Diagnosis of external ventricular drainage related infections with real-time 16S PCR and third-generation 16S sequencing

Abstract Objective Investigate the performance of real-time 16S PCR and third-generation 16S sequencing in the diagnosis of external ventricular drain related infections (EVDRI). Methods Subjects with suspected EVDRI were prospectively included at Uppsala University Hospital. Subjects were included into three groups: subjects with negative CSF culture with and without antibiotic treatment and subjects with positive CSF culture, respectively. CSF was analysed with real-time 16S PCR and third-generation 16S sequencing. Real-time 16S PCR positivity/negativity and number of 16S sequence reads were compared between groups. For culture positive subjects, species identification in third-generation sequencing and routine culture was compared. Results 84 subjects were included. There were 18, 44 and 22 subjects in the three groups. Real-time PCR was positive in 17 of 22 subjects in the culture positive group and negative in 61 of the 62 subjects in the two culture negative groups. The sensitivity and specificity for real-time 16S PCR compared to culture was estimated to 77% and 98%, respectively. Species identification in 16S sequencing and culture was concordant in 20 of 22 subjects. The number of 16S sequence reads were significantly higher in the culture positive group than in both culture negative groups (p < 0.001). There was no significant difference in number of 16S sequences between the two culture negative groups. Conclusions Real-time 16S PCR predict culture results with sufficient reliability. Third-generation 16S sequencing could enhance sensitivity and species identification in diagnostics of EVD-related infections. False negative culture results appear to be uncommon in patients with suspected EVDRI.


Introduction
Clinical symptoms and changes in cerebrospinal fluid (CSF) biomarkers associated with central nervous system (CNS) infections are unspecific and hard to interpret in neurosurgical patients; especially in critically ill patients with external ventricular drains (EVDs) and suspicion of EVD-related infection (EVDRI).This leads to high rates of empirical antibiotic treatment and missed diagnoses [1-3].The microbiological gold standard of bacterial culture on CSF may also lack in diagnostic performance in this setting as the sensitivity could be hampered by ongoing antimicrobial therapy.In addition, the common findings of bacteria that are considered potential contaminants with low virulence, such as Coagulase negative Staphylococci (CoNS), reduces specificity [4].
Polymerase chain reaction (PCR) is used to analyse if a specific deoxy-ribonucleic acid (DNA) or ribonucleic acid (RNA) sequence is present in a sample.The 16S rRNA gene in bacteria codes for RNA that is a component of the 30S subunit of the bacterial ribosome.This gene contains both regions that are highly conserved as well as those highly variable between different species of bacteria.This makes it a useful target for PCR since the conserved regions can serve as targets for PCR primers for amplification for a broad spectrum of bacteria, while sequencing of the variable regions in the resulting amplicon can often identify which species of bacteria that is present in the sample [5].Hence, 16S PCR has become a widespread diagnostic method for a wide range of bacterial infections.Next generation sequencing (NGS) based technologies such as Nanopore sequencing have greatly improved the capacity and bandwidth of nucleotide sequencing.Through direct detection of bacterial pathogens by sequencing of the 16S gene, sensitivity and species identification could potentially be improved, turnover times reduced and allow detection of polymicrobial infections [6].
CSF 16S PCR is used in many centres for the aetiological diagnosis of nosocomial CNS infections in neurosurgical patients, but analyses of the performance of PCR-based diagnostic methods in this setting are limited and the results are hitherto somewhat diverging.Generally, PCR results have shown good agreement to bacterial culture results in patients with positive CSF cultures but the rate of PCR positivity in culture negative samples has varied [7][8][9][10][11][12].The turn-around time for PCR-based methods is usually shorter than for culture results.This means that, if the results are reliable, the use of empirical antimicrobial therapy could be reduced while ensuring adequate and early therapy for patients with infections.A recently published study by Jang and co-workers suggests that the detection of 16S with the third-generation sequencing Nanopore technology enhances the detection and etiologic diagnosis of bacterial meningitis after neurosurgery [13].
The main objective of this study was to investigate the diagnostic performance of real-time 16S PCR and third-generation 16S metagenomic sequencing with the long-read Nanopore technology, on CSF from neurosurgical patients with culture confirmed EVDRI.Secondary aims were to investigate the rate of real-time 16S PCR positivity and the number of 16S sequence reads in Nanopore sequencing in CSF samples from patients with clinically suspected infection but negative CSF cultures and, furthermore, study if there were differences between subjects without and with ongoing antibiotic treatment at time of sampling, indicating a risk of false negative culture results in the latter.

