FACS-Proteomics strategy toward extracellular vesicles single-phenotype characterization in biological fluids: exploring the role of leukocyte-derived EVs in multiple sclerosis

Abstract

Background

The isolation and proteomics characterization of extracellular vesicles (EVs) from body fluids is challenging due to their vast heterogeneity. We have recently demonstrated that Fluorescence-activated Cell Sorting (FACS) efficiently isolates the whole EV circulating compartment directly from untouched body fluids enabling a comprehensive EV proteomics analysis.

Results

Here, we characterized, for the first time, a single-phenotype EV subset by sorting leukocyte-derived EVs (Leuko EVs) from peripheral blood and tears of healthy volunteers. Using an optimized and patented staining protocol of the whole EV compartment we identified and excluded non-EV particles, debris and damaged EVs. We further isolated, using an anti-CD45 antibody, Leuko EVs (CD45+ EVs), reaching a high level of purity (> 90%). Purified Leuko EVs were characterized using atomic force microscopy, nanoparticle tracking, and shotgun proteomics analysis revealing a similar coded protein cargo in both biological fluids. Subsequently, the same workflow was applied to tears from Relapsing–Remitting Multiple Sclerosis (RRMS) patients, revealing a Leuko EVs protein cargo enrichment that reflects the neuroinflammatory condition characteristics of RRMS. This enrichment was evidenced by the activation of upstream regulators TGFB1 and NFE2L2, which are associated with inflammatory responses. Additionally, the analysis identified markers indicative of endothelial cell proliferation and the development of enhanced vascular networks, with AGNPT2 and VEGF emerging as activated upstream regulators. These findings indicate the complex interplay between inflammation and angiogenesis in RRMS.

Conclusions

In conclusion, our combined FACS-Proteomics strategy offers a promising approach for biomarker discovery, analysing cell-specific EV phenotypes directly from untouched body fluids, advancing the clinical value of tears EVs and improving the understanding of EV-mediated processes in vivo. Data are available via ProteomeXchange with the identifier PXD049036 and in EV-TRACK knowledgebase with ID: EV240150.

Background

Isolation and characterization of extracellular vesicles (EVs) from biological samples remain a technological challenge even if several improvements have been recently achieved. Circulating EVs in biological fluids are an extremely heterogeneous compartment in terms of sizes and phenotypes since they are released from different cell types and it is well known that cells release EVs with specific molecular cargoes in term of proteins, lipids/metabolites, RNAs and DNA fragments, depending on their role and the individual physio-pathological condition [1, 2]. It is becoming increasingly clear that the intercellular information exchange mediated by EVs displays potent regulatory functions, mediated both by their surface receptors and their content [3]. EVs are emerging as relevant players in the inter-cellular crosstalk and in different pathophysiological conditions [4, 5], given that they convey materials dynamically related to the status of their parental cells. Furthermore, EVs were identified in all body fluids, such as cerebrospinal fluid (CSF), tears, saliva, urine, milk, and peripheral blood (PB) [2, 6,7,8]. Therefore, isolating EVs from body fluids and analyzing their biological content represents, today, a stimulating challenge to improve our knowledge on EV biology.

Nowadays, the characterization of proteome has been dramatically accelerated thanks to the technological improvements of the liquid chromatography coupled to tandem mass spectrometry (LC–MS/MS) platforms, which are useful to understand the phenotype and the role of EVs even in complex biological matrices unraveling key factors biological pathway modulations [9, 10].

Nevertheless, the EV isolation step, not yet standardized, represents one of the main issues in EV proteomics characterization, particularly affected by biological fluid soluble proteins cross-contaminations. Moreover, isolated EVs could be “coated” with proteins, glycoproteins or glycolipids likely to cause aggregation, fusion and co-sedimentation of similar vesicles, negatively affecting the recovery, purity, and further molecular analyses. Another relevant issue concerns the vast heterogeneity of the circulating EV population in terms of phenotypes since each single cell type releases specific EVs, biologically different from those released by other cell types [11]. Ideally, the best EV isolation method, not only for proteomics characterization, should be able to minimize soluble contaminants and allow to collect separately the different EV phenotypes based on their cell of origin.

Giving a brief overview of principal EV isolation protocols, the most used are based on differential or density gradient centrifugation [11, 12]. Other studies employed the use of chemical precipitation kits [13, 14], or several series of ultrafiltration combined with ultracentrifugation or size-exclusion chromatography (SEC) [15, 16], or immuno-isolation with magnetic beads like ExoChip [17], and microfluidic separations with ciliated micropillars [18]. Moreover, recently, various isolation and detection methods have been combined simultaneously [19, 20] improving final performances.

