Objective 1 (Optimize the performance of 16S HTS in CSF samples with low bacterial loads)
We used laboratory techniques to control systematic error associated with reagent contamination, DNA extraction efficiency and bias, and batch effects, comparing three such bench pipelines when extracting an experimentally-constructed suspension of bacteria commonly infecting CSF (a “mock community”). Two computational methods were then used to remove potential contaminant sequences from the resulting 16S HTS results. The refined approach was then applied to bacterial meningitis samples. (Figure 1) More complete details of each of these approaches are provided as follows.
Mock Community
The mock community was created by mixing equivalent cell numbers of type strains of bacteria known to be frequently isolated from CSF shunt infections.[10] (Table 1), yielding a total final concentration of 6.00 x 107 CFU/µl. Serial 1:10 dilutions were made of this mock community, ranging from 6.00 x 107 CFU/µl to 6.00 x104 CFU/µl.
Table 1
Bacterial species and genera present in the CSF mock community
Bacterial species and genera
|
CFU/µl
|
Acinetobacter baumannii
|
5.00E+06
|
Corynebacterium sp.
|
5.00E+06
|
Cutibacterium acnes
|
5.00E+06
|
Enterococcus faecalis
|
5.00E+06
|
Escherichia coli
|
5.00E+06
|
Klebsiella pneumoniae
|
5.00E+06
|
Proteus mirabilis
|
5.00E+06
|
Pseudomonas aeruginosa
|
5.00E+06
|
Staphylococcus aureus
|
5.00E+06
|
Staphylococcus epidermidis
|
5.00E+06
|
Streptococcus pyogenes
|
5.00E+06
|
Streptococcus salivarius
|
5.00E+06
|
Total CFU/µl
|
6.00E+07
|
We used this mock community to optimize and compare our sample processing and sequencing procedures prior to analyzing patient CSF samples and to identify and address biases introduced during DNA extraction, library construction, and taxonomic assignment. A total of 36 mock community samples (including dilutions) were included in our experiments comparing results of three different DNA extraction protocols (Table 2).
Control Samples
In addition to mock communities, each mock community analysis included several control samples to identify and quantify bacterial DNA contamination in DNA extraction kits and PCR reagents (reagent contamination). These controls included (1) samples in which the processing reagents were included, with no purposely input (“template”) DNA, from DNA extraction and before the PCR step required to construct sequencing libraries (“no-sample extraction control”), and (2) other samples with only the reagents used during and after PCR and library construction (“no-template library PCR control”). In both types of controls, water was used to make up for the sample volumes of CSF and mock community extraction experiments. A total of 35 no-template extraction controls were included in DNA extraction experiments, and 18 no-template controls were included in library experiments. (Table 2). The PCR reaction for a group was repeated if the no-template control yielded a visible EtBr-stained band on a 1% agarose gel. Otherwise, CSF sample amplicons and no-template controls were analyzed using identical protocols.
Table 2: Number of control samples used for each DNA extraction type and for library PCR
|
Controls
|
Purpose
|
DNA Extraction Method
|
Total
|
|
mag mini kit (MAG)
|
BiOstic Bacteremia DNA isolation Kit (BNC)
|
BiOstic Bacteremia DNA isolation Kit with carrier RNA
(BWC)
|
Mock community
|
Identification of possible biases introduced during DNA extraction, library construction, and taxonomic assignment
|
81
|
161
|
121
|
36
|
No-sample DNA extraction controls
|
Identification of potential contaminants from kit components
|
112
|
132
|
112
|
35
|
Sample No-template library PCR controls
|
Identification of potential contaminants during generation of 16S rRNA libraries
|
3
|
3
|
3
|
93
|
Mock Community No-template library PCR controls
|
Identification of potential contaminants during generation of 16S rRNA libraries
|
|
|
|
43
|
Extraction control No-template library PCR controls
|
Identification of potential contaminants during generation of 16S rRNA libraries
|
|
|
|
53
|
Sequencing negative control
|
Identification of mis-assigned reads during sequencing analysis
|
|
|
|
1
|
Total
|
90
|
1Number of mock community controls per method varied depending on number of kits used to extract the sample set, for required technical replicates, and to address day-to-day variability |
2Number of no-sample DNA extraction controls per method depended on number of kits used to extract the sample set and required technical replicates, which differed between methods |
3Total number of no-template library PCR control includes three such controls performed for each DNA extraction method (nine in total), four for the mock community analysis, five for the no-sample DNA extraction controls. |
Dna Extraction
To address methodologic variation in DNA extraction efficiency and bias, DNA was extracted and purified from all samples using three extraction approaches: the AGOWA mag Mini DNA isolation kit (AGOWA, LGC Genomics, Berlin, Germany), hereafter MAG; the BiOstic Bacteremia DNA isolation Kit (Qiagen), hereafter BNC; and BNC with the inclusion of carrier RNA (Qiagen) to increase DNA yield by preventing binding to plastic in the kits[11], hereafter BWC. Components of each kit were aliquoted before extraction in an AirClean® Systems PCR Workstation (USA Scientific) decontaminated with LookOut® DNA Erase (Sigma-Aldrich), according to the manufacturer’s instructions, followed by 15 min of UV-ray exposure to minimize environmental contamination.
