3.1. Sample Preparation and Instrumental Settings
Blood smeared over the ITO slide forms a single cell layer that is firmly attached to the surface even without a fixation agent (
Figure 4,
Figure 5 and
Figure 6). Zones of different cell dispersions/densities are clearly visible (
Figure 6). Among those, the zone containing the lowest surface density of blood cells that had not been clumped together was chosen for MSI. The 9AA and CHCA matrices deposited by sublimation lasting 5 min formed transparent layers over the smear (
Figure 4 and
Figure 5), but at least in the D0 case,
m/
z signals recorded using a negative ionization were relatively weak, which may be attributed to the low LASER energies and thin matrix layer (
Figure 3). The noisy appearance of spectra recorded using negative ionization/9AA matrix (
Figure 3a,b) may be associated with easily ionized free fatty acids (FFA) and phospholipids: both blood plasma and different blood cells contain significant quantities of these compounds [
4,
14,
19]. Even in the best-case scenarios provided by short exposure to EtOH and MeOH, there was a significant trade-off between the sample integrity and plasma removal (
Figure 1): careful visual inspection of
Figure 4 confirms the existence of plasma remnants in regions surrounding blood cells. All blood cells, while circulating in vascular system, are immersed in the plasma: plasma removal using more effective agents not only could cause profuse cell destruction but also could compromise the assessment of blood cell chemical composition. However, stronger heme and ATP
m/
z signals recorded in negative ionization/9AA mode using selected D1 in comparison to D0 settings suggest that oversampling may explain the noisy appearance of the spectra: smaller LASER spot sizes may leave blood cells intact, leading to domination of easily ionized plasma lipids covering the cell surfaces (
Figure 3a vs.
Figure 3b and
Figure 5a,b vs.
Figure 5e,f,
Supplementary Figures S3–S5) [
18]. The mass spectra given in
Figure 3a were recorded using the same slide as in
Figure 3b. A lower LASER energy in the D0 case is the most probable cause of different appearances in these figures. A D0 vs. D1 comparison using positive ionization/CHCA matrix, heme and IMP as cellular markers shows the same trend: weak intracellular marker signals are associated with nonoverlapping low energy LASER spots, i.e., D0 settings, while application of oversampling, i.e., D1 settings, produces stronger
m/
z signals probably due to the increased cell wall permeability and/or more desorbed material (
Figure 3c vs.
Figure 3d and
Figure 5c,d vs.
Figure 5g,h,
Supplementary Figures S2, S4, and S5).
Evaluation of the correspondence between the light microscopy image and LASER footprint preceded the MSI experiments. As shown in
Figure 4, total LASER footprints recorded using the selected D0 or D1 settings correspond well with the preselected ROI. Graphical representation of the following MSI experiments is given in
Figure 5. As expected, the strongest signal intensities coming from the cellular markers are located over or in close vicinity to the cells while the non-populated spaces correspond to the weaker signals. Lateral distribution of the cellular markers is consistent with the cell type specificity: while heme-associated
m/
z signals cover primarily RBC, the ATP-associated
m/
z signals are shared by WBC and RBC [
1]. IMP signals are too weak to be analyzed in detail, but even these may be detected over some RBC, as expected [
3]. Individual cells are not expected to have the same content of metabolites. Therefore, the different
m/
z signal intensities associated with individual cells is an expected outcome. However, some signals are “smeared” over the spaces surrounding cells: in the D1 case, this can be attributed to LASER spot overlap and to the leakage of cellular components as a result of the EtOH treatment or cell death. LASER spot overlap raises the possibility of mixing the mass spectra coming from pixels of interest with the spectra of its surroundings (
Figure 4). Besides, in some instances, the LASER spot center does not perfectly match the center of some particular cell. On the other hand, in the D1 case, most cells are covered by chosen
m/
z signals, but in the D0 case, chosen
m/
z signals cover a smaller fraction of existing cells (
Figure 5). Partial coverage of blood cells with
m/
z signals coming from the cellular markers in the D0 case probably does not provide the true cellular markers’ content: it is rather an artefact of weak desorption or ionization.
