Mass spectrometry: a revolution in clinical microbiology?

Principle (Figure 1)

The first description concerning the use of MS in bacteriology was in 1975 [3]. The technology was developed to study the biomarker profiles of some bacterial species. It used the ionization by fast atom bombardment and the association of gas chromatography and MS. The difficulty was to detect the release of the ribosomal and membrane proteins without destroying them in order to analyze the protein profiles and obtain mass pattern spectra. This encouraging study was not followed by any other development until 1996. At this time, the first matrix-assisted laser desorption/ionization time-of-flight MS (MALDI-TOF MS) experiment was successful in identifying bacteria directly from whole colonies based on protein content [4, 5]. However, the most important evolution corresponded to the system of detection by soft ionization techniques such as MALDI and electrospray ionization (ESI) allowing the analysis of biomolecules and large organic molecules, which tend to be fragile when ionized by the old other conventional ionization methods [6]. The continual development of the hardware provided increasing accuracy and resolution of the different proteins, and the MALDI-TOF MS was being used for the identification of bacteria in research settings [4]. Following this period, the new approaches for species identification were developed involving the use of a different matrix. The change of matrix allowed the ionization of mainly ribosomal proteins, which are more conserved than surface proteins [7]. This was considered to be more reliable for routine identification of bacterial species, as culture conditions seemed to have little effect on the results of identification [8]. Therefore, the MS became a revolutionary tool for bacteriology laboratories.

Matrix-assisted laser desorption/ionization involves the principle of the co-crystallization of the sample with a matrix. The couple sample-matrix is irradiated with photons of a laser whose wavelength is in the absorption band of the matrix. This radiation causes the ionization in the gas phase of molecules of the sample and matrix (Figure 1). The ions formed are then accelerated and sent in a vacuum flight tube where they are separated according to their speed. This speed depends itself of a mass/charge ratio peak (m/z). Ions characterized by a high m/z fly more slowly than those with lower m/z. The approach of protein identification is based on the accurate mass measurement of a group of peptides derived from a protein by sequence-specific proteolysis. Proteins of different amino acid sequence produce a series of peptides masses, which can be detected by the detector. The spectrum of identified peptide masses is unique for a protein specific to a bacterial species. The peptide profiles are generated from direct ionization of an intact colony or a bacterial protein extract after manual extraction. The spectrum obtained is then compared with spectra contained in the database according to the algorithm-specific software used. Identification occurs after a protein’s spectral signature is correlated to a database of spectra collected from reference strains. The results are returned with a scoring system, which appears to be conservative enough to avoid false-positive identifications with both systems [9, 10]. Different systems (software/database) have been created for routine identification of bacteria: the Bruker instrument provides its own solution, MALDI BioTyper (software, bioinformatic and database); the Shimadzu instrument uses also its own software (Launchpad) and the SARAMIS database developed by AnagnosTec GmbH and recently acquired by BioMérieux; Andromas (a French start-up) provides a different type of database and software for routine bacteriology, compatible with either Bruker or Shimadzu hardware. These databases currently available for both systems need to be optimized for certain species but are large and contain up to 2000 species (including bacteria, yeast and mycobacteria), with over 3000 spectra. Their performance is in any case higher than phenotypic identification systems [11–13]. We could note that no statistically significant difference was identified between the platforms for clinically relevant bacteria [11, 14].

Figure 1

Principle of MALDI-TOF MS and ESI-MS identification of bacteria.

For MALDI-TOF, laser impact causes thermal desorption of ribosomal proteins of bacteria embedded in matrix material and applied to the target plate (analytes shown as red, light blue, and orange spheres, the matrix is given as green spheres). In an electric field, ions are accelerated according to their mass and electric charge. The drift path allows further separation and leads to measurable differences in time-of-flight of the desorbed particles that are detected on top of the vacuum tube. From the time-of-flight, the exact mass of the polypeptides can be calculated. For ESI, the DNA amplicons are dissolved in a solvent and injected in a conductive capillary, where high voltage is applied, resulting in the emission of aerosols of charged droplets of the sample. The latter are sprayed through compartments with diminishing pressure, resulting in the formation of gas-phase multiple-charged analyte ions, which then are detected by spectrometer.

