Bloodstream infections (BSIs) are severe diseases caused by the presence of microorganisms, mainly bacteria, in blood and are characterized by high morbidity and mortality^1,2. The BSIs and associated organ dysfunctions (sepsis or septic shock) remain a life-threating disease due, partially, to the inability to rapidly detect and identify the responsible pathogens^3,4.

Staphylococcus spp. are Gram-positive facultative anaerobic bacteria that frequently colonize the human body^5,6. These pathogens are becoming increasingly resistant to antibiotics and are well-established in both community and healthcare environments, being commonly isolated in intensive care units (ICU)^6,7. Staphylococcus aureus is a common cause of a variety of infections, from superficial skin infections to life-threatening diseases, including necrotizing pneumonia⁸, infective endocarditis⁹ and BSIs¹⁰. Coagulase-negative staphylococci (CoNS) have also been described as harmful to humans, causing several infections, particularly in patients with implanted medical devices⁶.

The empirical antibiotic therapy remains the standard of BSIs treatments¹¹ and its correct use within the first hour after the recognition of the BSI is recommended by the Surviving Sepsis Campaign Guidelines¹¹ and was reported as having a great impact on the patient survival rate¹². Nevertheless, the extensive use of broad-spectrum antibiotics and the large number of patients having negative blood culture samples and thus receiving unnecessary antibiotic treatment, are important contributors to the increase of antimicrobial resistance^13–15. Thus, sensitive, rapid, cost-efficient and specific detection of pathogens in blood, followed by antimicrobial testing, is critical to de-escalate empirical antibiotic therapy and decrease the negative impact of BSIs^2,14,16.

Blood cultures remain the reference standard for the detection of bacteria causing sepsis¹⁷. Generally, blood samples are collected and aseptically inoculated in bottles with specific media for aerobic and anaerobic microorganisms. These bottles are then incubated either in manual or in automatic systems that continuously monitor microbial growth¹⁷.

The conventional culture methods for diagnosis of BSIs involve sub-culturing and Gram staining upon blood-culture positivity, followed by phenotypic methodologies for bacterial identification and antibiotic susceptibility testing. These procedures can be accurate and reliable but are laborious and time-consuming¹⁸. In the last decade, other detection techniques have emerged as alternatives to conventional culture methods for the detection of BSIs, directly from positive blood cultures or from whole blood, and have been improved the time needed for pathogen identification. These include the Polymerase Chain Reaction (PCR)^19,20, Peptide Nucleic Acid Fluorescence In Situ Hybridisation (PNA-FISH)^21,22, Matrix-Assisted Laser Desorption Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS)²³ and DNA microarrays²⁴. However, these methods present some drawbacks, namely: PCR-associated amplification problems (such as PCR inhibitors)²⁵, unspecific hybridization, which can be caused by the human DNA interference with primers and probes^25,26, infidelity in DNA replication, interference of non-microbial material^17,25, limited number of available probes¹⁸, the results obtained are complex and difficult to interpret²⁶, and are unable to distinguish between live and dead cells leading to the occurrence of false positives^25,26. Moreover, pathogen detection directly from blood samples remains a challenge due to the innumerable blood components that can interfere in the analysis^25,26 and to the low bacterial load normally present in the blood from patients with BSIs (1 to 100 CFU mL⁻¹)^26,27. Consequently, most of the detection methods for BSIs are dependent on blood cultures to increase the number of pathogens before the diagnostic test can be conducted¹⁷.

A promising approach for bacterial detection is the use of bacteriophages (phages) or phage-derived proteins as specific probing elements in conjugation with measurement techniques or biosensors. Phages are viruses that infect bacteria with high host specificity²⁸. At the end of their life cycle, phages produce enzymes, called endolysins, to degrade the bacterial cell wall for the release of progeny virions. These proteins have been considered valuable tools to detect and control bacterial infections^29–33. Endolysins from phages infecting Gram-positive bacteria present a modular structure composed of at least one enzymatic catalytic domain (ECD) and one cell binding domain (CBD)²⁹. Most staphylococcal phage endolysins have three distinct domains: two ECDs, namely an N-terminal cysteine histidine-dependent amidohydrolase/peptidase (CHAP) domain and a central N-acetylmuramoyl-L-alanine amidase (Ami_2 or Ami_3) domain; and a C-terminal bacterial src-homology 3 (SH3b) domain as CBD^34,35. Generally, the C-terminal CBD is responsible for directing the enzyme to its substrate by recognizing specific ligands on the bacterial peptidoglycan or other cell wall-associated molecules with extreme specificity and affinity^30,36. Therefore, CBDs have been used as probing elements for the rapid detection or concentration of bacteria^32,37,38, mainly foodborne pathogens³¹. The CBDs binding spectra are usually broader than the host ranges of the corresponding phage and can encompass entire bacterial genera³⁹ and their binding affinity to bacterial cells is equal or even higher than antibodies³⁶. Moreover, phage proteins have shown increased stability to temperature and pH variations, less propensity to aggregation and a faster, easier and less expensive production in comparison with antibodies^40,41.

