Fatty acid and proteomic signatures of circulating CD81 posi- tive small extracellular vesicles define the disease stage in melanoma patients

The early detection of cutaneous melanoma, a potentially lethal cancer with rising incidence, is key to increase survival and therapeutic adjustment. Especially in stages II-IV biomarkers are urgently needed for adjuvant therapy purposes after resection and for treatment of metastatic patients. We here investigated if fatty acid (FA) and protein composition of small extracellular vesicles (sEV) deriving from plasma of 0-I, II, and III-IV stage melanoma patients (n=38) could reflect disease stage and thus function as biomarker. The subpopulation of sEV expressing CD81 (CD81sEV) was isolated by an ad hoc immune affinity technique from microvesicle-depleted plasma. Biological macromolecules were investigated by gas chromatography and mass spectrometry in CD81sEV. A higher content of FA and a decrease in Saturation Index (C18:0/C18:1), already detectable in early stages, distinguished patients’ from healthy donor CD81sEV. Proteomics (identifier PXD024434) detected an exclusive and significant increase of CD14, PON1, PON3 and APOA5 in stage II and a significant decrease of Rap1b in stage III-IV CD81sEV. The FA and proteomic stage dependent CD81sEV signature strengthens the potential of circulating sEV studies in providing discriminatory information for early diagnosis, prediction of metastatic behavior and follow up of melanoma patients.


Introduction
Extracellular vesicles (EV) characteristics change depending on the status of the releasing cells. Cancer impacts the host's body at different levels and circulating EV may reflect the actual disease status [1]. Thus, EV monitoring represents a promising non-invasive strategy to obtain biomarkers of diagnosis, progression, response or resistance to therapy [2,3]. EV comprise different subfamilies including exosomes, lipid bilayer-surrounded endosomal-derived vesicles of 30-150 nm. The identification of biomarkers of early detection and prediction of recurrence risk has demonstrated the feasibility of subtyping breast cancer by proteomic analysis of serum EV [4]. Additionally, the integrin landscape of tumor EV allows anticipating the metastatic behavior [5]. Melanoma EV foster progression, immune suppression, epithelial-to-mesenchymal transition, promote the establishment of premetastatic niche and counter anti-cancer therapies [6]. At difference to healthy donors (HD), plasma of melanoma patients contains higher amounts of EV exposing melanoma inhibitory activity (MIA) and calcium binding protein B (S100B), together with CD63 and Caveolin-1 [7,8]. Cutaneous melanoma is an aggressive cancer that tends to early metastasis and with a worldwide increasing incidence. In advanced stages, despite the advent of targeted and immunotherapies, the prognosis remains generally poor due to tumor immune escape and intrinsic cell resistance [9][10][11]. Surgery is the treatment of choice in early stages, while advanced stages require systemic therapy [12]. Tumor thickness, ulceration, invasion of blood vessels or lymphatics and immune response are the predominant tools to determine prognosis [13]. Tumor, nodes, metastasis (TNM) classification and stage grouping criteria facilitate accurate risk stratification to guide patient treatment [14].. Early detection of melanoma and its metastases play a key role in increasing survival and therapeutic adjustment. The lack of specific biomarkers encourages the identification of new candidates, especially in stages II-IV, for adjuvant therapy purposes after resection, as well as for therapies in metastatic patients. Several molecules were proposed as prognostic markers: S100B, MIA, LDH, and BRAF mutation [15,16]. Cell-free nucleic acids and circulating tumor cells, detectable in 29-72% metastatic patients, are prognostic for overall survival [17]. Additionally, PD-L1 immune checkpoint, exposed by tumor and non-tumor EV, has been tested as response biomarker of immunotherapy with PD-1 checkpoint inhibitors [18]. In contrast, studies quantifying lipid species in cancer patients' body fluid EV are only starting to emerge, as shown for prostate cancer [19]. Exosomes are enriched in sphingolipids, cholesterol, phosphatidylserine and display a higher saturation level of the fatty acyl groups in phospholipids, with respect to their originating cells. Their composition is similar to lipid rafts, and exosomes have a higher lipid order and detergent stability than other EV types [20]. Exosomal lipid profiles may also reflect their phenotype and functions [21,22]. Here we combined fatty acids (FA) and proteomic studies to investigate potential differences of plasma small EV (sEV) profiles of different stage (0-IV) melanoma patients and compared them to HD. In particular, we focused on immunoaffinity-captured CD81 + sEV (CD81sEV) to investigate potential discriminatory information for early stage melanoma patients that contributes to early diagnosis and predicts metastatic properties. We also aim at understanding if proteome and FA signatures of circulating melanoma patients' CD81sEV can reflect the alterations induced by the tumor at liquid biopsy level.

