Extracellular vesicles from Zika virus‐infected cells display viral E protein that binds ZIKV‐neutralizing antibodies to prevent infection enhancement

Abstract

Mosquito‐borne flaviviruses including Zika virus (ZIKV) represent a public health problem in some parts of the world. Although ZIKV infection is predominantly asymptomatic or associated with mild symptoms, it can lead to neurological complications. ZIKV infection can also cause antibody‐dependent enhancement (ADE) of infection with similar viruses, warranting further studies of virion assembly and the function of envelope (E) protein‐specific antibodies. Although extracellular vesicles (EVs) from flavivirus‐infected cells have been reported to transmit infection, this interpretation is challenged by difficulties in separating EVs from flavivirions due to their similar biochemical composition and biophysical properties. In the present study, a rigorous EV‐virion separation method combining sequential ultracentrifugation and affinity capture was developed to study EVs from ZIKV‐infected cells. We find that these EVs do not transmit infection, but EVs display abundant E proteins which have an antigenic landscape similar to that of virions carrying E. ZIKV E‐coated EVs attenuate antibody‐dependent enhancement mediated by ZIKV E‐specific and DENV‐cross‐reactive antibodies in both cell culture and mouse models. We thus report an alternative route for Flavivirus E protein secretion. These results suggest that modulation of E protein release via virions and EVs may present a new approach to regulating flavivirus‐host interactions.

An optimized purification protocol shows that ZIKV‐infected cells secrete noninfectious extracellular vesicles that relieve antibody‐dependent enhancement of flavivirus infection.

Introduction

Zika virus (ZIKV), a mosquito‐borne flavivirus, was first isolated from a febrile rhesus monkey in the Zika forest, Uganda, in 1947 (Dick et al, 1952). Before the 21^st century, cases of ZIKV infection were sporadic in Africa and Southeast Asia (Faye et al, 2014). After earlier outbreaks in Micronesia in 2007 and subsequent epidemics in French Polynesia in 2013–2014, a ZIKV pandemic emerged in South and Central America, which evolved an international public health emergency (Duffy et al, 2009; Cao‐Lormeau et al, 2014; Musso et al, 2018). Although approximately 80% of cases are asymptomatic (Pierson & Diamond, 2018), and most others have only mild symptoms, rare neurological complications of ZIKV infection have been reported, including Guillain–Barré syndrome in adults (Oehler et al, 2014; Roze et al, 2016) and microcephaly in newborns (Calvet et al, 2016; Mlakar et al, 2016).

Antibody‐dependent enhancement (ADE) is another consideration in infections with flaviviruses such as Dengue (DENV; Guzman & Harris, 2015) and ZKV. Antibody responses to flavivirus infection mostly target the envelope protein (E), a surface‐displayed type II fusion protein that mediates viral entry. ADE is triggered by homotypic E‐specific antibodies in subneutralizing concentrations or cross‐reactive antibodies from heterotypic flavivirus infection, enhancing infection with related viruses or serotypes and exacerbating disease. ADE can occur with natural antibody responses and with application of therapeutic antibodies that otherwise protection against infection (Wang et al, 2016, 2019; Rey et al, 2018). Because of this, vaccine and therapeutic antibody development should consider ways to prevent or mitigate ADE (Williams et al, 2013; Barba‐Spaeth et al, 2016; Dejnirattisai et al, 2016; Katzelnick et al, 2017, 2020a,b; Fowler et al, 2018; Langerak et al, 2019; Shim et al, 2019), especially to avoid severe disease such as dengue haemorrhagic fever and dengue shock syndrome (DHF and DSS) in infants born to dengue‐immune mothers.

Extracellular vesicles are cell‐secreted enveloped particles that are predominantly 50–200 nm in diameter and are released from the plasma membrane and endosomal compartments. EVs carry proteins, lipids, and nucleic acids from the parent cell and play important roles in many aspects of human health and disease, including reproduction and development, immune responses and infection, cancer and neurodegeneration (Tkach & Thery, 2016; Pegtel & Gould, 2019; Kalluri & LeBleu, 2020). As an agent of cell–cell communication, EVs mediate the transport of signals and molecules (miRNA/mRNA/DNA/protein cargo) and may impact intracellular biological processes upon uptake by target cells (Mathieu et al, 2019; Mukhamedova et al, 2019; Nandakumar et al, 2019; Perez et al, 2019; Schmittgen, 2019). EVs also have a close relationship with viruses. Enveloped viruses use EV biogenesis pathways for their release, and EVs have been reported to block, enhance, or even mediate infection by certain viruses, induce innate immune responses, and help viruses avoid neutralizing antibodies (Crenshaw et al, 2018). Despite numerous recent studies of EVs in mosquito‐borne flavivirus infection (Nolte‐'t Hoen et al, 2016; Perez et al, 2019; Sung et al, 2019; Conzelmann et al, 2020; Freitas et al, 2020; Martinez‐Rojas et al, 2020), research based on rigorous separation of flavivirions and EVs, which overlap in size and density (Thery et al, 2018; McNamara & Dittmer, 2020), is still needed.

In this study, using ZIKV as a model virus, we combined differential ultracentrifugation and immuno‐capture (IC) through EV surface CD9 to extract EVs from culture supernatant of virus‐infected cells. Our data demonstrated that purified EVs do not mediate ZIKV infection but carry viral RNA and proteins. Abundantly loaded ZIKV E protein on the surface of EVs was recognized by ZIKV‐neutralizing antibodies. ADEs mediated by neutralizing antibodies and dengue‐infected patient sera were reduced in the presence of EVs derived from ZIKV or DENV infection, indicating that EVs display virion similar immunological landscape.

