Detection of microorganisms in blood specimens using matrix-assisted laser desorption ionization time-of-flight mass spectrometry: a review

Authors

  • M. Drancourt

    1. Unité de Recherche sur les Maladies Infectieuses Emergentes (URMITE) UMR CNRS 6236, IRD 198, IFR48, Université de la Méditerranée, Marseille, France
    Search for more papers by this author

Corresponding author: M. Drancourt, URMITE CNRS UMR 6236, IRD 198, IFR48, Faculté de Médecine, 27, Boulevard Jean Moulin, 13385 Marseille Cedex 05, France
E-mail:michel.drancourt@univmed.fr

Abstract

Clin Microbiol Infect 2010; 16: 1620–1625

Abstract

Matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) initiated a revolution in the identification of organisms grown on solid medium, including bacteria and fungi. Rapid identification of organisms responsible for septicaemia, which are typically grown in broth, is now expanding the field of application. Despite the fact that there are fewer than ten reports in the literature, published data indicate that MALDI-TOF MS yields accurate identification of blood-borne organisms in ≥80% of cases for inocula of >107 organisms/mL. A major current limitation is failure to accurately identify Streptococcus pneumoniae among viridans steptococci. Identification is achieved in <2 h, sharply reducing the turn-around time for communication of identification to the clinician. Further progress in handling protocols and automation, and extraction of antibiotic resistance data from the MALDI-TOF MS spectra, will further push this emerging approach as the standard one in the laboratory diagnosis of septicaemia, paving the way to application in further clinical situations and clinical specimens.

Introduction

The rapid detection and identification of bacteria in clinical blood specimens has been advocated in order to shorten the turn-around time for appropriate management of patients suffering from septicaemia [1,2]. Although the automated blood culture systems have indeed reduced the delay in detecting the presence of blood-borne bacteria and fungi, identification of such organisms still relies on microscopic examination of the organism morphology after Gram staining and further identification after subculturing of the organism on a solid support. Direct detection and identification of such organisms with the use of real-time PCR and a commercialized system yielded encouraging, but limited, results [3–5]. However, Gram staining lacks both specificity, because of false-positive results resulting from technical limitation and limited expertise of the reader, and because a true-positive result may not provide the clinician with enough information for accurate adaptation of the antibiotic treatment [6]. Furthermore, relying on subculture for the identification delays such identification by 12–72 h. Direct inoculation of the positive blood culture broth into identification automat has been proposed, but this procedure was limited to Gram-negative rods [7]. Direct hybridization of the positive blood culture broth with probes in the fluorescence in situ hybridization technique is limited to the identification of one or a few determined bacterial species [8–10]. Various PCR-based techniques, including real-time PCR and pyrosequencing [11], have been developed to speed the identification of blood-borne bacteria, and a commercially available real-time PCR has been introduced for the direct detection of bacteria in blood [4], but no such solution has actually been implemented as a routine technique in clinical microbiology laboratories.

Matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) has emerged as a potent tool for the rapid identification of bacteria and yeast grown on solid medium, comparing favourably with Gram staining [12]. In this situation, organisms form colonies that are easily detached from the medium surface, and MS analysis can be performed without any further preparation in most cases of bacterial identification [13]. For fungi, rapid treatment by ethanol washing and protein extraction with formic acid and acetonitrile (AN) or pretreatment with trifluoroacetic acid (TFA) has been proposed in order to standardize MALDI-TOF MS identification [14,15]. Organisms grown in blood culture broth have to be separated from the liquid phase and from the red cells to prevent interference during the MALDI-TOF MS analysis.

Protocols

There are a few reports indicating the possibility of direct identification of bacteria in blood culture specimens [16–21]. In these studies, the detection of organisms was performed with automated blood culture systems, and the identification of detected organisms was performed by MALDI-TOF MS analysis. In this situation, the organisms grew in a liquid medium of complex composition, owing to the presence of both medium proteins and the red blood cells. Therefore, in contrast to what happens with most cultured organisms, which are deposited onto the MS plate without any further preparation [22], organisms grown on liquid medium do have to be prepared before the MALDI-TOF MS analysis, using various protocols [16–18]. In particular, the haemoglobin must be discarded, as it yields strong peaks that hamper the interpretation of specific peaks (Fig. 1). Basically, these protocols comprise four steps: (i) specimen collection; (ii) removal of red cells; (iii) protein precipitation; and (iv) protein extraction and solubilization (Fig. 2).

Figure 1.

 Identification of bacteria in positive blood culture broth: process overview.

Figure 2.