Human subjects
Blood was obtained from fasting melanoma patients (n=38), stages 0-IV (Table S1), and from age and gender matched healthy donors (HD=17) routinely donating blood at Centro Trasfusionale Universitario of Azienda Policlinico Umberto I, Sapienza University of Rome, Italy. Subjects signed informed consent approved by Centro Trasfusionale Universitario and Clinica Dermatologica of Azienda Policlinico Umberto I, Sapienza University of Rome, Italy (Board resolution approval number #35/2017). Heparinized blood was diluted 1:2 with PBS1X, centrifuged at 1,200 x g for 20 min and plasma was collected and stored at -80°C.

Immunocapture of CD81sEV
To investigate a homogeneous EV population and exclude lipoprotein contamination we captured sEV expressing CD81 pan EV marker. Protein A+G Sepharose resin (Pierce, Thermo Fisher Scientific) was washed twice with sodium tetraborate 0.1 M pH 9.0 and incubated with monoclonal anti-CD81 antibody (B-11, Santa Cruz Biotechnologys Heidelberg, Germany) in rotation at RT. After washes, the resin was resuspended in 0.1M sodium tetraborate containing 20 mM dimethyl pimelimidate DMP (Pierce) crosslinker and incubated o.n. After washing twice with 50 mM Tris pH 7.5 the resin was incubated o.n. with plasma, previously diluted 1:1 and centrifuged for 45 min at 12,000 × g, to capture CD81sEV.

Nanoparticle tracking analysis of sEV
Total sEV were subjected to nanoparticle tracking analysis (NTA, NanoSight NS300, Malvern Panalytical, UK). Detection conditions: capture level 15, threshold 5, slider gain 366, capture duration 60 s. Five videos of 60 s were recorded and analyzed by NTA 3.0 software (Malvern Instruments).

FA Gas Chromatography of total sEV
Lipids were transmethylated with Boron trifluoride-methanol solution (Sigma-Aldrich) at 70°C. Fatty acid methyl esters (FAME) were extracted and analyzed by gas chromatography (Agilent, Palo Alto, CA), equipped with a fused silica capillary column (Omegawax 250, 30m x 0.25mm i.d. and 0.25 μm film thickness, Supelco, Sigma-Aldrich) and a flame ionization detector. The injector (split 5:1) temperature was 260°C, and the detector temperature was set at 280°C. The heating program began at 190°C, increased by 2°C per minute and was held at 240°C. FAME were identified by comparison with authentic standards (Supelco) and calculated as percentages of total FA. Data were analyzed by Prism (Graphpad).

FA GC-MS analysis of CD81sEV
Membrane phospholipids were extracted and transmethylated [25]. Fatty acid methyl esters were measured with an Agilent 7890B gas chromatograph coupled to a 5977B quadrupole mass selective detector (Agilent Technologies, Palo Alto, CA). Separations were carried out with an Agilent HP5ms fused-silica capillary column (30m × 0.25mm i.d.) coated with 5%-phenyl-95%-dimethylpolysiloxane (film thickness 0.25 μm) as stationary phase. Injection mode: splitless at a temperature of 280 °C. Column temperature program: 120 °C (1 min) then to 320 °C at a rate of 20 °C/min and held for 10 min. The carrier gas was helium at a constant flow of 1.0 ml/min. The spectra were obtained in the electron impact mode at 70 eV ionization energy; ion source 280 °C; ion source vacuum 10-5 Torr. MS analysis was performed simultaneously in TIC (mass range scan from m/z 50 to 600 at a rate of 0.42 scans s-1) and SIM mode. Data were analyzed by Prism (Graphpad).