Study design and setting
This was a prospective observational study of patients in the neuro-intensive care unit (neuro-ICU) and neurointermediary care unit (neuro-IMCU) with suspected infection at the time of sampling.Subjects were included between February 2018 and November 2021 at the neuro-ICU or neuro-IMCU at Uppsala University Hospital, Uppsala, Sweden.This regional centre for neuro-ICU care for parts of middle Sweden, with a catchment area housing approximately 2 million inhabitants, is supported daily by infectious disease specialists.
Inclusion and exclusion criteria are listed in Table 1.In addition to subjects with EVD, subjects with an externalised ventriculoperitoneal (VP) shunt were eligible for inclusion but no such subjects were included.Inclusion was conducted prospectively, with all study subjects with sufficient volume of CSF in the clinical sampling considered eligible for inclusion.However, inclusion rate was at times reduced during vacation periods due to capacity issues.When inclusion into the culture negative subject groups met their pre-allocated sizes, only culture-positive subjects were included.
Subjects were divided into three groups -(1) subjects with negative CSF culture in the study sample and no antibiotic treatment at sampling, (2) subjects with negative culture in the study sample and ongoing antibiotic treatment at sampling and lastly (3) subjects with positive CSF culture in the study sample.Subjects with a negative CSF culture at study sampling and a previous non-study CSF sample with positive culture were however excluded from the group assignment.Subjects with confirmed or suspected community acquired acute bacterial meningitis (ABM) were also excluded from the group assignment.On inclusion, a CSF sample of 2 ml taken at the time of clinical sampling was collected for this study.Samples were frozen in −20 � C within 4 h from sampling.The sample was shortly thereafter thawed, aliquoted and frozen in −70 � C.After completion of the clinical bacterial culture the subjects were allocated to the pre-defined study groups.For subjects with repeated study samples where several were culture positive, the first culture positive sample was used.Similarly, for subjects with repeated culture negative study samples, the first culture negative sample was used.

EVD catheters and CSF sampling
EVD insertion was performed in the operating room under sterile conditions.EVDs without anti-bacterial coating were used (HanniSet, Xtrans, Smith Medical GmbH, Glasbrunn, Germany).Catheters were inserted by a right sided frontal burr hole and tunnelled subcutaneously approximately five cm from the incision site.The EVD was connected to an external draining system with a pressure monitoring device (HanniSet, Xtrans, Smith Medical GmbH, Glasbrunn, Germany or VentrEX, Neuromedex, Hamburg, Germany).Two grams of cloxacillin was administered intravenously at the start of surgery.Clindamycin, 600 mg, was used if the subject was allergic to penicillin.The catheters were not routinely exchanged.CSF samples were taken under aseptic conditions in accordance with clinical routine.

Cultures, CSF cytochemistry and blood sample analysis
Cultures were performed at the Department of Clinical Microbiology and Hospital hygiene at Uppsala University Hospital according to local clinical procedure.All CSF samples were cultured both on solid agar and in broth.
Agar plates were incubated in 35 � C in CO 2 and anaerobically for 8 days.Broth cultures were performed through the inoculation of 2 mL of CSF in pediatric blood culture flasks (PF-BacT/ALERT PF Plus [bioM� erieux Inc., Durham, NC, USA]) supplemented with 4 mL of horse blood and incubated for 10 days at 35 � C in an incubator with automated growth detection (BacT/ALERT VirtuO[bioM� erieux Inc., Durham, NC, USA]).Species identification was performed using normal procedures at the laboratory, ie matrix-assisted laser desorption/ionizationtime of flight (MALDI-TOF) (Bruker Daltonics) and phenotypic tests.
All cytochemical analysis of CSF and blood/plasma was performed at the Department of Clinical Chemistry and Pharmacology at Uppsala University Hospital, as part of the standard clinical procedure.

