Optimisation and validation of a nanofluidic real-time qPCR assay for the detection and serotyping of Klebsiella pneumoniae

Abstract

ii ABSTRACT Introduction: Klebsiella pneumoniae is a leading cause of neonatal sepsis in Sub-Saharan Africa. Robust surveillance systems and methods that allow for fast and reliable serotype detection are needed to guide research priorities and direct interventions aimed at reducing the burden of K. pneumoniae induced disease. Method: A high throughput nanofluidic qPCR reaction set was developed to detect and serotype thirteen lipopolysaccharide (O) loci (OL: O1/O2v1, O1/O2v2, O1/O2v3, O3/O3a, O3b, O4, O5, O8, O12, OL101, OL102, OL103, OL104), five O-antigen serotypes (O1, O2a, O2ac, O2afg, O2aeh), and fifteen capsular loci (KL: KL2, KL3, KL10, KL15, KL22, KL25, KL28, KL30, KL37, KL39, KL62, KL102, KL110, KL149, KL171) of K. pneumoniae. Additional assays to discriminate K. pneumoniae from other strains in the Klebsiella Species Complex (KpSC) and Escherichia coli were also included in the nanofluidic qPCR reaction set. The optimised reaction set was validated using 77 invasive K. pneumoniae isolates from South African infants ≤90 days old, previously serotyped by WGS and the Kaptive v2.3 pipeline, and 83 archived colonisation swab samples (skin, rectal, nasal) enriched in brain heart infusion (BHI) broth. Isolates from the same BHI samples were serotyped using WGS and Kaptive v2.3. Statistical performance of the qPCR method was evaluated against WGS using sensitivity, specificity, and ROC-AUC metrics. Results: All assays within the nanofluidic qPCR reaction set effectively amplified all 37 targets within the prescribed efficiency range (92-107%). The variance from the linear equation of the regression curve was low (r2>0.99) while the analytical sensitivity was high (limit of detection: 10-100 gene equivalents). A blind analysis of the 77 invasive isolates and 83 colonisation swab samples enriched in BHI showed perfect concordance (kappa=1) between qPCR and WGS for the detection of all O-loci, O antigen types, and K-loci included in the nanofluidic qPCR panel. The nanofluidic qPCR method, which enables the detection of multiple K. pneumoniae strains, was applied directly to swab samples (rather than cultured isolates). This de novo testing identified concurrent colonisation with two or more K-loci and O-loci in 38% (16/42) and 40% (33/83) of swab samples enriched in BHI broth, respectively. Conclusion: The nanofluidic qPCR method is a promising method for detecting and serotyping K. pneumoniae across multiple sample types (i.e. bacterial isolates and direct swab samples - nasopharyngeal, skin, and rectal). The established qPCR assay may be useful in probing the epidemiology of K. pneumoniae, including comprehensive serotyping of temporal changes in colonising and invasive isolates should interventions be established against K. pneumoniae invasive disease.