LC-MS/MS proteomics of CD81sEV
Electrophoresis was performed on NuPAGE 4-12% (Novex, Invitrogen, CA), loading 8 µg for each CD81sEV deriving from a pool of 2 melanoma patients or 2 HD. For each stage, (0-I, II, III-IV) and HD we analyzed 3 pools, for a total of 6 samples/group. Each gel lane was almost completely cut in 10 slices, avoiding only areas close to 50 and 25 KDa to elude IgG and serum albumin contaminants. Coomassie (Colloidal Blue Staining kit, Invitrogen) stained proteins were enzymatically in gel digested after reduction and alkylation of cysteine residues [26]. The peptide mixture was desalted on a trap column (Acclaim PepMap 100 C18, Thermo Fisher Scientific) and separated on a 20-cm-long silica capillary (Silica Tips FS 360-75-8, New Objective), packed in-house through a reverse-phased capillary high pressure liquid chromatography using an UHPLC nano system (Ultimate 3000 Dionex, Thermo Fisher Scientific), with a 75 min long acetonitrile gradient with a flow rate of 250 nl/min [23]. Eluted peptides were in-line introduced to an Orbitrap Fusion Tribrid Mass Spectrometer (Thermo Fisher Scientific, CA). Orbitrap detection was used for MS1 measurements, at resolution of 60 K (at m/z 200), while MS2 was performed in the ion trap with a rapid scan rate. The most abundant quadrupole isolated precursors were selected in the range of m/z 350-1550 and a data-dependent MS/MS analysis was performed in top speed mode with a 3 s cycle time. Precursors were fragmented with HCD using 32% of normalized collision energy and dynamic exclusion was enabled for 60 s. A standard Automatic Gain Control (AGC) with 50 ms of maximum injection time was used for MS1, while 20% of normalized custom AGC with a dynamic maximum injection time mode was applied for MS2.

Computational MS data analysis of CD81sEV
MS raw files were analyzed by Proteome Discoverer 2.4 software (Thermo Fisher Scientific) and peak lists were searched against the human database from UniProtKB/Swiss-Prot database (Release 25 October 2017; 42253 sequences) by Sequest HT search engine. Peptide identification was obtained using the precursor and fragment tolerance of 10 ppm and 0.6 Da, respectively. Cysteine carbamidomethylation was used as fixed modification while methionine oxidation and N-acetylation on protein terminus were set as variable modifications; specific trypsin cleavage with two miscleavages were allowed. False discovery rate was set to 1% and determined by searching a reverse database using Percolator node, based on q values. Label-free quantification was based on precursor intensity using all peptides. The proteins were analyzed by VENN diagram (http://bioinformatics.psb.ugent.be/webtools/Venn/). Enrichment analyses including comparison with Vesciclepedia and Exocarta database (http://www.exocarta.org/ ) [27] were performed using FunRich 3.1.3 [28,29] http://www.funrich.org. Protein-protein interaction was analyzed by String v 11.0 (https://string-db.org/) [30]. MS data were deposited to the ProteomeXchange Consortium via the PRIDE [31] partner repository, dataset identifier PXD024434". Proteins were analyzed following the 'DEP package' [32] within the Bioconductor project [33] and based on the statistical programming language R. After quantification, data underwent: 1) filtering: proteins (n=526) were selected when showing at least the 80% of valid values; 2) variance stabilizing transformation for normalization; 3) imputation according to 'minDet' method replacing the missing entries with an estimated minimal value observed in each sample. To identify differentially expressed proteins, the three stages were compared to HD, by combined approach relying on protein-wise linear models and empirical Bayes statistics [34]. Proteins were marked as significant based on two thresholds: a p-adjusted lower than 0.05 and a fold change higher than 1.5. Correction was applied for multiple hypotheses testing through the algorithm of false discovery rate estimation as implemented in the 'fdrtool' package [35].