Despite several attempts, the best strategy for EVs isolation from biological fluids remains an open question for the scientific community, especially looking at their suitability for proteomics application. Isolated EVs may be used as liquid biopsy allowing a huge reduction of the high dynamic range of protein concentration detectable in whole biofluids, making them a promising candidate for biomarker discovery in the so called “dark proteome” [21]. Their circulating proteome may reveal an underappreciated world to investigate, suggesting that their molecular cargo may reflect pathophysiological conditions of the Central Nervous System (CNS) considering that distinct types of brain cells release EVs which function as shuttles to deliver their cargo among different cells [22]. For example, in the context of multiple sclerosis, EVs can drive both inflammatory processes and immune tolerance, which are typical characteristics of such immune-mediated demyelinating disorder [23]. Many studies have been conducted on multiple sclerosis EVs molecular cargo characterization, revealing specific membrane markers for endothelial-derived EVs (i.e., CD31 (PECAM-1), CD51 (integrin), CD54 (ICAM-1), CD62E (ELAM-1 or E-selectin), CD63E, and CD105 (endoglin) that increases the adhesion of inflammatory leukocytes to endothelial cells and promotes their transendothelial migration into the CNS), and leukocyte-derived EVs (C-C chemokine receptor type 3 (CCR3) and type 5 (CCR5), unravelled as specific markers of T cells in CSF) [24]. Moreover, patients have exhibited a strong increase of serum myelin basic protein (MBP) levels compared to healthy people, demonstrating a potential crosstalk between EVs released by oligodendrocytes and microglia with demyelinating processes [25]. Altogether, EV protein cargo reveals specific molecules involved in neurodegeneration and demyelination, for example acid sphingomyelinase (aSMase) was highly expressed in CSF of relapsing–remitting multiple sclerosis (RRMS) patients, correlating with EVs release [26]. Plasma kallikrein (KLKB1), dickkopf-related protein 3 (DKK3), complement component C6 (C6), and protein S100A9 were identified in EVs CSF of RRMS subjects confirming their role in wound healing, homeostasis, and coagulation, such as fibrinogens which can promote the autoimmunity and inflammatory demyelination disorders [27, 28].

Within such a complex and dynamic scenario, we have recently optimized an innovative protocol for the isolation and proteomics characterization of EVs directly from unprocessed biofluids. The FACS-based purification method takes advantage of the combination of the lipophilic cationic dye (LCD) and phalloidin for the identification of damaged membranes (patent codes: 10,201,800,000,398; EP19164567A·2019-03-22). By applying this protocol, a high level of repeatability in terms of EV counts was achieved. We previously demonstrated that EVs isolated by such a strategy were purified from the most abundant contaminants (i.e., non-EV particles), such as albumin and apolipoprotein when EVs from peripheral blood were purified [29]. As a result, we were able to probe and purify LCD+/Phalloidin- intact EVs suitable for proteomics characterization [29]. The first application of such an innovative “FACS-Proteomics” workflow was the study of total EV pools (label free) from untouched CSF and tears collected from RRMS patients, revealing a proteins cargo involved in inflammation, angiogenesis and immune response signalling in both biofluids [30].

We demonstrated that tear analysis can be extended beyond the ocular disease since they contain enriched biological molecules filtrated from circulatory system, representing a precious source of information in physiological and pathological conditions. This concept was elegantly strengthened by Hu et al. in a recent paper in which they found the genesis of tear EV proteins with 37 tissues and 79 cell types by using universal analysis in combination with the Human Protein Atlas consensus dataset [31].

Here, we reported a first application of the recently optimized FACS-Proteomics workflow in the proteomics characterization of EV single-phenotype (labelled) from untouched body fluids. As a proof of concept, we characterized the protein cargo of leukocyte-derived EVs (Leuko EVs) isolated from blood and tears of healthy subjects. Finally, we carried out a label free proteomics analysis on Leuko EVs purified from tears of RRMS patients, proving that the leukocyte’s role in a such autoimmune disease, may be, at least in part, EV-mediated.

Methods

Ethics statement

The study design was made following the protocol approved on 29 December 2020 by the Ethic committee of “G. d’Annunzio” University, in accordance with the Declaration of Helsinki (World Medical Association, 1997). All enrolled people were informed about the procedures and provided written informed consent to participate in the study.

Recruitment and sample collection

Subjects enrolled as Healthy donors (HCs) were selected with no prior history of systemic disease or administration of immunosuppressing or immunomodulating drugs (n = 10). RRMS patients were enrolled at the Multiple Sclerosis Centre of “SS. Annunziata” Hospital, Chieti (Italy) after that their clinical diagnosis was confirmed by magnetic resonance imaging (MRI) studies and by the presence of oligoclonal bands (OCBs) in CSF fulfilling the 2017-revised McDonald diagnostic criteria. 23 RRMS patients with the same characteristics (i.e., diagnosis of RRMS according to the 2017-revised McDonald criteria) were focused on the proteomics characterization of Leuko EVs cargo (Table 1). All RRMS patients were in clinical remission (i.e., they did not experience relapses within the 30 days prior to sample collection nor were treated with steroids in the month before the enrolment). At the time of tear collection, 5 patients (21.74%) were receiving moderate-efficacy disease-modifying treatments (DMTs), 15 patients (65.22%) were on high-efficacy DMTs, and 3 patients (13.04%) were untreated [32]. Exclusion criteria for tears collection in both RRMS patients and HCs were ocular diseases in the previous 12 months, history of topical therapy, and the use of contact lenses. Briefly, as reported by Rossi et al. [33], tears were collected on graduated Schirmer’s strips asking the patients to look up and pulling the lower lid gently downward for 5 min and then extracted in 0.01 M PBS, Phosphate Buffered Saline (Sigma-Aldrich, St. Louis, MI, USA).

Table 1 Clinic-demographic characteristic of subjects enrolled for sorting Leuko-derived EVs for proteomics assessment

On the other hand, PB samples were used for protocol optimization and were obtained from 7 healthy Caucasian HCs who gave written informed consent. They did not declare themselves to be under chronic therapies, nor to be affected by chronic conditions.

