Single Extracellular Vesicles (EV) Proteomic Proling Altered and Identies Co-Localization of SARS-CoV-2 Nucleocapsid Protein with CD81/Integrin-Rich EV Subpopulation in Sputum Samples of COVID-19 Patients

Understanding the pathogenesis of SARS-CoV-2 is crucial to respond to the current coronavirus disease 2019 ( COVID-19 ) pandemic. Sputum samples from 20 COVID-19 patients and healthy controls were collected, respectively. During the isolation of infectious SARS-CoV-2 virus, EV-like vesicles were associated with virions under a transmission 16 electron microscope. Next, the expression of IL6 and TGF- β increased in EVs derived from the 17 sputum of patients, and these were highly correlated with the expression of the SARS-CoV-2 N 18 protein. Further, proximity barcoding assay (PBA) was used to investigate the immune-related 19 proteins in the EVs, and the relationship between EVs and SARS-CoV-2 N protein in 20 COVID-19 patients’ samples. Particularly, to investigate the differential contribution of the 21 specific EV subsets, the protein expression of a single EV was detected and analyzed for the 22 first time. Among the 40 EV subpopulations, 18 were found to have significant differences. The EV subpopulation regulated by CD81 were most likely to correlate with the changes in the pulmonary microenvironment after SARS-CoV-2 infection. This study provides evidence on the association between EVs and the SARS-CoV-2 virus, give a deep insight into the possible pathogenesis of SARS-CoV-2 infection and the possibility of nanoparticles drug intervention in viral infection.


