The MALDI spectrum of a tryptic digest of isolated wild-type bacteriophage K at 5 × 108 pfu/mL is shown in Fig. 1, with peaks at m/z that correspond to peptides discovered by LC-MS/MS and database searching highlighted (see Electronic Supplementary Material (ESM) Table S1). Intense peaks at m/z 2111.1, m/z 2399.3, and m/z 2609.4 correspond in mass to the bacteriophage K capsid protein tryptic peptides (K)SFQTGYGITPDTQIDAGALR(R), (K)LSINVNAMYQQQPQFVSIYR(Q), and (K)GFGTATDAYMPIGVHADFVNSILGR(Q), respectively. TOF/TOF analysis of each respective peak is shown in Fig. 2, and database searching of each fragment ion spectrum against a bacteriophage K sequence database resulted in a singular, significant database identification that matched the amino acid sequences of the respective MS peaks. Further evidence that these peaks are tryptic peptides belonging to bacteriophage K is revealed upon comparison of wild-type bacteriophage K with 15N-labeled bacteriophage shown in Fig. 3. The mass difference between the peak at m/z 2111 and m/z 2136 corresponds with the exact number of nitrogen atoms (25) in the peptide (K)SFQTGYGITPDTQIDAGALR(R). TOF/TOF analysis of the peak at m/z 2136 reveals a fragment ion spectrum identical to the spectrum derived from the peak at m/z 2111 after accounting for the number of nitrogen atoms in each fragment. This similar observation can be made for the pairs of peaks at m/z 2399 and m/z 2428 and at m/z 2609 and m/z 2640.
The detection limit for bacteriophage K was determined by a 50% serial dilution series beginning at 5 × 108 pfu/mL down to 7.8 × 106 pfu/mL. Fig. S1 (see ESM) shows the MALDI spectrum of relevant dilutions, revealing that S/N falls below detection levels at approximately 1 × 108 pfu/mL. Thus, any bacteriophage amplification event must produce at least 1 × 108 pfu/mL of phage K in order to generate enough MALDI-MS signal to be deemed a successful infection and the presence of S. aureus established.
A detection experiment was conducted such that four strains of S. aureus (ATCC-12598, ATCC-27694, BAA-1720, and BAA-1750) at 2.5 × 107 cfu/mL were infected with an inoculum of 15N bacteriophage at an initial concentration of 1 × 108 pfu/mL. Tryptic peptide peaks associated with newly propagated bacteriophage should contain predominantly 14N, because the broth used to culture the bacterial host consists of nutrients with nitrogen isotopes at natural abundances. Figure 4 displays the MALDI spectrum of each respective amplification event, showing the appearance of the established three tryptic bacteriophage peaks at m/z 2111, m/z 2399, and m/z 2609. Mass spectra obtained from controls consisting of only bacteriophage K (no bacteria) and bacteria only (no labeled phage K) carried through the infection and digest procedures revealed no amplified phage peaks and no peaks that may interfere with peaks associated with newly amplified bacteriophage (ESM Fig. S2). Thus, the presence of S. aureus in the respective cultures can be established after a 5-h infection process based upon the detection of phage amplification.
Antibiotic susceptibility of the four strains of S. aureus to 4 μg/mL cefoxitin was determined by carrying out a bacteriophage amplification experiment in the presence of the antibiotic with comparison to a no-antibiotic phage amplification infection as demonstrated above. Because bacteriophage can replicate only in living bacteria, no phage amplification—and therefore no phage proteins—is expected in antibiotic susceptible S. aureus strains. Figure 5 shows the side-by-side comparison of each respective S. aureus strain after phage amplification with the no-antibiotic sample and in the presence of 4 μg/mL cefoxitin. For the two methicillin-sensitive S. aureus (MSSA) strains—ATCC-12598 and ATCC 27694—significant amplification can be seen in the no-antibiotic samples as shown in Fig. 4, while no phage-based tryptic peaks can be detected in the sample with cefoxitin; phage replication did not occur to produce the 14N-labeled progeny phage, and the 15N parent phage is below the limit of detection of the MALDI-TOF-MS. Therefore, not only can the presence of S. aureus can be determined in these samples, but they are also susceptible to cefoxitin. The methicillin-resistant S. aureus (MRSA) strains BAA-1720 and BAA-1750 show significant phage amplification in both the no-antibiotic and the cefoxitin-containing samples, indicating that live S. aureus bacteria in the cefoxitin-containing sample were responsible for the phage amplification. Thus, the presence of S. aureus is established in these two samples, and their susceptibility to cefoxitin is confirmed.
While the above experimentation for PAD combined with MALDI was conducted on a research grade instrument with TOF/TOF capabilities in reflector mode, MALDI instrumentation in clinical laboratories often consists of instruments that function in linear ion mode without a reflectron. Because of the expanding use and distribution of these commercial instruments, it would be useful to establish the use of PAD as detected on contemporary instrumentation such as the Bruker Biotyper or the BioMerieux VITEK MS that have been approved by the FDA for routine bacterial identification in clinical laboratories. Figure 6 shows the MALDI-MS spectra obtained on a Bruker Biotyper of the phage amplification tryptic products of the four S. aureus strains used in this study incubated in the no-antibiotic sample and in the presence of 4 μg/mL cefoxitin. While the MSSA strains ATCC-12598 and ATCC-27694 show clear phage amplification as determined by the presence of the peaks at m/z 2111, m/z 2399, and m/z 2609 in the sample with no antibiotic, the spectra of the phage amplification products in the presence of 4 μg/mL cefoxitin show no peaks indicative of a successful phage amplification event which require live bacteria. Thus, these strains would be accurately identified as S. aureus and would be deemed susceptible to cefoxitin. The MRSA strains BAA-1720 and BAA-1750 show phage amplification in both the no-antibiotic and 4 μg/mL preparations, successfully identifying each strain as S. aureus and determining its susceptibility to cefoxitin.