16S rRNA gene analysis
Analysis of the 16S rRNA gene was performed at the Department of Laboratory Medicine, Clinical Microbiology at € Orebro University Hospital after study inclusion was completed.In addition to the study samples, 25 negative controls (anonymized clinical CSF samples with leukocyte count <4 � 10 6 /L) were included in the analysis.DNA was extracted from 200 mL CSF and pretreated with 100 units of mutanolysin (Sigma-Aldrich, St Louis, MO, USA) at 37 � C for 30 min before extraction with the MagDEA Dx kit on MagLEAD 12gC (Precision System Science Co., Ltd., Chiba, Japan).The elution volume was set to 50 mL.In each extraction run, a positive control (containing Staphylococcus haemolyticus suspended in NaCl, for the Nanopore sequencing) and a negative control (containing only reagents) was included.
Library preparation for long read 16S metagenomic sequencing of the entire 1500 bp 16S rRNA gene, including variable regions (VR) 1-9, was performed using the 16S Barcoding kit 1-24, SQK-16S024 (Oxford Nanopore Technologies, Oxford, England) with a slightly modified protocol: the PCR annealing temperature was lowered from 55 � C to 52 � C and the number of PCR cycles were increased from 25 to 40.The PCR amplification was performed on a Veriti TM Thermal Cycler (Thermo Fischer Scientific, Waltham, MA, USA).A Qubit fluorometer (ThermoFisher Scientific) was used to quantify the barcoded libraries.Depending on the DNA concentration, either 0.5 mL (>10 ng/mL), 1 mL(1-10 ng/mL) or 2 mL (<1 ng/mL) of the barcoded libraries were pooled.From this pool 10 mL was used for sequencing.Sequencing was run for 12 h on a R9.4.1 flow cell in a GridION instrument (Oxford Nanopore Technologies) using Super-accurate basecalling.The sequence data was uploaded to the 1928 platform (1928 Diagnostics, Gothenburg, Sweden) for taxonomic classification.All sequencing reads were trimmed and filtered on amplicon length (1200-1700 bp for V1-V9) and compared against each other to determine strain-level representatives, which were subsequently mapped against the SILVA (v138.1)reference database for taxonomic assignment.Up to 100 000 reads were used for taxonomic classification and the results were presented as species level and relative abundance.
Each PCR reaction (20 mL) contained 2 mL of LightCycler FastStart DNA Master SYBR Green I (Roche Diagnostics), 4 mM MgCl 2 , 0.3 mM of the respective primers and 5 mL of DNA template.A positive PCR control (DNA from Streptococcus pneumoniae) and the negative extraction control was included in each PCR run.Samples with a crossing point (Cp) value that was lower (1.5 cycles or more) than the negative extraction control were considered PCR positive.All samples were tested for inhibition by adding a spike-in control (Streptococcus pneumoniae DNA), 1 mL to 4 mL sample, in a separate reaction.Sequencing for species identification was not performed on these samples as the main reason for performing this analysis was to evaluate the bacterial load.

Clinical data
Results from CSF cultures and cytochemistry (polymorphonuclear-and monomorphonuclear leukocytes, erythrocytes, CSF/plasma-glucose ratio, lactate and albumin), blood cultures and cultures from normally sterile materials such as extracted EVDs or ventriculoperitoneal shunt components were extracted from electronic patient records.Additionally, results from analysis of biomarkers of inflammation and organ dysfunction in blood samples (C-reactive protein (CRP), leukocytes, thrombocytes, aminotransferase (ALAT) and creatinine) collected in connection to CSF sampling, were also extracted.Finally, information regarding gender, age, comorbidities, cause of admission, duration of neuro-ICU care, duration of EVD treatment, number of CSF samples collected, and antimicrobial treatment were also collected from the electronic patient records.Whether a positive culture was regarded as infection or colonization/contamination during clinical management was defined based on whether the prescribed antibiotic treatment indicated a clinical decision to treat an EVD-related infection.Cases where no antibiotic treatment in appropriate CNS dosing directed at the bacteria cultured in CSF was initiated, or such treatment was discontinued after <7 days, were defined as clinically assessed colonization/contamination.Cases where antibiotics directed against the cultured bacteria in appropriate CNS dosing was prescribed for �7 days were defined as clinically assessed infection.