Leukocyte-derived EVs staining and acquisition

To identify, phenotype and enumerate Leuko EVs, PB and tear samples, an already optimized and published flow cytometry protocol [34,35,36,37] adhering to all current ISEV flow cytometry guidelines, including MIFlowCyt-EV and MISEV2023 recommendation [38,39,40]. Briefly, samples were stained using 1 μL of LCD, 0.2 μL of fluorescein isothiocyanate (FITC)-conjugated phalloidin and 1 μL of CD45 Brilliant Violet 510 (BV510)-conjugated, all from BD Biosciences (San Jose, CA, USA). Samples were incubated at RT for 45 min in the dark and then diluted in PBS 1X before being filtered with 0.70 μm filter pore size. CD45+ events were acquired by flow cytometry (FACSVerse, BD Biosciences, San Jose, CA, USA) from RRMS and HC, as already reported [29]. Data were analysed using FACSDiva v 6.1.3 software (BD, BD Biosciences, San Jose, CA, USA). EV concentrations were obtained by volumetric count [29].

Leukocyte EVs purification by fluorescence-activated cell sorter

Tears and PB Leuko EVs were separated by instrumental cell sorting (100 μm nozzle, FACSAria III, BD Biosciences), based on the gating strategy shown in the scheme of Fig. 1, panels B–G. In detail, the trigger threshold was placed on the APC channel and, for all parameters, the height (H) signals, as well as bi-exponential or logarithmic modes were selected. The post-sorting purity was assessed by using the same instrument (FACSAria III) and the same setting applied for EVs separation, resulting anytime higher than 90% (not shown) [29, 30].

Fig. 1

Workflow, identification and isolation of tear and PB EVs. A EVs subtypes counts scheme in tears and PB samples of healthy subjects (HCs). The image was created by BioRender.com. B The scatter area containing EVs was firstly gated on FSC-H/SSC-H dot-plot; C Events positive to LCD (y-axis) and negativity to Phalloidin (x-axis) were identified as EVs; D EVs (LCD+/Phalloidin- events) were subtyped to identify leukocyte derived-EVs as events positive to CD45 (x-axis). E–G The same gating strategy was used to identify and separate PB EVs. Data are representative of all analysed samples

Nanoparticle tracking analysis

EVs isolated by FACS (350,000 EVs/sample) were analyzed using ZetaView (Particle Metrix GmbH, Meerbusch, Germany) equipped with a blue laser (488 nm, 40mW) and a long pass emission filter (LWP) for blue excitation laser (cut-off at 500 nm). The autofocus was adjusted to avoid indistinct particle visualization. Post-acquisition settings were optimized and maintained across sample analyses: for each sample 8–10 fields were analyzed, and a video of 30–60 s was recorded using ZetaView software. For each measurement a report for all analyzed parameters including mean, mode, and median size of EVs together with an estimation of concentration was obtained.

Atomic force microscopy

Atomic Force Microscopy (AFM) analysis was performed to investigate the size distribution and the morphology of the isolated total and Leuko EVs by using MultiMode 8 AFM microscope with Nanoscope V controller (Bruker, Billerica, Massachusetts, US). The samples were prepared by depositing a drop of diluted suspension of the isolated exosomes on SiO₂ wafer followed by drying in the oven at 37 °C for 2 h and then at room temperature overnight. The obtained samples were scanned by the silicon ScanAsyst-Air probe (triangular geometry, cantilever resonance frequency 70 kHz and nominal spring constant 0.4 N/m) in ScanAsyst in Air mode. Images of 512 × 512 pixels were collected with different scan sizes and were elaborated using NanoScope Analysis 1.8 software. The Particle Analysis tool of the software was used to detect and measure the size of the vesicles and the data were collected from the analysis of tens of particles captured from several AFM images.

Flow cytometry analysis of EV surface antigens

EV subtyping

Tears and PB EVs were identified, counted and subtyped using the panel detailed in Table S1. Briefly, 100 µL of tears or 5 µL of PB samples were stained using the reagent mix listed in Table S1. Samples were incubated for 45 min in the dark at RT, and diluted using 500 µl of PBS, as already reported [29].

EV marker detection

EVs were analyzed for the expression of CD63, and CD81, as detailed in Table S2. Samples were incubated for 45 min in the dark at RT, and then diluted using 500 µl of PBS.

Flow cytometry EV acquisition

At least 1 × 10⁶ EVs/sample were acquired using a CytoFLEX (Beckman Coulter, Sacramento, CA, USA) or a FACSVerse (BD Biosciences) flow cytometer, using a low (600–800) trigger threshold on the channel in which the LCD emits (APC channel) [37, 41]; CompBeads (BD Biosciences) and single stained samples were used to assess the compensation [42]. Fluorescence Minus One controls were used as recommended [34].

