Abstract
The aim of this study was to develop an easy-to-use culture-free diagnostic method based on next generation sequencing (NGS) of PCR amplification products encompassing whole 16S-23S rRNA region to improve the resolution of bacterial species identification. To determine the resolution of the new method 67 isolates were subjected to four identification methods: Sanger sequencing of the 16S rRNA gene; NGS of the 16S-23S rRNA region using MiSeq (Illumina) sequencer; Microflex MS (Bruker) and VITEK MS (bioMérieux). To evaluate the performance of this new method when applied directly on clinical samples, we conducted a proof of principle study with 60 urine samples from patients suspected of urinary tract infections (UTIs), 23 BacT/ALERT (bioMérieux) positive blood culture bottles and 21 clinical orthopedic samples. The resolution power of NGS of the 16S-23S rRNA region was superior to other tested identification methods. Furthermore, the new method correctly identified pathogens established as the cause of UTIs and blood stream infections with conventional culture. NGS of the 16S-23S rRNA region also showed increased detection of bacterial microorganisms in clinical samples from orthopedic patients. Therefore, we conclude that our method has the potential to increase diagnostic yield for detection of bacterial pathogenic species compared to current methods.
Similar content being viewed by others
Introduction
Timely, appropriate treatment of infection depends on rapid and specific identification of causative microorganisms. To date, identification of bacterial species highly depends on culture or molecular tests, including 16S rRNA gene Sanger sequencing1, 2. Significant limitations of culture methods are that some bacteria are slow-growing or fastidious, making identification of them complicated and time and resource consuming3. Although molecular tests that target specific microorganisms are more rapid and sensitive than culture methods they need an a priori knowledge of the likely pathogenic species that could be present in a sample. If bacteriological identification by culture methods fails, Sanger sequencing of the variable 16S rRNA gene is used for identification. The gene sequence has been proven to be a reliable genetic marker as it is present in all bacteria and its function has not changed over time4. However, unequivocal identification is not always possible due to the high sequence similarities of the 16S rRNA gene in some bacterial species5. Although Sanger sequencing of the 16S rRNA gene can be applied directly on clinical materials, in polymicrobial samples it usually cannot identify more than one species simultaneously or at least this process is challenging by sorting out the ambiguous signals from mixed chromatograms using a computer program6.
Next generation sequencing (NGS) allows culture free detection of a theoretically unlimited number of pathogens and thus provides insight in the full microbiome. Since the introduction of benchtop sequencers, NGS is likely to become a diagnostic tool within the next few years in microbiological laboratories3, 7, 8. Metagenomics will be the ultimate approach in detecting all microorganisms (e.g. bacteria, viruses, fungi) in a sample. Unfortunately, analysis of large datasets requires a combination of bioinformatics skills and computational resources that is nowadays mostly absent in diagnostic (medical) microbiological laboratories. Furthermore, metagenomics approaches are time consuming as the turnaround time is approximately 4–5 days. To fill the gap between the conventional methods (culture and PCR) and metagenomics, a culture free approach using targeted NGS will be an excellent approach to detect and identify bacterial species as, compared to metagenomics, it is less complicated, and cheaper and therefore more likely to get implemented in diagnostic laboratories within a short timeframe.
The aim of the present study was to develop a rapid and easy-to use culture free diagnostic method based on NGS of PCR amplification products encompassing the whole 16S-23S rRNA region to improve the resolution of bacterial species identification. Moreover, the new method was compared with three commonly used identification methods. Finally, the feasibility of the new identification method to detect and identify bacterial species in clinical specimens was evaluated.
Results
Comparison of identification potential of the tested methods
The results of bacterial identification obtained by the four tested methods are shown in Supplementary Table S1. The rates of accurate identification to the species level using NGS of the 16S-23S rRNA region, Sanger sequencing of the 16S rRNA gene, Microflex MS and Vitek MS methods were 92.5%, 56.7%, 73.1% and 64.2%, respectively. The rates of accurate identification to the genus level using NGS of the 16S-23S rRNA region, Sanger sequencing of the 16S rRNA gene, Microflex MS and Vitek MS methods were 100%, 94.0%, 88.1% and 83.6%, respectively. At the species and genus level, assessment of the statistical significance of the differences in accuracy of the four methods for assigning species and genus showed statistically significant differences between NGS of the 16S-23S rRNA region versus all other tested methods for all bacterial isolates (P < 0.05). The only exception was observed between the two sequencing methods at the genus level (P = 0.2482). Moreover, the Sanger sequencing of the 16S rRNA gene was significantly more discriminative at the genus level than Vitek MS (P < 0.05).