Application of the
t-test to MSI sets produced differentially expressed
m/
z for each cell type, each matrix, and each instrumental setting evaluated (
Table 1): RBC and D1 settings produced fewer differentially expressed
m/
z in comparison to WBC and D0 settings, respectively, while the CHCA matrix enabled detection of more differentially expressed
m/
z then the 9AA matrix. The first trend is supposed to be associated with the fact that RBCs have less complex metabolic networks and more resistant membranes. This result may also mean that RBCs show greater cell-to-cell variability: 5 RBCs vs. only 1 WBC per blood donor were included in differential
m/
z expression analysis. The second trend is supposed to be associated with mixing of the neighboring pixel’s spectra: in the case of D0 experiments, which did not cause LASER spot overlap, neighboring LASER spot spectra mixing is less probable (
Figure 4). However, due to the stronger signals, only the application of D1 settings in positive ionization mode produced significant differences in the heme-associated
m/
z signal between the RBC and plasma. The third trend reflects the greater fraction of neutral and basic metabolites that are prone to positive ionization and the somewhat stronger signals coming from CHCA-treated samples.
3.2. Human Metabolome Database Search
Differential expression analysis of the
m/
z signals coming from the RBC, WBC, or blood plasma was performed using
t-test. The cells were compared to plasma, and significantly different
m/
z signals entered the HMDB and KEGG search using ±0.025 Da tolerance. Single hits coming from this search that correspond to endogenous or essential compounds were integrated over all evaluated chemical and instrumental settings, while multiple search hits, including isobaric compounds, were eliminated. In comparison to the total count of differentially expressed
m/
z signals given in
Table 1, the total count of differentially expressed metabolites given in
Table 2 is substantially lower due to the excluded multiple database search hits and due to the
m/
z values that did not correspond to any of existing database entries.
The results presented in
Table 2 corroborate the association of differentially expressed metabolites with appropriate blood cell types. Among the identified metabolites, the ones in which cellular content decreased in comparison to plasma predominated. RBC and WBC contents of lipids like acetyl-CoA are seemingly decreased, but this primarily reflects the fact that blood plasma represents the major medium for their transport. The same stands for some hormone metabolites like 3,5-diiodo-
l-tyrosine and small peptides like
N-acetylaspartylglutamic acid. The energetic demands of cells, especially WBC, are reflected in the decreased content of NADPH, NADHX, etc. The number of metabolites for which content is increased in cells in comparison to plasma is not too many, but these metabolites may provide more convincing evidence of blood smear MSI performance. Contents of dATP were increased in both cell types in comparison to plasma. In WBC, dATP is a substrate and inhibitor of DNA synthesis, and in RBC, it may serve as a secondary source of energy in case of ATP depletion [
20]. This metabolite inhibits DNA synthesis in WBC by inhibition of ribonucleotide reductase, but it also inhibits
S-adenosylhomocysteine (SAH) hydrolase in both cell types. This in turn leads to depletion of deoxynucleotides like dCDP in WBC and to accumulation of demethylated metabolites like
l-homocysteic acid accompanied by depletion of methylated products like 5-methyldihydrofolic acid and (6R)-5,10-methenyltetrahydrofolate detected in RBC and WBC, respectively. Among the metabolites which content is increased in both cell types,
l-glutamic acid 5-phosphate stands out: it is required for the removal of ammonia from the cells, and as a precursor for the glutathione synthesis, it is essential for the proper response to oxidative stress. Therefore, it is expected that
l-glutamic acid 5-phosphate’s content is increased in both cell types in comparison to plasma. On the other hand, the accumulation of heme and thiamine diphosphate is specific to RBC: heme plays a crucial role in RBC-mediated oxygen transport, while thiamine pyrophosphate plays a crucial role in the RBC-specific pentose phosphate pathway. Dihydroneopterin phosphate accumulates in WBC as expected due to its role in WBC-mediated inflammatory response [
21]. Selenoproteins and molybdate listed in
Table 2 are endogenous compounds physiologically present in the blood cells, and their roll in cell metabolism is described in the literature [
22,
23,
24]. The association of the rest of the metabolites (
Table 2) with different cell types is omitted since they are either redundant or there is a lack of evidence of their cell-specific role and expression.