Extracellular vesicles isolated from serum by size exclusion chromatography were characterized by their size and the presence of specific biomarkers. The SEC fraction #8 was enriched in vesicles, which size was estimated in a range between 50 and 150 nm by the DLS measurement (with the maximum at 100-120 nm) (Figure 1A). The size of the isolated vesicles was confirmed by transmission electron microscopy (TEM) (Figure 1B). Furthermore, the presence of exosome biomarkers, tetraspanins CD63 and CD81, was confirmed in the same fraction by Western blot analysis (the same proteins remained undetected in the whole serum) (Figure 1C). Considering their specific size and the presence of exosome-specific biomarkers, vesicles present in the analyzed fraction were called exosomes for simplicity, yet other subpopulations of the small EVs could be also present in this material.
The GC-MS-based approach was used for profiling of metabolites in the whole serum and the corresponding serum-derived exosomes of HNC patients, in either pre-treatment (A) and post-treatment (B) samples, or samples of matched healthy controls (C). In general, the untargeted approach allowed the identification of 182 metabolites in serum samples and 46 metabolites in exosome samples, of which 33 metabolites overlapped; the complete list of 195 identified compounds is presented in Supplementary Table S1. Figure 2 illustrates the distribution of different classes of small metabolites identified by GC-MS in serum and serum-derived exosome samples. Among the most numerous classes of metabolites common for serum and exosomes were fatty acids, sugar alcohols, and carboxylic acids (22%, 15%, and 12% of all identified compounds, respectively). Noteworthy, amino acids that were the most numerous group of metabolites in serum samples were markedly less frequent in exosome samples (21% vs. 7% of all identified compounds, respectively, which corresponded to 40 and 3 compounds). All identified metabolites were used to perform unsupervised clustering of samples. The metabolite composition of the whole serum enabled the relatively good separation of all three groups of samples using either the principal component analysis (Figure 3A) and the hierarchical cluster analysis (Figure 4A). Noteworthy, control samples C were more similar to cancer pre-treatment samples A than to cancer post-treatment samples B, which indicated the additional putative treatment-related differential component. In contrast, neither the PCA nor the HCA type of analysis allowed the separation of corresponding groups when samples of serum-derived exosomes were analyzed (Figure 3B and Figure 4B).
In the next step, specific metabolites were detected whose abundances were significantly different between groups. First, we looked for compounds that differentiated cancer patients (pre-treatment samples A) from healthy individuals (control samples C). There were 27 compounds whose serum levels were markedly different (large effect size) between control and cancer samples. These included 12 metabolites upregulated (4 amino acids, 4 fatty acids, 2 purines, 1 glycerolipid, and lactose) and 15 metabolites downregulated (3 carboxylic acids, 3 purines, 3 sugars, 2 fatty acids, serotonin, acetyl-hexosamine, isoleucine, and phosphate) in cancer samples, which were listed in Table 1. Furthermore, there were 18 cancer-upregulated and 38 cancer-downregulated compounds where differences showed a medium effect size (Supplementary Table S1). On the other hand, there were only a few compounds whose abundance was significantly different in serum-derived exosomes from healthy controls and cancer patients. 1-Hexadecanol was markedly upregulated while citric acid, 4-hydroxybenzoic acid, and propylene glycol were markedly downregulated (large effect size) in exosomes from cancer patients (Table 1). Moreover, there were 7 metabolites where differences showed medium effect size, including myo-inositol, linoleic acid, succinic acid, and glyceric acid downregulated in cancer samples (Supplementary Table S1). Furthermore, metabolites whose levels were different between control and cancer samples (either large effect size or medium effect size) were annotated with their corresponding metabolic pathways. We observed that overrepresented pathways associated with metabolites discriminating cancer patients and healthy controls in both whole serum and serum-derived exosome samples included ones involved in energy production (citric acid cycle, Warburg effect, pyruvate metabolism, mitochondrial electron transport chain) and inositol metabolism. Pathways associated specifically with serum metabolites included metabolism of amino acids, sugars, and lipids. On the other hand, pathways associated with metabolites specific for serum-derived exosomes included oxidation of fatty acids and ketone body metabolism (Figure 5A).
In the second step, we looked for metabolites whose abundance was different in serum and serum-derived exosomes of cancer patients between pre-treatment samples A and post-treatment samples B, which allowed detection of changes related to RT. There were 12 compounds whose serum levels were markedly different (large effect size) between pre-treatment and post-treatment cancer samples. These included 4 metabolites upregulated (including hypotaurine and serotonin) and 8 metabolites downregulated in post-RT serum samples, which are listed in Table 2. Furthermore, there were 29 RT-upregulated and 12 RT-downregulated compounds where differences showed a medium effect size (Supplementary Table S1). In marked contrast, only two metabolites detected in serum-derived exosomes (glycerol and cholesterol) showed reduced levels (medium effect size) in post-RT samples. Finally, metabolites whose abundance was different in pre-RT and post-RT samples (either large effect size or medium effect size) were annotated with their corresponding metabolic pathways. Over-represented pathways associated with metabolites whose serum levels were affected by RT included those involved in the metabolism of different classes of compounds (amino acids, sugars, nucleotides, lipids, and biogenic amines), which indicated multifaceted effects of radiation on serum metabolome profile (Figure 5B).