Total sEV of melanoma patients and HD
Our principal aim was to verify if in early stages including stage 0-I (in situ melanoma) and stage II we could detect discriminating characteristics with respect to stage III-IV advanced patients and in comparison to HD. We isolated total sEV by differential centrifugation and our protocol included a 45 min 12,000 x g step followed by 0.45 μm filtration to eliminate microvesicles. SEM ( Figure 1A, panels a and b) and TEM ( Figure 1A, panels c and d) showed the presence of vesicles displaying the typical structure and dimensions of exosomes, as depicted for representative HD and melanoma stage II sEV. The expression of CD81 and TSG101 proteins confirmed their exosomal-like nature ( Figure 1A, panels e and f HD; g and h patient; CD81 and TSG101, respectively). NTA profiling highlighted differences in particle concentration and size (mode and mean), as shown by the representative spectra of a HD and different stage patients ( Figure 1B). Of note, the number of circulating sEV augmented with progression and differences between stage 0-I and stage II and stage III-IV were statistically significant ( Figure 1C). NTA also revealed that circulating sEV of patients were smaller (mean 93.8+/-9.4; mode 67.2 +/-8.2) than those of HD (mean 107.7 +/-10; mode 83.8+/-8.0) and this difference was maintained throughout the stages ( Figure 1C).

FA signature in total and CD81sEV
GC FA analysis of total sEV showed a prevalence of saturated FA, essentially palmitic (C16:0) and stearic acids (C18:0) (Figure 2A). Compared to HD, patient sEV showed a decrease of saturated FA (C16:0, C18:0, C20:0, C22:0) and the increase of unsaturated FA (C18:1, C18:2n6, C20:4n6). These differences were maintained throughout the stages 0-I, II, III-IV ( Figure S1). The saturation index (SI=C18:0/C18:1) (36) of sEV evidenced a decrease in patients' sEV ( Figure 2B). Capturing CD81sEV led to the loss of long/very long chain FA and evidenced differences in FA profiles. In contrast to total sEV the interstage differences became evident in CD81sEV. Patients displayed a higher content of FA compared to HD, particularly evident for C12 lauric, C14 myristic, C16 palmitic, C18 stearic and C18:1 oleic acids ( Figure 2C). The SI (18:0/18:1) of CD81sEV revealed a strong reduction in stage II and III-IV patients ( Figure 2D). The single FA in CD81sEV revealed the expression of C12:0 and C14:0 FA at high level in advanced stages, although these FA are usu-ally expressed at very low level in human cells. Moreover, C18:0 and C18:1 levels were significantly different between early and late stages and even more in comparison to HD ( Figure 2E).

Proteomic analysis of CD81sEV
CD81sEV of 0-I, II and III-IV stages and HD were analyzed by LC-MS/MS. We identified a total of 899 proteins quantifying 826 of them (Proteome Xchange with identifier PXD024434). We found small differences among the proteins detected in each biological replicate (Supplementary Figure S2A), with an average of 631 proteins detected in each sample (Supplementary Figure S2B). The biggest portion (average 87%) of proteins was detected at least in two independent samples ( Figure S2C and D). Venn diagram ( Figure 3A) showed a small number (56 corresponding to 7% of the total) of proteins uniquely detected in each sample (healthy or melanoma stage) and almost all of these were identified only one time, indicating that these were likely low abundant and thus close to the sensitivity threshold or specific to single samples. Only 6 proteins were identified at least two times uniquely in specific samples and comprised Filamin B for HD, Integral membrane protein 2B and Palmitoyl-protein thioesterase 1 for stage 0-I, Insulin-like growth factor-binding protein 5 and Syntenin-1 for stage II and Dyslexia-associated protein KIAA0319-like protein for stage III-IV (Figure 3, Table B). Comparison with Vesciclepedia using the FunRich tool evidenced 97% of the proteins detected in our samples were present in the database comprehensive of all vesicle types and 95% in that specific for ex-osomes (Supplementary Figure 3E, F). Moreover, 81% of the top 100 Vesciclepedia proteins were detected in our dataset (Supplementary Figure 3G). Similarly, the comparison with Exocarta database evidenced 83% of proteins as common (Supplementary Figure 3H). Cellular component analysis also showed the highest enrichment in exosomes ( Figure 3C), confirming the good quality of our preparations. The comparison with the FunRich database of site of protein expression showed a significant enrichment of our protein set in melanoma and in immune cells, including neutrophils, monocytes, B cells and CD4 and CD8 T cells ( Figure 3D). The analysis of the biological processes involving the detected proteins showed a significant enrichment particularly in protein metabolism, cell growth and/or maintenance and immune response ( Figure 3E). The proteins detected by mass spectrometry were also quantified by label free approach. Five proteins were found as significantly differentially expressed ( Figure 4A): the proteins APOA5, PON1, PON3 and CD14 were increased in stage II CD81sEV in comparison with HD ( Figure 4B), while the protein Rap-1b decreased in stage III-IV patients vs HD. We detected the highest difference between HD and stage II, with respect to the other pairs. Protein-protein interaction among was analyzed by STRING ( Figure 4C) and three proteins (PON3, PON1, and APOA5) had direct interactions and two biological processes were enriched (FDR<0.05): PON1 and PON3 are involved in the lipoxygenase pathway while APOA5, Rap-1b and CD14 are involved in the regulation of vesicle-mediated transport.