Statistical analysis
Results are presented as medians (interquartile ranges [IQR]), medians (range), absolute number (%) or means as appropriate.The Mann-Whitney U-test was used for all group-wise comparisons.P-values below 0.05 were considered statistically significant.Statistical analysis and graphic visualisation were performed using R version 4.2.2andR studio version 2022.07.2

Ethical considerations
All procedures performed involving human participants in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.The study was approved by the regional ethics review board in Uppsala (2017/177).Subjects were included in the study after informed consent or, when this was not possible due to the condition of the subject, after consultation with next of kin.Written information was given both to study subjects and, when needed, to next of kin.If additional CSF collection at clinical sampling was considered a risk, inclusion and study sampling were not performed.Results from real-time 16S PCR and 16S Nanopore sequencing were not available to the treating physician and, hence, did not affect the clinical management of the study subjects.Personal data was collected only by JW and pseudonymized before data analysis to minimise the compromise of subjects' integrity.

Study population
In total, 96 subjects were included in the study.After screening of electronic patient records 12 patients were excluded from the group allocation.Out of the 84 remaining subjects, 22 were culture positive and 62 were culture negative.Out of these, 18 had no ongoing antibiotic treatment at time of sampling, while 44 had.The group allocation process is illustrated by a flow chart in Figure 1.
The clinical characteristics and levels of CSF and blood biomarkers of inflammation and organ dysfunction in the study groups are summarised in Table 2.

Microbiology
The microbiological findings of the subjects with positive CSF culture are summarised in Table 3.
Results of 16S Nanopore sequencing and real-time 16S PCR (positive/negative) for the groups are displayed in Figure 2. Using real-time 16S PCR, 17 of the 22   culture positive subjects were positive and 61 of the 62 culture negative subjects were negative.The resulting sensitivity and specificity was 77.3% (95% CI 54.6-92.2%)and 98.4% (95% CI 91.5-100%), respectively.Median number of 16S sequence reads were significantly higher in the culture positive group than in the culture negative groups without antibiotics and the culture negative group with antibiotics (p < 0.001).There was no statistically significant difference in number of 16S sequence reads between the negative control group and the culture negative without antibiotics and the culture negative with antibiotics (p ¼ 0.59 and p ¼ 0.78, respectively).

Characteristics of culture positive samples
Species identification by 16S Nanopore sequencing was in agreement with species identification by MALDI-TOF from cultured colonies in 20 out of 22 culture positive subjects.Discrepant findings were one case where Staphylococcus aureus was identified in culture and Nanopore sequencing showed high numbers of Staphylococcus haemolyticus and Staphylococcus epidermidis sequences and one case where Micrococcus luteus was identified in culture and Nanopore sequencing showed low numbers of unclassified and Acinetobacter tandoii (A gram-negative rod normally isolated from sludge) sequences.
Real-time 16S PCR was positive in 17 of 22 samples.The samples that were negative in real-time PCR had significantly lower numbers of 16S reads in Nanopore sequencing compared to real-time 16S PCR positive samples (median 2935 and 99930, p ¼ 0.0016).These samples were also to a larger extent (4/5) only positive in broth cultures compared to samples positive in realtime 16S PCR (5/17).
Based on prescribed antibiotic therapy, culture findings were assessed as infection in 17 of 22 subjects.In these subjects, real-time 16S PCR was positive in 15 of 17 subjects and 16S Nanopore sequencing showed reads approaching 100 000 in 16 of these 17 subjects.In the five subjects where culture findings were clinically assessed as contaminations, real-time 16S PCR was positive in two subjects and 16S Nanopore sequencing showed high numbers (100 000) of sequence reads for one subject, low numbers for two subjects (<500) and intermediate numbers (8864 and 19 043) in two subjects.Details of species identification of culture positive samples are displayed in Table 4.