Patients with COVID-19 secreted more proteins in individual EV, and EVs participated in the 98
immune response 99 The EV and virus co-expression proteins were identified using the PBA method. The 100 scheme of the workflow is illustrated in Figure 2A. The antibody-conjugated oligonucleotides 101 were brought into the proximity on the same EV due to the protein-antibody interaction, 102 thereby obtaining the same EV tag barcoding [13]. EVs obtained from different sources were 103 characterized by the presence of specific combinations of surface proteins and their abundance, 104 allowing each EV to be quantified in the mixed samples, to serve as markers for specific 105 engagement in the disease. After library construction and sequencing, the original data were 106 obtained in fastQ file format. After quality control and tag extraction, the file of the identified 107 individual EVs and detected proteins are summarized for each sample. 108 COVID-19 infection (nCOV, n=20), healthy controls (HC, n=20) and PBS negative controls 115 (PBS,n=4). As shown in Figure 2C, SARS-CoV-2 N protein signals could be detected in EVs 116 obtained from COVID-19 patients ( Figure 2C). In the control group, the protein signal of 117 some individuals was slightly higher than that of PBS, which was considered as an acceptable 118 systematic error and antibody nonspecific binding. In the subsequent analysis, the 119 SARS-CoV-2 N protein was centralized, and the data for SARS-CoV-2 N expression of the 120 control group was set as 0 for the data processing of the nCOV group. By the summation of signals of each detected protein on all EVs of the sample, the EV associated protein expression 122 of each sample was obtained. 123 Consistent with previous reports on the expression of cytokines in serum [14], the 124 expression levels of IL6 and TGF-β were also increased in EVs of  Furthermore, this elevation was highly correlated with the SARS-CoV-2 N protein expression 126 ( Figure 2D). We also identified other proteins that were significantly increased after 127 SARS-CoV-2 infection, including T-cell activation marker CD26, human leukocyte antigen 128 HLA-A, and adhesion molecule MAdCAM-1 (mucosal addressing cell adhesion molecule-1), 129 which were overexpressed in inflammatory mucosal tissues (Fig. S1). These results show that 130 EVs were involved in the immune response to although 131 immunoglobulin A (IgA) might be higher than IgM (consistent with the findings in serum 132 [15]), there was no significant difference in the total expression of IgA and IgM in sputum 133 EVs of healthy controls and patients with COVID-19 ( Figure 2E). After TMM normalization, 134 protein expression heatmap ( Figure 2F) showed that nCOV patients have a general shift of 135 EV proteomic profile compared to HC samples, although with exceptions. Differentially 136 expression proteins were analyzed in a volcano plot after normalizing TMM protein 137 expression data and then generated the dot plot (Fig. S2). Compared to HC group, the 138 abundance of TROP-2, CD36, EGFR, IgA and IgM decreased in nCOV group while CDH1 139 (E-cadherin1), ZEB-1 and ZO-1 increased. 140 141 EV subpopulations atlas and the featured change in patients with  The algorithm FlowSOM was used to analyze the behavior of all markers on all individual 143 EVs, and the clusters of EVs were generated using a self-organizing map. The clusters, which 144 were the EV subpopulations, were determined according to proteomic fingerprints of each EV. 145 The investigators detected 9,377,119 EVs with an average of 234,428 EVs per sample ( Figure 2B). The dimensionality reduction indicated the substantial phenotypic similarity and 147 differences between patients with COVID-19 and controls. The t-distributed stochastic 148 neighbor embedding (tSNE) plot for each sample is shown in Figure 3A, which identified 40 149 clusters. The clustering of individual EVs obtained from all samples was displayed in the tSNE 150 plot, in which 40 clusters were color labelled ( Figure 3B). 151 Next, a modeling approach was employed to detect the features that distinguish healthy 152 individuals from infected individuals. Figure 4A shows the similarity and differences in EV 153 proteomics between the sputum sample of the nCOV and HC groups, in which all EVs in the 154 nCOV group were colored green, while EVs in the HC group were colored red. Through these 155 red and green markers, the different subpopulations of EVs between these two groups can be 156 more intuitively distinguished ( Figure 4A). The proteomic similarity of EVs was observed in 157 the tSNE plot. The proportion of each subpopulation was quantified. Among the 40 158 subpopulations, 18 clusters were found with significant differences. These were cluster 2, 3, 4, 159 6, 7, 9, 10, 12, 13, 14, 32 and 34, which had a significantly elevated ratio of EV subpopulations 160 in the nCOV group, while the ratio for the subpopulation decrease in cluster 16, 21, 22, 26, 27 161 and 33 (Figure 4B and S3). 162 Among these, cluster 2, 3, 4, 12, 13, 34 and 33 accounted more in the distribution of the 163 subpopulation with differences, and the difference was particularly significant. We further 164 analyzed the seven groups, and the proteomic fingerprints for each subpopulation were profiled. 165 For each differentially expressed EV subpopulation, the location in the total EVs and the top 7 166 featured proteins are shown in Figure 5. First, we were concerned about cluster 2, which 167 constituted 4.92% of all EVs in the nCOV group, and only 0.55% of all EVs in the HC group 168 ( Figure 5A and S3). The EVs in cluster 2 contained a large amount of  These should be the EVs directly contacted with SARS-CoV-2 or secreted by the cells 170 responsible for viral replication. These EVs highly expressed exosome biomarker CD81, and the following cell adhesion molecules: epithelial cell adhesion molecule (EpCAM), CDH1, 172 ITGB4, ITGA5, SNAI2, etc. Figure 5B shows the proteomic characteristics of cluster 2 (100 173 proteins). Clusters 3, 4, 12, 13 and 34 are EVs that increased in the nCOV group, with highly 174 expressed protein CLEC2A, CD81, ITGB3, CD151 and ITGB2, respectively ( Figure 5C). In 175 contrast, cluster 33 was reduced, and comprised 3.37% of the EVs in the nCOV group, while 176 this comprised of 14.12% of EVs in HC group (Fig. S3). Cluster 33 featured a higher 177 expression of EGFR and IgA ( Figure 5D). The proteomic profiles are shown in Figure 5D. individuals. The quantity of each possible pair of co-expressed proteins was obtained as the 185 protein combination dataset, and this was used as input variables for the abundance and 186 differential analysis ( Figure 6A). The protein combinations exhibited a universal increasing 187 trend in the nCOV group, except for the combinations of EGFR and IgA. The co-expression 188 between the integrin subgroups significantly increased. 189 To investigate the co-localization of viral protein with EVs, we further analyzed the 190 combinations of SARS-CoV-2 N protein with other proteins on individual EVs. Among the 191 markers that regulate EVs, the co-expression of CD9, CD63, CD81 and Alix with 192 SARS-CoV-2 N were calculated ( Figure 6B), and found that the EVs regulated by CD81 were 193 more likely to bind to the SARS-CoV-2 N protein ( Figure 6C). In addition, we found that 194 cluster 2 ( Figure 5B), cluster 4 ( Figure 5C), cluster 6, cluster 7, cluster 12 and cluster 34 were 195 the EVs that were highly expressed after SARS-CoV-2 infection, while CD81 was highly expressed in both clusters. In particular, the protein matrix of cluster 4 shows that the 197 expression of CD81 is the most and abnormally high ( Figure 6D). These results suggest that 198 the EVs regulated by CD81 are the most likely subpopulations of EVs that cause the changes in 199 the pulmonary microenvironment after SARS-CoV-2 infection. The protein expression 200 distribution of CD81(red) and SARS-CoV-2 N (blue) in all EV and the green part represented 201 the co-expression region ( Figure 6E). 202

