Proteomic and phosphoproteomic landscape of salivary extracellular vesicles to assess OSCC therapeutical outcomes

Circulating extracellular vesicles (EVs) have emerged as an appealing source for surrogates to evaluate the disease status. Herein, we present a novel proteomic strategy to identify proteins and phosphoproteins from salivary EVs to distinguish oral squamous cell carcinoma (OSCC) patients from healthy individuals and explore the feasibility to evaluate therapeutical outcomes. Bi‐functionalized magnetic beads (BiMBs) with Ti (IV) ions and a lipid analog, 1,2‐Distearoyl‐3‐sn‐glycerophosphoethanolamine (DSPE) are developed to efficiently isolate EVs from small volume of saliva. In the discovery stage, label‐free proteomics and phosphoproteomics quantification showed 315 upregulated proteins and 132 upregulated phosphoproteins in OSCC patients among more than 2500 EV proteins and 1000 EV phosphoproteins, respectively. We further applied targeted proteomics by coupling parallel reaction monitoring with parallel accumulation‐serial fragmentation (prm‐PASEF) to measure panels of proteins and phosphoproteins from salivary EVs collected before and after surgical resection. A panel of three total proteins and three phosphoproteins, most of which have previously been associated with OSCC and other cancer types, show sensitive response to the therapy in individual patients. Our study presents a novel strategy to the discovery of effective biomarkers for non‐invasive assessment of OSCC surgical outcomes with small amount of saliva.

of effective biomarkers for non-invasive assessment of OSCC surgical outcomes with small amount of saliva.

K E Y W O R D S extracellular vesicles, proteomics, phosphoproteomics, oral squamous cell carcinoma, prm-FASEF INTRODUCTION
Oral Squamous Cell Carcinoma (OSCC) is the most common type of oral cancer with approximately 50% mortality rate [1]. Typically, OSCC is diagnosed in patients when they enter into the middle and last stages of cancer [2]. At that stage, surgical resection is the most effective way to save the patient's life [3]. Following the oral surgery, postoperative assessment is critical to provide effective follow-up therapies and a convenient and noninvasive evaluation of the operation outcomes is highly desirable.
However, as a complex biofluid, saliva contains a large number of electrolytes, high-abundant proteins, enzymes, lipids, and other molecules which present a significant obstacle to salivary biopsy [19]. To overcome the challenges of liquid biopsy using clinical body fluids as the source, a new strategy has emerged based on the profiling of extracellular vesicles (EVs) isolated from body fluids such as saliva, urine, tears, and plasma [20][21][22]. Cancer-specific moieties have been identified in cancer-derived EVs that promote cancer growth, invasion and metastasis, tumour microenvironment transformation, and angiogenesis [23].
Versatile membrane structure of EVs can maintain the stability of its cargoes and effectively reduce the complexity of body fluids, which make the analysis of EVs attractive and promising in early detection and monitoring disease status [24][25].
Despite recent efforts and good progresses, there is still no effective and practical method for salivary EV isolation and downstream cargo analysis. Current EV isolation methods include ultracentrifugation, density gradient centrifugation, immunoaffinity interaction, precipitation aggregation, ultrafiltration, microfluidic isolation, and size exclusion [26], but each method has its challenges and barriers such as low membrane integrity, highly time-consuming, high-cost, poor recovery, and low purity. Recent attempts based on chitosan [27] and EXODUS [28] were made, but their EV capture efficiency was still low in saliva. As a result, the reported biomarker discovery methods based on the proteome of salivary-EVs only had limited success with relatively low protein identification [29][30][31], and to the best of our knowledge, no study has been reported based on more demanding phosphoproteome of salivary EVs for biomarkers screening. The development of saliva biopsy based on salivary EV proteomics and phosphoproteomics is severely hampered by the inability to efficiently, quickly, and reproducibly capture saliva-derived EVs.
Herein, we introduce a rapid and non-invasive strategy based on salivary EV proteomic and phosphoproteomic analysis to distinguish OSCC patients from healthy controls and explore the feasibility of assessing the therapeutical performance of OSCC patients ( Figure 1).
We isolated salivary EVs based on a novel bi-functionalized magnetic beads. The high isolation efficiency and specificity were demonstrated using spectroscopic methods, immunoassay and mass spectrometry (MS), enabling downstream EV proteomic and phosphoproteomic analysis. After quantitative proteomic and phosphoproteomic profiling of salivary EVs from OSCC patients and healthy controls, targeted proteomics based on parallel reaction monitoring with parallel accumulation-serial fragmentation (prm-PASEF) was performed to measure panels of proteins and phosphoproteins from salivary EVs collected before and after surgical resection, resulting in six proteins and phosphoproteins with sensitive response to oral surgery.
The strategy and analytical platform provides a powerful tool for clinical applications such as evaluating surgical outcomes on individual patients. bodies, and large aggregates, and the operation was repeated once. All the samples were collected between 9:00 am and 10:00 am, and the post-surgical saliva of patients were obtained at about 1 month later after oral operation. The final supernatants were collected and frozen at −80 • C until they were used.

