2. Current Insights
Reprogramming of cellular energy metabolism is one of the hallmarks of cancer [
35]. Changes in cancer metabolism should be reflected in the metabolite composition of cancer tissue or blood samples of patients, which has been shown for several cancer entities [
36,
37,
38,
39]. However, metabolic profiling of body fluids such as plasma is challenging since the metabolite content is influenced by factors such as diet, microbiome and exposome [
40], and, in particular, by the stability of the compounds in an aqueous solution. Identifying (onco) metabolites in the more sheltered matrix of EV, which originate at least in part directly from the tumor cells, therefore seems enticing. Additionally, since EV from breast cancer patients have been shown to be critical for conferring invasiveness [
8], it can be assumed that the most important compounds for these effects are associated with EV. Nevertheless, surprisingly few studies [
20,
21,
22,
41] have focused on the metabolome of EV in cancer patients, most of them characterizing either EV as a whole or exclusively sEV. In contrast, our analysis was performed on the larger MV, which confer the crucial advantage of rapid and simple harvest from peripheral blood [
25].
Since we were interested in a high-throughput tool, the first question we asked was whether the AbsoluteIDQ
® p180 kit (Biocrates) was suitable for this matrix. Our study is the first to demonstrate the application of the AbsoluteIDQ
® p180 kit (Biocrates, Innsbruck, Austria) for analyzing MV. Validation by parallel analysis via an untargeted metabolomics approach on a QToF mass spectrometer showed that most of the significant metabolites were also detected with the second method. Although the kit detected a broad variety of compounds, we found mostly lipids, such as (lyso)PCs and SMs, as the predominant metabolites. This is not surprising, since these compounds represent key players in energy housekeeping, membrane structure and signal transduction [
42,
43] and are to be expected in high abundance in the large membrane-enclosed MV. In contrast, the absence of other metabolites that are highly abundant in blood plasma, e.g., amino acids and carnitines, underlines the quality of our preparation protocol for EV. However, some metabolites of low abundance, but nevertheless of functional importance, may have been missed using the standardized approach, suggesting that the application of specifically optimized mass spectrometry methods would be necessary.
Through metabolic profiling of blood plasma several authors have previously identified cancer-specific metabolites that differentiate between healthy subjects and lung [
44,
45,
46] or breast cancer patients [
11,
12]. Now, we demonstrate for the first time that the metabolome of blood-derived MV from breast cancer patients differs significantly from that of healthy controls. We describe eight lipids that are present in higher concentrations in cancer patients. The profile in our study is partially superimposable with the metabolic profile identified by Qiu et al. [
12] in plasma samples of breast cancer patients (6/8 metabolites). Likewise, 11/24 metabolites in the cohort luminal B vs. controls were identical with the Qiu data. This is hardly surprising, given that plasma contains EV in addition to soluble molecules. Depending on the sample preparation before mass spectrometry, EV can also be lysed and included in subsequent analyses. The fact that we observed higher lipid concentrations of all metabolites than Qiu et al. is in line with this assumption.
Some of the identified compounds were not only suitable discriminators between cancer patients and controls, but also markers of poor clinical outcome. Higher concentrations of the phosphatidylcholine (PC) “PC aa C38:5” as well as the lysoPC “lysoPC a C26:0” were associated with significantly shorter OS, thus underlining the prognostic relevance of this finding. In accordance, Guo and coworkers had previously described that deviations of lysoPCs were associated with cancer progression [
44]. Interestingly, this association differs between cancer entities. Prostate cancer patients showing high amounts of lysoPCs before the onset of disease seem to have a lower risk for advanced disease at the time of diagnosis [
47].
Comparison of the MV metabolome from breast cancer patients with different molecular subtypes revealed distinct features predominantly for the two large groups of luminal A and B versus each other as well as luminal B versus controls. Despite some variations regarding single metabolites, we could not detect significantly different profiles for the Her2-enriched and the basal-like cohort, although their diverging biological characteristics would strongly suggest such differences. However, both subcohorts comprised only few patients (basal-like: 9 patients; Her2-enriched: 4 patients); thus, the statistical power to detect significant differences was not reached. The reason for the small numbers was that only a minority of the patients in these clinically aggressive subgroups could be included in our study, since most of them were under cancer-specific (chemo)therapy that had been established as an exclusion criterion to avoid contamination with apoptotic bodies.