NGS of 16S-23S rRNA region was the only method which correctly identified to the species level Shigella sonnei, Streptococcus oralis, Streptococcus mitis, Streptococcus vestibularis, Abiotrophia para-adiacens and Achromobacter marplatensis (Table S1). Sanger sequencing of 16S rRNA gene was the only method which did not have a potential to identify Corynebacterium ulcerans, Escherichia coli, Enterococcus faecium, Streptococcus parasanguinis, Staphylococcus aureus, Staphylococcus capitis, Staphylococcus epidermidis, Lactobacillus rhamnosus, Delftia acidovorans and Klebsiella pneumoniae. Vitek MS showed identical and Bruker MS almost identical potential compared to NGS of the 16S-23S rRNA region for identification of the Staphylococcus isolates to the species level. Moreover, Bruker MS was the only method which correctly identified B. cereus and L. paracasei to the species level, and Pseudomonas aeruginosa was identified only by Vitek MS. Both Mass Spectrometry methods could not identify Acinetobacter bereziniae, Corynebacterium afermentans, Corynebacterium pseudogenitalium, Salmonella enterica, Streptococcus tigurinus and Moraxella lacunata. Moreover, Vitek MS could not identify Acinetobacter parvus, Acinetobacter baumannii, Corynebacterium amycolatum, Corynebacterium coyleae, Streptococcus dysgalactiae and Actinomyces graevenitzii, and Bruker MS had a problem with identification of Staphylococcus lentus to the species level. No method could unambiguously identify S. pasteuri to the species level (Table S1).
Urine samples
Sixty fresh urine samples from patients suspected for urinary tract infections (UTIs) were subjected to conventional culture-based methods and culture-independent NGS of the 16S-23S rRNA region. Bacterial culture reported the growth of predominant one or two microorganisms at ≥104 CFU/ml as possible cause of the suspected UTI. Growth of <104 CFU/ml or growth without predominant microorganism was considered inconclusive for UTI. NGS of the 16S-23S rRNA region identified from 1 to 47 different bacterial species in each PCR-positive sample (Table 1). There was an observed association between the increasing number of bacterial genera/species identified using NGS of 16S-23S rRNA region and decreasing CFU value, median: ≥105 CFU – 2 species/genera; 104 CFU – 5 species/genera; 103 CFU – 5 species/genera; 102 CFU – 20 species/genera. Twenty-nine samples were reported as clinically significant. Among them NGS of 16S-23S rRNA region identified only one bacterial species in 11 samples (UR34, UR35, UR37-UR44, UR50). In the majority of these cases (n = 10) pathogenic species identified by NGS of 16S-23S rRNA region were also identified as the cause of infection by conventional culture methods. The only exception was sample UR44, for which NGS of the 16S-23S rRNA identified solely K. pneumoniae, but culture also identified Proteus mirabilis with 103 CFU/ml. In 12 samples, in which 2 or more species were identified by NGS of the 16S-23S rRNA region (UR3-UR5, UR7-UR9, UR33, UR46, UR48, UR49, UR53, UR55), contigs representing the culture identified pathogenic species consisted of the highest number of reads (53.7–99.1% of all reads). For 3 samples (UR2, UR9, UR59) NGS of the 16S-23S rRNA region showed improved identification compared to conventional culture. In sample UR2 conventional identification showed only the presence of P. mirabilis, however the NGS-based method also revealed the presence of the genetically related species; Proteus vulgaris. In the sample UR9, Klebsiella oxytoca was identified by NGS of 16S-23S rRNA region additionally to E. coli. Colonies produced by these two microorganisms can be difficult to distinguish due to similar morphology on agar plates. In the sample UR59 possible misidentification by conventional culture method could occur. In this sample K. pneumoniae was identified by Vitek MS and Klebsiella variicola by NGS of 16S-23S rRNA region. It was shown very recently, since K. variicola is closely related to K. pneumoniae, it is difficult to distinguish between these two species by commonly used methods, including Vitek MS9. In the remaining 4 positive samples (UR1, UR36, UR52 and UR54) bacterial species that were identified by conventional culture methods as the cause of infection were also found by the NGS of 16S-23S rRNA method but were not predominant. Usually, only contigs containing commensal organisms consisted of higher number of reads. The urine samples classified by culture as “no clinical significance” (n = 31) contained numerous bacterial species, which probably represented the commensal flora. PCR-negative samples (n = 13) were only those samples with a growth density of 102 CFU/ml.
