Specific and rapid reverse assaying protocol for detection and antimicrobial susceptibility testing of Pseudomonas aeruginosa based on dual molecular recognition

The worldwide emergence and spread of antimicrobial resistance is accelerated by irrational administration and use of empiric antibiotics. A key point to the crisis is a lack of rapid diagnostic protocols for antimicrobial susceptibility testing (AST), which is crucial for a timely and rational antibiotic prescription. Here, a recombinant bacteriophage tail fiber protein (TFP) was functionalized on magnetic particles to specifically capture Pseudomonas aeruginosa, while fluorescein isothiocyanate-labeled-magainin II was utilized as the indicator. For solving the magnetic particles’ blocking effects, a reverse assaying protocol based on TFP recognition was developed to investigate the feasibility of detection and AST of P. aeruginosa. P. aeruginosa can be rapidly, sensitively and specifically detected within 1.5 h with a linear range of 1.0 × 102 to 1.0 × 106 colony forming units (CFU)⋅mL−1 and a detection limit of 3.3 × 10 CFU⋅mL−1. Subsequently, AST results, which were consistent with broth dilution results, can be obtained within 3.5 h. Due to the high specificity of the TFP, AST can actually be conducted without the need for bacterial isolation and identification. Based on the proof-of-principle work, the detection and AST of other pathogens can be extended by expressing the TFPs of their bacteriophages.

Antimicrobial resistance (AMR) is a worldwide health crisis resulting in growing economic burden and increasing mortality 1,2 . The most important strategies to minimize AMR are the prudent antibiotic use and the development of new antibiotics 3,4 . For the correct antibiotic use, the development of rapid and accurate diagnostic protocols for antimicrobial susceptibility testing (AST) facilitates the timing of prescription of effective antibiotics. Therefore, numerous efforts are exerted on developing rapid, sensitive and acute detection and AST for bacterial infections.
Traditional bacterial growth-based protocols are considered as the gold standard methods for bacterial detection and AST. These protocols ideally need to show repeatability, high standardization and good reliability. However, the isolation and identification procedures for bacterial detection usually require a tedious process of 24-48 h 5 . The subsequent broth dilution or disk diffusion tests for AST require bacterial culturing with given concentrations of various antibiotics, which demands almost another 24-48 h 6,7 . Lack of timely results of bacterial AST results in frequent empiric antibiotic therapy, causing irrational antibiotic use and the development of AMR 8 . Therefore, some other bacterial growth-based protocols are reported to aim at shortening the time of AST, such as microscopy detection [9][10][11] , electrochemical sensor [12][13][14] , phase-shift spectroscopy detection 15 , fluorescent detection 16 , microfluidic devices based on the slipchip technique 17 , and surface-enhanced raman scattering 18 . Nevertheless, due to their lack of capacity of identifying the given bacterial species, they also require time-consuming pretreatment procedures for bacterial culturing, isolation and identification. Polymerase chain reaction-based protocols gradually attract attention for bacterial detection and AST due to the advantages of rapidity, sensitivity and because it is a culture-free process [20][21][22] . However, they suffer from the prerequisite of precise resistant gene information and frequent gene mutation.
With the increasing global AMR, bacteriophages have regained interest as antimicrobial agents since they are the natural enemies of the bacteria. Bacteriophages highly specifically (sometimes up to the strain level) recognize their target bacteria even in harsh environments. A typical lifecycle of a virulent bacteriophage involves the following steps: specific adsorption on the bacterial cell wall, deoxyribonucleic acid (DNA) injection into the bacterial cell, the formation of new phage particles and the lysis of bacteria cell for releasing its progeny. Bacteriophage receptor binding proteins such as tail fiber proteins (TFP), tailspike proteins (TSP) and the endolysin are the essential recognition elements which are responsible for adsorption, injection and lysis, respectively 23 . Therefore, receptor binding proteins are the ideal molecular recognition attributes for a high specificity, robustness, good anti-interference capability and universality to each bacterium 23 . Unfortunately, just like the bacteriophage entity as a whole, the endolysin and TSP both have an inherent lytic activity which could be unfavorable for bacterial capture and sample manipulation.
Previously, we showed that the TFP (NCBI accession number YP 007236480.1) of Pseudomonas aeruginosa phage PaP1, recombinantly produced in Escherichia coli, can specifically recognize P. aeruginosa. This recombinant TFP can specifically recognize P. aeruginosa without displaying a lytic activity. To investigate the application of this TFP in AST, TFP-functionalized magnetic particles (MPs) were utilized to specifically capture P. aeruginosa, and fluorescein isothiocyanate (FITC) labeled magainin II was utilized as the fluorescent tracer. A reverse assaying protocol (RAP) combined with magnetic separation was developed to conduct specific, rapid and sensitive detection and AST of P. aeruginosa. In the RAP, quantitative excess FITC labeled magainin II is mixed with a suspension of P. aeruginosa. After P. aeruginosa is captured and magnetically separated by TFPfunctionalized MPs, the supernatant solution of remaining fluorescent tracer is utilized to quantify P. aeruginosa. The change of P. aeruginosa concentrations under various antibiotics can indirectly respond to the susceptibility of P. aeruginosa to the antibiotics.