There are only limited data on the breast cancer metabolome in the literature and most of these are derived from tissue analysis. In accordance with our results, Ide and coworkers [
14] described an elevated content of saturated and unsaturated PC in breast cancer tissue, in particular PC aa C32:1, PC aa C34:1 and PC aa C36:0, which were significant metabolites in the luminal B subtype profile we observed. Similarly, PC aa C32:1 was found by other authors in the tissues of breast cancer and oral squamous carcinoma [
13,
27]. Only two recent studies aimed to identify metabolome signatures specific for the various molecular subtypes by analyzing blood plasma [
11,
48]. Similar to our data, Dìaz-Béltran et al. found several candidates that were able to discriminate between the subtypes; however, there is no overlap between our profile for luminal B and the one proposed by that study. Fan et al. were able to distinguish between molecular subtypes using eight metabolites. However, none of these metabolites overlapped with our data. A distinct classification of molecular subtypes may explain these findings, in addition to the fact that our source was MV rather than plasma. On a second note, we identified several lipids not previously described by metabolic studies investigating tissue samples. This is to be expected, since MV extracted from the blood represent a mixture of MV from platelets, red blood, endothelial and immune cells, with only a minority of tumor MV. Cancer cells and their MV influence the microenvironment by reprogramming the innate and adaptive immune cells [
49,
50]. Thus, our approach offers the advantage to detect specific metabolites not only in the tumor MV but also to characterize metabolic changes in the whole blood MV “reactome”, allowing a broader, holistic view on cancer metabolism.
Lipids are critically involved in cancer-related signaling cascades. Accordingly, pathway analysis of our detected metabolites yielded two significantly altered metabolic pathways. Distinct changes in the glycerophospholipid metabolism were found in breast cancer MV samples in comparison to healthy controls as well as in the luminal B subtype MV versus controls. Within this pathway, particularly the conversion of the PC with the KEGG identifier C00157 (e.g., PC aa C32:1) into lysoPCs with the KEGG identifiers C04230 and C04233 (e.g., lysoPC a C16:0) and back were involved. These reactions are catalyzed by the enzymes lecithin–cholesterol acyltransferase (LCAT) and lysophosphatidylcholine acyltransferase (LPCAT), respectively. An accumulation of PC (16:0/16:1), a potential isomer of PC aa C32:1, and the dysregulation of the responsible enzyme has also been observed by others in colorectal cancer tissues [
31]. The second significantly involved pathway, the ether lipid metabolism, did not discriminate between the whole cohort of cancer patients and the controls but predominantly between the luminal B subtype and the controls. The proposed relevant metabolites were the KEGG identifiers C05212 and C04598, which comprise the compounds PC ae C36:0 and PC ae C34:0 found in our analyses. These isomers are substrates for phospholipases and the enzyme LPCAT2. Changes in ether lipid metabolism have recently been described on the genomic and metabolic level in vitro as well as in vivo for breast cancer patients [
32,
33,
34]. There are first pre-clinical efforts to target ether lipid metabolism in cancer cells [
51].
Taken together, it's demonstrated that metabolomic profiling of blood-derived MV is feasible using a standardized high-throughput tool for mass spectrometry (AbsoluteIDQ
® p180 kit; Biocrates, Innsbruck, Austria). Analysis of MV and other vesicles, in contrast to whole plasma, is of particular interest for the understanding of breast cancer progression, since we have shown earlier that vesicles, and not soluble plasma compounds, mediate the invasiveness of breast cancer cells [
8]. The identified lipid biomarkers did not only discriminate between breast cancer patients and controls but were also prognostic markers of clinical outcome. Additionally, a distinct metabolic profile was described for the luminal B subtype, which was very different from the controls and the luminal A cohort. This is in line with the biological behavior of these subtypes, the luminal A group representing the most favorable variant with the highest similarities to the original ductal tissue [
30]. Assumedly, the aggressive Her2-positive and triple-negative subtypes display distinct metabolic features but our subcohorts were too small for statistically reliable conclusions.
In summary, metabolic profiling of MV yields promising biomarkers for diagnosis and prognosis of breast cancer. Further studies are necessary to determine whether they are indeed of functional relevance and to clarify their potential mode of action.