Blood stream infections
Among the 23 positive blood culture samples, conventional culture methods and the NGS of the 16S-23S rRNA region approach produced the same identifications for 20 samples (Table 2). For 3 samples (BC11, 19 and 22) identification of bacterial organisms to the species level was improved by the NGS-based method. In case of sample BC11 conventional culture methods were only able to identify the microorganism to the genus level of Bacteroides. In the other two samples, the contigs representing Streptococcus lutetiensis (BC19) and Bacteroides dorei (BC22) showed 99.9% (4283/4284) and 100% (4433/4433) sequence homology, respectively, during BLAST analysis of the 16S-23S rRNA region. Only a 16S rRNA sequence for Streptococcus infantis was available in the GenBank database showing 95.4% (1413/1481) similarity and for Bacteroides vulgatus the whole 16S-23S rRNA region shared only 96.5% (4286/4443) similarity to the obtained contig sequence.
Orthopedic infections
Of the 21 clinical samples of orthopedic patients 18 were found to be culture negative and three samples were not cultured. In 5 samples (KM2, KM7, KM12, KM16 and KM20) the NGS-based method was able to detect only eukaryotic DNA and in 2 samples (KM8 and KM10) it yielded non-interpretable results most likely because of template degradation. In the remaining 14 samples the number of microorganisms detected in the orthopedic samples ranged from 1 to 3 different genera/species (Table 3). The only exception was sample KM9 in which 11 different genera/species were found. In our study, orthopedic samples had a low amount of starting material, so they were especially prone to be swamped by the contaminating DNA and result in misleading results. All negative controls contained Pseudomonas fluorescens. In a single negative control, Propionibacterium acnes was also found. Therefore, in orthopedic samples P. fluorescens (present in KM11 and KM19) and P. acnes (KM1, KM5, KM9, KM11, KM13, KM14, KM18 and KM19) were regarded as contamination introduced during sample preparation. Moreover, several bacterial genera previously reported as contamination of DNA extraction kits, PCR and other laboratory reagents were absent in all negative controls but present at orthopedic samples, including Acinetobacter, Bacillus, Chryseobacterium, Enhydrobacter, Paracoccus, Psychrobacter and Undibacterium 10. These bacteria can also be potentially considered as contaminations of the 16–23S rRNA NGS process introduced during sample processing.
Reproducibility
To assess the reproducibility of the method based on NGS of the 16S-23S rRNA region, 6 urine samples were used: 2 samples with bacterial count level at 102 CFU/ml, single samples with bacterial count levels at 104 and 103 CFU/ml, and 2 samples with bacterial count level at ≥ 105 CFU/ml (Table 4). The samples were analyzed in triplicate by 3 different operators, including independent PCR amplification, PCR product purification, and NGS library preparation. Subsequently, two independent MiSeq sequencing runs were conducted. The first replicate of each sample was sequenced in the first run, while the second and third replicate of the tested samples were sequenced in parallel in the second run. When an organism identified was represented by 5% or more reads in at least one replicate of a sample, this organism was always detected in its remaining two replicates (Table 4). The only exception was Undibacterium oligocarboniphilum, previously reported as a contaminant in DNA extraction kits, and in PCR and other laboratory reagents10. In the samples with bacterial count level at ≥105 CFU/ml, the method achieved 100% of reproducibility with respect to the bacterial composition. Moreover, in all replicates of a sample with a bacterial count level at ≥103 CFU/ml always the same organism was predominant and represented by 50.9–100% of reads.
