct assignment, length, and the number of ambiguous nucleotides 4.

Changes in the taxonomic description of bacteria could also lead researchers into misidentification of isolates 4, as illustrated by the classification of genus Moraxella into the new family Moraxellaceae, from Neisseriaceae 17. Furthermore, the 16S rDNA sequencing method essentially uses polymerase chain amplification to amplify DNA fragments for sequencing, and it is subject to various sources of errors. Errors in PCR polymerases and sequencing can significantly impact downstream sequence data by introducing false nucleotides 11.

Currently, there is no single quintessential method to identify all bacteria species. Findings from this study suggest that 16S rDNA sequencing should complement existing phenotypic identification systems to accurately identify clinically relevant, fastidious, and rare bacteria isolates. A full bacterial genome sequencing 19 could also be feasible if the cost becomes less prohibitive. In conclusion, we showed here nonetheless that the 16S rDNA sequencing is a useful method for the identification of rare and difficult to identify bacteria species. This method, however, does have its limitation for species‐level identification of some bacteria.