Experimental setup
Instrumentations. Scanning electron micrographs (SEM) were recorded by an S-3000 N scanning electron microscope (Hitachi, Japan). Fluorescence (FL) signals were obtained using an Infinite M200 PRO microplate reader (TECAN, Switzerland). FL micrographs were recorded by using a NI − U FL microscope (Nikon, Japan).

Results and discussion
The principle of RAP for P. aeruginosa detection. As shown in Fig. 1, we developed a RAP for P. aeruginosa detection. Recombinant TFP was functionalized on MPs to specifically capture P. aeruginosa through a specific interaction between TFP and its receptor on the bacterial cell wall 24,25 . FITC labeled magainin II anchoring on the cytoplasmic membrane of both Gram-positive and Gram-negative bacteria through electrostatic and hydrophobic interactions was utilized as the fluorescent tracer 26,27 . When P. aeruginosa was captured by the TFPfunctionalized MPs to form a bacteria-MPs complex, quantitative excess FITC labeled magainin II was added to this complex. After FITC-magainin II-bacteria-MPs complex was magnetically separated, the supernatant solution was transferred into the microplate to obtain the FL intensity. The changed values of FL intensity (ΔFL) were utilized to quantify P. aeruginosa, in which the ΔFL was defined as: ΔFL = FL blank sample -FL sample . Compared to a direct assaying protocol, a RAP can solve the blocking effects of the MPs to the intensity of FITC since the size of MPs is much bigger than that of FITC. When the FITC-magainin II-bacteria-MPs complex was formed, only the FL tracer of the complex on the side of exciting light can be triggered to emit the FL signals.
To assay the coinstantaneous binding capability of TFP and magainin II to P. aeruginosa, TRITC labeled TFP and FITC labeled magainin II were simultaneously mixed with P. aeruginosa to stain the cells of P. aeruginosa. As shown in Fig. 2, red FL from TRITC and green FL from FITC can be both observed on the surface of P. aeruginosa cells. This phenomenon demonstrates that TFP and magainin II can simultaneously bind with P. aeruginosa at different sites to form the sandwich complex. Specificity. The specificity of RAP was investigated by selecting three Gram-negative bacteria (E. coli, P. solanacearum and S. Typhimurium) and three Gram-positive bacteria (S. aureus, S. epidermidis and S. mutans).
The concentrations of these interference bacteria were all 1.0 × 10 5 CFU⋅mL −1 for the specificity investigation. The specificity of the RAP was calculated by the designed interference degree (ID) values of the above interference bacteria in the following equation.
As illustrated in Fig. 4, the ID values of the tested interference bacteria were all below 5.26%. For the further investigation of potential interference to P. aeruginosa detection, Mixture A was composed of all the six interference bacteria and Mixture B was prepared by mixing P. aeruginosa with Mixture A. The ID value of Mixture A was 4.37%. Compared to that of P. aeruginosa, the ΔFL intensity of Mixture B showed the minor difference (3.14%).
Twenty five strains of P. aeruginosa with different characteristics (Table S1) were tested to investigate the strain specificity of this protocol. As shown in Table S2, the FL signals from the 25 strains of P. aeruginosa showed (1) ID = �FL interference bacteria /�FL P. aeruginosa × 100%  www.nature.com/scientificreports/ minor differences from -0.78% to 0.66% in comparison with those signals from the host strain. Therefore, this RAP for P. aeruginosa detection showed good specificity of species.