Discussion
As defined in the CLSI guidelines, a species identification can be assigned when the max score is 99% or higher and if the sequence similarity between best and second best species are greater than 0.5% using DNA target sequencing11. However, species identification using CLSI’s criteria for the 16S rRNA gene sequences was often weak because the criteria of distance scores greater than 0.5% to the next closest species was not met for most strains. In such instances, only identification to the genus level was feasible. Therefore, Park et al.12 proposed a modified CLSI (mCLSI) method which was more practical and pragmatic for identification of species based on 16S rRNA sequences than the CLSI method. The mCLSI method assigns bacterial species when the similarity score is 99% or higher but irrespective of the similarity score differences. In our study, we applied the similarity score differences with the next closest species as ≥0.2%, which reflected at least 3 and 9 nucleotides difference by sequencing the 16S rRNA gene or the 16S-23S rRNA region, respectively. It allowed elimination of misidentification of very closely related species like S. oralis and S. mitis (Table S1) which in their 16S rRNA gene sequence differ only by 1 nucleotide. Using Sanger sequencing of the 16S rRNA gene, species was assigned with the similarity score 99.4% or higher (Table S1). Using NGS of the 16S-23S rRNA region, in a great majority of cases (n = 61), species were assigned with a similarity score above 99.2%. The exceptional 6 species with lower scores showed a similarity ranging from 96.2% to 98.8%. This could be caused by the facts that (i) a limited number of 16S-23S rRNA sequences were available in databases and (ii) the CLSI guidelines were developed for comparison of gene sequences and did not include the intergenic regions which for some genera can be endowed with higher variation. We believe that with increasing number of deposited 16S-23S rRNA sequences, it will be always possible to assign bacterial species with similarity scores of 99% or higher.
We assigned bacteria to the genus level by NGS of 16S-23S rRNA region in clinical samples when the similarity score was at least 90%. This value was determined based on the results produced during comparison of the four identification methods on pure cultures (Table S1). The 16S-23S rRNA sequence alignments produced the lowest identity for second closest match within the same genus for: Actinobaculum 88% (Table S1; sample 6), Lactobacillus 89% (sample 30), Corynebacterium 91% (samples 14, 17 and 18). We applied the uniform cut-off equal to 90% for all bacteria in clinical samples. However, defining 16S-23S rRNA sequences for all or at least majority species within a genus will create interpretive criteria for defining the genus and probably will vary according to the queried microorganism.
The 16S-23S rRNA sequences can differ by length even among highly genetically related species. It is caused by size variation in 16S-23S rRNA intergenic spacer region (ISR). Staphylococcus aureus clonal complex 75 has been recently renamed as Staphylococcus argenteus and now is a novel species of the Staphylococcus genus13. S. argenteus showed identical or nearly identical 16S and 23S rRNA gene sequences to S. aureus but differed substantially by the ISR length. The scoring system of BLAST had been designed not to allow large gaps. The BLAST algorithm produced a set of smaller separate alignments, with the longest alignment encompassing the 23S rRNA gene (about 2, 5 kb in size). The lack of an alignment for the full-length sequences allowed distinguishing S. aureus from S. argenteus.
In our study all negative controls contained P. fluorescens. We, however, only found this microorganism in the orthopedic samples (Table 3) and not in the urine and blood samples (Tables 1 and 2). Orthopedic samples were characterized with lower starting microbial material compared to that of the urine and blood samples. This showed that the impact of contaminating sequences is greater in low biomass samples. Moreover, P. fluorescens had not been associated previously with human infections and its presence can be clearly regarded as contamination. However, in a single negative control, which was introduced to monitor the impact of contaminations on identification procedure of the orthopedic samples, P. acnes was also identified. P. acnes is a common human skin-associated organism and had been previously shown as the cause of orthopedic infections14, 15. It will be highly important to limit the impact of such contaminations as P. acnes during samples handling as this is also a clinically significant microorganism and its contribution to an infection can be misinterpreted.
NGS of the 16S-23S rRNA region will provide enhanced information on the presence of microbial DNA within a clinically relevant time frame which is necessary for timely and accurate treatment of infections and therefore key for proper infection and antibiotic resistance control. Data of this study showed that this method correctly identified bacterial species that were identified as the cause of infection by conventional culture methods. In general, in the culture-positive samples a few additional species/genera per sample were identified by the NGS method. However, contigs representing the culture identified pathogenic bacterial species had the highest number of sequence reads. To assess the clinical relevance of the identified species/genera in samples with a low amount of bacterial DNA, a prospective clinical validation study will be carried out including samples of complex patient groups (e.g. orthopedic patients with a clinical suspicion of a prosthetic joint infection) and samples of control groups of patients without a clinical suspicion of an infection.
Currently, NGS of the 16S-23S rRNA region suffers from two major limitations. First limitation is a lack of reference database with the 16S-23S rRNA sequences and a complementary software allowing easy and reliable species identification. Second major limitation of the method is a lack of reference sequences for many bacterial species. We currently work on the development of a reference data set for assigning clinically relevant bacterial species based on the 16S-23S rRNA sequences. The quality and amount of data accumulated in the databases is particularly important for the performance of bacterial identification using sequencing analysis of the 16S-23S rRNA region. Also, this lack of reference 16S-23S rRNA sequences in the GenBank database might have potentially introduced bias in the results when comparing 16S rRNA with 16S-23S rRNA sequences.