While detection of bacteria using PAD detected by MALDI-MS has been shown previously, to our knowledge, this is the first study showing the ability to ascertain antibiotic susceptibility by PAD as detected by MALDI-MS. Moreover, this is the first study using tryptic peptides of amplified phage proteins to detect phage amplification by MALDI-MS. In our experience, high-mass (>20 kDa), intact proteins derived from bacteriophage cannot always be detected by MALDI-MS, regardless of phage concentration. For example, while the capsid protein for S. aureus bacteriophage 53 is readily detected by MALDI-MS with basic sample preparation techniques , we have been unable to detect the intact major phage capsid protein for phage K by MALDI-TOF-MS regardless of sample preparation techniques, which include a reduction of disulfide bonds and sonication. However, with all bacteriophages in our laboratory, we have been able to detect tryptic peptides of phage proteins following trypsin digestion regardless of the bacteriophage in question.
We believe that demonstration of PAD on MALDI-MS instrumentation to detect bacteria and determine antibiotic susceptibility is a significant and novel step forward in integrating PAD for everyday clinical use. The use of trypsin digestion prior to MALDI-MS analysis means that many hard-to-detect (by MALDI-MS) bacteriophages using whole, intact virus analysis can likely be used in a PAD scheme where MALDI-MS is used as a detector. In our experience, we have been able to detect proteins from all bacteriophages at significant quantities using trypsin digestion prior to MS analysis, whereas MALDI-MS of whole, intact viruses often fails to detect any virus-derived proteins. We anticipate the continued acceptance of commercial MALDI-MS instrumentation for bacterial diagnostics and believe that PAD reagents and methodologies can be integrated into instruments already in place in microbiology and clinical laboratories. While MALDI-MS instrumentation is already capable of identifying bacteria from pure culture, PAD by MALDI-MS adds the component of antibiotic susceptibility determination to the MALDI-MS platform. As we have already demonstrated, PAD can be utilized to detect susceptibility to multiple antibiotics . PAD in clinical samples can be conducted in multiple-well formats, lending itself to the processing of multiple samples in parallel and subsequent spotting on MALDI plates prior to analysis. Because MALDI-MS functions as a universal detector, obviating the need to develop and validate antibody-based detection systems, PAD kits can be added directly to existing MALDI-MS-based detection systems as they are developed and validated.
Several advantages combining PAD with MALDI-MS to identify bacteria and determine antibiotic resistance are apparent. First, while MALDI-MS currently relies upon the isolation of bacterial colonies prior to analysis, PAD can identify bacteria in mixtures , thereby significantly shortening the time to detection (5 h) because extended incubation times to visualize and obtain colonies can be avoided. Second, multiple antibiotics can be added to PAD workflows to obtain antibiotic susceptibility information against a panel of antibiotics. As phage replication is dependent upon a viable bacterial host, any agent that arrests growth or kills bacteria can be tested using PAD. Third, if a phage cocktail is utilized that contains typing phages, MALDI-MS can identify by mass which respective phage is amplified, thereby providing additional epidemiological data.
The single major barrier to implementation of a phage-based assay, in our opinion, is obtaining a bacteriophage or formulating a phage cocktail that has the requisite sensitivity and specificity. Phage detection systems do exist that incorporate a single bacteriophage; for example, the gamma phage lysis test to detect Bacillus anthracis is widely used in cases of anthrax outbreaks . Contrariwise, the KeyPath MRSA/MSSA blood culture test incorporates a phage cocktail to obtain suitable sensitivity and specificity [16, 17]. Phage K, utilized in the present study, has a broad host range for many strains of S. aureus yet has been shown to also infect a small number of Staphylococcus epidermidis strains . Thus, if phage K was to be incorporated into an assay, it would have to be augmented with other S. aureus bacteriophages, and the cross-reactivity to S. epidermidis would have to be mitigated, for example by the addition of the iron chelator deferoxamine to inhibit the growth of S. epidermidis strains .
While this study lays the methodological foundation for a PAD assay using MALDI-TOF-MS to identify S. aureus and determine antibiotic susceptibility, a robust validation must occur before moving to clinical implementation. Many more strains of bacteria including MSSA, MRSA, and common contaminating bacteria such as S. epidermidis must be tested to ensure broad-enough coverage while minimizing false positives. Importantly, S. aureus strains with MICs near the CLSI cefoxitin susceptibility (<4 μg/mL) and resistance (<8 μg/mL) values should be represented. Because of the ease of modifying this method, moving to a cutoff of 8 μg/mL cefoxitin may be warranted as more S. aureus strains are tested to ensure strains of intermediate resistance are properly assigned accurate susceptibility status. For the purposes of a PAD assay, the proper concentration of cefoxitin used must be determined empirically. For example, the FDA-approved KeyPath phage amplification detection methodology used 2 μg/mL cefoxitin as the cutoff point in its assay [16, 17].