Flow cytometry is a multiparametric high-throughput technique which enables the detection and quantification of cells or particles individually in a flow, allowing a rapid and effective analysis of their physical and chemical properties within a population⁴². The cells are often labelled with fluorescent markers, therefore allowing the discrimination of different populations in a sample⁴². In the past years, this technique has been widely used for pathogen detection purposes, although the assays that have been described present some drawbacks, such as being laborious and expensive⁴³ or lack the ability to detect pathogens in biological samples^44,45.

In this study, we describe a novel flow cytometry assay based on phage-derived proteins for the detection of Staphylococcus in blood. For this, we exploited different domains at the C-terminus of the S. aureus phage endolysin E-LM12⁴⁶ (Amidase, SH3 and Amidase+SH3) in order to select the truncation that presents the higher binding affinity and specificity for Staphylococcus spp. to be used as a specific probe in the detection assay. To the best of our knowledge, this is the first detection method described in the literature, in which a CBD was used as a recognition molecule in a flow cytometry assay, enabling the sensitive and specific detection of Staphylococcus in blood.

BSIs and sepsis are the leading causes of human mortality among hospitalized patients and thus have a great impact on healthcare systems. Staphylococci are frequently isolated bacteria from BSIs and therefore their rapid identification from blood would allow the early diagnosis of a possible septicaemia^7,10.

In this study, a new endolysin-based flow cytometry assay was developed for the early detection of Staphylococcus in blood. Firstly, different domains at C-terminus of S. aureus E-LM12 endolysin were cloned fused to GFP, expressed and functionally analysed. The results revealed that the GFP-SH3 protein was able to bind to S. aureus cells, which is in accordance with previous studies that reported that SH3 is the binding domain of endolysins in general^29,39,51 and of Staphylococcus phage endolysins in particular^39,52. Interestingly, GFP-AMI_SH3 protein, which contains the amidase and SH3 domains, presented a higher binding efficiency to S. aureus cells than the GFP-SH3, suggesting that the amidase domain contributes to a more efficient binding of the SH3. This might be explained by the PGRP domain that was predicted in the same region of the Amidase_2, and that comprises conserved pattern recognition molecules that bind to bacterial peptidoglycan⁵³. However, when the Amidase_2 domain (GFP-AMI) was used alone with S. aureus cells did not present any binding capacity. A previous study also showed that the amidase domain improved the binding affinity of a staphylococcal CBD to target bacterial cells, which resulted in an enhanced lytic activity of the endolysin⁵⁴.

The specificity and sensitivity assays performed with the GFP-AMI_SH3 revealed that this protein was able to bind to all Staphylococcus spp. tested. This constitutes an advantage of this biorecognition molecule once these pathogens have been frequently isolated from ICU infections⁵⁵, being the CoNS and S. aureus responsible for 20.5% and 8.7% of ICU-acquired BSIs in Europe, respectively⁵⁵. Other authors also reported a binding spectrum at the genus level for other CBDs from Staphylococcus endolysins^39,52,56. Some studies suggested that CBDs recognize conserved binding ligands such as the glycine-rich inter-peptide bridge common to most staphylococcal strains^57–59, as it has been reported for the SH3b-like cell wall targeting domain of lysostaphin⁶⁰, which can explain their broad host range. Curiously, the GFP-AMI_SH3 protein demonstrated some binding affinity to two Enterococcus strains and one S. pneumoniae strain. Although the binding affinity of GFP-AMI_SH3 was not high and only few Enterococcus and Streptococcus cells were recognized, it can be hypothesized that the GFP-AMI_SH3 binds to ligands of peptidoglycan in the bacterial cell wall that are conserved among these bacteria genera. In fact, although these bacterial species are members of two distinct phylogenetic orders namely Bacillales for Staphylococcus and Lactobacillales for Streptococcus and Enterococcus, they present peptidoglycan structures relatively similar apart from their cross-bridges that vary widely in composition and length⁵⁹. Other authors also reported a broader binding affinity of endolysins and CBDs, showing recognition to other bacterial species^61,62, such as the enterococcal PlyV12 endolysin which kills enterococcal, streptococcal and staphylococcal strains⁶² and the Staphylococcus SH3b domain of endolysin LysF1 that binds to Staphylococcus and Streptococcus species⁵⁸. We can also hypothesise that the amidase domain plays a role in the binding of SH3b to other bacterial species. Further investigations into the nature of the SH3 receptor molecules and the amidase domain are needed, which may be important for the understanding of the binding mechanisms of other endolysin CBDs as well.