As a proteomic tool for microbial identification, MALDI-TOF MS is superior to NGS-based methods in cost and speed. Currently, the total costs of all reagents and consumables for NGS of 16S-23S rRNA region (not including labor) amount to ~70 € per sample with turnaround time of ≤4 days. However, NGS of the 16S-23S rRNA region was culture independent and more significantly discriminative at the species and genus level than MALDI-TOF MS approaches. Moreover, NGS of 16S-23S rRNA region, unlike all other mass spectrometry methods described above, can be extremely useful for identification of rare or unknown bacteria and bacteria with unusual phenotypic profiles.
In summary, the main objective of the present study was to develop a new diagnostic method based on NGS of 16S-23S rRNA region and assess its identification potential. Its resolution power was found to be superior to the other identification approaches commonly used in routine clinical microbiology laboratories; moreover, the method was easy in use. The method correctly identified urinary tract pathogens and blood stream pathogens previously identified as the cause of UTI and blood stream infection (BSI) with conventional culture. NGS of the 16S-23S rRNA region also showed increased sensitivity in the diagnosis of bacterial microorganisms in samples collected from patients suspected of orthopedic infections. In the analyzed samples several bacterial species had been previously reported as the cause of orthopedic infections, including Haemophilus parainfluenzae 16, Moraxella osloensis 17 and S. epidermidis 18. Therefore, we conclude that our approach has the potential to increase diagnostic yield and will decrease time to result for detection of unexpected bacterial pathogens and bacterial species compared to current methods, thereby improving targeted antibiotic treatment. Furthermore, there is a huge potential of this method for detection of bacterial pathogens that can not be cultured at all, due to VNC state or due to antibiotics prior to collection of sample. Finally, with our method it will be possible to streamline processes in the laboratory and to implement it in several disciplines, like clinical, environmental and veterinary microbiology. However, this approach needs further validation and determination of its sensitivity. Furthermore, studies focused on the clinical relevance are necessary for determining the applicability of this NGS-based approach in routine diagnostics.
Materials and Methods
Ethics
All procedures were carried out according to guidelines and regulations of Certe concerning the use of patient materials for the validation of clinical methods, which are in compliance with the guidelines of the Federation of Dutch Medical Scientific Societies (FDMSS). The project was approved by the Certe medical staff under project submission 3305-0037. Every patient of the Certe is informed that samples taken may be used for research and publication purposes unless they indicate that they do not agree to it. Informed consent was obtained from all individuals or their guardians prior to study participation. All samples were used after performing and completing a conventional microbiological diagnosis and were coded to protect patients confidentiality.
Bacterial isolates
Pure cultures of 44 difficult to identify clinically bacterial isolates were included. Additionally, a selection of 23 isolates was used, including 15 ATCC strains (Table S1).
Clinical samples
Sixty fresh urine samples from patients suspected of urinary tract infections (UTI), 23 BacT/ALERT (bioMérieux) positive blood culture bottles from patients suspected of BSI and 21 clinical samples of orthopedic patients were collected. Culture was performed as part of routine diagnostics by the department of Bacteriology at Certe.
Vitek MS (bioMérieux)
Vitek MS slides were prepared and interpreted according to the manufacturer’s instructions. Strains which did not yield an identification due to unreliable results or bad spectra were repeated once.
Bruker MS (Bruker)
Bruker Microflex slides were prepared using on-slide extraction and interpreted according to manufacturer’s instructions. Strains which did not yield an identification due to unreliable results or bad spectra were repeated once.
Bacterial isolates DNA extraction
Blood agar plates were used for culturing the bacterial strains. DNA isolation from pure cultures was performed using the UltraClean Microbial DNA Isolation Kit (Mobio) according to the manufacturer’s instructions.
Urine DNA extraction
DNA isolation of urine samples was performed using the UltraClean Microbial DNA Isolation Kit (Mobio). Briefly, 500 µl of urine was centrifuged for 30 seconds at 10,000 g. The supernatant was discarded and the pellet reconstituted in 300 µl Microbead solution. Since then the manufacturer’s instructions were followed for DNA extraction.