Characteristics of culture negative samples with high numbers of 16S sequence reads
An analysis of culture negative samples with number of 16S reads exceeding the third quartile of the negative controls (3369 reads) was performed to assess whether there were signs of culture negative infections among subjects with or without antibiotic treatment at the time of sampling.There were 17 culture negative subjects with more than 3369 16S sequence reads, 12 were treated with antibiotics at the time of sampling and 5 were not.Real-time 16S PCR was negative in all cases except one with a Cp 3.68 lower than negative control.This case also had the highest number of 16S sequence reads (30 258) of all culture negative cases.The sequences in this sample were identified to numerous bacterial species where Variovorax paradoxus and Massilia eurypsychrophila sequences showed the highest relative abundance (20.4% and 18.8% respectively).Details are summarised in Table 5.

Characteristics of samples excluded from group allocation
Real-time 16S PCR and Nanopore 16S sequencing were also performed on subjects excluded at group allocation.To further explore the performance of Nanopore 16S PCR in diagnosis of CNS infection, these results were analysed and presented in Supplementary table 1.

Discussion
This study investigates the potential of real-time 16S PCR and 16S Nanopore sequencing for diagnosing EVDRI.The data shows that samples positive in CSF culture were to a large extent also positive in real-time 16S PCR and yielded high numbers of 16S DNA sequence reads in Nanopore sequencing.In cases with positive culture and negative real-time 16S PCR, the number of 16S sequence reads in Nanopore sequencing were generally low, which could potentially indicate a false positive culture as the cases were also more likely to be assessed as contaminations by the treating clinician.Subjects with negative culture were almost exclusively negative in real-time 16S PCR and had significantly lower numbers of 16S DNA sequences in Nanopore sequencing, regardless of whether antibiotic treatment was ongoing or not at time of sampling.Hence, our findings do not support a common occurrence of false-negative CSF cultures in this setting.Also, species identification with Nanopore sequencing was generally concordant with species identification with phenotypic methods from cultured bacteria.
To this date, there have been few studies of the role of real-time 16S PCR in diagnosing EVDRI and other healthcare-associated CNS infections.Despite this fact, the method is commonly used in clinical practice, but the lack of scientific evaluation could make interpretation of its results challenging for the clinician.Earlier studies have used PCR assays not fully comparable to modern 16S real-time PCR assays and showed conflicting results [7-10].To our knowledge, the most recent publication on real-time 16S PCR for diagnosing bacterial meningitis in neurosurgical patients was published in 2021 by Perdigão and co-authors [12].In this study, a self-developed 16S PCR protocol was used to analyse clinical CSF samples from 43 subjects with varying degree of clinical suspicion of EVDRI.The rate of 16S PCR positivity was between 40-60% in culture negative subjects and the authors concluded that real-time 16S PCR could be used to identify non-cultivable microorganisms in bacterial meningitis post neurosurgery.Our results differ from those of Perdigão and co-authors in the fact that we observed only one instance of real-time 16S PCR positivity in the 62 culture negative subjects in our study.This illustrates the fact that depending on the protocol used for DNA-extraction and PCR amplification, the sensitivity and specificity of a PCR assay can vary.It could be argued that the protocol for real-time PCR in our study could be lacking in sensitivity, explaining the However, the fact that most culture positive subjects were also positive in real-time 16S PCR speak against this.Also, the use of Nanopore sequencing in our study adds another dimension to the data that provides insights regarding true or false PCR and culture negativity.Most PCR negative samples displayed low 16S sequence read numbers comparable to negative controls and the reads were often classified to several bacterial species with low pathogenicity.Contrastingly, most PCR positive subjects exhibited maximum 16S sequence read numbers, generally from a single bacterial species concordant with the cultured species.We believe this data support that real-time 16S negativity reflects true negativity in most cases in our study.There were five culture positive cases that were negative in real-time 16S PCR which reduces the sensitivity to 77%.However, the number of 16S sequences in Nanopore sequencing were significantly lower in these samples than in other culture positive cases and the culture findings were clinically assessed as contamination in three out of five cases, suggesting that one or several of these cases might be false positive cultures reflecting contamination during sampling or sample handling.If this is the case, 16S PCR sensitivity for infection would be higher if using a better gold standard for comparison.