As we previously noted, MS needs the use of a matrix to facilitate the ionization of proteins. They allow a burst of microorganisms and the release of proteins that migrate performing a true chromatography. Depending on the matrix, we obtain a spectrum of proteins of specific molecular weight ranges. Alpha-4-cyano-4-hydroxycinnamic acid (HCCA) induces the formation of small spherical crystals with more uniform distribution. It is not suitable for some taxons and requires fewer laser shots and allows the production of 80 to 150 peaks per spectrum [15]. The UV absorbing matrices used were found to be highly specific to bacterial Gram type: HCCA for Gram-negative bacteria and 5-chloro-2-mercaptobenzothiazole for Gram-positive bacteria [16]. The latter matrix system enhances the sensitivity of the analysis of bacterial endotoxins (lipid A) by more than 100-fold and provides tolerance to high concentrations of reagents (such as sodium dodecyl sulphate, sodium chloride and calcium chloride) [17]. The 2,5-dihydroxybenzoic acid (DHB) allows the formation of long crystals from the periphery to the center of the deposit. It is suitable for a majority of taxons and requires more laser shots to obtain 100 to 200 peaks per spectrum and many signals whose the m/z is >10 kDa [18]. The manipulation of this matrix is trickier. Finally 3,5-dimethoxy-4-hydroxycinnamic acid (sinapinic acid or SA) is a more recent matrix. It allows the analysis of proteins of higher molecular weight than the HCCA and DHB [19].

Advantages of MALDI-TOF MS (Table 1)

The MALDI-TOF MS is the most promising technology for the present and the future in the microbiology laboratories. This is due to its ability to analyze whole bacterial cells with virtually no sample preparation or no batching and the improvement in the identification time of a positive culture (10–20 s for acquisition of the protein spectra and 15–30 s for the comparison in the databank), starting from a colony (Figure 2). MALDI-TOF MS is also widely used because of its high accuracy, low running cost and low maintenance needs. Indeed this technology requires only the medium to grow the organism and a small quantity of matrix. It has already replaced most of the biochemical tests currently used for bacterial identification in routine laboratories (e.g., catalase, oxydase, identification card or API gallery, latex test, agglutination tests) because it does not require prior knowledge about the organism [68]. This new method has now proven to be reliable and safe for the identification of the clinically relevant bacteria (e.g., Enterobacteriacae, the non-fermenting bacteria, staphylococci or streptococci) [12, 69–72]. Continuous improvement of the database is performed and updates are released every 3–6 months. In this way, the identification of most rarely isolated bacteria such as some potential bioterrorism agents (Brucella sp. [73], Coxiella burnetii, Francisella tularensis and Bacillus anthracis [74]), some Gram-negative bacilli (Pasteurellaceae [75], Acinetobacter baumannii group [76, 77], Yersinia sp. [78], Legionella sp. [79]), anaerobes [80–82], or bacteria difficult to identify after Gram staining (Leptospira sp. [83], Mycobacterium sp. [84, 85]) have been reported.

Figure 2

Typical workflow of new and old methods used in a clinical microbiology laboratory. The time to identification, typing and resistance analysis is noted.

MALDI-RE, matrix-assisted laser desorption ionization resequencing; MALDI-TOF MS, matrix-assisted laser desorption ionization time-of-flight mass spectrometry; MLST, multilocus sequence typing; PCR-ESI MS, electrospray ionization mass spectrometry.

All these developments are promising notably since the publication by Gaillot et al. on cost-effectiveness of MS. The authors reported that phasing out of conventional techniques in favor of MS resulted in the overall saving of $177,090 in 1 year [86].

Inconveniences of MALDI-TOF MS (Table 1)

The speedy identification of bacteria by MS clearly presents a major advantage for clinicians in the management of antibiotic treatment. However, even if the protein mass pattern spectra can be analyzed for identification of bacteria to the genus and species level, many results are rarely available to the subspecies level. This means that this tool is not completely efficient to identify all bacteria (e.g., difficulty in distinguishing between Escherichia coli and Shigella sp.) and does not provide help in epidemiological studies to follow crossed transmission of bacteria. Moreover, the major problem is the incomplete current databases. The composition and quality of these databases is crucial for a correct identification. They still need implementation and expansion [68].

Adjustment of the default MALDI-TOF MS database allowed the identification of all members of the A. baumannii group as well as other Acinetobacter spp. with similar accuracy, as was reported by Espinal et al. [76].

A large number of bacterial cells are required for identification. Usually a whole intact colony is used for analysis, limiting the ability to rapidly identify microorganisms directly from biological fluids where the bacterial count is expected to be relatively low. Research is currently being performed to mitigate some of these requirements.