Salivary EVs isolation by BiMBs
The frozen saliva was thawed at 37 • C and diluted with four times volume of PBS, 0.1% NP40/TritonX-100 buffer and designated amount of BiMBs for 1 h incubation at room temperature. Then, beads were separated with a magnet and the supernatant was discarded. Subsequently, these beads were washed three times with PBS. Finally, add 200 μL (200 mM) of TEA to elute the EVs from beads (vortex oscillation 10 min) and then collect the elution into a new tube. The elution was dried using freeze dryer and used for further study.

LC-MS/MS analysis
The separation of extracted peptides was done by 30 cm long homepacked column having an inner diameter of 75 μm using C18 resin For prm-PASEF acquisition mode, all the samples were analysed using 90 min gradient at a flow rate of 300 nL/min.

Data Processing and Bioinformatics Analysis
Using PEAKS Studio X+ software, the raw files were scanned directly on serine, threonine, or tyrosine residues. The peptides evaluated from the sequence database were searched using complete trypsin/P digestion, allowing for a total of two missing cleavages. The false discovery rates (FDRs) of peptides, and phosphopeptides were all set to 1% (proteins were confined by both -10lgP ≥ 20 and ≥1 unique peptide), and PEAKs Studio X+ was also used to localize the phosphosites. There

Significance Statement
We reported here the first time to large-scale profile proteins and phosphoproteins in saliva extracellular vesicles (EVs) and demonstrate the feasibility of monitoring EV phosphoproteins to assess therapeutic outcomes with small amount of saliva. Salivary EVs were isolated by a novel bi-functionalized magnetic beads (BiMBs) with high efficiency and specificity, facilitating downstream proteomic and phosphoproteomic analyses and longitudinal measurement of individual patients before and after oral surgery.
were no repetitive entries among the protein and peptide identifications, only distinct peptides/phosphopeptides and distinct master proteins were identified.
The intensity of peptides and phosphopeptides were extracted with initial precursor mass tolerance at 15 ppm, minimum number of isotope peaks as 2, maximum ΔRT of isotope pattern multiplets -0.2 min, and peptide-spectrum match (PSM) confidence FDR of 1%.
The top three peptides were summed to calculate the intensity of a protein. As for the intensity calculation of phosphoproteins, only respective phosphopeptides were measured. The label-free quantitation (LFQ) method was used to compare proteomic, phosphoproteomic differential expression, EV proteins, as well as free salivary proteins within different samples. Based on these results, the differential expression analysis of whole proteins and phosphoproteins in clinical samples were conducted using Perseus software to generate the volcano plot and heatmap (p-value < 0.05, t test S0 = 0, |log 2 (Fold change)| > 1 was regarded as differential proteins) and the missing values were reduced by ID-transfer (match between runs) and supplemented according to the normal distribution of protein intensity in the sample.

Quantification analysis based on prm-PASEF
During the discovery stage, 315 up-regulated proteins and 132 upregulated phosphoproteins were identified by the LFQ method. Three pooled samples including healthy control pooled saliva, OSCC pooled saliva before and after surgery were analysed in DDA mode to screen for suitable unique peptides and phosphopeptides (the identification results by PEAKS Studio were exported for Skyline software to create spectral library, the peptides with FDR ≤ 1% were selected), and the choice of peptides and phosphopeptides were based on their good reproducibility, product ion coverage, and biological relevance. We also excluded known highly abundant plasma proteins in the list. In view of the timsTOF Pro mass spectrometer, with the help of ion mobility information, we can well understand the phosphorylation site isomers in the same phosphoprotein. Some phosphopeptides that could not distinguish the isomers well were eliminated, and only the phosphopeptides Overview of the strategy to the discovery of protein and phosphoprotein biomarkers from saliva EVs to assess surgical outcomes. Saliva samples from 30 OSCC patients and 30 matched healthy controls were randomly assigned to three groups (ten samples each). Two millilitre pooled saliva sample from each group (200 μL from each sample) was used for proteomic and phosphoproteomic differential expression analysis. For the prm-PASEF experiments, the pooled samples of healthy control group, OSCC patients before surgery, and OSCC patients after surgery were first tested to establish a prm-PASEF method to monitor the intensity of peptides and phosphopeptides representing the selected upregulated proteins and phosphoproteins in the preoperative samples and postoperative samples of ten OSCC patients without isomers or the phosphopeptides that can clearly distinguish the isomers will be retained. To establish the prm-PASEF method well, the pre-test method further eliminates unstable targets (precursor were also subjected to prm-PASEF experiments to further refine the target list and test its feasibility.
All of the prm-PASEF data were imported into Skyline-daily to compare the same peptide across runs and adjust the RT location manually.
The precursor charges were set to +2, +3 and the ion charge was set to +1 with b, y, p ion types. The product ions from precursor to three product ions to pick the top three most intense product ions.
All of the matched transitions were auto-selected. Prior to statistical analysis, the intensities of quantified unique peptides and phosphopeptides were normalized according to the total ion current (TIC) intensity.
Subsequently, the summation of intensities from its corresponding top three most intense product ions were calculated to represent the intensities of each unique peptides and phosphopeptides. Each protein's intensity was quantitated using the summation of intensities from all of its corresponding peptides.