Discussion
Despite representing less than 5% of all cutaneous malignancies, melanoma accounts for the majority of skin cancer deaths [37]. Although the advent of targeted and immune therapies have improved the median survival for metastatic patients, biomarkers of surveillance and early detection of initiation and progression are urgently needed to guide clinical intervention. In this context, the protein content of sEV isolated from plasma may be informative [38]. Lipid metabolic alterations in tumor cells have attracted major interest [39], as they are essential for progression and aggressiveness. We and others have previously found that advanced melanoma patients have increased amounts of circulating EV compared to HD [8], the protein amount is higher in late versus early stages and decreases in response to therapy [40]. Here we show that increased sEV concentrations are detectable already in stages 0-II ( Figure 1C) and sEV showed a 20% size reduction. Such characteristics may facilitate EV distribution to lungs and crossing the blood-brain barrier, promoting pre-metastatic niche formation [41]. On the other hand we cannot rule out that the increase in concentration of sEV, isolated as bulk sEV by ultracentrifugations, might derive from lipoproteins precipitating during the ultracentrifugation step despite our sEV isolation protocol included a 12,000 x g centrifugation of 45 min to reduce such contaminations [42]. Similarly, bulk sEV contained long and very long-chain FA that were not detectable in CD81sEV potentially deriving from lipoprotein contaminants precipitating during EV isolation (Figure 2A, B) [42,43]. FA analysis of CD81sEV revealed a higher content of the FA C12 lauric, C14 myristic, C16 palmitic, C18 stearic and C18:1 oleic acids ( Figure 2C). The SI (18:0/18:1) decerased in stage II and III-IV patients, suggesting a link with disease progression ( Figure 2D). The SI was proposed as malignancy marker, consistently with the reduced SI in neoplastic cells and in red blood cell (rbc) membranes of cancer patients [36,44]. In fact, FA analysis of rbc membranes from cancer patients presented a lower relative percentage of saturated FA, higher levels of total MUFA and PUFA and a lower SI compared to controls [45]. Our data show that in Stage 0-I, II and III-IV both C16:0 and C18:0 are significantly higher than in HD, while C18:1 increases in stages II and III-IV ( Figure 2E). The increase of C18:1 typically correlates with the metastatic potential [46], as increased membrane fluidity is a key physical property determining adhesion and migration of tumor cells. The higher abundance of lauric (C12:0) and oleic acids we measured in stage III-IV CD81sEV may be explained by the accelerated metabolism of advanced cancer [47]. Proteomics of CD81sEV identified 899 proteins, 97% of which were comprehensive of all vesicles and 95% specific for exosomes and 81% of the top 100 Vesiclepedia proteins were detected in our dataset, while 83% were common to Exocarta, supporting the efficiency of our CD81-capture. The integral membrane protein 2B (ITMB2) is involved in p53-independent apoptotis and has been identified in stage 0-I [48]. Palmitoyl-protein thioesterase 1 (PPT1), also identified in stage 0-I, is increased in cancer and associated with poor survival [49]. This indicates that even in very early melanoma relevant changes in protein composition of circulating CD81sEV may occur, justifying the rare cases of metastatisation [50]. Insulin-like growth factor-binding protein 5 (IGFBP5) and Syntenin-1 were identified in stage II CD81sEV. The presence of IGFBP5, a tumor suppressor [51], in circulating sEV suggests its discharge from cells, likely favouring tumour progression. Syntenin can support tumor cell proliferation and migration [52] and its presence in sEV could mediate tumor progression. When we quantified the proteins detected by mass spectrometry by label free approach five proteins were