Other problems could also be noted: 1) the use of MS is challenged by the high costs of the instruments and by the long period of maintenance (up to 2 days) requiring another solution during this period; 2) this technology must be completely adapted to the new constraints of diagnostic laboratories, in particular, the traceability of all the tests; 3) Anderson et al. has demonstrated the effects of some selective solid-medium type on the rate of identification of bacterial isolates by MS [87]. For example, Staphylococcus spp. from colistin-nalidixic acid agar medium exhibit low identification rates whereas the same bacteria from blood medium were perfectly identified. In addition to that, it has also been found that protein extraction enhances identification rates and is recommended for colonies grown on different media. However, this extraction increases the time of the experiment; 4) there is also an important need to obtain isolated bacterial colonies to avoid the growth of microorganisms from potentially contaminated material because the technique’s ability to resolve mixtures is lacking; and 5) the uncultured bacteria detected by this technology are not mandatory pathogens and must be evaluated with the clinical signs.

It is probable that all these minor inconveniences could be corrected in the future developments. However, one of the most important objectives support improved antibiotic prescription. In fact, this is one of the missing features of this technology and bacterial cultures are still required for antimicrobial susceptibility testing. Therefore, accurate measures and identification of resistance factors could represent the future evolution of this technology.

Detection of resistance

Recently some resistance markers to one or more antimicrobial agents have been detected by MALDI-TOF MS. Reports suggest that MS has the ability to differentiate methicillin-susceptible S. aureus from methicillin-resistant S. aureus (MRSA) strains [93, 94], and also detect carbapenem resistance activity based on the detection of degradation of β-lactam antimicrobials [20, 21, 95]. Basically, the MS follows the enzymatic hydrolysis of the antimicrobial agent. The sensitivity and specificity of this approach is high (97% and 98%, respectively) [21] and the results are available in <3 h [95]. However, this approach must be validated in routine laboratories to consider replacing the conventional techniques (such as cefoxitin disks, or PCR to detect the mecA gene). We can speculate that the time frame of MS can be further shortened as suggested by Hooff et al. [20].

Even if the detection of subtle protein alterations will probably be difficult to assess by MALDI-TOF MS, recently, some reports have demonstrated it is possible such as the detection of rpoB mutations in Brucella sp. [22]. There have also been reports of the detection of bacterial enzymes targeting antibiotics, such as β-lactamases or carbapenemases in E. coli and A. baumannii [21, 23–25, 95], the CfiA carbapenemase in Bacteroides fragilis [26]. Other resistance mechanisms recently reported to be detected by this technology including porin defects and expression of efflux pumps [27].

The ability to use MS technology to rapidly detect the resistance mechanisms produced by a bacterial pathogen will be a key element that will revolutionize clinical microbiology.

Detection of virulence factors

Differences in virulence profiles for bacterial isolates can be based on the selective determination of the presence or absence of m/z peaks in the MALDI-TOF MS instrument. This approach allows not only the identification of the bacterial species, but also to show the presence of some key surface-associated molecules or some well-known virulence factors involving a rapid management of the infection. In this way, the MALDI-TOF MS technology could detect a m/z peak specific to Panton-Valentine leukocidin-producing S. aureus strains, a well-known virulence factor in the development of acute severe S. aureus infection [28]. However, a recent publication calls into question the previous work. Szabados et al. claim that protein peaks of 4448 and 5302 Da are not associated with the presence of Panton-Valentine leukocidin [29].

Some efforts have been made to find biomarkers to differentiate between infectious and non-infectious causes of systemic inflammatory response syndrome (SIRS). Discriminatory peaks have been detected suggesting a direct link between infectious-related protease activity and a sepsis-specific diagnostic pattern for discrimination of patients with SIRS [30]. An interesting example is represented by the detection of the staphylococcal delta-toxin which was found associated with the acute infection [31].

Detection and identification of quorum sensing signals, immune-modulatory proteins and the binding of host factors including antibodies are targets for future research in this field.