Leukocyte-derived EVs proteomics characterization

At first, we optimized the FACS-proteomics protocol on PB and tears EV samples for a single-phenotype characterization by sorting 2.0 × 10⁶ (PB) and 1.4 × 10⁵ (tears) of pooled CD45+ EVs from HCs. Meanwhile, in the second part of the study, the same number of Leuko-derived EVs (CD45+ EVs, 1.4 × 10⁵) were sorted from tears of RRMS patients. As already published, the number of separated EVs was used as a normalization parameter for protein label-free identification and quantification [30]. In both cases, samples were prepared for FASP protocol by lysing Leuko EVs through sonication on ice (Sonicator U200S control, IKA Labortechnik, Staufen, Germany) at 70% amplitude in a lysis buffer (urea 6 M in 100 mM Tris/HCl, CHAPS 2% and Triton X-100 1% at pH = 7.5) using Amicon® Ultra Centrifugal Filters with 10 KDa molecular mass cut-off (MilliporeSigma Headqua, Massachusetts, USA). After that an overnight tryptic digestion was carried out at 37 °C in a buffer containing urea 8 M, Tris HCl 100 mM at pH = 8.5 on 10 KDa filters (Nanosep®, Membrana: Omega™, VWR International, Milan, Italy). EV tryptic peptides were analyzed in triplicate by nanoLC-MS/MS using the UltiMate™ 3000 UPLC (Thermo Fisher Scientific, Milan, Italy) chromatographic system coupled to the Orbitrap Fusion™ Tribrid™ (Thermo Fisher Scientific, Milan, Italy) mass spectrometer, as previously published in our works [36, 43]. Briefly, the flow rate was set at 300 nL/min, with a total run time of 65 min using a chromatographic gradient from 2 to 90% of phase B (phase A: water, phase B: acetonitrile, both with 0.1% formic acid). Orbitrap analyzer was used for MS1 scans with 240,000 of resolution and the signal intensity threshold at 1 × 10⁴ for MS2 sequencing look at the ion trap performing a HCD fragmentation. Subsequently, MaxQuant, version 1.6.6.0, (Max-Planck Institute for Biochemistry, Martinsried, Germany) was used for processing raw MS/MS data with UniProt database (released 2020_06, taxonomy Homo Sapiens, 20,588 entries). All processing parameters and settings were reported in detail in our published works [36, 44]. Briefly, the false discovery rate (FDR) was set to 1% both for protein and for peptide levels. Match-between-runs (MBR) algorithm was used to transfer the peptide identifications from one LC–MS/MS run to all others using its default settings (match window of 0.7 min and alignment time of 20 min) [45]. Nonspecific cleavage to both ends of the peptides were allowed with maximum of 2 missed cleavages. In this regard, MS search output for protein quantification (i.e., number of peptides, proteins groups etc.) was detailed in raw file of Supplementary Table S9. For the second part of study and clinical application of FACS-Proteomics workflow, LFQ Intensity values were used for bioinformatics analyses performing by Perseus software, version 1.6.10.50, (Max-Planck Institute for Biochemistry, Martinsried, Germany) by fixing the minimum number of valid LFQ Intensity values at 2 in at least one clinical group (CV% was calculated between replicates to assess proteomics analysis variability showing a CV% < 50 for most of the quantified proteins, see Table S9) in order to evaluated not only the different protein modulation, but also the presence (fold change = 100) and absence (fold change = 0.01) of proteins between Leuko EV tears RRMS compared to HC ones. Subsequently, protein ratios (RRMS/HC) were used for functional enrichment analysis through “Core Analysis” of Ingenuity Pathway Analysis software (IPA, Qiagen, Hilden, Germany) [46]. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [47] partner repository with the dataset identifier PXD049036 and https://doiorg.publicaciones.saludcastillayleon.es/10.6019/PXD049036.

Moreover, to define the differential proteins significantly expressed between pathologic group and control one, a univariate statistical analysis was performed with a p-value threshold of 0.05 visualizing results as Volcano Plot.

STRING version 12.0 (https://string-db.org/) analysis was used for evaluation of protein–protein interaction (PPI) networks and Gene Ontology (GO) enrichment analysis for cellular component (CC) together with Retrieve/ID mapping through UniProt SEARCH BOX [48].

FunRich (Functional Enrichment analysis tool, http://www.funrich.org/) version 3.1.4 release on 2020 was used to easily match our Leuko EV datasets with Vesiclepedia filtering the search to Homo Sapiens as species and using the whole deposited studies on biological fluids (e.g., blood, urine, saliva, plasma, serum, seminal fluids, amniotic fluids, aqueous humor and epididymal fluid) and vesicle types [49,50,51].

We have submitted all relevant data of our experiments to the EV-TRACK knowledgebase (EV-TRACK ID: EV240150) [52].

Results

Leukocyte-derived EVs identification, count and purification

In such a study, we focused on the analysis of CD45+ Leuko EVs single phenotype purification from biological fluids, given that they are particularly abundant both in tears and in PB.

Leuko EVs represent an EV subtype released from a quite heterogeneous population of immune cells, including granulocytes (neutrophils, eosinophils, and basophils), monocytes, and lymphocytes (T cells and B cells) recognised as CD45+ events. Thus, total and intact EVs of tear and PB samples were identified as LCD+/Phalloidin− events as reported in the gating strategies of Fig. 1 panels B–D (tears) and panels E–G (PB samples). As also endorsed by ISEV guidelines [11], CD45+ EVs events were reclassified as Leuko EVs (Fig. 1 panels D and G, respectively), detected and isolated by instrumental cell sorting, according to the reported gating strategy (Fig. 1). Moreover, such a population of EVs was purified from lacrimal fluids of patients affected by RRMS as described below. Therefore, we focused on the proteomics characterization of Leuko EVs purified from tears and PB samples.

The workflow shown in Fig. 2 summarizes the study design and the proteomics results highlighting both the samples and techniques used to characterize Leuko EV populations in terms of protein cargoes in the first optimizing phase of FACS-Proteomics method (tears and PB of HC) and in the second phase focusing on clinical application to RRMS tears.

Fig. 2

Proteomics study workflow. PB: peripheral blood, Leuko EVs: leukocyte-derived EVs, HC: healthy controls, RRMS: relapsing–remitting multiple sclerosis, DEPs: differential expressed proteins. The image was created by BioRender.com

Characterization of total and CD45+ EV features

Total EVs both from PB (Fig. 3A) and tears (Fig. 3B) of healthy volunteers were analysed by flow cytometry for the expression of CD63 and CD81 specific EV markers. As shown in Fig. 2, all the analysed EV populations displayed significant levels of tetraspanins, confirming their EV nature.