 Confusing haemoglobin peaks (*) in matrix-assisted laser desorption ionization time-of-flight mass (MALDI-TOF) mass spectrometry analysis of blood-borne organisms.

Positive blood culture broth is collected immediately after the positive vial is taken out of the automat [16,17] or after the positive vial has been held for 3–10 h at room temperature [18], and 0.2–5 mL of blood culture broth is collected into a standard sterile tube [16,17] or into a serum separator tube with a clot activator [18].

Red blood cells are further removed by combining cell lysis, by using distilled water [13], ammonium chloride [17,18] or saponin [19], and low-speed centrifugation and sterile water washing. Washing also allows the elimination of blood and broth proteins, including blood albumin. SDS has been used in the particular case of Candida spp. [21]. Proteins are further precipitated by washing with 70% ethanol [17,18], and are further solubilized and extracted by incubation in the presence of TFA combined with AN [16], or formic acid combined with AN [16,18] or TFA [21]. Comparison of the latter protocols indicated that the combination of formic acid with AN was superior to the combination of TFA with AN [16].

The following steps are not specific, and comprise MALDI-TOF MS plate deposition, matrix deposition and drying, and MS analysis, with the exception that absolute ethanol can be poured onto spots before the matrix is added (Fig. 2). Bacterial proteins are spotted onto a MALDI-TOF MS MTP (MALDI target plate) plate, overlaid with 2 μL of matrix solution (saturated solution of α-cyano-4-hydroxycinnamic acid in 50% AN and 2.5% TFA), and crystallized by air-drying at room temperature before protein analysis by MS.

Performance

Three studies have investigated the performance of MALDI-TOF MS for the identification of bacteria in positive blood culture broths [16–18] (Table 1). It must be noted that the performance obtained with bottles artificially seeded with microorganisms may not reflect the actual performance obtained in routine detection, probably because of the variations in the inoculum, as discussed below (Table 1). Likewise, there is no evidence suggesting that the broth had an impact on the performance of MALDI-TOF MS identification; four of five studies used the Bactec 9240 (Becton Dickinson, Le Pont-de-Claix, France) apparatus and either standard aerobic/anaerobic broths [18] or the Bactec Plus broths [17,20], whereas one study used the Bact/Alert (BioMérieux, Marcy-l’Étoile, France) apparatus [19]. With regard to identification of bacteria detected in blood culture bottles, a 5-month study analysed 584 positive blood cultures, including 562 specimens containing a single bacterial species and 22 specimens containing more than one bacterial species [16]. This study found a clear difference in the identification of Gram-negative and Gram-positive organisms, with 94% of Gram-negative organisms being identified and only 67% of Gram-positive organisms being identified. In particular, this study found that viridans streptococci were not identified at all. The study also found that the use of AN for the preparation of the specimen was deleterious [16]. A further 9-week study indicated that an identification could be obtained in 78.7% of 122 positive blood cultures, including identification at the species level in 98.95% of cases and identification at the genus level in the remaining case. Identification failure was observed mainly for Streptococcus and Staphylococcus. A clear limitation was the failure to identify 8/10 Streptococcus pneumoniae isolates; the two other S. pneumoniae isolates were identified at the species level, albeit with a low score [17]. Also, the encapsulated organisms Klebsiella pneumoniae and Haemophilus influenzae were among the few Gram-negative organisms exhibiting a low (<1.7) identification score [17], a limitation that had not been observed in the previous study (26/28 well-identified K. pneumoniae isolates and 1/3 well-identified H. influenzae isolates) [16]. A third study of a total of 212 positive cultures was examined, representing 32 genera and 60 species or groups. The identification of bacterial isolates by MALDI-TOF MS was compared with biochemical testing, and discrepancies were resolved by gene sequencing. No identification (spectral score of <1.7) was obtained for 42 (19.8%) of the isolates, most commonly because of insufficient numbers of bacteria in the blood culture broth. Among the bacteria with a spectral score of ≥1.7, 162 (95.3%) of 170 isolates were correctly identified. All eight isolates of Streptococcus mitis were misidentified as S. pneumoniae isolates. This method provides a rapid, accurate, definitive identification of bacteria within 1 h of detection in positive blood cultures, with the caveat that the identification of S. pneumoniae would have to be confirmed by an alternative test [18]. In bottles spiked with bacteria, Staphylococcus aureus was distinguished from non-aureus staphylococci in 100% of cases [19]. In bottles spiked with S. pneumoniae/S. mitis organisms, agglutination correctly identified S. pneumoniae organisms [19].