MAG extractions were performed as follows without carrier RNA due to manufacturer concerns that this RNA could displace sample DNA during extraction (personal communication). A 100 µl volume of each sample was aliquoted into a sterile low binding microfuge tube (Axygen, Catalogue Number (CN): 31104081), to which 20 µl of 20mg/mL Proteinase K (Invitrogen, CN: 25530-049) was added. The mixture was vortexed for 20 seconds and incubated at 55°C for 10 minutes. After incubation 250 µl of Lysis buffer was added to the tube and vortexed gently for 15 seconds. The mixture was transferred to a clean 2 ml tube (Sarstedt, CN: 72.693.005) containing 0.3 g of 0.1mm zirconia/silica beads (Biospec Products Bartlesville, OK, USA, (Biospec) CN: 11079101z). Using a Mini-Beadbeater-16 (Biospec, CN: 607) the sample was mechanically disrupted by bead-beating for 3 minutes, followed by a 1-minute incubation on ice and a further 3 minutes of bead-beating. The sample was centrifuged at 4,000 rpm for 10 minutes. The resulting supernatant was transferred to a new low binding microfuge tube. To this, 325 µl of Binding buffer and 10 µl of mag particle suspension (mag-particles) were added, vortexed for 15 seconds, and incubated at room temperature for 30 minutes with continuous mixing on a VWR Tube Rotator (VWR, CN: 10136-084). After the incubation step DNA extraction proceeded according to the manufacturer’s instructions as described.[12]
In both BIOstic extraction methods, 100 µl of sample was mixed by gentle vortexing either with or without (depending on the method) 1 µl of added carrier RNA at a stock concentration of 1 µg/µl of RNA. DNA was then extracted from each sample according to the manufacturer’s instructions and as described.[12]
Batch Effects
To control for batch effects, all CSF samples were randomized using a random number generator program.[13] Extractions were performed by research staff blinded to the sample identification key.
Bacterial Quantification
A quantitative PCR (qPCR) assay targeting the 16S rRNA genes was used to measure the total bacterial load in each CSF sample as described earlier.[12]
Bacterial 16S rRNA gene amplification, sequencing and analyses (16S HTS)
Amplicon library construction was carried out using a one-step PCR amplification targeting the 16S rRNA gene V4 region. The amplicon library primers [14] each contained the specific Illumina adapters, an 8-bp index sequence to allow for multiplex sequencing of the samples, and the 16S rRNA gene specific primer.[15] Library construction, pooling and sequencing proceeded as described for 600 cycles on an Illumina MiSeq desktop sequencer using the Miseq Reagent Kit v3.[12]
16s Hts Analysis
Sequencing data were processed using the denoising program DADA2 [16] (version1.6.0) as described,[17] and aligned to the SILVA 16S reference database (v. 132)[18] to produce a 16S amplicon taxa table for downstream computational analysis. All samples, regardless of sample type or extraction protocol, were run on the same Illumina flow cell to reduce inter-run variation.
Post-sequencing Contaminant Removal
Two analytic strategies were used to identify and remove contaminant sequences. Using the decontam R package installed from GitHub (https://github.com/benjjneb/decontam),[19] contaminant sequences were identified as those either with decreasing abundance at higher concentrations (threshold p < 0.10) or that were more prevalent in kit control samples than in CSF samples (threshold p < 0.50). Additionally, all sequences detected in extraction kit controls were considered contaminants and were removed from the CSF samples. Results from each computational approach was compared with the known mock community composition for objective 1 and microbiological culture results for objective 2.
Objective 2 (Perform refined 16S HTS on CSF samples from bacterial meningitis)
Study Subjects
Children ≤18 years old undergoing treatment for conventional culture-confirmed meningitis at St. Louis Children’s Hospital in St. Louis, Missouri were eligible for enrollment in this study. Enrollment occurred from 2010 to present. Meningitis was defined as identification of organisms on microbiological culture of CSF fluid obtained from a lumbar puncture in a child without an existing neurosurgical device. All study subjects’ CSF underwent routine microbiological processing and testing in the St. Louis Children’s Hospital Microbiology Laboratory. For this study, we examined CSF obtained from the subset of 40 children whose conventional microbiological cultures recovered bacterial organisms and for whom there was 300 ul of banked CSF available. The study received Institutional Review Board approval from the Seattle Children’s Research Institute, St. Louis Children’s Hospital and Children’s Hospital Los Angeles.
Clinical Data
For each child we collected culture information, specifically organism recovered in CSF and blood if applicable.
Csf Specimen Collection
Sterile conditions were standard practice throughout recovery and storage of CSF. CSF samples were obtained from lumbar puncture. The CSF sample analyzed in this study either was left over from this first diagnostic sample or was obtained under sterile conditions. After being collected, CSF samples were stored at 4°C for up to 5 days. CSF was then aliquoted into vials of ~1000 µl for the study and stored at -70°C. After identification for this study, samples were shipped overnight on dry ice for analysis.
Conventional Microbiological Culture Identification Of Bacteria
All CSF samples were tested using routine aerobic culture techniques in hospital-certified laboratories at SLCH. Conventional cultures are the traditional diagnostic approach used to detect typical pathogens in infectious diseases and were performed in a clinical microbiology laboratory following Clinical and Laboratory Standards Institute guidelines; however, conventional culture approaches do not detect all bacteria present in human disease.[20, 21] During analysis, the laboratory team remained blinded to the samples’ culture results.
Data Analyses
Sequence incidence was calculated for each sample as the ratio of sequence reads divided by total reads. While sequences have been linked to genera to aid in clinical interpretation, multiple sequences may be associated with the same organism or with unique variants; therefore, it is unclear whether adding read counts from unique sequences associated with common genus provides an accurate estimate of the incidence of that genus. For visual clarity, sequences representing less than 10% of reads across all samples, extraction methods and post-processing steps were aggregated to a single ‘other’ category in figures.