MALDI-TOF MS monitoring of PBMC activation status in sepsis

The study was approved by the Ethics Committee of the Assistance Publique-Hôpitaux de Marseille, France. Blood was collected after informed and written consent of healthy donors and septic patients.

Blood was collected in EDTA-containing tubes and PBMCs were isolated using Ficoll cushions (MSL, Eurobio). After centrifugation, PBMCs were washed in sterile phosphate buffered saline (PBS) without Ca²⁺ and Mg²⁺, and 1 × 10⁶ cells were suspended in 10 μL of PBS and frozen at − 80 °C for 2 to 3 days before analysis. In some experiments, PBMCs from healthy donors (1 × 10⁶ cells in 6-well plates) were incubated in 2 mL of RPMI 1640 containing 10% fetal calf serum (FCS) and stimulated with 20 ng/mL IFN-γ (PeproTech), 20 ng/mL IL-4 or IL-10 (R&D Systems) for 18 h [22]. PBMCs were also stimulated with LPS from Escherichia coli (1 μg/mL, Sigma-Aldrich), peptidoglycan (PGN) from Bacillus subtilis (10 μg/mL, Sigma-Aldrich), CpG-ODN (2 μg/mL, InvivoGen) and poly I:C (25 μg/mL, InvivoGen). Finally, PBMCs were stimulated with heat-inactivated bacteria (10 bacteria per cell). Bacteria included oxacillin-sensitive S. aureus, community strain group B streptococcus, Pseudomonas aeruginosa (ATCC 27853) and E. coli (ATCC 25922). PBMCs stimulated for 18 h were pelleted in 10 μl of PBS and frozen as unstimulated PBMCs for MALDI-TOF MS analysis.

After thawing of the PBMCs, 1 μL of cell suspension was added to 1 μL of matrix solution (saturated solution of α-cyano-4-hydroxy-cynnamic acid in a mixture of 50% acetonitrile, 25% trifluoroacetic acid and water) as previously described [17, 20, 23]. The mixture was deposited on the MALDI target. The evaporation that gradually took place at room temperature allowed the formation of α-cyano-4-hydroxy-cynnamic acid crystals containing the dispersed samples. Measurements were performed using an Autoflex II mass spectrometer (Bruker Daltonics, Wissembourg, France) equipped with a 337-nm nitrogen laser. Each sample was irradiated with a laser for desorption and ionization. Each spectrum resulted from the sum of positive ions obtained after 525 laser shots in different regions of the analyzed spot (automatic mode). All the positive-ion mass spectra were acquired in the linear mode at an acceleration voltage of 20 kV in the delayed extraction mode. A signal-to-noise of 3.0 was selected to define peaks, with a maximum of 100 peaks per spectrum. Spectra were automatically acquired with a mass/charge (m/z) ranging from 2000 to 20,000 Da using FlexControl and FlexAnalysis 2.4 software (Bruker Daltonics). The x-axis of spectra represented the m/z ratio (in daltons) of ionized molecules, and the y-axis indicated the intensity (relative abundance) of these ions.

Analyses and graphical outputs were performed using the free and open source statistical analysis software R (version 2.13), along with specific analysis libraries (MALDIquant) as previously described [20]. The gel view representation indicates the reproducibility of the spectra. A hierarchical clustering with a ward algorithm for agglomeration and a dissimilarity matrix based on the Jaccard distance were used to classify the spectra. The MALDI Biotyper 3 software (Bruker Daltonics) was used to create an average reference spectrum for each PBMC sample, corresponding to at least 10 individual spectra. The Biotyper software realigns acquired spectra and automatically creates an average spectrum using default Biotyper software settings provided by the manufacturer, and we created a database as previously described [20]. The Biotyper software also allows the identification of unknown spectra as shown in clinical samples by comparison with reference spectra, for the identification and classification of microorganisms [14]. The score values proposed by the manufacturer have been used for microorganism identification. The score values between 0.000 and 1.699 do not allow reliable microbe identification; the values between 1.700 and 1.999 allow probable cell identification and score values higher than 2.0 are considered statistically significant; they allow the confident identification of different microbe species. We extended this method to assess PBMC activation status in septic patients. As the score values provided by Bruker Daltonics ranged from 0.000 to 2.000 when we used control samples and reference spectra, we considered that medians of matching scores higher than 1.5 allowed reliable identification of the activation profile of patient PBMCs and could be clinically relevant. Differences between conditions were tested with the Mann-Whitney U test and a cutoff value of 0.05 was chosen to consider a difference statistically significant.