A more recent study by Jang and co-workers have used 16S Nanopore sequencing in diagnosis of bacterial meningitis after neurosurgery [13].In this prospective study of 178 neurosurgical patients, bacterial culture, 16S PCR with subsequent Nanopore sequencing on PCR positive samples were performed on 285 CSF samples.In 14.4% of samples, it was determined that bacterial CNS infection was present and in 56.1% of these the presumed causative pathogen was found in 16S Nanopore sequencing but not in culture, indicating that the culture was false negative.The authors concluded that 16S PCR followed by Nanopore sequencing enhanced detection of bacterial meningitis post neurosurgery.The difference in findings and resulting conclusions compared to our study is interesting and may have several explanations.Firstly, the culture method used was not described leaving questions on the reference method.Secondly, the sensitivity of the PCR protocol used by Jang et al. may be more sensitive even though our data from Nanopore sequencing do not implicate this.Thirdly, determining whether infection was present or not when PCR/sequencingresults were available might result in an overestimation of the diagnostic performance of the PCR/sequencing.
In the study by Jang et al. 16S PCR of the complete 16S gene was performed and Nanopore sequencing was performed for 2-3 h on positive samples.Although attractive from a cost point of view, their method seems to require extensive hands-on laboratory work and thus, might not be as rapid in real life clinical settings as in the study setting.In our study, 16S Nanopore sequencing were performed on all samples, directly from sample, which simplifies the workflow and reduces the risk of laboratory contaminations.However, further optimisation of the workflow, definition of cut-offs and interpretation guidelines are needed to allow implementation in a clinical setting.
The principal strengths of this study are the prospective inclusion, the relatively large proportion of culture positive cases and the combined use of real-time 16S PCR and Nanopore sequencing which gives insights on the performance of both methods as well as bacterial culture regarding sensitivity and specificity.There are also several limitations.One being the fact all culturepositive samples were included in the culture positive group without regard to whether they were clinically assessed as contaminations or genuine infection.This is evident for example in the case with Microccocus luteus in culture.This is seldom a relevant finding and were not regarded as such by the treating clinician.Including such subjects in the culture positive group underestimates the sensitivity of real-time 16S PCR and 16S Nanopore sequencing, but to reduce the risk of bias in selection of which subjects to include, we opted for a protocol including all culture positive cases without regard to their interpretation.
Another point of concern is that there is no standard criterion for what a positive or negative 16S Nanopore result is.Nanopore sequencing gives quantitative information regarding the number of DNA sequence reads.This number is influenced by several factors like DNA-extraction, PCR amplification and sequencing time.Also, the number of 16S gene copies varies between bacterial species which can also affect number of 16S sequences in the PCR product [14].However, the number of 16S sequence reads still reflects the relative amount of bacterial DNA in the original sample and this information combined with the information from the species identification could be used by clinical microbiologists to interpret whether a sample is positive or negative.The interpretation will also need to be made in relation to the clinical specimen analysed and the presence of contaminating DNA (the 'Labome').Thus, establishing a cut-off for positivity is challenging and it is likely that some results will need to be interpreted as uncertain and the interpretation of 16S Nanopore results will require specific competence from both clinical microbiologists and treating clinicians.Further studies and method evaluation can clarify how the method should be implemented and combined with other diagnostic methods for maximised clinical utility.

Figure 1 .
Figure 1.Flow chart of patients in the study.EVD ¼ external ventricular drain.

Table 1 .
Inclusion and exclusion criteria.
� Only applied to culture negative samples.

Table 2 .
Epidemiological and laboratory parameters the study groups.

Table 3 .
Microbiology of culture positive subjects.

Table 4 .
Characteristics of culture positive samples.Klebsiella aerogeneswas formerly named Enterobacter aerogenes.�� In the follow up samples 16S Nanopore sequencing identified the sequences as originating from Staphylococcus aureus.��� The subject developed CNS infection with Enterococcus faecalis despite ongoing intravenous and intrathecal vancomycin.

Table 5 .
Details of culture negative samples with high number of 16S reads.Bacteria with a relative abundance of �10% are listed.