Fig. 3

Protein expression markers detection. Typical EV markers (CD63 and CD81) were analysed on the whole EV compartment both in PB (panel A) and in tear (panel B) samples

The isolated pure CD45+ EV fraction, both from tears and PB HC samples was characterized using Nanoparticle tracking analysis (NTA) and Atomic Force Microscopy (AFM). According to NTA, the whole population of EVs derived from PB or tear samples showed a median diameter of 120.3 nm and 105.4 nm respectively (Fig. 4A), while CD45+ EVs derived from PB or tear showed a median diameter of 127.5 nm and 145.8 nm, respectively (Fig. 4B). NTA results were in line with the size range detected by AFM for all EV samples. The AFM images of the 2D topography of total (Fig. 4A) and CD45+ EVs (Fig. 4B) revealed the presence of isolated vesicular structures characterized by different sizes. In line with NTA analysis results, pure total EV population sorted from PB and tears exhibited average diameters of 76.7 ± 29.6 nm and 109.3 ± 60.2 nm, respectively, while the mean particle sizes for Leuko-derived EVs isolated from blood and tears were 104.9 ± 50.9 nm and 82.7 ± 28.7 nm. The height profiles of representative EVs reported in Fig. 4 also confirmed the globular and flattened shape, likely due to the alteration induced by the drying step.

Fig. 4

Characterization of Leukocytes-derived EVs from healthy controls biofluids. A NTA tracking of size (diameter/nm) and concentration (particles/mL) of pure total EVs population sorted from PB (left) and tears (right). EVs derived from total PB or total tear showed a median diameter of 120.3 nm and 105.4 nm, respectively, falling within the size range detected by AFM reported in the figure. The height cross sectional profiles of representative vesicles of total PB (blue arrow) and total tear (red arrow) are also reported. B NTA tracking of size (diameter/nm) and concentration (particles/mL) of pure Leuko EVs sorted from PB (left) and tears (right). Leuko EVs derived from PB or tears showed a median diameter of 127.5 nm and 145.8 nm respectively, falling within the size range detected by AFM indicated in the figure. The height cross sectional profiles of representative Leuko EVs of PB samples (green arrow) and tears (orange arrow) are also reported

In this context, the NTA analysis showed that CD45+ EVs were characterized by an increased median diameter when compared to total EVs samples, significant only for EVs derived from tear samples (Table S3); on the other hand, AFM analysis highlighted statistical evidence only in the EVs samples derived from PB (Figure S1). Despite the slight discrepancy between AFM and NTA analyses, related to some variabilities between the two used techniques, the AFM measurements show a good level of agreement with NTA regarding the vesicles sizes, both of which are under 200 nm.

Detailed information about NTA measurements and results were listed in Supplementary Table S3.

Functional annotation of single-phenotype EVs proteomics in blood and tears

Purified Leuko EVs from pooled HC tears and PB were further analysed by proteomics. In detail, 1.4 × 10⁵ and 2 × 10⁶ events were purified as Leuko EVs from tears and PB, respectively, and, despite the limited amount of tear sample, we identified 78 proteins from tears Leuko EVs against 155 proteins form PB Leuko EVs. First, we compared the identified proteins to the reference EV platform Vesiclepedia (http://www.microvesicles.org) [49, 51], filtering the deposited studies on biological fluids, obtaining 88.67% and 92.31% of matching for Leuko EVs isolated from PB and from lacrimal Leuko EVs, respectively, as reported in Fig. 5A. All the identified proteins matched and unmatched in Vesiclepedia repository are reported in Supplementary Tables S5 and S6. Interestingly, going deep into the unmatched proteins, we found that they are mainly referred to as keratinocytes for tear fluid and immunoglobulins for PB, while about 90% of identified proteins are referred to the EV compartment. By comparing the identified proteins in Leuko EVs from both PB and tears, 17% of such proteins were found to be in common between the two biological fluids, as shown in the Venn diagram of Fig. 5B. Moreover, to study the Leuko EV protein cargoes from a functional point of view, we performed a GO annotation processing of proteomics data to find putative network of interaction of the Leuko EV-enriched proteins identified in both biological fluids. They were analyzed for CC compartment showing that Leuko EVs proteomes have a strong enrichment of “Extracellular Exosome” (GO:0070062) associate proteins (83.77% and 76.92% for PB and tears, respectively, as shown in Supplementary Table S7). Meanwhile, cytoplasmic proteins are poorly represented because only 6 (3.90%) and 8 (10.25%) proteins were uniquely reclassified as cytosolic proteins for PB and tear EV preparations, respectively (protein IDs highlighted in red in Table S7). Between them desmoglein-1 and protein S100-A7 were identified in common between Leuko EVs purified from two biological fluids. In this regard, we reported the interaction network of the identified proteins build by the STRING software in which the most significant cellular components referring to each protein have been highlighted by a color code for the proteins identified in Leuko EVs both from PB (panel C) and tears (panel D). In detail, the proteins that belong to the “Extracellular Exosome” (GO:0070062) component are highlighted in red with a FDR of 6.55 × 10^–89 (for PB Leuko EVs) and FDR of 1.23 × 10^–35 (for tears Leuko EVs), while the proteins that belong to the “vesicle” component (GO:0031982) are highlighted in blue with a FDR of 2.07 × 10^–63 (for PB Leuko EVs) and 2.22 × 10^–30 (for tear Leuko EVs), as reported in Fig. 5 panels C and D. White dots represent proteins not identified with the GO classifications. For the list of detail proteins see Supplementary Table S7, sheets 4 and 5. This phenotypic characterization corroborated simultaneously the isolation of intact EVs by instrumental cell sorting, as previously demonstrated in Marchisio et al. [29], and the EV sample preparation in term of purified proteins for proteomics purposes.