Blood culture bottle DNA extraction
DNA from 1.8 ml from positive blood culture bottles was extracted using the BiOstic Bacteraemia DNA Isolation Kit (Mobio) according to the manufactures instructions.
DNA extraction of the clinical samples of orthopedic patients
The Purelink Genomic DNA purification kit (Invitrogen) was used for DNA extraction of the orthopedic samples. Initial lysis was performed using lysis buffer (0.25 M NaCl, 0.5% SDS, 5x TE, 2.25 U/ml Proteinase K (Roche)). For tissue samples a small piece was digested in 200 µl lysis buffer. For fluidic samples 200 µl was added to 200 µl lysis buffer. Digestion was performed in a thermoshaker at 56 °C under light shaking for 18 hours. 200 µl Purelink Genomic lysis/binding buffer was added to 200 µl of lysed sample and vortexed to create a homogenous solution. 200 µl 96% ethanol was added and the DNA purification protocol was followed according to the manufacturer’s instructions.
16S rRNA gene amplification and Sanger sequencing
The DNA of the bacterial isolates and controls were amplified with 0.3 μM primers [LPW57 5′-AGTTTGATCCTGGCTCAG-3′ (nucleotide position 10–27) and LPW58 5′-AGGCCCGGGAACGTATTCAC-3′ (nucleotide position 1370–1389)]19 using GoTaq® MDx Hot Start Polymerase and Colorless GoTaq® Flexi Buffer (Promega). PCR amplification was performed on a PTC-100 thermocycler (Bio-Rad) using the following conditions: an initial incubation at 94 °C for 2 min, 25 cycles of 94 °C for 30s, 58 °C for 30s, and 72 °C for 1 min, followed by a final incubation at 72 °C for 5 min. PCR products were purified using DNA Clean & Concentrator (Zymo Research). DNA sequencing was performed by the GATC company (Cologne, Germany) using the PCR primers. 16S rRNA gene sequences were subjected to BLAST analysis against the NCBI nucleotide database. The sequencing data analysis allowed assigning species when the similarity score was 99% or higher and the similarity score differences with the next closest species were equal to or greater than 0.2%, which reflected 3 or more nucleotides. The genus was assigned when the similarity score was 97% or higher. The BLAST analysis of the 16S rRNA gene sequences was verified against the curated leBiBi-QBPP database (https://umr5558-bibiserv.univ-lyon1.fr/lebibi/lebibi.cgi) not revealing any differences in identification.
Amplification of the 16S-23S rRNA region
The 16S-23S rRNA region was amplified by PCR using forward primer 27F (5′-AGAGTTTGATCMTGGCTCAG-3′), specific for the 16S rRNA gene and adapted from the study by Sergeant et al.20, and reverse primer 2490R (5′-GACATCGAGGTGCCAAAC-3′), specific for the 23S rRNA gene and slightly modified by truncation of a single nucleotide compared to the original publication by Hunt et al.21. The amplification of the 16S-23S DNA region was carried out in a 25 µl reaction consisting of 1x Phire hotstart buffer (Thermofisher), 5 mM dNTP’s (Roche), 0.5 µl Phire hotstart II DNA polymerase (Thermofisher), 600 nM of each primer and 5 µl of DNA template. PCR was performed using a Biorad PTC-200 thermocycler. An initial denaturation of 98 °C at 30 sec was followed by 25 cycles of 98 °C for 30 sec, 66 °C for 30 sec and 72 °C for 2 minutes with a final extension of 72 °C for 1 min. To increase PCR sensitivity in clinical samples the number of cycles was increased to 35. PCR products were purified using the Qiaquick PCR purification kit (Qiagen) according to the manufacturer’s instructions.
NGS library preparation and Illumina MiSeq sequencing
For library preparation, the Nextera XT DNA Sample Preparation Kit (Illumina) was used according to the manufacturer’s instructions. Briefly, the purified PCR amplicons quantified with a Qubit 2.0 Fluorometer (Thermofisher) were diluted to 0.2 ng/µl and a total of 1 ng of DNA was tagmented at 55 °C for 5 min. PCR amplification to introduce Illumina index sequences was performed in PCR strip tubes in a BioRad T100 thermocycler. Size distribution of fragments was estimated with a 2200 TapeStation using the Agilent D1000 High Sensitivity kit according to the manufacturer’s instructions. Fragments of 200 to 1000 bp were obtained. The library DNA fragments were size selected and purified using AMPure XP beads (Beckman Coulter, Inc.). The indexed libraries were normalized, pooled and loaded onto an Illumina MiSeq reagent cartridge using MiSeq reagent kit v3 and 600 cycles. The paired-end 2 × 300 bp sequencing was run on an Illumina MiSeq sequencer.