Results

EVs harvested by ultracentrifugation display ZIKV infectivity

It has been proposed that EVs from flavivirus‐infected cells can mediate virion‐free viral infection (Ramakrishnaiah et al, 2013; Vora et al, 2018; Zhou et al, 2018, 2019). However, interpretation of such results is difficult since ultracentrifugation pellets contain both virions and EVs, and the two are difficult to separate from each other because of overlapping size and density. Human umbilical vein endothelial cells (HUVECs) support both robust ZIKV infection and high expression of CD9, which is generally present on EV surface. For these reasons, we select HUVECs to study the role of EVs in ZIKV infection. We first revisited the molecular composition of EVs harvested by ultracentrifugation (UC‐EVs; Fig 1A). Prior to EV collection, sequential centrifugations at 700 g, 2,000 g, and 10,000 g were performed to remove dead cells, cell debris, and any large vesicles (Appendix Fig S1A). These steps rarely remove ZIKV infectious virions from viral supernatant (Appendix Fig S1B). Electron microscopy (EM) performed on UC‐EVs (100,000 g pellets) showed that their sizes ranged from 50 to 200 nm in diameter (Fig 1B), similar to EVs isolated from other mammalian sources (Hoog & Lotvall, 2015; Zabeo et al, 2017). Ultracentrifugation recovered < 5% infectivity, and the rest of infectivity remains in the post‐centrifugation viral supernatant (Appendix Fig S1B). Western blot analysis showed that ZIKV structural proteins (E, capsid) were present in both infected cells and UC‐EV pellets. Alix, HSP70, TSG101, syntenin, CD63, CD81, and CD9 were detected in UC‐EV pellets (Fig 1C), suggesting successful recovery of EVs. For finer separation of EVs from virions, we next subjected concentrated ZIKV culture supernatant (loaded on the bottom to avoid interference of concentrated serum proteins) and UC‐EVs (loaded on the top) to iodixanol gradient ultracentrifugation. From the concentrated ZIKV culture supernatant (Fig 1D and E), the viral genome and infectivity were enriched in the heavy fractions (1.133–1.217 g/ml), whereas the EV markers were located in fractions with densities ranging from 1.072 to 1.33 g/ml. ZIKV E protein was present in a broader fraction range (from 1.049 to 1.245 g/ml). Based on infectivity versus genome ration, the heavy fractions (from 1.191 to 1.217) are more infectious, and they are devoid of significant co‐recovery with EV markers. Virions from these fractions were termed as EV‐free virons. From UC‐EVs (Fig 1F and G), the ZIKV genome and infectious particles were detected in the fractions ranging from 1.026 to 1.189 g/ml; additionally, ZIKV E was present in the fractions (1.043–1.119 g/ml) in which the EV markers were located. To summarize, EV preparations obtained by using sequential ultracentrifugation from ZIKV culture supernatant exhibit ZIKV infectivity, and the density range of EVs (1.04–1.12 g/ml) overlaps that of ZIKV infectious virions (1.05–1.25 g/ml). To note, sequential ultracentrifugation method adapted for EV enrichment could not recover the majority of infectious virions from viral supernatant.

Figure 1. EVs harvested by ultracentrifugation exhibit ZIKV infectivity.

• A
  Schematic diagram of viral supernatant concentration and EV isolation by traditional ultracentrifugation and further purification by density gradient centrifugation. HUVEC culture supernatant (240 ml) was harvested at 72 h post ZIKV infection (MOI = 0.01). For viral supernatant concentration, post a prior filtration through 0.45 um filter, 100 ml ZIKV‐containing supernatant was concentrated to 1 ml by ultrafiltration (100 kDa cutoff). For preparation of UC‐EVs, 240 ml viral supernatant were centrifuged at 700 g for 10 min, 2,000 g for 20 min, and 10,000 g for 30 min to deplete debris and large vesicles, followed by ultracentrifugation (100,000 g for 2 h). The pellet‐containing EVs were resuspended in 0.6 ml of PBS. The viral concentrates and UC‐EVs were loaded on odixanol/PBS gradients, followed by ultracentrifugation (250,000 g for 18 h). Density fractions (0.3 ml each) were harvested and subjected to subsequent tests.
• B
  Transmission EM (TEM) observations of negatively stained UC‐EVs. The size distribution of vesicles was evaluated by counting the number of EVs in each size range. N indicates total number of counted EVs.
• C
  Western blots of UC‐EVs derived from ZIKV‐infected cells. Antibodies against ZIKV structural proteins (capsid, E) and EV components were used. The total cell lysates served as a control.
• D, E
  Density gradient centrifugation analysis of virus‐containing supernatant. Forty milliliters of culture supernatant was harvested at 72 h post ZIKV infection (MOI = 0.01) and concentrated to 1 ml, followed by gradient centrifugation as described in the methods. Fractions were analyzed by Western blot with the indicated antibodies (D). The ZIKV titer of each fraction was measured by plaque assay, and the viral genome was determined via RT‐PCR (E). Data represent the means from three biological replicates; error bars represent standard deviations from the means. Dashed line represents infectivity detection limit.
• F, G
  Density gradient centrifugation analysis of UC‐EVs. UC‐EVs (0.1 ml) were loaded onto the top of a density gradient in a centrifugation tube as described in the methods. The fractions were obtained from the top of gradient and analyzed by Western blot with the indicated antibodies (F), the ZIKV titer of each fraction was measured by plaque assay, and the viral genome was quantified via RT‐PCR (G). Data represent the means from three biological replicates; error bars represent standard deviations from the means. Dashed line represents infectivity detection limit.

Source data are available online for this figure.

Affinity‐purified EVs carry viral RNA but do not transmit infection

Ultracentrifugation‐harvested EV preparations from ZIKV‐infected cells exhibit infectivity. However, due to the evident overlapping density ranges of EVs and infectivity, infectious virions could be precipitated after ultracentrifugation. To determine whether EVs are able to transmit ZIKV infection, a method supporting specific purification of EVs from ZIKV culture supernatant and avoiding co‐harvested ZIKV virions is required. CD9 is highly expressed in HUVEC cells and is located on both cell membrane and multivesicular bodies. Nanoflow cytometry analysis of UC‐EVs from naïve HUVEC cells showed that approximately 50% particles are successfully labeled with CD9 antibody, suggesting more than 50% of EVs display CD9 (Appendix Fig S2A and B). Therefore, UC‐EVs were further subjected to immuno‐capture (IC) by using a homemade monoclonal antibody against large extracellular loop of CD9, followed by CD9‐competing peptide mediated elution. From HUVECs, UC‐EVs were subjected to anti‐CD9 affinity capture. After six washes to deplete the unbound materials, IC‐EVs were harvested via elution with CD9‐competing peptide (Fig 2A). We first tested this combined method for EV purification from uninfected cells. Characterization of these purified EVs was conducted by Western blot, EM analysis, and density gradient ultracentrifugation (Appendix Fig S2C–E). The results demonstrated successful capture of EVs, as evidenced by efficient EV markers' depletion from UC‐EV suspension after CD9‐capture, robust recovery of EV markers in IC‐EV fraction, EV morphology, and the density profile (Appendix Fig S2C–E). Next, we performed an analogous purification and harvested IC‐EVs from ZIKV infection culture supernatant (Fig 2). The results showed that purification of EVs from ZIKV culture supernatant was achieved. These EVs exhibit morphology and density profiles similar to those from uninfected cells (Fig 2B–D), suggesting that ZIKV infection did not drastically alter the production of EVs. Classical flavivirus virion structure (Fig 2C) was not observed in IC‐EVs through negative staining or cryo‐EM analysis. Gradient density analysis showed that EV markers were in fractions ranging from 1.06 to 1.117 g/ml.