Practical sample detection.
To investigate the potential application of this RAP for P. aeruginosa detection, 5% glucose injection, rat plasma and human urine were spiked with P. aeruginosa suspension at given concentrations. As shown in Table 1, the recovery values ranged from 90.1% to 104.2%, with the RSD all below 5.0%. These results demonstrated the reliability of RAP for detecting P. aeruginosa in a complex matrix.
AST of P. aeruginosa. AST of P. aeruginosa was evaluated by detecting the ΔFL signals of 1.0 × 10 5 CFU⋅mL −1 P. aeruginosa cultured with serial concentrations of antibiotics. According to the guidance of CLSI M100S, the four antibiotics of group A including PIP/TAZ, CAZ, TOB and GEN and one antibiotic of group B selected as LVX were utilized to validate the AST of P. aeruginosa to demonstrate its reliability. After P. aeruginosa was cultured with the absence (blank group, BG) and the presence (test group, TGs) of serial concentrations of antibiotics for 2 h at 37 °C, the ΔFL signals of P. aeruginosa were calculated and compared. The same amounts of P. aeruginosa suspension stored at 4 °C were detected as the control group (CPs). Since P. aeruginosa at 4 °C grew extremely slowly, the concentrations of P. aeruginosa was considered as remaining almost unchanged. The results of AST were obtained through comparing the ΔFL signals of TGs with those of CGs and TGs. For the AST of P. aeruginosa to PIP/TAZ, at the concentration ranging from 16/4 to 128/4 μg⋅mL −1 , the ΔFL signals of TGs were about 99.7% and 29.7% of those of CGs and BGs, respectively (Fig. 5A). As shown in Fig. 5B-D, similar results were also found for CAZ (8-32 μg⋅mL −1 ), TOB (4-16 μg⋅mL −1 ) and GEN (4-16 μg⋅mL −1 ). These results demonstrate that under these antibiotics concentrations the growth of P. aeruginosa was significantly inhibited. The minimum inhibitory concentrations (MICs) of PIP/TAZ, CAZ, TOB and  www.nature.com/scientificreports/ GEN were < 16/4, < 8, < 4 and < 4 μg⋅mL −1 , respectively. According to the guidance of CLSI M100S (Table S3), P. aeruginosa was susceptible to these four antibiotics ( Table 2). For the AST of P. aeruginosa to LVX, when the concentrations were 1 and 2 μg⋅mL −1 , the ΔFL signals of TGs were about 312% and 93.2% of those of CGs and BGs, respectively (Fig. 5E). These results demonstrated that the growth of P. aeruginosa was slightly inhibited by LVX in comparison with the normal growth of P. aeruginosa (BGs). However, the concentration of LVX reached 4 μg⋅mL −1 , the ΔFL signals of TGs reduced to about 99.7% and 29.7% of those of CGs and BGs, respectively. It  www.nature.com/scientificreports/ shown that the growth of P. aeruginosa was significantly influenced by LVX at the concentration of 4 μg⋅mL −1 . Therefore, the MIC of LVX was 4 μg⋅mL −1 and P. aeruginosa was resistant to LVX ( Table 2). The AST results for all the testing antibiotics were consistent with the provided data of the document of CLSI 100S. These results demonstrated that the RAP protocol showed good reliability for the AST.
Other bacteriophage-based biosensors for P. aeruginosa detection. As listed in Table 3, several bacteriophage-based biosensors were previously reported for P. aeruginosa detection. None of these biosensors were utilized to perform the AST of P. aeruginosa. The recognition elements of these biosensors included the living entities of the bacteriophages 28,29 and the proteins 30 or peptides 31-34 displaying on the coat proteins of the bacteriophages. As the recognition elements for P. aeruginosa detection, bacteriophage entities offered several advantages such as a relatively cheap and easy production and their stability to pH and temperature variations. However, there still remained a great challenge of immobilization of the bacteriophages on the sensor surface in the right orientation for the optimized capture of P. aeruginosa 35 . Another disadvantage of bacteriophage as a whole was the fact that they were biological active and infectious against the host P. aeruginosa 35 . TFP as the recognition element, shows ideal innate non-lytic activity for bacteria capture and sample manipulation. Moreover, TFP can also be recombinantly produced for the fixed orientation by expressing biotinylated proteins as the recombinant fusion protein on the N-terminal containing the avi-tag of GLNDIFEAQKIEWHE could be biotinylated specifically at the site of lysine residues (K).

Conclusion
In conclusion, a rapid, sensitive and specific RAP using TFP and magainin II as dual recognition elements was developed to perform the detection and AST of P. aeruginosa. Since TFP can specifically recognize the target cells of P. aeruginosa from other interference bacteria, the results of AST can actually be obtained within 4 h without the time-consuming process of bacterial isolation and identification, which can facilitate the decreasing frequency of irrational empiric antibiotic therapy. Based on this proof-of-principle work, the detection and AST of other bacteria can be facilely completed by the expression of the TFP of their bacteriophages. In the future work, we will focus on further reduce the detection time of AST based on TFP recognition through other detection technique such as microfluidic system or single-cell imaging.