Rapid bacterial identification from formalin‐fixed paraffin‐embedded neuropathology specimens using 16S rDNA nanopore sequencing

Bacterial infections of the central nervous system (CNS) can be severe, life-threatening diseases. Early detection of causative pathogens is crucial for the rapid administration of a tailored antibiotic regime. However, microbiological cultures are negative in one third of bacterial CNS abscesses [1], and routine neuropathology diagnostics are typically limited to histological staining techniques such as Gram or Warthin-Starry. Moreover, tissue might not have been submitted for microbiological culture at the time of surgery in cases without a preoperative suspicion of CNS infection. Molecular diagnostics using pathogen-specific PCR provides an alternative diagnostic approach and has been successfully applied to formalin-fixed paraffinembedded (FFPE) tissue samples [2] but is hypothesis driven and thus requires a priori suspicion of the causative pathogen. Metagenomic sequencing of the bacterial 16S rRNA gene represents an unbiased alternative and has been successfully applied to study the bacterial metagenome of brain abscesses based on DNA extracted from unfixed specimens [3, 4]. Here, we introduce a rapid and scalable approach for bacterial pathogen detection from FFPE specimens using nanopore sequencing of 16S rDNA amplicons. FFPE samples of 32 bacterial CNS infections (31 brain abscesses and 1 subdural empyema) and 3 control samples (reactive brain tissue) were retrieved from our archive. Clinical data and microbiological culture results were compiled by reviewing patient records. After DNA extraction, multiplex PCR with four primer sets covering variable regions 3–7 of the 16S rRNA gene was performed (see the supporting information). PCR products were sequenced on a nanopore Mk1c device. After basecalling, raw fastq files were uploaded on the EPI2ME platform (Metrichor Ltd., Oxford, UK), and results were loaded into the R environment for further analysis (supporting information). Raw sequencing files are available under the BioProject Accession No. PRJNA899355 (ncbi.nlm.nih.gov/bioproject/899355). The median age of the 8 females and 24 males was 57 years (range 13–84 years). Preoperatively, a diagnosis of bacterial CNS infection has been suspected in 24 patients (75%), whereas a malignant tumour was radiologically considered in six cases (19%); in two patients, the aetiology of the lesion had been unclear (Table S1). On Gram staining, bacteria were identified in 23 samples (72%). All cases had positive bacterial cultures, and 10 samples (31%) showed mixed infections with up to three taxa. Deep 16S rDNA nanopore sequencing yielded 31,706 classifiable reads per sample (median, interquartile range: 18,255–40,901). To control for potential sources of contamination, we additionally sequenced four no-template PCR controls (NTC) and three reactive brain tissue samples (median depth: 27,077 reads). Taxonomic classification of the controls revealed Brachybacterium, Brevundimonas and Bradyrhizobium as the most prevalent genera accounting for >90% of reads (Figure S1). To further evaluate these taxa’s influence on metagenomic classification, we sequenced five serial dilutions (range from 1:10 to 1:10) of a sample with a monoinfection of Nocardia abscessus (sample 1). With subsequent dilutions, metagenomic profiling revealed an increase in contaminating reads Received: 30 September 2022 Revised: 9 November 2022 Accepted: 13 December 2022