Figure 2. Purification of ZIKV‐infected cell‐derived EVs.

1. Schematic diagram of EV purification by ultracentrifugation followed by IC. UC‐EVs (0.6 ml) were purified via beads coated with an anti‐CD9 antibody. After six washes, CD9 peptide (1 mg/ml) was used to elute EVs (termed IC‐EVs, 0.2 ml).
2. Western blots of ZIKV‐infected cell‐derived UC‐EVs or IC‐EVs. Samples during purification were subjected to western blot analysis with the indicated antibodies against cellular proteins. Total lysate of ZIKV‐infected cells was served as positive control.
3. TEM and cryo‐EM observations of IC‐EVs derived from ZIKV‐infected cells. The size distribution of EVs was evaluated by counting the number of EVs in each size range. N indicates the total number of counted EVs. EV‐free virions were shown to mark the size and morphological difference between EVs and virions.
4. Density gradient centrifugation analysis of IC‐EVs. ZIKV‐infected cell‐derived UC‐EVs (0.1 ml) were subjected to density gradient analysis similarly. Fractions were obtained from the top of gradient and analyzed by Western blot with the indicated antibodies.

Source data are available online for this figure.

To evaluate infectivity of IC‐EVs, we performed infection assays. Results showed IC‐EVs were not infectious and CD9‐IP did not extract infectious ZIKV from UC‐EV suspension (Fig 3A). To note, neither CD9 capture nor CD9 peptide addition harms viral infectivity (Fig 3B). In comparison, E‐capture led to significant reduction of viral core protein, E protein, and infectivity from UC‐EV suspensions (Fig 3C–E), indicating the presence of contaminated infectious virions in UC‐EVs. Density fractions of IC‐EVs were not infectious but contained ZIKV genome RNA (Fig 3F). CD9‐IP capture approximately 5% of ZIKV RNA from UC‐EV suspensions (Appendix Fig S3A), but transfection of IC‐EV derived RNA to naïve Vero E6 cells did not lead to ZIKV genome replication, suggesting that IC‐EV encapsulated ZIKV RNA was not fully functional (Appendix Fig S3B). Taken together, these results indicate that a strategy combining sequential ultracentrifugation and subsequent CD9 capture is a straightforward method for EV isolation and that IC‐EVs derived from ZIKV‐infected cells cannot mediate ZIKV infection. We applied this strategy to capture large EVs from ZIKV infection derived culture supernatant. Partial extraction of large EVs was achieved, but these CD9 captured large EVs were not infectious either (Appendix Fig S4).

Figure 3. IC‐EVs do not mediate functional ZIKV infection.

• A
  Detection of infectivity in UC‐EVs, IC‐EVs, and samples during the purification process by immunofluorescence. Ten microliter of each sample was directly inoculated on a Vero E6 cell monolayer for 4 h. Forty‐eight hours post inoculation, ZIKV infection was evaluated by immunofluorescence staining against ZIKV E.
• B
  Effect of CD9 competing peptide addition to ZIKV infectivity. CD9 peptide was added to ZIKV culture supernatant. Quantification of infectivity in the absence and in the presence of CD9‐peptide was conducted. Data are shown as the means with standard deviation from three biological replicates.
• C–E
  Detection of protein level (C) and infectivity (D, E) of UC‐EVs pre‐ and post CD9‐capture or E‐capture. UC‐EV was subjected to CD9‐capture and E‐capture, respectively. After 1 h incubation, ZIKV infectivity of preincubation and post‐incubation samples was measured. Protein (E, capsid and CD9) abundances in pre‐, post‐capture UC‐EV and CD9‐beads were analyzed by Western blot. NA: not applicable. Data are shown as the means with standard deviation from three biological replicates.
• F
  The ZIKV infectious titer of each fraction was measured by plaque assay, and the number of viral genomes was determined via RT‐PCR. Data represent the means from three biological replicates; error bars represent standard deviations from the means. Dashed line represents infectivity detection limit.

Scale bars in panel A and panel D represent 200 μm.

Source data are available online for this figure.

EVs derived from ZIKV‐infected cells contain abundant E proteins on their surface

Although the morphological characteristics of EVs were not significantly altered by ZIKV infection, the overall protein composition of EVs might be modified. Increasing evidence has shown that EVs can deliver viral proteins to recipient cells and activate various signal transduction pathways (Meckes et al, 2010; Mukhamedova et al, 2019; Perez et al, 2019). To examine the protein composition difference between control EVs and those from infected HUVECs, we performed comparative mass spectrometry (MS) analysis of IC‐EVs from naïve or ZIKV‐infected cells (naïve IC‐EVs or ZIKV⁺ IC‐EVs). Compared with IC‐EVs from naïve cells (Appendix Fig S2), ZIKV‐infected cell‐derived IC‐EVs showed the same marker proteins, morphology, and density (Fig 2). Silver staining demonstrated that the overall protein profiles of naïve IC‐EVs or ZIKV⁺ IC‐EVs were similar (Appendix Fig S5A). MS‐based proteomics identified approximately 1,080 proteins in ZIKV⁺ IC‐EVs and 1,000 proteins in naïve IC‐EVs (Fig 4A). Among the common proteins identified in both IC‐EVs, 22 proteins were found relatively enriched (log2FC > 2.1) in ZIKV⁺ IC‐EVs (Fig 4B). We pool the these 22 proteins with those proteins only present in ZIKV⁺ IC‐EVs and performed biological process and cellular localization enrichment analysis (Appendix Fig S5B and C). No significant enrichment of specific biological process or cellular compartment was identified. Viral proteins (capsid, prM, E, NS1, and NS5) were specifically identified in association with ZIKV⁺ IC‐EVs. Highly abundant ZIKV envelope protein E was detected, which was evidenced by both Western blot and highest coverage via MS analysis (Fig 4C and D).