Characterization of salivary EV isolation by BiMBs
We have recently introduced bi-functionalized magnetic beads (BiMBs) with Ti (IV) ions and a lipid analog, 1,2-Distearoyl-3-snglycerophosphoethanolamine (DSPE) for urine EV isolation [32]. We were interested in applying BiMBs for the isolation of salivary EVs for the first time. The high isolation efficiency and specificity due to its synergistic functionalities of EV capture were demonstrated using spectroscopic methods, immunoassay and mass spectrometry (MS), enabling downstream EV proteomic and phosphoproteomic analysis. Salivary EV isolation by BiMBs was monitored by TEM, NTA, Western Blotting, and MS, step by step ( Figure 2). Firstly, salivary EVs were captured by the functionalized magnetic beads and verified by the clear immobilization of EV membrane structure with BiMBs, as shown in Figure 2A. Isolated salivary EVs were then eluted off BiMBs to remain intact as the pristine membrane and three-dimensional structure were clearly observed by TEM shown in Figure 2B. The NTA data showed that the eluted salivary EVs have an average diameter of 139.5 nm and 8.0 × 10 10 EVs from 1 mL of saliva ( Figure 2C).
And by Western blotting assay, the intensity of EV markers, CD9, TSG101, and CD81 from EVs isolated using BiMBs was more than 4 times of which by the ultracentrifugation method ( Figure 2D). EV proteins and phosphoproteins were analysed by LC-MS/MS, and the number of unique peptides, protein groups, phosphopeptides, and phosphoproteins identified from the BiMB method was almost twice of the ultracentrifugation method, as presented in Figure 2E (and Table   S1-4). Furthermore, the selectivity and specificity of BiMB isolation was evaluated by comparing the intensity ratios of known EV proteins identified by BiMB and ultracentrifugation method, including exosome proteins in ExoCarta and other proteins reported in MISEV2018 [33], and three free salivary proteins as contaminating proteins. The enrichment of EV proteins by BiMB method enhanced the intensities of the selected EV proteins more than 15 times on the average and about 38 folds higher in case of protein ACTN4 relative to ultracentrifugation, however contaminating proteins' intensity only increased to 1.8 times on the average ( Figure 2F). These results demonstrated the efficient capture of salivary EVs by BiMBs, and the method has clear advantages for the downstream proteomics and phosphoproteomics analyses.  (Table S5 and Among these up-regulated total proteins, a number of proteins such   as IGF-II, MMP1, MMP3, MMP13, EBP-1, uPA, FSCN1, CA-II, CD44, HMGCS2, PAI-1, and so on have previously been reported to be associated with OSCC [11,[34][35]. In addition, some of the proteins have also been previously discovered to have association with lung cancer, bladder cancer, SARS, oral submucous fibrosis, colorectal cancer, cervical cancer, and kidney cancer [36][37][38][39][40][41].