Detection directly from samples

Molecular methods have revolutionized clinical microbiology. Indeed a growing range of rapid diagnostic tests that can be performed at the point-of-care has been implemented such as real-time PCR assays [32]. These tools considered to be expensive and time-consuming have clearly evolved and represented a suitable evolution for routine identification. For example, different tests (e.g., GeneXpert™ system or BD GeneOhm™ StaphSR) have been developed for rapid management of contagious diseases such as Clostridium difficile, Bordetella pertussis or Neisseria meningitidis (33, 34). These tests also allow MRSA to be detected from different sample types (e.g., blood, skin and soft tissue, nasal swabs) and to prevent unjustified prescriptions (e.g., detection of Enterovirus or Streptococcus agalactiae in pregnant women [35, 36]).

In this way, the use of MALDI-TOF MS for microorganisms’ identification in clinical samples has become essential in the future development of this technology. Some reports have been published recently to detect microorganisms directly from blood or urine samples [37–47]. Methods (combining centrifugation steps, the use of serum separator tubes or ammonium chloride lysis) for the processing of positive blood culture samples have been proposed to increase the sensibility of the technique but especially in the case of direct testing of other materials (e.g., urine specimens), consensus still has not been reached [37–47]. The analytical sensitivity in blood culture varied between 66% and 76% with a major precision in the identification of Gram-negative bacteria (around 90%) compared to Gram-positive bacteria (<50%). It is of note that current MALDI-TOF MS data software analysis is not able to reliably identify all microorganisms present in mixed cultures. Direct identification of pathogens in urine samples has also been evaluated. The results are not yet satisfactory, the most promising result was obtained for urine containing more than 100,000 CFU/mL [44] and other developments still seem necessary. Different protocols have been used (e.g., concentration step, membrane filtration and magnetic separation) to improve the sensitivity of MS [43, 48]. The use of automation of urine analysis (e.g., urines flow cytometry) in the laboratory in order to eliminate negative samples might render downstream use of MALDI-TOF MS more efficient.

All these evolutions are attractive for the future. Vlek et al. have demonstrated that the direct performance of MALDI-TOF MS on positive blood culture broths reduced the time until species identification by 28.8 h and was associated with an increased proportion of patients receiving an adequate antibiotic treatment within 24 h [49].

PCR/ESI-MS (Table 1)

Other MS solutions have been recently developed to increase the interest of MS technology. The PLEX-ID system (Abbott™) is a nearly fully automated system that associates broad-spectrum PCR (targeting ribosomal and housekeeping protein genes) with electrospray ionization-MS (ESI-MS) (Figure 1) [10]. It delivers broad microbial screening to semiquantitatively identify all organisms present in a sample. It is capable of running multiple human identification targets such as mitochondrial DNA, short-tandem repeat (STR) or single-nucleotide polymorphism (SNP). The principle is to measure the m/z of amplicons, generated by multiplex PCRs that target several loci within bacterial or fungal genomes. The method targets both conserved and species-specific genetic regions to identify microbes based on amplicon base compositions relative to a known database of microorganisms. To date this tool has demonstrated its ability to directly detect and identify bacteria [50] and associated antibiotic resistance genes, such as drug-resistance M. tuberculosis, carbapenemase-producing A. baumannii or K. pneumoniae and quinolone resistance in A. baumannii from isolates [51–55] (Figure 2). Indeed specific primers may be added to the assay to screen the mecA gene for methicillin resistance, the vanA and vanB genes for vancomycin resistance in enterococci, and the bla_KPC gene for resistance to carbapenems [56–58]. Most recently, the method produced highly accurate results when used to identify bacterial and yeast pathogens directly from the clinical specimen (blood culture) [59], in particular, to detect Erlichia isolates [60].

It is important to highlight the technology also offers extended utility for epidemiological surveillance and infection control [51, 52, 54]. It provides quick results (<6 h) and can identify mixtures of up to three to four microorganisms but requires the batching of six samples at a time [10].

In contrast, the first and most limiting step of this technique is DNA extraction from clinical samples. This induces an increase of the costs (2 to 3 times compared to MALDI-TOF) due to the cost of consumables, software package and the need to use DNA extraction reagents (buffers, enzymes, and primers) for PCR.

Although this technology needs to be validated more extensively, there are some recent publications on the detection of Aspergillus terreus from bronchio-alveolar lavage [61] and a panel of respiratory viruses from nasopharyngeal aspiration that represents an important target for the future in clinical microbiology [62].

iPLEX MassARRAY^® system (Table 1)

The MassARRAY iPLEX single-nucleotide polymorphism (SNP) typing platform uses and the MS technology coupled with single-base extension PCR to analyze amplicons of PCR for rapid and accurate molecular identification of microorganisms [63]. This system is commercialized by Sequenom™ (San Diego, CA, USA).