Figure 4. Proteomic profiling of IC‐EVs derived from naïve or ZIKV‐infected cells.

1. Venn diagram of proteins identified in IC‐EVs isolated from naïve or ZIKV‐infected cells. Total 1,084 and 1,000 proteins were identified by MS in naïve IC‐EVs and ZIKV⁺ IC‐EVs, respectively.
2. Relative abundances of proteins identified in both control IC‐EVs and ZIKV infection‐derived IC‐EVs were analyzed. CD9, Actin, Heat shock cognate 71 kDa protein and fatty acid synthase were selected as general EV markers to draw an EV baseline. In comparison to control IC‐EVs, upregulated proteins (log2 fold change > 2.2) were marked in red.
3. Peptide spectra matched (PSMs) and coverage for viral proteins in IC‐EVs by MS.
4. Western blots of ZIKV‐infected cell‐derived UC‐EVs or IC‐EVs using indicated antibodies against viral proteins and EV proteins. Total lysate of ZIKV‐infected cells was the positive control.

Source data are available online for this figure.

To corroborate the association between ZIKV E protein and purified EVs, we performed gradient ultracentrifugation, immunoprecipitation, and immuno‐EM analysis (Fig 5). The results showed that the density gradient profile of the ZIKV E protein completely overlapped with that of CD9 (Fig 5A). Purified EVs can be immunoprecipitated by using ZIKV‐neutralizing antibody (2B10)¹² specifically recognizing E protein DIII (Fig 5B). Finally, immuno‐EM demonstrated that the ZIKV E protein was located on the surface of purified EVs (Fig 5C). In agreement to the observations from negative staining of EV‐free virions and purified EVs (Fig 2C), the mean diameter of ZIVK E+ EVs is around 100 nm, whereas the diameter of EV‐free virions is around 50 nm. Approximately 46% of purified EVs were recognized by an anti‐ZIKV E antibody (Fig 5D). Taken together, these results indicated that ZIKV E protein was coated on the surface of IC‐EVs derived from ZIKV‐infected cells.

Figure 5. ZIKV E was exhibited on the surface of EVs.

1. ZIKV E and CD9 distribution in a density gradient. After density gradient centrifugation, all fractions were analyzed by Western blotting with anti‐ZIKV E and anti‐CD9 antibodies. CD9 blots were reused from Fig 2D.
2. Re‐capture of IC‐EVs derived from ZIKV‐infected cells by anti‐ZIKV E (2B10) antibody‐coated beads. An anti‐HCV antibody (AR3A) was used as a negative control. After 3 h of co‐incubation with IC‐EVs, the beads were washed and analyzed by Western blot with the indicated antibodies.
3. Immuno‐EM analysis of IC‐EVs and EV‐free virions derived from ZIKV‐infected cells. 15 μl IC‐EV suspension was dropped on the nickel grid and anti‐E monoclonal antibody (2B10) was used as the primary antibody. A secondary antibody conjugated to 10 nm gold spheres was used to detect E on IC‐EVs and images were examined by TEM.
4. Quantification of the number of ZIKV E proteins on the EV surface. The percentage of 10 nm gold sphere‐labeled IC‐EVs was counted based on the immuno‐EM images of IC‐EVs. N indicates the total number of counted IC‐EVs.

Source data are available online for this figure.

ZIKV‐infected cell‐derived EVs compete with ZIKV virions for binding to neutralizing antibodies

Knowing that ZIKV‐infected cells produce E protein‐coated EVs, we hypothesize that these EVs might interact with neutralizing antibodies and could hinder antibody function in ZIKV infection. ZIKV interact with entry receptors through E and phosphatidylserine (PS) on virion surface. ZIKV E protein‐coated EVs might interfere with the interaction between ZIKV virions and entry factors due to the presence of abundant E protein and PS, which might cause a possible inhibitory effect on ZIKV entry in normal infection. To avoid this possibility, we used an in vitro ADE model of ZIKV infection based on an FcγRII‐expressing K562 cell line that is poorly permissive to ZIKV or DENV infection in the absence of E protein‐specific antibodies (Castanha et al, 2017). To better characterize the function of ZIKV E protein‐coated EVs, we prepared ZIKV infectious virions without EVs (Fig 1D and E; 1.191–1.217 g/ml), which was termed EV‐free ZIKV. We hypothesize that the ADE effect could be affected by ZIKV E‐coated EVs due to their capability for antibody binding. To define an optimal subneutralizing concentration for ADE, we first titrated the 2B10 antibody. The results demonstrated that ZIKV infection of K562 cells was not efficient (ZIKV infection‐positive cells < 0.3%) in the absence of ZIKV E‐specific antibodies, whereas the addition of the neutralizing antibody 2B10 could lead to significantly enhanced infection. In the presence of 2B10 at 0.25 μg/ml, approximately 63% of cells were infected with ZIKV (Fig 6A and B), indicating a strong ADE of K562 cells. Next, 2B10‐mediated ADE experiments were performed in the presence of EVs (Fig 6C). The results showed that naïve HUVEC‐derived IC‐EVs did not impact ADE of ZIKV infection and that ZIKV IC‐EVs did not mediate infection of K562 cells. In contrast, IC‐EVs from ZIKV‐infected HUVECs attenuated ZIKV ADE in K562 cells in a dose‐dependent manner. At the peak enhancement of ADE‐mediated ZIKV infection, the rate of ZIKV ADE was reduced from the highest at 220 to 65 after the addition of IC‐EVs (10 μg, approximately 3.7 × 10⁹ EVs). Moreover, at lower concentrations of 2B10 (0.13 μg/ml, 0.06 μg/ml), the same doses of IC‐EVs almost blocked ADE‐mediated ZIKV infection. 2.5 μg EV treatment inhibited ADE when lower anti‐E concentrations (0.06, 0.13 μg/ml) were applied. However, 2.5 μg EV treatment had no inhibitory effect to ADE when the antibody concentration reached 0.25 μg/ml. These data suggested that ZIKV RNA encapsulated in IC‐EVs could not trigger general and significant antiviral effects in K562 cells. By using another ZIKV E‐specific neutralizing antibody (Z3L1) (Wang et al, 2016), targeting a tertiary epitope spanning DI and DII, we also found that the addition of ZIKV E‐coated EVs could reduce ADE in K562 cells (Appendix Fig S6). To further corroborate these findings, three lines of additional experiments were performed. Syntenin knock down led to reduced release of EV and ZIKV E protein (Appendix Fig S7A). ZIKV culture supernatant derived from low‐syntenin cells containing less EV‐associated E protein was more sensitive to ADE (Appendix Fig S7B). We further extracted IC‐EVs from subgenomic replicon cells and prM‐E expressing cells (Appendix Fig S7C). Only IC‐EVs from HUVEC cells expressing prME could also attenuate ADE in vitro (Appendix Fig S7D). IC‐EVs derived from 24–48 h.p.i. viral supernatant contained more E protein and could attenuate ADE more efficiently than those derived from 0 to 24 h.p.i. viral supernatant (Appendix Fig S7E and F). In summary, these results demonstrated that ZIKV E‐coated EVs attenuate the ADE effect of ZIKV infection induced by ZIKV E‐specific neutralizing antibodies at sub‐neutralizing concentrations.

