Use of MALDI-TOF mass spectrometry in the battle against bacterial infectious diseases: recent achievements and future perspectives

KEYWORDS:

• MALDI-TOF/MS
• microbial identification
• antibiotic resistance
• machine learning
• biomarkers

1. Introduction

A number of studies have been carried out in order to demonstrate the usefulness of the use of MALDI-TOF as a reliable alternative method for rapid identification of microorganisms and for antibiotic susceptibility testing. MALDI-TOF MS contributes to a reduced mortality and length of stay of patients resulting in a decrease on cost spent per patient, together with the impact on public health.

Mass spectrometry (MS) has led toward a more sophisticated antibiotic use, examples of which are Escherichia coli and Klebsiella pneumonia, through shortening the initiation time of the correct antibiotic action. Furthermore, the cost analysis per sample still remains well below alternative techniques of classical microbiology and of molecular biology.

In a recent review by Sandalakis et al. [1], in 2017 an attempt was made to comment on as many of the different areas of use of MALDI-TOF MS as possible. Since then, a number of studies have been performed on the field.

2. Identification of microorganisms: some recent advances

An outstanding improvement on the identification of mycobacteria has been done since in a recent study MALDI-TOF MS correctly identified as many as 92% of the M. tuberculosis isolates and 68% of M. bovisisolates. Opposite results are still obtained when different systems are used, especially in the case of M. fortuitum, M. kansasii, M. marinum and M. terrae [2]. In fact, the use of VITEK MS with the concomitant use of the IVD v3.0 software was shown as being a reliable method for identifying non-tuberculosis mycobacteria to the species level directly from culture-positive Mycobacteria Growth Indicator Tube broths without the need for subculture. Nevertheless, it is still recommended that the detection of M. tuberculosis should be performed using immunochromatographic assays, since they are easy to perform and the final result is not affected by the presence of other bacteria in the specimen. It seems that the application of MALDI-TOF MS in clinical pathogenic mycobacteria identification still laps behind of what a satisfactory level would be considered and certainly there is a lot of space of improving the efficiency of the method to provide correct identification especially at the species level.

As regards fungi, experiments on dermatophytes (Trichophyton rubrum, T. violaceum, T. tonsurans, T. equinum and T. interdigitale) have shown that an acceptable identification can be achieved down to the species level at least for three of these species (T. rubrum, T. violaceum and T. interdigitale), while all isolates associated with these three species have been identified correctly down to the strain level [3]. The resulting ability to differentiate them at the strain level is of vital importance since it can play a decisive role on the improvement of patient outcome, through for example, the early detection of strains with limited sensitivity to anti-fungals.

Of great concern pose infections from the pathogenic fungus Histoplasma capsulatum. To date, diagnosis of histoplasmosis is based on the isolation of the microorganism from cultures together with its intracellular visualization in tissues. However, cultures are time-consuming and a level 3 facility is usually required together with an additional confirmation test. In a recent study performed on H. capsulatum, a reference database was built allowing for the identification from the microorganism yeast forms and early mycelial cultures [4].

Similar attempts have been carried out on the identification of molds, where, however, improvement of sample extraction protocols and upgrading of the existing database is required in order to include local representative strains.

3. Typing

It generally seems that biomarkers can potentially differentiate biothreat bacteria from less-pathogenic near-neighbor species [5] and their introduction in the use of MS may prove an excellent step onwards. A representative example is that of Salmonella serotyping. It has been shown that the use of ribosomal proteins S8, L15, L17, L21, L25, and S7, Mn-cofactor-containing superoxide dismutase (SodA), peptidyl-prolyl cis-trans isomerase C (PPIase C), and protein Gns, and uncharacterized proteins YibT, YaiA, and YciF as biomarkers can allow the serotyping of Salmonella enterica subspecies enterica [6].

In fact the use of ribosomal subunit proteins seems quite promising as a typing technique for other pathogens as well; recently, quite promising results were retrieved when trying to identify and discriminate fungal strains and in particular Aspergillus species, using this approach [7].

Other research groups have used essential microbial glycolipids as chemical fingerprints; it seems that their subsequent analysis in negative ion mode to obtain glycolipid mass spectra has assisted on the identification of the ESKAPE pathogens [8].

Recently, another approach has been introduced and relies on the idea of processing the samples without the need to depend on the MALDI-TOF MS library. According to this approach, proteins from free access databases, such as UniProt, can be used to get a match with the spectra produced by MS. In such a study carried out, a choice of 10 most informative genes was made and they were used as protein panels with the identification accuracy exceeding 80%. Certainly there is much more to do on this approach, but it still remains an effort to identify bacteria using alternative means [9].