Fig. 5

Proteomics assessment of Leuko EVs sorted from PB and tears of healthy subjects. A Bar diagrams depict the matching between Leuko EV proteins and Vesiclepedia, an online EV proteome repository. Only 8.92% and 5.2% of the Leuko EV proteins were unmatched in Vesiclepedia for the pooled PB and tear samples, respectively. Details of all proteins are shown in Tables S5 and S6. B The Venn diagram shows the common proteins between PB Leuko EVs or tears Leuko EVs. C, D Gene Ontology (GO) classification of proteins are reported as red and blue dots to underline “extracellular exosome” and “vesicle” as the most significant cellular component in PB and lacrimal Leuko EVs with an FDR of 6.55 × 10^–89 and 2.07 × 10^–63 for PB Leuko EVs (C), and 1.23 × 10^–35 and 2.22 × 10^–30 for tears Leuko EVs (D), respectively. White dots represent proteins not identified with the GO classifications (see Table S7, sheets 4 and 5, for protein details). E, F Mechanistic networks of “leukocyte migration” downstream in PB Leuko EVs (p-value = 6.31 × 10^–25) and tears Leuko EVs (p-value = 5.95 × 10^–8). Shapes and symbols of proteins are shown in Figure S4

Going in deep to the leukocytes origin of sorted CD45+ EVs, the same protein dataset was used for the expression analysis through IPA tool to predict the downstream effects associated to the protein expression found in Leuko EVs in the two analysed biological fluids. An overview of functional results is displayed in Table 3 depicting the main significant downstream effects expected by Leuko EV proteins from PB (on the left) and tears (on the right). In Fig. 5 we reported a mechanistic network of “leukocyte migration”, resulting significantly covered by the protein dataset for PB Leuko EVs with an overlap p-value of 6.31 × 10^–25 (panel E) and for tears Leuko EVs with a p-value of 5.95 × 10^–8 (panel F). Notably, proteins involved in “leukocyte migration” in tear Leuko EVs overlapped to 23.2% of those in the PB demonstrating as in both body fluids the molecular information carried by Leuko EVs toward the recipient cells can be recognized and investigated. Among these, protein S100-A7, A8 and A9 were identified and connected each other because of their pro-inflammatory role and their contribution in leukocyte adhesion, migration and recruitment.

Among the significant downstream predicted by IPA software in Leuko EVs isolated from both biofluids, reported in Table 2, an overlap of 55.2% was observed. Moreover, most of the significant biological functions covered by the protein dataset in both biological fluids (reported in bold in Table 2) are associated to processes of immune system cells including granulocytes (neutrophils, basophils, and eosinophils) and mast cells. In addition to “leukocyte migration”, we found also other downstream directly related to the origin cells, such as the “adhesion of neutrophils” (p-value = 1.08 × 10^–19 for PB Leuko EVs and 5.54 × 10^–4 for tears Leuko EVs) and “cell movement of neutrophils” (p-value = 7.80 × 10^–18 for PB Leuko EVs and 4.12 × 10^–5 for tears Leuko EVs).

Table 2 List of the main downstream effects predicted to be modulated by Leuko EVs protein from PB and tears of healthy subjects

We compared the GO terms and downstream observed in PB Leuko EVs to the total circulating EVs isolated from PB of the same pooled HC samples. Data were reported in Supplementary Table S4 and S7 and supported the specificity of the proteomics cargo obtainable by the isolation of a specific EVs phenotype.

For all functional GO analyses the complete list of downstream effects and the relative overlap p-values was reported in Table S8.

Leuko EVs protein expression in tears of multiple sclerosis

Our described FACS-proteomics approach was used to study Leuko EV protein cargo in tears from RRMS patients (n = 23) compared to HCs (n = 10). As pre-analytics may change between HCs and patients, we assessed CD45+ EVs purified from lacrimal fluids of RRMS subjects with NTA and AFM techniques showing a median diameter in line with those obtained by HC tears, as shown in Figure S2 and Table S3. Subsequently, 1.4 × 10⁵ Leuko EVs, isolated from pooled tears samples, underwent to proteomics analysis to explore their proteome cargo in RRMS patients compared to HCs. Protein numbers were consistent with those obtained in previous analyses [2, 30, 53]. Venn Diagram in Fig. 6A showed the summary of proteomics investigation in term of the common and unique proteins in the two analysed clinical conditions. 77.42% of proteins were in common between RRMS and HCs, while 22.58% of the proteins were identified and quantified only in Leuko EVs isolated from RRMS and/or HCs tears, as listed in Table 3. In particular, 15 proteins were quantified as unique proteins in RRMS Leuko EVs, and they showed significant interactions between them and were related to “Immune System HAS-168256 Reactome Pathway” (FDR = 2.07 × 10^–5, yellow dots) as demonstrated by STRING analysis (PPI enrichment p-value = 1.06 × 10^–6) and reported in Fig. 6B. Alongside these unique proteins, we compared the expression of common proteins between RRMS and HC Leuko EVs as reported in the Volcano plot of Fig. 6C highlighting 19 significantly up-regulated (red dots) and 14 down-regulated (green dots) proteins in RRMS, as listed in Table 4.