Figure 6. IC‐EVs attenuate the ADE of ZIKV mediated by potent neutralizing antibodies at sub‐neutralizing concentrations.

1. ADE of EV‐free ZIKV with different concentrations of 2B10. The 2B10 antibody was mixed with EV‐free ZIKV in indicated concentrations. After 1 h incubation, the mixtures were used to inoculate K562 cells. After brief washing, the cells were further incubated in fresh medium for 48 h, and followed by staining with anti‐ZIKV E antibody and subsequent FACS. The fold changes were presented. Data represent the means from three biological replicates; error bars represent standard deviations from the means.
2. FACS‐based quantification of ZIKV ADE. 0.25 μg/ml 2B10 antibody was used.
3. Detection of ADE‐induced ZIKV infection in the presence of the ZIKV‐specific neutralizing antibody 2B10 and IC‐EVs derived from ZIKV‐infected cells by ADE assay. K562 cells were incubated with a mixture of EV‐free ZIKV, neutralizing antibody, and IC‐EVs. Forty‐eight hours post incubation; cells were stained for ZIKV E and analyzed by FACS. Data represent the means from three biological replicates; error bars represent standard deviations from the means. Differences between treatments were analyzed using the two‐tailed, unpaired Student t test available in the GraphPad Prism 5 software package. P values of < 0.05 (indicated by asterisks) were considered statistically significant. The following categories were used: ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05.

Flavivirus E protein‐coated EVs attenuate DENV patient serum‐mediated ADE

Despite our findings with ZIKV, it is not clear whether the flavivirus E protein on EVs possesses broader virion‐analogous epitopes to attenuate ADE mediated by cross‐reactive antibodies. Due to the clinical significance of dengue ADE, especially in infants born to dengue‐immune mothers (Guzman et al, 2016), we investigated DENV as well. DENV‐infected Vero E6 cells produced CD9‐positive EVs coated with DENV2 E protein (Appendix Fig S8), showing that more than one flavivirus envelope protein can be displayed on EVs. We analyzed whether DENV2 E on the EV surface could affect ADE mediated by DENV3 patient sera (Zhang et al, 2014), which contain an antibody pool induced by natural infection. EV‐free DENV2 and CD9‐positive DENV2 E‐coated EVs were prepared from DENV2 culture supernatant for subsequent experiments. The results demonstrated that serum from healthy donors (HS) could not mediate DENV2 ADE, whereas DENV3 infection patient serum could significantly enhance DENV2 infection through ADE (Fig 7A). As shown in Fig 7C, the addition of DENV2 E‐coated EVs from DENV2‐infected cells reduced ADE in K562 cells in a dose‐dependent manner. It has been shown that DENV patient serum mediates ZIKV ADE through cross‐reactive antibodies (Bardina et al, 2017; Castanha et al, 2017). Using the K562 cell ADE model, we confirmed that DENV3 patient serum could significantly enhance EV‐free ZIKV infection through ADE, although at relatively high concentrations for cross‐reaction (Fig 7B). Moreover, we found that the addition of ZIKV E‐coated EVs led to significant attenuation of ZIKV ADE (Fig 7D). Collectively, these results demonstrated that flavivirus‐coated EVs could reduce ADE mediated by patient serum, suggesting that flavivirus E protein on EVs displays diverse functional epitopes that could be recognized by a natural antibody pool targeting the E protein.

Figure 7. IC‐EVs decreased the ADE effect caused by DENV3 infection patient serum.

• A, B
  Infection of K562 cells with DENV2 (A) or ZIKV (B) in the presence of serially diluted sera from DENV3 infection patients. Data represent the means from three biological replicates; error bars represent standard deviations from the means.
• C, D
  IC‐EVs affected the ADE of DENV2 (C) or ZIKV (D) mediated by DENV3 patient serum. Diluted sera were mixed with EV‐free viruses (DENV2 or ZIKV) and IC‐EVs derived from DENV2‐ or ZIKV‐infected cells and then incubated with K562 cells as described in the methods. At 48 h post incubation, cells were stained with an antibody against flavivirus E and analyzed by FACS. Data represent the means from three biological replicates; error bars represent standard deviations from the means. Differences between treatments were analyzed using the two‐tailed, unpaired Student t test available in the GraphPad Prism 5 software package. P values of < 0.05 (indicated by asterisks) were considered statistically significant. The following categories were used: ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05.

EVs relieve ADE‐induced exacerbation of ZIKV pathogenesis in vivo

Maternally acquired flavivirus‐specific antibodies play both protective and pathogenic roles during primary infection in infants, depending on the concentration of neutralizing antibodies. To mimic ADE in infants and to evaluate whether flavivirus E‐coated EVs impacted ADE in vivo, A6 mice (Ifnar1 ^−/− C57BL/6, 6 weeks old) were infused with different doses of antibody against ZIKV (2B10) to define an optimal concentration for ADE and later challenged with 2 × 10⁵ PFUs of EV‐free ZIKV. We found that 0.1 μg per mouse was the optimal infusion dosage to mimic ADE mediated by neutralizing antibodies at subneutralizing concentrations in vivo (Appendix Fig S9). To evaluate the role of ZIKV E‐coated EVs in ADE in vivo, after antibody infusion, mice were challenged with 2 × 10⁵ PFUs of EV‐free ZIKV mixed with different amounts of ZIKV E‐coated EVs (Fig 8A). We observed that mice that received 2B10 (red line) died before the control mice (PBS group, black line). However, the survival rate of mice in the antibody‐receiving groups infused with the indicated amounts of IC‐EVs derived from ZIKV‐infected cells was significantly increased (green and blue lines) (Fig 8B). The ZIKV E‐coated EV‐only treatment group (gray line) showed no mortality or weight loss in mice, suggesting that ZIKV E‐coated EVs cannot transmit ZIKV infection, whereas the other ZIKV‐challenged groups both exhibited significant weight loss (Fig 8C). At the late stage of ZIKV challenge, surviving mice injected with 2B10 and ZIKV E‐coated EVs (20 μg) showed weight gain and recovery (Fig 8C and D). Consistent with the in vitro results, antibody‐induced ADE of ZIKV infection in vivo was reduced by ZIKV E‐coated EVs that were not infectious in vivo.