Bacterial infections of the central nervous system (CNS) can be severe, life-threatening diseases. Early detection of causative pathogens is crucial for the rapid administration of a tailored antibiotic regime. However, microbiological cultures are negative in one third of bacterial CNS abscesses [1], and routine neuropathology diagnostics are typically limited to histological staining techniques such as Gram or Warthin-Starry. Moreover, tissue might not have been submitted for microbiological culture at the time of surgery in cases without a preoperative suspicion of CNS infection. Molecular diagnostics using pathogen-specific PCR provides an alternative diagnostic approach and has been successfully applied to formalin-fixed paraffinembedded (FFPE) tissue samples [2] but is hypothesis driven and thus requires a priori suspicion of the causative pathogen. Metagenomic sequencing of the bacterial 16S rRNA gene represents an unbiased alternative and has been successfully applied to study the bacterial metagenome of brain abscesses based on DNA extracted from unfixed specimens [3,4]. Here, we introduce a rapid and scalable approach for bacterial pathogen detection from FFPE specimens using nanopore sequencing of 16S rDNA amplicons.
FFPE samples of 32 bacterial CNS infections (31 brain abscesses and 1 subdural empyema) and 3 control samples (reactive brain tissue) were retrieved from our archive. Clinical data and microbiological culture results were compiled by reviewing patient records. After DNA extraction, multiplex PCR with four primer sets covering variable regions 3-7 of the 16S rRNA gene was performed (see the supporting information). PCR products were sequenced on a nanopore Mk1c device. After basecalling, raw fastq files were uploaded on the EPI2ME platform (Metrichor Ltd., Oxford, UK), and results were loaded into the R environment for further analysis (supporting information). Raw sequencing files are available under the BioProject Accession No. PRJNA899355 (ncbi.nlm.nih.gov/bioproject/899355).
The median age of the 8 females and 24 males was 57 years (range 13-84 years). Preoperatively, a diagnosis of bacterial CNS infection has been suspected in 24 patients (75%), whereas a malignant tumour was radiologically considered in six cases (19%); in two patients, the aetiology of the lesion had been unclear (Table S1) and three reactive brain tissue samples (median depth: 27,077 reads).
Taxonomic classification of the controls revealed Brachybacterium, Brevundimonas and Bradyrhizobium as the most prevalent genera accounting for >90% of reads ( Figure S1). To further evaluate these taxa's influence on metagenomic classification, we sequenced five serial dilutions (range from 1:10 to 1:10 5 ) of a sample with a monoinfection of Nocardia abscessus (sample 1). With subsequent dilutions, metagenomic profiling revealed an increase in contaminating reads while the proportion of reads supporting Nocardia concurrently decreased ( Figure S2). To rule out contamination of laboratory materials, we cultured laboratory reagents and consumables (see the supporting information) but did not identify bacterial growth. Since Brachybacterium, Brevundimonas, and Bradyrhizobium were deemed environmental contaminants and are not known to be associated with CNS infectious diseases, all reads mapping to these genera were filtered out, resulting in a median of 14,553 classifiable reads per sample (interquartile range: 6380-33,588; Figure S3). Bacterial taxa accounting for >1% of filtered reads and at least 100 supporting reads were kept in the final metagenomic profile. Employing these criteria, 5 of 32 samples contained one bacterial taxon, whereas the remaining samples comprised at least two (up to 21) taxa (Table S2). Alignment with microbiology results revealed that all cultured bacteria of monoor mixed infections were identified in 26/32 (81%) metagenomic profiles ( Figure 1). In addition, at least one cultured taxon could be detected in four specimens with mixed infections, whereas in two samples (6 and 21), none of the cultured bacteria was identified by metagenomics. Both samples had significantly higher numbers of contaminating reads (both >90%) compared to other samples (Wilcoxon test, p = 0.02), potentially indicating low bacterial load and/or low sample DNA integrity. Of note, in 6/9 cases without evidence of bacteria on Gram staining, all cultured taxa were identified; in 2/9 cases, at least one taxon was detected. Across all samples, metagenomics identified 94 bacterial taxa that were not found by culture including anaerobic bacteria that are difficult to culture, such as Fusobacterium, Finegoldia, and Prevotella (Table S2).
Taken together, nanopore sequencing represents a rapid method to identify clinically relevant bacteria in FFPE specimens. Diagnostic applications include cases where bacterial culture is negative or has not been performed but also the rapid determination of taxonomic compositions that are characteristic of specific anatomical locations indicating the potential focus of CNS infections [3]. Most metagenomic studies have employed Illumina sequencing platforms, with

Key Points
• Nanopore sequencing of 16S rDNA amplicons represents a method to identify bacteria from formalin-fixed and paraffin-embedded neuropathology specimens.
• The protocol enables rapid detection of clinically relevant bacteria with a DNA-to-answer time of <8 h.
• The workflow is cost-efficient and easily applicable to smaller neuropathology labs.
F I G U R E 1 16S rDNA sequencing of formalin-fixed paraffin-embedded tissue samples. Barplots representing bacterial taxa identified by metagenomic 16 rDNA sequencing. The green colour indicates that the bacterial genus has also been detected in the bacterial culture. All genera identified by culture are listed next to the plot with the percentage of supporting reads in brackets.
sample-to-answer turnaround times of 48-72 h and the requirement of high case numbers for batchwise processing. In contrast, nanopore sequencing can detect bacterial pathogens within minutes of starting sequencing and a library preparation time of <8 h following DNA extraction [5]. Given the possibility of flow cell reuse, adjustable sequencing duration and barcoding for parallel sequencing of multiple samples (1-16 samples in our experiments), nanopore sequencing is highly scalable for neuropathology diagnostic purposes. Moreover, the approach applies well to archival FFPE tissues (oldest sample in our cohort: 12 years) and thus enables sequencing of retrospective cohorts to address scientific questions such as the role of bacterial pathogens in neurological disease or brain tumours [6]. By amplifying >50% of the bacterial 16S rRNA gene using short amplicons, our approach increases coverage and resolution for the detection of bacterial taxa compared to the widely used V3-V4 or V4 amplification [3,4,7]. Shortcomings of nanopore sequencing include the higher error rates relative to Illumina sequencing, although these limitations are continuously improving due to refinements of sequencing chemistry [8] as well as basecalling algorithms [9] and can be addressed by increasing sequencing depth. Moreover, reagent and environmental bacterial contamination play an important role in 16S rDNA sequencing analysis [10] and should therefore be continuously monitored throughout each sequencing run. Due to negligible capital costs for the nanopore sequencing device and minor additional requirements (see the supporting information), the workflow is easily applicable to smaller neuropathology labs or lower-infrastructure locations. In conclusion, 16S rDNA nanopore sequencing represents a rapid and valuable method for detecting bacteria in FFPE specimens.

ACKNOWLEDGEMENTS
Open Access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST
The authors declare that they have no competing interests.

ETHICS STATEMENT
The use of biopsy specimens for research upon anonymization was in

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1111/nan.12871.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in NCBI at ncbi.nlm.nih.gov/bioproject/899355, reference number PRJNA899355.