Sequencing of the 16S-23S rRNA region and data analysis
NGS generated 25,000–50,000 or 1–2 million sequencing reads for pure culture or clinical sample, respectively, to obtain a minimum coverage of 1000 per sample. The 300-nucleotide paired-end reads were de novo assembled into contigs with the SeqMan NGen software (DNASTAR) using parameters: mer size 31 nucleotides and minimum match percentage 93%. The sizes of resulting contigs produced during analysis of pure cultures ranged from 4183 bp (Bacillus cereus) up to 4856 bp (Acinetobacter parvus) with an average size of 4423 bp (Table S1). In case of clinical samples, most often an identified bacterial species/genus was represented by a single contig of expected size, around 4.5 kb. However, in some instances only smaller contigs (ranging between 1 and 3 kb in size) represented the same bacterial species/genus present in a sample. In those cases all contigs belonging to the same organism were combined and their reads were added up. Species identification was based on alignment of contig sequences with 16S-23S rRNA sequences deposited in the GenBank database using nucleotide BLAST (Basic Local Alignment Search Tool, http://www.ncbi.nlm.nih.gov/BLAST/). When a reference 16S-23S rRNA sequence was not available in the database, a reference 16S rRNA gene sequence was used for a species identification. The contig sequences were submitted via the website and bacterial species was assigned when the similarity score was 99% or higher and the similarity score differences of the first match with the next closest species was equal to or greater than 0.2%. This reflected a 9 or more nucleotides difference for a sequence of 4423 bp (the average size of the 16S-23S rRNA amplicon). When the similarity score was between 90% and 99%, the genus could be assigned. The score below 90% was interpreted as an unidentified organism.
Statistical analysis
The four bacterial identification methods were compared using McNemar’s test, a test of paired proportions. Only a single strain per species was taken (from Table S1) for calculation of the accuracy of bacterial identification at the species or genus level according to McNemar’s test. When the P value was less than 0.05, we concluded that there was a significant difference between the methods. Statistical analysis was performed using the GraphPad software.
References
Law, J. W., Ab Mutalib, N. S., Chan, K. G. & Lee, L. H. Rapid methods for the detection of foodborne bacterial pathogens: principles, applications, advantages and limitations. Front. Microbiol. 5, 770 (2014).
Rudkjobing, V. B. et al. Comparing culture and molecular methods for the identification of microorganisms involved in necrotizing soft tissue infections. BMC Infect. Dis. 16, 652 (2016).
Hasman, H. et al. Rapid whole-genome sequencing for detection and characterization of microorganisms directly from clinical samples. J. Clin. Microbiol. 52, 139–146 (2014).
Patel, J. B. 16S rRNA gene sequencing for bacterial pathogen identification in the clinical laboratory. Mol. Diagn. 6, 313–321 (2001).
Kalia, V. C., Kumar, R., Kumar, P. & Koul, S. A Genome-Wide Profiling Strategy as an Aid for Searching Unique Identification Biomarkers for Streptococcus. Indian J. Microbiol. 56, 46–58 (2016).
Kommedal, O., Karlsen, B. & Saebo, O. Analysis of mixed sequencing chromatograms and its application in direct 16S rRNA gene sequencing of polymicrobial samples. J. Clin. Microbiol. 46, 3766–3771 (2008).
Junemann, S. et al. Updating benchtop sequencing performance comparison. Nat. Biotechnol. 31, 294–296 (2013).
Loman, N. J. et al. Performance comparison of benchtop high-throughput sequencing platforms. Nat. Biotechnol. 30, 434–439 (2012).
Berry, G. J., Loeffelholz, M. J. & Williams-Bouyer, N. An Investigation into Laboratory Misidentification of a Bloodstream Klebsiella variicola Infection. J. Clin. Microbiol. 53, 2793–2794 (2015).
Salter, S. J. et al. Reagent and laboratory contamination can critically impact sequence-based microbiome analyses. BMC Biol. 12, 87 (2014).
Clinical Laboratory Standards Institute. Interpretive criteria for identification of bacteria and fungi by DNA target sequencing; approved guideline. MM18-A. Wayne, PA CLSI (2008).