203
Discussion 204 In the current study we isolated and identified EVs from the sputum of COVID-19 patient 205 to investigate EV inflammatory and immune responses in COVID-19 patients. We found 206 EV-like vesicles that coexisted alongside virions (Figure 1A), and the mean number of EVs 207 showed an increasing trend after the SARS-CoV-2 infection ( Figure 2B). The nucleocapsid 208 protein of SARS-CoV-2 (SARS-CoV-2 N) is an important structural protein, which is located 209 in the core part of the virus particle, and binds to the viral RNA, playing an important role in the 210 process of virus packaging and other process [16,17]. As expected, the SARS-CoV-2 N 211 protein can be detected in EVs obtained from patients with COVID-19 infection: when the 212 mean signal value in the control group was taken as the baseline (the blue dotted line in Figure  213 2C), 19 of 20 individuals were detected for SARS-CoV-2 N protein in EVs; when the 214 maximum signal value in the control group was taken as the baseline (the red dotted line in 215  (Figure 2D), which is consistent with the previous results in 227 peripheral blood [19]. Furthermore, the expression of IL-6 and TGF-β was highly consistent 228 with that of the SARS-CoV-2 N protein in EVs. Meanwhile, we found that most integrins and 229 other adhesion molecules were also upregulated ( Fig. S1), which could jointly influence the 230 interaction of immune cells with the local microenvironment [20,21]. All these indicate that 231 EVs are engaged in the immune response to COVID-19 infection. Secretory immunoglobulin 232 A (IgA) play an important role in the protection and homeostatic regulation of the respiratory 233 mucosal epithelium, which is referred to as "immune exclusion" [15]. However, there was no 234 significant difference in the total expression of IgA in sputum EVs before and after the 235 infection with SARS-CoV-2 ( Figure 2E). It was considered that there should be differences 236 in IgA in some EV subpopulations, but this could not be reflected in the analysis of the total 237