Figure 8. IC‐EVs relieved ADE‐induced exacerbation of ZIKV pathogenesis in vivo .

• A–D
  (A) A schematic representation of the in vivo ADE assay. Five‐ to six‐week‐old A6 mice (ifnar ^−/− ) were injected with a neutralizing antibody (2B10, 0.1 μg per mouse) via the intraperitoneal route. One hour later, all mice were challenged with ZIKV (EV‐free, 2 × 10⁵ PFUs per mouse) with or without ZIKV‐infected cell‐derived IC‐EVs by subcutaneous injection into the back. Infected mice were monitored daily for survival (B), body mass (C), and clinical score (D). Data represent the means from three biological replicates; error bars represent standard deviations from the means. Five mice were used for each treatment.

Discussion

Flaviviruses are widely distributed around the world and are considered major human pathogens (Holbrook, 2017). As a mosquito‐transmitted flavivirus, ZIKV remains a concern because of its potential global health threat and unique clinical features, including microcephaly in virus‐infected fetuses, persistence in immune‐privileged sites, and diverse virus transmission (Pierson & Diamond, 2018). Recently, it was reported that flaviviruses, including ZIKV, could achieve viral transmission through virion‐free EVs (Ramakrishnaiah et al, 2013; Vora et al, 2018; Zhou et al, 2018, 2019; Martinez‐Rojas et al, 2020). Ultracentrifugation‐obtained EV preparations from ZIKV‐infected cells carry viral components and play roles in inducing monocyte differentiation and mediating transmission in cortical neurons (Zhou et al, 2019; Martinez‐Rojas et al, 2020). In agreement with these findings, the present study confirmed that UC‐EVs were infectious. However, IC‐EVs without virion contamination from ZIKV infected HUVECs were not infectious. Independent from virions, secretion of flavivirus E protein through EVs is observed. These EVs carry abundant ZIKV E proteins that possess epitopes similar to those on infectious virions.

Approximately 10% of extracellular E protein was associated with EVs, which are lighter than classical flavivirus virions. Capsid, NS1, and NS5 were also found in association with EVs. However, whether these viral proteins were present in the same or separate populations of EVs is not known. Nano‐Flow profile with satisfactory combination of detection means will be required to solve this open question. Apart from viral proteins, more than 200 cellular proteins were found relatively enriched in ZIKV infection derived EVs. Comparative proteome analysis in this study only involved IC‐EVs. ZIKV infection induces cytopathic effects to infected cells including ER stress, subcellular compartment modification, and alteration of global protein expression profile. Comprehensive proteome analysis of both cell and IC‐EVs is required to answer whether this cellular protein enrichment in infection derived IC‐EVs comes from distinct EV cargo protein loading process caused by infection modified subcellular compartments or simply from the significantly altered global cellular protein expression profile in infected cells.

The major challenge that we addressed in this study was the separation of EVs and ZIKV virions because of their overlapping size and density ranges (Zabeo et al, 2017; McNamara & Dittmer, 2020). Although Tian et al demonstrated that ultracentrifugation was an effective method for EV purification compared with other methods, such as polymer precipitation, membrane filtration, and size exclusion chromatography (Tian et al, 2020), ultracentrifugation alone cannot separate EVs from virions. EV preparations obtained through ultracentrifugation were reported to transmit HCV RNA, possibly via cell‐to‐cell contact, avoiding receptors (Ramakrishnaiah et al, 2013). However, without a complete set of assembly machinery, seeding of subgenomic HCV replicon cells in naïve Huh7.5.1 cells did not lead to successful transmission (Zhao et al, 2017). Combining sequential ultracentrifugation and IC against CD9, we successfully obtained EVs without viral infectivity. CD9 capture recovered EV markers, such as syntenin, CD81, CD9, CD63, and TSG101, from UC‐EV preparations. ZIKV E and genome RNA were present in IC‐EVs. EVs from infected cells carry HCV, DENV, and YFV RNA and stimulate plasmacytoid dendritic cells to secrete interferons (Dreux et al, 2012; Decembre et al, 2014; Sinigaglia et al, 2018). This mode of viral RNA loading is likely independent of viral structural protein. In the present study, RNA was also found in ZIKV infection derived IC‐EVs which hardly contain sufficient capsid protein, suggesting that viral genome RNA loading into EVs lacks a viral and host protein‐concerted encapsidation process governing the integrity of the RNA genome (Xie et al, 2019; Tan et al, 2020). Transfection of EV‐derived ZIKV RNA could not lead to successful initiation of viral replication, which confirmed that EV‐encapsulated RNA was not viral. Moreover, our data unambiguously proved that RNA‐loaded EVs could not transmit ZIKV infection in cell culture or in a mouse model. Infection of hepatotropic flavivirus HCV was shown to produce envelope protein‐coated EVs independent from infectious lipo‐viral particles morphogenesis (Deng et al, 2019). Unlike flaviviruses, enteroviruses, including hepatitis A virus, poliovirus, coxsackievirus, rotavirus, and norovirus, employ extracellular vesicles to pack single virion or virion clusters, resulting in efficient infection in the presence of neutralizing antibodies (Feng et al, 2013; Chen et al, 2015; Santiana et al, 2018). This divergence indicates that multiple modes of virus–host interaction in the interface spanning virus assembly and EV biogenesis occur and determine distinct functions of EVs in infection.