As compared to phenotype studies that require panels of antibodies to characterize circulating cells, MALDI-TOF MS enables simple and rapid measurement of cell status. First, we wondered if PBMCs from healthy controls were characterized by a specific whole-cell MALDI-TOF MS signature. The spectra of PBMCs from ten healthy donors were composed of numerous peaks ranging from 2000 to 15,000 Da, with a major peak at 4961 Da. Note that the spectra of the ten donors were highly reproducible (Fig. 1a). Second, we found that the signatures of PBMCs from ten septic patients were similar but they were distinct from those of PBMCs from healthy controls. Because MALDI-TOF MS profiles were specific for PBMCs from healthy controls and septic patients, the dendrogram constructed by Ouedraogo et al. [17] was implemented with reference spectra of PBMCs from two healthy donors and two septic patients. As expected, the PBMCs from the two healthy donors were similar and clustered with T lymphocytes and, to a lesser extent, with monocytes and polymorphonuclear cells (PMNs). Likewise, the PBMCs from the two septic patients were similar and formed a specific cluster. This cluster was close to that of monocytes and PMNs but was away from the T lymphocyte cluster (Fig. 2). This specific pattern underlines the prominent role of the innate immune response in sepsis. Taken together, these results demonstrated that PBMC fingerprints distinguished patients with sepsis from healthy controls.

We wondered if the analysis of individual peaks enables stratification of patients with sepsis, microbiologically documented or not. When the spectra were analyzed by comparing the presence or absence of peaks, we found that 10 peaks were present only in controls, 12 peaks were common to healthy donors and septic patients and 18 peaks were specific in septic patients (Table 1). Taking into account the limitations related to the limited number of patients in this preliminary study, among these 18 specific peaks, 15 were common to all septic patients independently of microbiological documentation. Concerning the three other peaks, the peak at 5415 Da was found in patients with gram-negative bacillus bacteremia (4/4) and one patient without microbiological documentation. The peak at 3329 Da was present in 3/4 patients with gram-negative bacillus bacteremia and the same patient without bacteriological documentation, suggesting that these two peaks could be associated with gram-negative bacillus bacteremia and that this patient without microbiological documentation could have gram-negative bacillus sepsis. Finally, the peak at 2942 Da was found in the two patients with S. aureus bacteremia and in one patient without microbiological documentation, suggesting that this peak could be related to S. aureus bacteremia (Table 1). Despite the limits of the study, these results show that the signature of septic patients is very similar, regardless of documented or undocumented infection.

Because the differences between the signatures of healthy and septic PBMCs may be related to their activation state, we stimulated PBMCs with various agonists to determine if PBMC activation results in specific profiles. The stimulation of PBMCs with IFN-γ, LPS, IL-4 and IL-10 led to specific and reproducible responses, as shown using a virtual gel view representation (Fig. 3). The spectra from all stimulated samples were clearly separated from those of unstimulated PBMCs. Each type of stimulation led to specific fingerprints. The spectra from the PBMCs stimulated with inflammatory molecules (IFN-γ/LPS) clustered together. Similarly, immunoregulatory cytokines induced the clustering of PBMCs. The existence of these three major clusters (unstimulated, IFN-γ/LPS-, IL-4/IL-10-stimulated PBMCs) suggests that MALDI-TOF MS may be used to analyze the inflammatory response of PBMCs in the clinical setting.

We also investigated the responses of PBMCs to PAMPs and bacteria often found in sepsis. For that purpose, we used unstimulated PBMCs as controls and we compared the fingerprints induced by PAMPs and bacteria to these controls. Two very different clusters were clearly identified: one including unstimulated and poly I:C-stimulated PBMCs and the other with PBMCs stimulated with bacteria and other PAMPs (Fig. 4). Among the bacterial signatures, we observed that the signatures induced by P. aeruginosa and E. coli (Gram negative bacteria) were coupled to the signature induced by LPS from E. coli. Similarly, the signatures induced by S. agalactiae group B and S. aureus (gram-positive bacteria) were associated and close to the signature induced by PGN from B. subtilis. These results suggest that MALDI-TOF MS may be useful to assess a series of specific responses of PBMCs to varied agonists.

Finally, we used matching scores to compare the different modes of PBMC activation after having created averaged spectra of stimulated PBMCs. The MALDI-TOF MS profiles of PBMCs from septic patients were then compared with the database. None of the septic patients matched with the averaged spectrum of healthy donors, while the spectra of septic patients significantly matched with IFN-γ and IL-10 spectra regardless of whether the infection was documented (n = 6) or not (n = 6), confirming that sepsis is characterized by both inflammatory and immunoregulatory features (Fig. 5 and Additional file 1: Figure S1). It is noteworthy that the scores obtained by comparing the spectra of E. coli- and S. aureus-infected patients with the reference spectrum of E. coli- and S. aureus-stimulated PBMCs were significantly lower (p < 0.05) than those obtained with IL-10- and IFN-γ-specific spectra, respectively (see Fig. 5a, b), suggesting the prominence of the activation profile as a specific response to pathogens. Importantly, we found that the spectra of PBMCs from septic patients significantly (p < 0.05) matched with CpG-ODN, independently of a documented infection. Clearly, the activation profile found in the patients with unknown infection was similar to that of the patients with documented infection (see Fig. 5c), suggesting that these patients had a bacterial infection. Taken together, these results are consistent with that septic PBMCs were activated in a specific way and suggest a bacterial infection in septic patients without documented infection.