protein. 238
Toobtain the protein expression of a single EV, the algorithm FlowSOM was applied to 239 analyze the behavior of all markers on all individual EVs, and generate the clusters of EVs 240 using a self-organizing map. The clustering of individual EVs obtained from all samples was 241 displayed in the tSNE plot, in which 40 clusters were color labelled ( Figure 3B). After 242 quantifying the proportion of each subpopulation, we found that there were significant 243 differences in 18 clusters (Fig. S3). 244 The EVs in cluster 2, which constituted 4.92% of all EVs in the nCOV group and only 245 0.55% of all EVs in the HC group ( Figure 5A and S3), contained a large amount of protein SARS-CoV-2 N. It was considered that these are EVs that directly transport SARS-CoV-2. In 247 cluster 2, epithelial cell adhesion molecule (EpCAM), CDH1, ITGB4, ITGA5, SNAI2, CD81,  248   ITGB2 The present results merely confirm that these were all highly expressed in cluster 2. Therefore, 257 we boldly speculate that the high expression of adhesion proteins, such as EpCAM and 258 CDH1 (Figure 2F and S2 35]. We found that the EVs regulated by CD81 were more likely to bind to the SARS-CoV-2 N 282 protein ( Figure 6C). In addition, it is known that cluster 2 (Figure 5B), cluster 4 ( Figure 5C), 283 cluster 6, cluster 7, cluster 12 and cluster 34 (Fig. S3) are EVs with an upregulated protein 284 expression after SARS-CoV-2 infection, and that CD81 is highly expressed in all these clusters. 285 These results suggest that the EVs regulated by CD81 are the most likely subpopulations of 286 EVs that cause the changes in the pulmonary microenvironment after SARS-CoV-2 infection. 287 Furthermore, it was found that hepatitis C virus (HCV), which have been extensively 288 studied, enters the host cell through interactions with a cascade of cellular factors，such as 289 CDH1, claudin-1(CLDN1), and occludin (OCLN) [22]. Subsequently, it can be be recognized 290 and absorbed by recipient cells. Unexpectedly, these transmission processes may be similar to 291 those of SARS-CoV-2, the EVs regulated by CD81 also highly expressed CDH1. The 292 difference is that EGFR is not required for the transmission of the SARS-CoV-2 ( Figure 6A). 293 HCV uses a dynamic and multi-step process to engage and enter host cells, in which EGFR is 294 necessary for internalization [36].
In conclusion, we found that EVs (mostly regulated by CD81) can carry the 296 SARS-CoV-2 N protein, and its expression is highly correlated with that of inflammatory 297 factors in EVs. These results demonstrate that EVs derivedthe from sputum of patients may 298 participate in the infection and immune response of COVID-19. The mechanism of the HCV 299 infection, and subsequently internalizing these into recipient cells, might have some 300 similarities to the relationship between EVs and SARS-CoV-2 infection, giving us many 301 hints. This can provide some information for the further study of COVID-19 and promote our 302

Patient and healthy donor selection and inclusion criteria 308
The present study was approved by the Ethics Committee of the First Affiliated Hospital of 309 Guangzhou Medical University (Guangzhou, China). Written informed consent was obtained 310 from all study participants. A sample ID was applied to ensure sample tracking with 311 confidentiality on sample donor identity. The healthy control (HC) group included 20 healthy 312 donors without symptoms, such as cough, allergy, respiratory tract discomfort, and so on. The 313 nCOV group included 20 patients with RT-PCR confirmed infection of SARS-CoV-2. The 314 clinic-pathological conditions of patients included in the present study are shown in Figure 1A. 315 The HC group consisted of 15 males and 5 females, with an average age of 56.6 years old. The 316 nCOV group consisted of 14 males and 6 females, with an average age of 56.7 years old. 317

Detection of SARS-CoV-2 for the diagnosis of COVID-19: 319
The presence of the SARS-CoV-2 was detected by real-time RT-PCR methods [11]. 320 Nucleic acid was extracted from respiratory samples and sputum using a Viral RNA extraction 321 kit obtained from Daan Gene Co., Ltd. (Guangzhou, China). The RNA extraction from sputum 322 and blood was performed using a total RNA extraction kit obtained from Sangon Biotech 323 (Shanghai, China). The real-time PCR assay kit for targeting the SARS-CoV-2 RdRp and N 324 gene regions was provided by Daan Gene Co., Ltd. 325 326