Quantitative measurement of EV proteins and phosphoproteins in OSCC patients
The rest of peptides from each sample were processed to for phosphopeptide enrichment, followed by LC-MS/MS analyses. Collectively, 1119 phosphoproteins were identified from healthy groups as depicted in Venn diagram ( Figure S3A-B), while 1095 phosphoproteins were identified in the patient groups as shown in Venn diagram ( Figure S3C and Table S7 Table S10) and PPI analysis ( Figure S4). These 132 phosphoproteins are mostly related to the plasma membrane, cell junction, tau protein binding, activation of MAPK activity, MAP / protein tyrosine/ protein serine/threonine kinase activity, regulation of telomerase activity, and associate with the regulation of early endosome to late endosome transport. The KEGG pathway analysis indicates these phosphoproteins participate in the regulation of proteoglycans/ choline metabolism in cancer, and Rap1/ErbB/Chemokine/PI3K-Akt/mTOR/HIF-1 signalling pathway, and even take part in several cancer types. Last but not least, some of the upregulated phosphoproteins such as FLNA, IBP-3, S100-A9, and others were proposed as potential biomarkers for OSCC screening in previous studies [42][43].

Evaluation of OSCC surgical outcomes through prm-PASEF
To explore the feasibility of using proteins and phosphoproteins from saliva EVs for therapeutic evaluation, we performed targeted proteomics and phosphoproteomics by prm-FASEF [44][45]. With the accumulation of ionic intensity in the scheduled PRM mode, we could simultaneously measure a large number of peptides in a single prm-FASEF run with suitable targets/per PASEF event and PASEF events/per MS cycle. An independent cohort of 10 OSCC patients were selected to collect their saliva samples before surgery and about 1 month after the surgery (Figure 1).  Table S11) and 112 phosphopeptides (represents for 115 precursor ions, 55 phosphoproteins, Table   S12) were monitored using prm-PASEF methods, respectively. Quantitative measurements of these selected peptides and phosphopeptides are shown in Table S13 and S14 and the proteins were quantified as the intensity summation of their corresponding peptides (Table S15 and S16).
We used all the quantified full proteins and phosphoproteins selected in prm-PASEF results to screen for the feature proteins ( Figure 5A), and out of these we identified ten proteins (green colour) with the distinctive change from pre-surgical to post-surgical states ( Figure 5B). An initial ROC analysis indicated effectiveness of the panel of proteins in evaluating clinical salivary samples ( Figure S5). Among them, several proteins have previously reported as potential biomarkers for OSCC and other cancers. To further evaluate the effectiveness of saliva proteins for surgical outcomes by taking into account the individual differences of the samples, we quantified the change of these ten proteins before and after operation in each individual patients and the corresponding heatmap of the quantified intensity of these F I G U R E 5 Targeted proteomics to evaluate therapeutic outcomes. (A) Feature proteins (including whole proteins and phosphoproteins) assessment and selection to distinguish pre-surgical samples from post-surgical samples. (B) Ten proteins showing distinctive changes before and after surgery. (C) Heat map illustrating changes in the relative intensity of these ten feature proteins in individual patients. (D) Box-whisker plots of the intensity changes of selected proteins and phosphoproteins with significant intensity reduction among more than 7 of these 10 OSCC patients after surgery and high correlation with OSCC or other oral cancers ten proteins are shown in Figure 5C. Among the ten proteins, six proteins including three full proteins (HEP2, NHERF-2, and MMP25) and three phosphoproteins (PGM 1, ACLY, and KPCD), were sensitive to the surgery in at least 8 out of the 10 patients. The Box-whisker Plots show the intensity changes of four selected proteins and phosphoproteins before and after surgery in individual patients ( Figure 5D). HEP2 (Heparin cofactor 2), a previously reported biomarker for OSCC [34] and other cancers [46], has the most sensitive response to the operation and all ten patients had decreased intensity after surgery, compared to the intensities detected in pre-surgical samples. MMP25 (Matrix metalloproteinase-25) anchors to the membrane of EVs via a glycosyl-phosphatidyl inositol (GPI) anchor [47]. While it belongs to the metalloproteinase family, it, along with MMP17, has a different set of functions and regulatory mechanisms from other members of the MMP family, and has been reported to be highly associated with modulating immune infiltration levels and survival outcomes in patients with head and neck cancer [48]. Phosphorylation of ACLY (ATP-citrate synthase) has been shown to affect its activity which is critical in cholesterol and fatty acid synthesis and tumour cell growth [49]. KPCD (Protein kinase C delta type) and its phosphorylation was shown to have influence on the interaction of Prostaglandin E2/EP1 signalling pathway to enhance the oral cancer cell motility [50].

CONCLUSION
Saliva is an intriguing source of biomarkers for in vitro diagnosis of diseases such as OSCC. In this study, we developed bi-functionalized magnetic beads for efficient isolation of saliva EVs, facilitating downstream proteomics and in particular, highly demanding but rich infor- EVs from 10 OSCC patients before and after operation, we identified at least six proteins, three total proteins and three phosphoproteins, that are sensitive to most individuals. These potential total proteins and