The endolysin E-LM12 demonstrated capability to lyse all the Staphylococcus strains tested, including CoNS. These results are in agreement with other studies that report that Staphylococcus phage endolysins present a broad lytic activity covering all the Staphylococcus genus and commonly their spectrum is similar to the one from their CBD^56,63. Also, the amino acid sequence of E-LM12 is very conserved, being closely related to several staphylococcal endolysins, namely LysK, which was described to be effective against S. aureus and CoNS⁶⁴.

Regarding the LM12 phage, it exhibited a lytic effect against most of the Staphylococcus spp., although in some strains “lysis from without” was observed, which indicates the absence of phage infection⁵⁰. On the other hand, the GFP-AMI_SH3 protein bound to all the staphylococcal strains tested, demonstrating its higher potential as a recognition probe for Staphylococcus spp. when compared with the phage particle.

Regarding the flow cytometry assays, the results indicate that this methodology was able to specifically detect all the Staphylococcus species tested and no false positives or false negatives were obtained. In this study, the developed assay revealed to be capable of detecting 10⁵ CFU mL⁻¹ of S. aureus cells without enrichment, which was expected since the minimal concentration of bacteria detectable by flow cytometry is reported to be in the range of 10³ to 10⁴ CFU mL⁻¹, using viability stains⁶⁵. The detection around these limits affects the precision of the measurements due to the relatively poor analytical sensitivity⁶⁵.

The flow cytometry assay had to be adapted and optimized for the detection of Staphylococcus in blood samples since the number of microorganisms present in circulation during a BSI ranges from 1 to 100 CFU mL^{−1 26,27}. Therefore, a sample preparation method was designed to include a pre-enrichment step, allowing the effective detection of 1–5 CFU of Staphylococcus in 10 mL of blood. The enrichment procedure also assures the detection of viable cells⁶⁶, preventing the occurrence of false positives. Moreover, the developed method avoided the occurrence of positive signals derived from the autofluorescence of blood components^26,43,67 and revealed a good analytical reproducibility with results that were easily and objectively interpreted.

The conventional culture methods for the detection of Staphylococcus in blood are time-consuming and laborious. In fact, the mean time to blood culture positivity of S. aureus is reported to be between 12 to 16 hours^68–70 and after this, the phenotypic tests must be implemented to identify the specific pathogen which takes about 12 to 36 hours¹⁷. The assay described herein took less than 2 hours to perform after a pre-enrichment step, and therefore provides a rapid, highly sensitive and reliable methodology for the detection of pathogens in blood, outstanding the standard method of BSI diagnosis. Moreover, in this case, the extensive time required for bacterial enrichment was due to the initial bacterial load of 1 CFU per 10 mL of blood. Considering that the time to blood culture positivity is inversely proportional to the initially inoculated concentration⁷¹, the turnaround time of our assay can be significantly reduced if a higher concentration of Staphylococcus is present in patient blood samples.

Very few phage-based assays have been described in the literature for the detection of bacteria directly from blood^72–74. Our implemented method presents several advantages over others which require extensive sample pre-treatment^74–76, genetic manipulation of phage genomes⁷³, complex instrumentation^22,75–77, expensive biorecognition molecules^22,43,77,78 or entail nucleic acids extraction to complete the detection^74–76.

In summary, in this study, we have identified the C-terminal fragment of a Staphylococcus endolysin that presents a high binding affinity and specificity, which is composed not only of the obvious SH3b domain but also of the Amidase_2 domain. Moreover, we developed a novel and highly reliable endolysin-based flow cytometry assay that allows the rapid and specific detection and identification of Staphylococcus spp. from blood cultures.

The designed flow cytometry assay presents several advantages over other bacterial detection methods described in the literature, such as the high specificity rendered by the use of specific phage proteins, simple and inexpensive sample preparation, and the possibility to be tailored to detect other pathogens by using target-specific CBDs. Furthermore, the method can be adapted to a multiplex assay, if specific CBDs are fused with different coloured fluorescent proteins, allowing the simultaneous detection of pathogens prevalent in BSIs. The developed method can also be easily adapted to different clinical samples and types of bacterial infections, by adjusting the sample preparation methodology.

This work is a proof-of-concept that this novel phage endolysin-based flow cytometry assay can be employed in blood samples for the specific detection of bacterial cells. We envisage its validation in blood samples from septic patients in order to be implemented in a clinical environment.