Sputum sample collection and pretreatment 327
The sputum of patients was collected during a pulmonary exacerbation, and in a stable 328 condition. Similarly, the sputum of normal people (HC group) was induced by the inhalation of 329 hypertonic (NaCl 5%) or isotonic (NaCl 0.9%) saline. The sputum samples were observed 330 under a microscope to ensure the qualified samples. The standard is that the ratio of white 331 blood cells to squamous cells is greater than 2.5. The sputum samples were dispersed with PBS 332 at a ratio of 1:3, and centrifuged at 500 g to remove the cells, cell debris, and aggregates. All 333 operations were performed in biosafety laboratories, and the samples were aliquoted and 334 stored at -80°C until analysis. 335 336

Virus Isolation and Transmission Electron Microscopy 337
Vero E6 cells were used for virus isolation. A quantitative reverse transcription PCR 338 (qRT-PCR)-positive sputum swab specimen was saved in viral transport media (DMEM 339 containing 1% bovine serum albumin, 15 µg/mL amphotericin, 100 units/mL penicillin G, and 340 100 μg/mL streptomycin). Before virus isolation, the sample was filtered with a 0.45-μm 341 strainer and diluted 1:10 with DMEM containing 2% FBS and antimicrobial drugs. Cells were 342 infected at 37°C for 1 h. The inoculum was removed and replaced with a fresh culture medium.
after 72 h but not in mock-infected cells. Culture supernatant was negatively stained and 345 visualized by transmission electron microscopy. 346

Characterization of extracellular vesicles in the sputum sample 347
After collecting the induced sputum, the EVs were isolated from the sputum supernatant 348 using a standard ultracentrifugation protocol after initial extraction using the EV extraction kit 349 EVs. Then, the wells of the plate were rinsed for three times with PBST before further tests. 365 subpopulation for the nCOV and HC groups. Compared to the HC group, the nCOV group had 480 a significantly high ratio of EV subpopulations, which was indicated as cluster 2, 3, 4, 6, 7, 9, 481 10, 12, 13, 14, 32  Nanoparticle Tracking Analysis of particles in the sputum samples of COVID-19 patients (50μl of EVs were extracted from 100μl of sputum, and detected after 200-fold dilution). In the inset, the size distribution (black lines) of EVs is depicted (no magnet was employed). The error bars (in red) indicate the standard error of the mean.

Figure 2
Patients with COVID-19 secreted more proteins in individual EV and EVs participated in the immune response. (A) The Proximity Barcoding Assay (PBA) for the analysis of the protein pro le on a single EV level. A lipid membrane binding layer captured the EVs of each sample; that is, streptavidin-biotin-CTB, coated on the wall of a 0.2ml well. After xation with formaldehyde and permeabilization with Triton X, the 100-antibody panel of the barcoded antibodies (SI. Table 1) were applied to interact with the proteins on individual EVs. The PBA templates endow each EV with a speci c 16-bp EV tag through extension reaction. The DNA sequence results were prepared into the sequencing library, and read out through high throughput sequencing using Illumina NextSeq 500. (B) The quanti cation of EVs and proteins detected in the PBA, and the ratio of proteins per EV from COVID-19 patients (nCOV group) and healthy controls (HC group) were plotted. (C) SARS-CoV-2 N protein signals can be detected in EVs obtained from nCOV, HC and PBS negative controls.   EV subpopulations atlas and the featured change in patients with COVID-19. (A)The similarity and differences in EV proteomics between the sputum samples in the COVID-19 and HC groups are shown, in which all EVs in the nCoV group were colored green, while the EVs in the HC group were colored red. (B) The quanti cation of the EV subpopulation for the nCOV and HC groups. Compared to the HC group, the nCOV group had a signi cantly high ratio of EV subpopulations, which was indicated as cluster 2, 3, 4, 6,