significantly and differentially expressed in CD81sEV. APOA5, PON1, PON3, CD14 were increased in stage II, an interesting finding since stage II is at borderline between tumor in situ (stage 0-I) and distant metastasis (stage III-IV). APOA5, expressed by platelet-derived EV (ExoCarta) and podocyte-specific CR1-immunocaptured urine exosomes, modulates intra/extracellular triacylglycerol metabolism [53,54] and this may relate to the differences measured in FA composition. PON1, is implicated in eliminating carcinogenic lipid-soluble radicals. In esophageal squamous cell carcinoma increased oxidative stress is associated with decreased antioxidant PON1 activities [55]. PON1 levels may act as indicator of oxidative stress in cancer [56]. PON3, overexpressed in tumors, promotes cell death resistance. Moreover, PON3 can impair ER stress-induced apoptotic MAPK signalling [57]. Of note, APOA5, PON1 and PON3 displayed a direct interaction in STRING analysis. PON1 and PON3 are involved in the lipoxygenase pathway ( Figure 4C), leading to the generation of eicosanoids impacting cancer development, progression and immune responses [58]. Finally, of the four proteins increased in stage II CD81sEV, CD14 monocyte marker displayed the highest expression. Myeloid cell alterations including the accumulation of myeloid-derived suppressor cells, are typical in advanced stages [59]. However, routinely performed blood counts have evidenced survival associated alterations of peripheral blood leukocytes, including an increase in monocytes, also in early-stage melanoma patients [60]. In stage III-IV CD81sEV we found a decrease of Rap1b compared to HD. This was surprising since Rap1b is involved in MAPK and integrin activation in melanoma, [61], but was also identified as pan-EV marker [62]. Furthermore, Rap1b is expressed by platelet-derived sEV, thus physiologically present in circulating EV [62]. The decrease we measured in CD81sEV may thus potentially depend on the expansion of other EV populations, for instance deriving from immunosuppressive immune cells accumulating during progression.
The alteration of specific proteins and FA in CD81sEV of early-stage disease, especially stage II, may be of clinical significance. Indeed, stage II patients may have a worse melanoma-related survival compared to advanced patients [63]. Future research, combining the study of multiple biologic macromolecules in circulating EV, such as lipids, proteins, nucleic acids and carbohydrates, could be useful in terms of diagnosis, treatment and follow up of cancer patients.

Conclusions
Our results could contribute to intercept those patients who remain at high risk of relapse and thus may benefit from adjuvant therapy, currently intended only for stage III patients. Considering that liquid biopsy prognostic markers are usually searched for in stage III-IV patients while earlier stages are only beginning to be unveiled, stresses the importance of our findings [63,64]. We propose that the combined investigation of FA and protein alterations in plasma CD81sEV could represent a new and useful approach to improve the diagnosis and treatment of cutaneous melanoma, especially in early stages.

Supplementary Materials:
The following are available online at www.mdpi.com/xxx/s1, Table S1. Clinicopathological features of melanoma patients

Funding:
This research was funded by Associazione Italiana per la Ricerca sul Cancro (AIRC) (IG25078 to VH) and was supported by the Current Research Funding and WFR GR-2011-02351400 provided to LL from the Italian Ministry of Health and by Fondi 5x1000 Ministero della Salute 2015 (D/17/1VH to VH).

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethical Committee of Azienda Policlinico Umberto I, University Sapienza, Rome, Italy (Board resolution approval number #35/2017) .

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
Proteomic data are deposited in Proteome Xchange database with identifier PXD024434.