In summary, our study reported a clear‐cut method for the isolation of EVs derived from flavivirus culture supernatant. These purified EVs are lighter than flavivirus virions. They carry abundant flavivirus E protein on their surface but do not transmit infection in cell culture or in mouse models. Importantly, flavivirus E protein‐coated EVs could bind to antibodies against E and reduce the ADE effect. K562 cells lack complete antiviral machinery for the absence of functional interferon receptor (Colamonici et al, 1994) and ADE infection suppress RIG‐I/MDA5 signal pathway (Ubol et al, 2010), which ruled out the possible contribution of antiviral effect triggered by EV‐encapsulated ZIKV RNA. E proteins on infection derived EVs share common antigenic landscapes to those from infectious virions, which suggested that these EVs are potentially competent in both antibodies response and inhibition of virus entry. Due to the capability of E coating EVs to absorb ADE antibodies, modulation the E secretion balance between EVs and virions might be further studied for EV's potential in erasing maternally acquired antibodies mediated risk of severe flavivirus infection in more physiologically relevant models.

Materials and Methods

Cell lines, virus and antibodies

Cells and viruses

Human umbilical vein endothelial cells (HUVEC, CRL‐1730, American Type Culture Collection) and African Green kidney cell line Vero E6 (CRL‐1856, American Type Culture Collection) were maintained in DMEM (Invitrogen); the human erythroleukemia cell line K562 was grown in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco). All media were supplemented with 10% fetal bovine serum (FBS, Gibco), nonessential amino acids (Gibco), 100 U/ml penicillin (Gibco), 100 μg/ml streptomycin (Gibco) at 37°C with 5% CO₂. For EV collection, cells were grown in DMEM with EV‐depleted FBS (Shelke et al, 2014).

ZIKV and Dengue virus 2 were rescued by stabilized infectious cDNA clone pZL1 (Zhao et al, 2018), and pFK‐DVs (Fischl & Bartenschlager, 2013). ZIKV and DENV2 produced in HUVEC or Vero E6 cells were titrated on Vero E6 cells by plaque‐forming unit (PFU) assay.

Rabbit polyclonal antibodies to ZIKV viral proteins (C, E, NS1, NS5) were generated in‐house. The following commercial antibodies were used in this study: mouse monoclonal antibody 4G2 cross‐reactive with flavivirus E protein (Merck Millipore), mouse anti‐tubulin (Santa Cruz Biotechnology), mouse anti‐actin (Santa Cruz Biotechnology), and goat anti‐mouse IgG conjugated with Alexa Fluor 488 (Thermo Fisher Scientific). EV‐related polyclonal antibodies (anti‐human Alix/HSP70/ TSG101/ Syntenin/CD9/CD63/CD81), anti‐LC‐3 (autophagy marker), anti‐Calnexin (endoplasmic reticulum marker) were provided by Proteintech. ZIKV‐specific neutralizing antibodies 2B10 and Z3L1 were provided by Dr. Lingqi Zhang at Tsinghua University and Beijing ZEPING Bioscience & Technologies Co., Ltd, respectively. Dengue III patient serum samples were provided by Dr. Fuchun Zhang at Guangzhou Medical University.

Additional information on primers applied for quantitative PCR and specifics on key reagents were available in supplementary file (Appendix Tables S1 and S2).

Isolation of EVs by concentration, ultracentrifugation and immunocapture

Extracellular vesicles derived from naïve cells or virus‐infected cells (HUVEC for ZIKV infection and Vero E6 cells for DENV infection) were first isolated via ultracentrifugation and further purified by immunocapture (IC). Briefly, cells (HUVEC or Vero E6) grown to 80% confluence in 150 mm dishes were infected with ZIKV or DENV (MOI = 0.01). 4 h post‐infection, unbound viruses were removed, and fresh medium with 2% EV‐depleted FBS was added for further incubation. 3 days later, EV‐containing culture supernatant (total 240 ml) was harvested and spun at 700 g for 10 min at 4°C to remove cells (Sorvall ST16R, Thermo Fisher) and centrifuged at 2,000 g for 20 min at 4°C to remove debris and apoptotic bodies (Sorvall ST16R, Thermo Fisher). Next, the supernatant was spun at 10,000 g for 60 min at 4°C to deplete the large extracellular vesicles (5811, Eppendorf AG). Then, the supernatant was concentrated (Ultracel®‐100 K, Merck Millipore), transferred into ultracentrifuge tubes (330901A, HITACHI), and centrifuged at 100,000 g for 2 h at 4°C (CP80WX, HITACHI). The resulting EV pellets were resuspended in 600 μl cold PBS and termed UC‐EVs or pre‐incubation (the sample before IC).

For IC, 400 μl beads (Dynabeads™ Protein G, Invitrogen) were mixed with anti‐Human CD9 monoclonal antibody (generated in‐house, 0.25 mg/ml), and dynamic (rotating) incubation for 4 h at 4°C. After washing with PBST (PBS containing 0.02% Tween‐20), anti‐CD9‐coated beads were incubated with UC‐EVs in rotating tubes overnight at 4°C. After washing six times with 2 ml PBST, captured EVs were eluted with 200 μl of a CD9 competing peptide corresponding to the large extracellular loop of CD9 (1 mg/ml, dissolved in PBS) for 1 h on ice. Recovered material was termed IC‐EVs. Protein concentrations of EV suspensions were measured by BCA protein assay kit (Beyotime Biotechnology) according to the manufacturer's protocols. 1 mg/ml CD9 peptide samples was applied as blank to remove the contribution of CD9 peptide to EV samples.

For re‐capture of IC‐EVs, beads coated with specific anti‐ZIKV envelope antibody 2B10 or anti‐HCV envelope antibody AR3A (Prof. Sun Bing, Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, China) were incubated with IC‐EVs in rotating tubes for 4 h at 4°C. After washing with PBST, captured EVs were collected and stored at −80°C for further experimentation.

Density gradient centrifugation analysis of virus particles and UC/IC‐EVs

To characterize the density profile of viral supernatant, 40 ml ZIKV‐containing supernatant was centrifuged at 10,000 g for 30 min at 4°C to remove debris and concentrated to 1 ml by ultrafiltration (Ultracel®‐100 K, Merck Millipore). The concentrated virus sample was mixed with ice‐cold iodixanol for a final 80% iodixanol (OptiPrepTM, Axis‐shield) solution and transferred to the bottom of centrifugation tube, followed by addition of iodixanol/PBS (80–0%) gradients. To characterize the density profile of EV preparations obtained by sequential ultracentrifugation or immune‐capture, iodixanol/PBS density medium was prepared. 0.1 ml UC/IC‐EVs resuspended in ice‐cold PBS were loaded on the top of density gradient. Concentrates or EVs were subjected to ultracentrifugation at 34,000 rpm (250,000 g) for 18 h at 4°C using a P55ST Swinging rotor (HITACHI) to generate continuous gradient. Twelve individual fractions of 0.3 ml were collected from the top; the density of each gradient fraction was measured by using a refractometer (WAY‐2D, Shanghai optical instrument factory). The levels of viral/host proteins, viral RNA, and infectivity titer were measured by using western blot, Q‐RT‐PCR, and plaque assay, respectively.