The assay consists of an initial locus-specific PCR reaction, followed by single base extension using mass-modified dideoxynucleotide terminators of an oligonucleotide primer which anneals immediately upstream of the polymorphic site of interest. Using MALDI-TOF MS, the different mass of the extended primer identifies the SNP allele. The starting point of the protocol is the amplification of a target region of interest. T7- and SP6- promoter tagged primers are used to amplify the template. After treatment, in vitro transcription provides RNA transcripts which are base-specifically cleaved. The resulting RNA cleavage products are analyzed by MALDI-TOF MS.

The spectra are compared with the simulated spectra of the reference sequences as published for MLST. Due to the distinct mass of each nucleotide base, the results are as good as those of conventional dideoxy sequencing [64].

This technology has two main applications in clinical microbiology (Figure 2): the comparative sequence analysis, and the SNP genotyping. These two approaches provide a powerful tool in phylogenetic investigation, epidemiology (molecular typing) and surveillance of crossed transmission of bacteria [63, 65, 84]. An interesting example using the MassARRAY technology is the study performed by Syrmis et al. [63] related with the genotyping of MRSA.

MassARRAY iPLEX is more efficient than a sequencing method; however, the analysis by MALDI-TOF MS is much faster than the analysis by capillary electrophoresis, requiring a few seconds for the former one and up to several minutes for the latter one [85]. The iPLEX assay is suitable for high-throughput analysis, as either 96 or 384 samples can be analyzed on the same chip. The major drawback of this technology lies in the requirement for specific equipment and the cost of this equipment.

Other technologies

Other systems for future applications could be developed. One solution includes the surface enhanced laser desorption/ionization time-of-flight (SELDI-TOF) MS system (Table 1). This technology is a specific MALDI-TOF application that combines a chip-based chromatographic enrichment of proteins with TOF-MS [66]. The combination of SELDI (to generate protein profiles and identify significant peaks from large sample sets) and MALDI (to obtain sequence identity of significant peaks) can be extremely powerful for the rapid identification and validation of biomarkers. The SELDI technology incorporates sample prefractionation and binding to the active surface of a ‘ProteinChip’ array providing more information about the protein of interest than just size, with inferences based on the surface chemistry of the ‘chip’ (e.g., hydrophobic, reverse-phase, cation-exchange). This design feature of SELDI markedly decreases the complexity of protein-rich fluids such as serum and permits quantitative comparisons of peak intensities between samples using large sample sets [67, 96]. SELDI platforms are specifically designed for the rapid high-throughput comparative analysis of multiple biological samples, increasing the chance of finding proteins with consistently altered expression during disease development, progression or following treatment [96]. The technology allows assessment of the performance of individual biomarkers and to evaluate combinations of biomarkers with potential diagnostic. The two main limitations of SELDI are its relative imprecision in its assignment of molecular mass to any given peak [67] and its high cost. Very few studies are available concerning bacterial detection increasing the detection limit to the subspecies [66, 97].

A second solution is the liquid chromatography coupled to electrospray ionization triple quadrupole (LC-ESI-QqQ) MS, LC coupled to ESI-Q-TOF (LC-ESI-Q-TOF MS) or MALDI triple quadrupole coupled to MALDI-TOF (Table 1). The LC-ESI-QqQ in selected or multiple reactions way has been used for routine detection of small molecules including metabolites and drugs [31]. More recently, the LC-ESI-QqQ has been suggested as a replacement for classical ELISAs for the quantitation of proteins in complex matrices [98]. The MALDI triple quadruple measures enzyme-mediated, time-dependent hydrolysis of the β-lactam ring structure of penicillin G and ampicillin and inhibition of hydrolysis by clavulanic acid for clavulanic acid susceptible β-lactamases. This assay represents the basis for future investigations of β-lactamase activity in various bacterial strains [20]. These MS technologies have already been used in research settings extensively and it will be just a brief time before this technology can be introduced in the routine clinical microbiology laboratories. Noteworthy, the most popular routine diagnostic procedure to date was developed two decades ago to facilitate the analysis of solid next to the customary volatile compounds [99]. The technology must be used on liquid and could detect proteins especially peptides. It is completely adapted to quantification of proteins, but, until now, no development has been made in clinical microbiology.