Park, G. et al. Evaluation of four methods of assigning species and genus to medically important bacteria using 16S rRNA gene sequence analysis. Microbiol. Immunol. 59, 285–298 (2015).
Holt, D. C. et al. A very early-branching Staphylococcus aureus lineage lacking the carotenoid pigment staphyloxanthin. Genome Biol. Evol. 3, 881–895 (2011).
Bossard, D. A. et al. Optimal cultivation length to isolate Propionibacterium acnes in suspected bone and joint infections is >7 days. J. Clin. Microbiol. 54, 3043–3049 (2016).
Lavergne, V. et al. Clinical impact of positive Propionibacterium acnes cultures in orthopedic surgery. Orthop. Traumatol. Surg. Res. doi:10.1016/j.otsr.2016.12.005 (2017).
O’Neil, C. R., Wilson, E. & Missaghi, B. Bone and Joint Infections due to Haemophilus parainfluenzae: Case Report and Review of the Literature. Can. J. Infect. Dis. Med. Microbiol. 2016, 4503025 (2016).
Sugarman, B. & Clarridge, J. Osteomyelitis caused by Moraxella osloensis. J. Clin. Microbiol. 15, 1148–1149 (1982).
Morgenstern, M. et al. Biofilm formation increases treatment failure in Staphylococcus epidermidis device-related osteomyelitis of the lower extremity in human patients. J. Orthop. Res. 34, 1905–1913 (2016).
Woo, P. C., Cheung, E. Y., Leung, K. & Yuen, K. Identification by 16S ribosomal RNA gene sequencing of an Enterobacteriaceae species with ambiguous biochemical profile from a renal transplant recipient. Diag. Microbiol. Infect. Dis. 39, 85–93 (2001).
Sergeant, M. J., Constantinidou, C., Cogan, T., Penn, C. W. & Pallen, M. J. High-throughput sequencing of 16S rRNA gene amplicons: effects of extraction procedure, primer length and annealing temperature. PloS one 7, e38094 (2012).
Hunt, D. E. et al. Evaluation of 23S rRNA PCR primers for use in phylogenetic studies of bacterial diversity. Appl. Environ. Microbiol. 72, 2221–2225 (2006).
Acknowledgements
This study was partly supported by the InterregIVa- and InterregVa-projects EurSafety Health-net (InterregIVa/III-1-02 = 73) and EurHealth-1Health (InterregVa/202085), as part of the euregional diagnostic reference activities within the Dutch-German cross-border network supported by the European Commission, the German Federal States of Nordrhein-Westfalen and Niedersachsen, and the Dutch provinces of Overijssel, Gelderland, and Limburg. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. We thank Julia Berends, Emmy de Paus and Circe van der Heide for analysis of some urine samples during the study. We also thank Milena Brasca from Institute of Sciences of Food Production, Milan, Italy, for providing Enterococcus lactis strain.
Author information
Authors and Affiliations
Contributions
A.M.D.K.-S., E.Z., G.W., K.S. and R.F.B. designed the project, provided the strains, clinical material and epidemiological data. A.J.S., V.A., E.Z., G.W., and K.S. performed the experiments. A.J.S. and E.Z. carried out de novo assemblies. All authors interpreted the data. A.J.S. wrote the manuscript. All authors reviewed the manuscript.
Corresponding author
Ethics declarations
Competing Interests
The authors declare that they have no competing interests.
Additional information
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Sabat, A.J., van Zanten, E., Akkerboom, V. et al. Targeted next-generation sequencing of the 16S-23S rRNA region for culture-independent bacterial identification - increased discrimination of closely related species. Sci Rep 7, 3434 (2017). https://doi.org/10.1038/s41598-017-03458-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-017-03458-6
- Springer Nature Limited
This article is cited by
-
Harnessing the intragenomic variability of rRNA operons to improve differentiation of Vibrio species
Scientific Reports (2024)
-
Application of metagenomics for diagnosis of broilers displaying neurological symptoms
BMC Veterinary Research (2023)
-
Genetic and phenotypic assessments for the safety of probiotic Bacillus clausii 088AE
3 Biotech (2023)
-
Bacterial diversity in water from Xifeng Hot Spring in China
Brazilian Journal of Microbiology (2023)
-
Unveiling the microbial diversity and functional dynamics of Shiv Kund, Sohna hot spring, India through a shotgun metagenomics approach
Archives of Microbiology (2023)