Table 1.   Performance in identifying bacteria in positive blood culture broth
Nature of specimen (n = number of specimens)OrganismPercentage of interpretable spectra (%)Percentage of correct identification: genus level (%)Percentage of correction identification: species level (%)Major identification failuresReferences
  1. aIncluding spiked bottles.

Positive blood culture broth (n = 599)Bacteria947676Streptococcus spp.16
Positive blood culture broth (n = 126)Bacteria977957Streptococcus pneumoniae17
Positive blood culture broth (n = 179)
Spiked bottles (n = 33)
Bacteria100a80a80aS. pneumoniae
Propionibacterium acnes
18
Spiked bottles (n = 312)Bacteria989889S. pneumoniae 
Positive blood culture broth (n = 388)Candida spp.969891S. pneumoniae19
Positive blood culture broth (n = 304)Bacteria94.78787Uncommon species20
Spiked bottles (n = 48)Candida spp.100100100 
Positive blood culture broth (n = 1)Candida albicans10010010021

A delay of 20 min has been measured once the blood culture bottle has been detected positive [19]. Another study found an overall turn-around time of 75–140 min, with a hands-on time of 10–70 min [20].

Unresolved Issues

The identification of viridans Streptococcus spp. organisms in blood culture bottles remains problematic in most studies [16,17], in agreement with previous observations made with the use of agar medium-grown colonies [13]. Pooling of data from currently published reports indicates that 16/20 (80%) S. pneumoniae strains were not identified solely by MALDI-TOF MS. The lack of an appropriate database may partially explain this fact, which also illustrates difficulties in the taxonomy of this particular group of organisms. Therefore, complementation of the MALDI-TOF MS identification with an agglutination test is mandatory in the case of viridans Streptococcus sp. identification.

Identifying mixed organisms in the same blood culture specimen remains problematic. In one study, 15 mixed bacteria bottles were tested, and yielded one or two bacteria; the authors observed that use of a specific database improved the identification score [19]. Most studies report that only one of the two organisms is identified by MALDI-TOF MS [17,20]; moreover, mixed flora may result in a lack of identification or false species identification [16].

The optimal protocol for blood culture broth processing remains to be determined and evaluated; every one of the six major available studies listed its own protocol. Companies will probably develop and sell their own solutions for red cell lysis and protein extraction, thus standardizing the blood specimen processing and paving the way towards an automated process. Also, the optimization of laboratory processes, combined with the use of last-generation mass spectrometers and software, would further reduce the turn-around time for identification of blood-borne organisms. Mass spectrometers are, indeed, amenable to integration into an semi-automated or fully automated laboratory.

Future Directions

The capacity of MALDI-TOF MS to identify organisms in blood clearly depends on the inoculum [19]. Spiking experiments with Staphylococcus aureus and Escherichia coli indicated that organisms present at 107–108 CFU/mL were correctly identified, whereas 106 CFU/mL yielded signals that were indistinguishable from the background of negative controls, indicating that it was necessary to deposit at least 104–105 CFU on the plate in order to obtain an identifying spectrum [20]. In that study, it was determined that a positive blood culture broth contained a median of 108 organisms/mL [20]. Increasing the sensitivity of MALDI-TOF MS detection may open the way to the fast, direct detection and identification of blood-borne organisms and introduction of the procedure for point-of-care diagnosis of septicaemia after bedside collection of blood. The nature of the MALDI-TOF MS plate was not always precisely defined in reports; nevertheless, testing additional MALDI-TOF MS plates may increase the sensitivity of identification. Also, further analysis and interpretation of MALDI-TOF MS spectra derived from blood-borne organisms may provide information about antibiotic resistance [23] or phenotypic traits such as the Panton–Valentine toxin in Staphylococcus aureus [24], which directly influence the choice of optimal antibiotic treatment.

The expertise gained in studying blood-borne bacteria may prompt further studies on the detection and identification of bacteria in other normally sterile fluids. One study [19] addressed this topic by inoculating 46 fluids, including graft conservation liquid, joint fluid, deep abscess specimens and bone puncture specimens, in blood culture broth, and found that 44/46 (96%) were identified at the species level, the remaining two being identified at the genus level [19]. Such encouraging results should prompt further studies on the rapid MALDI-TOF MS identification of microorganisms in normally sterile fluids such as urine, with or without an enrichment step.

Finally, the true impact on the management of patients of decreasing the delay in communication of positive blood culture by the use of MALDI-TOF MS technology remains to be evaluated, as data obtained with previous methods are somewhat contradictory.

Transparency Declaration

This work was financed by URMITE UMR 6236, Marseille, France. The author declares no conflicting interests.

Ancillary