Fig. 6

Clinical application of single phenotype EV proteomics workflow to tear fluid of Multiple Sclerosis patients. A Venn diagram of quantified proteins in Leuko EVs sorted from tears of RRMS patients and healthy volunteers. B Proteomics PPI assessment of RRMS unique proteins. Yellow dots were reclassified as proteins involved in “Immune System HAS-168256 Reactome Pathway”. C Volcano Plot of proteins graphed by fold change (Difference) and -Log(p-value) by the comparison of RRMS vs HC subjects. Grey dots represent proteins that were not differentially expressed in the comparison carried out; red dots represent proteins that were significantly up-regulated and green dots indicate proteins that were significantly down-regulated in RRMS. D Regulator Effects network highlights the hypothesis of how “migration of endothelial cells” is activated by transforming growth factor beta 1 (TGFB1) as upstream regulator. E Networks of activated upstream regulators: Angiopoietin-2 (ANGPT2) and vascular endothelial growth factor group (VEGF). F Downstream effect analysis reveals the upregulation of “angiogenesis”. Orange and blue shapes represent predicted activation or inhibition, respectively. Instead, red and green shapes represent increased or decreased measurements of identified proteins, respectively, whose expression value is reported in the figure. Continuous lines represent direct relationships, dotted ones represent indirect relationships, whereas grey and yellow lines indicate the non-predicted and inconsistent relationships. Figure S4 reports an interpretation of IPA networks

Table 3 Unique proteins identified in the Leuko EVs of two different clinical conditions (RRMS vs HC)

Table 4 Differential proteins in Leuko EVs isolated from tears of RRMS patients compared to HCs

LFQ Intensity values for each identified protein obtained through MaxQuant software were used as quantification parameter, and the LFQ ratio (RRMS/HC) was further employed as input for functional analysis by IPA tool. Figure 6D–F report the main proteomics functional results obtained by comparing the protein expression in the two-condition analysed. Tear Leuko EVs from RRMS patients have a protein expression able to trigger information related to a deep increase of neuroinflammation and angiogenic processes in RRMS patients.

Regulator Effects feature of IPA software can generate hypotheses about potential mechanisms that determinate the biological functions according to the related genes up- or down-regulated as well as upstream. In fact, network represented in Fig. 6D shows how “migration of endothelial cells” (p-value = 7.08 × 10^–8, z-score = 2.00) function is significantly up-regulated via Leuko EVs proteins by Transforming growth factor beta 1 (TGFB1) (p-value = 4.33 × 10^–8, z-score = 2.28) as activated upstream regulator in RRMS. This data corroborates the complex role of Leuko EVs as shuttles in the Immune System machinery in response to neuroinflammation and neurodegeneration scenario in RRMS. Moreover, endothelial cell migration is an essential event to trigger angiogenesis process, which is one of the most significant downstream in RRMS. Actually, Fig. 6 panel E indicates Angiopoietin 2 (ANGPT2) (p-value = 1.17 × 10^–2) and Vascular endothelial growth factor group (VEGF) (p-value = 1.62 × 10^–1) as significantly activated upstream regulators in RRMS Leuko EVs with z-scores of 2.00, involved in “vascular remodelling” and “angiogenesis” resulted deeply activated in RRMS Leuko EVs (p-value = 2.87 × 10^–8, z-score = 1.92), as highlighted in the network of Fig. 6 panel F. The complete lists of downstream effects and upstream regulators for comparison RRMS vs HC were reported in Supplementary Table S8, sheets 3 and 4.

Discussion

In our previous reports we applied a new experimental workflow for isolating total EVs suitable for proteomics analysis, directly from untouched body fluids [2, 8, 30, 53, 54]. The method was applied for isolating and studying the total EV population in different pathological context and took advantage of an optimized FACS sorting isolation protocol coupled to mass spectrometry-based proteomics characterization [2, 8, 30, 33]. However, the study of ex-vivo cell-specific EV subsets should be considered the main target in this setting but, at the same time, represents an ongoing challenge [55]. EVs from cell cultures are a good way to access at single EV phenotypes, however, scientific results obtainable couldn’t represent the picture of the in vivo real EV behaviour, especially because of the EV mediated intercellular crosstalk complex machineries and should be taken with caution. Here, we reported an update on the EV characterization, by exploiting, for the first time, the intrinsic purification capability of FACS sorting based on labelled-recognition events and focusing on the isolation and proteomics characterization of a specific sub-type of circulating EVs. The first single EV phenotype proteomics application was carried out on Leuko EVs from PB and tears of enrolled volunteers by sorting CD45+ EVs. We are aware about the heterogeneity of such a population released from different immune cell types and the reason of the Leuko EVs sub-type choice arose from the need of a large population as possible due to the extremely low amount of material for proteomics analysis and then for our interest in a deeply characterization of the EV-mediated immune system machinery in RRMS [56]. Based on our previously published results in which we demonstrated the purity and the size distribution of isolated EVs [29], we were quite confident about the efficiency of the FACS sorting protocol in term of purity and quality of the sorted Leuko EV samples, after confirmed by the obtained proteomics results. Actually, Leuko EVs in PB and tears were identified as LCD+ /Phalloidin-/CD45+ events, isolated by fluorescence-activated cell sorting and characterized by AFM and NTA, revealing an EV mean diameter < 200 nm (small EVs according to MISEV 2023 classification). The post-sorting purity was assessed by using the same setting applied for EVs separation, resulting anytime higher than 90% (not shown). Based on our knowledge, this is the first time that Leuko EVs were recognized and isolated from PB and lacrimal fluid of the same volunteers paving the way to single-phenotype EV characterization from biological fluids. Proteomics results demonstrated a coded protein cargo of Leuko EVs found in the lacrimal fluid reflecting that observed in PB, opening the door to a major investigation perspective in the study of EVs from tears. Likely due to the major number of events sorted in PB, we found a greater number of identified proteins in PB Leuko EVs than those identified in tears Leuko EVs. Despite this discrepancy in numbers, a similar protein cargo conveyed in tears and PB Leuko EVs was found, which fits into a context of common leukocytes biological functions. Moreover, we identified a specific significant cluster of 20 proteins in lacrimal Leuko EVs associated to “RNA binding” function, indicating as Leuko EV information about RNA loading conveyed into target cells can be better recognised in tears [57, 58] (the same function in PB was not significant and referring to 7 of the 20 proteins found in tears, see Table S7, sheets 4 and 5).