Summing up with the identification of microorganisms, it is generally anticipated that MALDI-TOF MS shows much better performance on the detection of Gram negative bacteria than Gram positive bacteria at the species level, while different pretreatment strategies may influence the accuracy of the technique. Moreover, although the thick cell wall of fungi may influence the accuracy of identification, there have been studies that have shown some promising results both at the species and genus levels; however, this is not always the case with molds.

4. Sample preparation

Various attempts have been made in an attempt to achieve the most suitable initial specimen. The protocols vary in specific details between the different groups of organisms, while with respect to fungi, methods further vary between yeasts and molds or even among different mold genera if they lyse in different fashions.

The introduction of a lysis-centrifugation-wash procedure to prepare a bacterial pellet from positive blood cultures can help toward the elimination of the negative consequences of charcoal presence. Other groups have used a two-stage centrifugation with gravitational acceleration at 600 g and 3000 g, after which a lysis buffer is added and another centrifugation is performed at 3000 g. This procedure seems to work fine in the case of monomicrobe cultures [10]. Formic acid/acetonitrile lysis procedures have let to the optimization of urine specimen processing increasing the analytical sensitivity and have provided accurate and rapid detection and identification of bacterial pathogens directly from this type of sample. In case of positive blood cultures, a simple incubation for only four hours on solid media, has led to better identification scores compared to overnight growth, disregarding the instrument used [11].

Generally, the accuracy and precision exhibited following different pretreatments seems to be largely influenced by the microorganism to be tested.

5. Antibiotic resistance determination

A field of vast importance is the rapid and accurate determination of the antibiotic susceptibility profiles of the infectious agents in patients with bloodstream infections since this poses a critical step toward choosing an effective and targeted antibiotic regimen. Incubation of positive blood culture samples on blood agar plates for six hours results in a one day gain compared to the conventional antimicrobial susceptibility testing method [12]. Semi-quantitative MALDI-TOF MS has been used to analyze growth rate and to build a resistance profile independent of the resistance mechanism. This approach has been successfully applied to Gram-negative bacteria, mycobacteria and to Staphylococcus aureus. The parallel use of MALDI-TOF MS and of a vector machine model as a supporting tool for the generation of a reliable algorithm (ClinProTools) has allowed the rapid identification of Bacillus fragilis and the differentiation of cfiA-positive and cfiA-negative subgroups [13].

Recently a completely different approach has been introduced incorporating the development of a MALDI-TOF MS panel based on nucleotides in an attempt to identify M. tuberculosis resistance. Based on this so called MTBID panel, resistance to eight antibiotics caused by 45 chromosomal mutations was studied. Although this platform cannot allow de novo mutation identification, the introduction of specific probes in the spectrum may be the solution [14].

On the other hand, MS has proven unsuitable for the distinction between cefotaxime resistant and susceptible strains and has not allowed the discrimination of beta-lactamase-producing Capnocytophaga strains in clinical samples [15]. Moreover, it does not seem reliable on differentiating non-outbreak from outbreak clones and has proven of limited usefulness on the management of a VREfm outbreak [16].

6. Future trends of MS

The fields of use of MALDI-TOF MS will continue to expand. However, apart from improving the new instruments and their software, it is also crucial of MS to be combined with other sophisticated techniques; a number of attempts with promising results have been performed during the last 1–2 years.

Machine learning (ML) methods are multivariate analysis algorithms. These models can predict unknown cases by learning the pattern of known multivariate data. Examples of robust ML methods are the decision tree (DT) induction algorithm, the support vector machine (SVM), and the k-nearest neighbor (KNN) [17]. ML methods have been used for the construction of templates of specific types in order to facilitate the generation of predictive models of methicillin-resistant Staphylococcus aureus strain typing; based on the results of this attempt, the algorithm generated proved as a cost-effective and promising tool that may provide preliminary strain typing information on the basis of MALDI-TOF spectra [17]. ML could prove of great importance since some protein expression differences among strains of bacteria may be also present in the non-ribosomal proteins; such issues can only be solved by the generation of more detailed information.

More sophisticated software has also been built that allow the user to manage large numbers of data. An example is the lately introduced wide-ranging bacterium classifier tool, called Ribopeaks; it was created using r-protein data from the Genbank and now consists of ~30,000 bacterial taxonomic records. The principle of its use is that it compares the incoming m/z data from the analysis using its own models in order to taxonomically classify the bacteria [18].

A great percentage of microorganisms remain to be discovered (isolated, identified, typed, characterized, etc.). Culturomics has come into play into this field that appears as a step in the void since no one knows what to expect. The use of MADLI-TOF MS on this challenge of finding undiscovered microorganisms is more than crucial [19].

The adoption of MS in disciplines other than clinical diagnostics, such as in agriculture, food safety and quality testing, or ecology, will reveal many new opportunities for research and study.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.