Moreover, as a proof of concept, among the cellular functions significantly described by both protein cargoes there was “leukocyte migration”, demonstrating how the here proposed FACS-Proteomics approach is highly selective in isolating and characterizing the protein cargo of circulating Leuko EVs from blood and even from lacrimal fluid (see Table S4 where many biofunctions related to Leuko EVs biogenesis were listed for total PB EV protein dataset showing less significant p-values than PB Leuko EV proteins).

As a main aspect of these results there is the opportunity to consider the lacrimal fluid EVs as a container of extractable biological information for improving molecular mechanisms knowledge and discover novel biomarkers and therapeutics targets across a wide range of pathological conditions especially looking at central nervous system affecting diseases. Previously, we determined tear total EVs protein cargo in RRMS subjects highlighting a pathological protein profile in EVs that likely referred to immune cell activation [30]. Here, we carried out the first EVs single-phenotype proteomics investigation in RRMS through a label free proteomics approach on Leuko EVs purified from tears of RRMS patients. Proteomics results demonstrated as Leuko EVs in RRMS are enriched with a specific protein cargo reflecting the activation of many factors like TGFB1 as main upstream in the parental cells that exploits EVs to spread out their biological information through proteins messengers of inflammation and angiogenesis processes as main downstream predicted by bioinformatics analysis. According to IPA prediction and hypothesis, its overexpression could determine an increase of the “migration of endothelial cells”. Blood–brain-barrier (BBB) is composed of cerebral endothelial cells, which have a central role in trans-BBB leukocyte diapedesis during normal immune surveillance as well as neuroinflammatory conditions like RRMS [59, 60], where the integrity of BBB is impaired [61]. Our data demonstrated that protein cargo of lacrimal Leuko EVs reflects on the one hand the neuroinflammatory condition of RRMS (TGFB1 and NFE2L2, upstream regulators activated) and on the other hand endothelial cell proliferation together with an increase of vascular networks (AGNPT2 and Vegf, upstream regulators activated) during RRMS due to pro-angiogenic processes. In this regard, new vessel formation has been studied in many pathological conditions including RRMS and in its animal model (i.e., EAE, experimental allergic encephalomyelitis) [62, 63]. This phenomenon is directly connected to leukocyte infiltration through the BBB and its increased permeability is driven by VEGF that regulates vessel growth and, in the meantime, has chemotactic functions against lymphocytes and monocytes, promoting neuroinflammation. Together with Vegf, another important regulated upstream gene resulted activated by a set of RRMS Leuko EV proteins is ANGPT2, which was found increased in neurons, glia and inflammatory cells during EAE [64]. In agreement with these results, we revealed an activation of angiogenesis function which have been highlighted as up-regulated downstream effects also in PI3K/AKT signaling pathway (p-value = 6.93 × 10^–3, z-score = 2.00, Figure S3).

These functional proteomics results seem to be quite similar to those obtained by proteomics investigations on total EVs sorted from tears in RRMS subjects [30], unveiling that most of proteomics data in the analysis of total tears EVs come from circulating Leuko EVs, likely due to the immune system involvement in RRMS, besides the fact that they are one of the most abundant populations of EV phenotypes in tears, influencing the total protein expression.

Here, we provided a proof of concept and a perspective on the potential and challenges of the FACS-Proteomics isolation/characterization of single EV subtypes that can be further characterized from whole biological fluids without any kind of pre-analytical manipulation step, emphasizing the use of tears, collected using the Shirmer’s strip kit usually adopted in ophthalmology routine, which have proved to be a precious source of biological information not only in ocular pathologies [26]. Of note, proteomics results could be affected (especially in blood) from the formation of protein corona that forms spontaneously around many, if not all, nanoparticles in biological fluids [65]. Due to their importance as an in vivo address code on the EV surface specificity [66], this aspect will be considered in a more focused and future investigation. Actually, the peculiar EV isolation approach and the different matrix microenvironment of different biological fluids may result in different proteins being identified as potential EV corona proteins. However, the question if these proteins were contaminants or inherent components of EVs is still open. Due to the limited number of proteins obtainable from a single tear sample, quantitative proteomics data were carried out by pooling specimens. It is also noteworthy that a larger number of purified EVs should establish deeper and more complete proteomic coverage of CD45+ EVs from tears, but this would require pooling many more samples. In fact, pooling patient samples is the main limitation of our exploratory study, as it may not accurately reflect the EV proteins pattern of individual subjects. Moreover, we believe that the pooled samples were more representative of the majority that are taking high-efficacy DMTs. Therefore, to better investigate the role of Leuko EVs in such a multifactorial and autoimmune disease, the identified modulated inflammatory pathways could be further and deeply validated in a cohort of independent RRMS patients, taking into account not only the clinical outcomes but also the different treatments.

However, the rapid progress in mass spectrometry technologies and PFC tools promises to push the limits towards on single biological sample and to enable up to many orders of magnitude increased throughput in the proteomics characterization even of less abundant circulating EVs phenotypes, giving us the opportunity to improve our knowledge of real EVs behaviour in vivo.

In this context, the potential of mass spectrometry to provide sensitive identification of biomarkers in single-phenotype EVs, as shown in this study for Leuko EVs, could be integrated and supported by other high-throughput techniques, such as quantitative polymerase chain reaction (qPCR), to ensure the accuracy and reliability of the identified protein expression profiles for each individual clinical patient profile towards EV clinical applications and personalized medicine.