Immunoblot analysis

Lysates from cells, UC‐EV, IC‐EVs, and samples during EV purification were denatured in Laemmli buffer (125 mM Tris/HCl, 2% (w/v) SDS, 5% (v/v) 2‐mercaptoethanol, 10% (v/v) glycerol, 0.001% (w/v) bromophenol blue, pH 6.8), heated to 90°C for 30 min before loading. The samples were separated by SDS 12% polyacrylamide gel electrophoresis. Proteins were transferred onto a polyvinylidene difluoride membrane (PVDF, Merck Millipore). The membrane was blocked for 1 h in PBS supplemented with 0.5% Tween (PBS‐T) and 5% no‐fat dry milk (PBS‐M) at room temperature and then incubation for 2 h with primary antibody diluted in PBS‐M. Membrane was washed 3 times with PBS‐T and incubated for 1 h with horseradish‐peroxidase‐conjugated secondary antibody diluted 1:10,000 in PBS‐T with 1% dried milk. Bound antibodies were detected after three times washing with the ECL reagent (PerkinElmer). The signal was captured by Tanon 5500.

Silver staining analysis

Protein distribution of lysed cells or IC‐EVs was analyzed by using a PierceTM Silver Stain for Mass Spectrometry kit (Thermo Scientific) according to the manufacturer's protocols. Briefly, the samples were separated on 12% SDS‐PAGE gels. Then, the gels were washed twice by ultrapure water for 5 min and then treated with fix buffer (60% water, 30% ethanol, and 10% acetic acid) overnight at RT. After washed twice with 10% ethanol and ultrapure water (5 min for each time), the gels were treated with sensitizer working solution for exact 1 min, and washed twice by ultrapure water immediately. Later, added the silver stain enhancer solution to the container loaded with gels for 5 min. After washing twice with ultrapure water, the gels were treated with developer working solution, and the stop buffer (5% acetic acid) was added upon the bands appeared. Images were captured by Tanon 5500.

Total protein extraction and digestion for mass spectrometry

After EM verification on the integrity of IC‐EVs, control IC‐EVs and ZIKV‐infection derived IC‐EVs were dissolved with protein extraction buffer (8 M urea, 1% SDS), which contains protease inhibitors. The mixture was then treated by ultrasound for 2 min and following splitting for 30 min. After centrifugation at 12,000 g for 30 min, the concentration of protein supernatant was determined by using BCA Protein Assay Kit (Pierce, Thermo, USA). 13 μg of EV protein sample was treated with TCEP (tris (2‐carboxyethyl) phosphine, final concentration of 10 mM) for 60 min at 37°C. Following add IAM (Iodoacetamide) to the final concentration of 40 mM and react for 40 min at room temperature under dark conditions. Add precooled acetone (acetone: sample v/v = 6:1) to each sample and to incubate for 4 h at −20°C. After centrifugal for 20 min at 10,000 g, the sediment was collected and add 100 μl 100 mM TEAB solution to dissolve. Finally, the mixture was digested with Trypsin overnight at 37°C. The peptides were vacuum dried and then resuspended with 0.1% TFA. Samples were desalted with HLB and vacuum dried. Loading buffer was added to each tube to prepare samples for mass spectrometry analysis, and the concentration of each sample was 0.25 μg/μl.

LC–MS/MS analysis

Trypsin‐digested peptides were analyzed by nano flow liquid chromatography tandem mass spectrometry performed on an EASY‐nLC 1200 system (Thermo, USA) connected to a Q Exactive HF‐X quadrupole orbitrap mass spectrometer (Thermo, USA) through a nanoelectrospray ion source. Briefly, the C18‐reversed phase column (75 μm × 25 cm, Thermo, USA) as equilibrated with solvent A (A:2% ACN with 0.1% formic acid) and solvent B (B: 80% ACN with 0.1% formic acid). The peptides were eluted and separated at a flow rate of 300 nl/min. The Q Exactive HF‐X instrument was operated in the data‐dependent acquisition mode (DDA) to automatically switch between full scan MS and MS/MS acquisition. The survey of full scan MS spectra (m/z 350–1,500) was acquired in the Orbitrap with 60,000 resolution. The automatic gain control (AGC) target at 3e6 and the maximum fill time was 20 ms. Then the top 20 most intense precursor ions were selected into collision cell for fragmentation by higher energy collision dissociation (HCD). The MS/MS resolution was set at 15,000 (at m/z 100), the automatic gain control (AGC) target at 1e5, the maximum fill time at 50 ms, and dynamic exclusion was 18 s.

Protein identification

The RAW data files were analyzed using ProteomeDiscoverer (Thermo Scientific, Version 2.4) against Uniprot (Homo sapiens) database. The MS/MS search criteria were as follows: Mass tolerance of 20 ppm for MS and 0.02 Da for MS/MS tolerance; trypsin as the enzyme with two missed cleavage allowed; carbamido methylation of cysteine as fixed modification and methionine oxidation as dynamic modification, respectively. False discovery rate (FDR) of peptide identification was set as FDR ≤ 0.01. A minimum of one unique peptide identification was used to support protein identification.

Before GO term and subcellular localization analysis, relative abundances of proteins identified in both control IC‐EVs and ZIKV‐infection derived IC‐EVs were analyzed. CD9, Actin, Heat shock cognate 71 kDa protein and Fatty acid synthase were selected as general EV markers to draw an EV baseline. In comparison to control IC‐EVs, upregulated proteins (log2 fold change > 2.1) were selected and pooled to the proteins specifically identified in ZIKV‐infection‐derived IC‐EVs. Subsequently, GO term and subcelluar localization analysis were performed on proteins enriched in ZIKV‐infection‐derived IC‐EVs using the Diamond software (Version 2.0.11) using Uniprot database (Version 2021.09).

RNA transfection

Transfection of total RNA from IC‐EVs or ZIKV in vitro transcripts was performed by using Lipofectamine MessengerMAX (Thermo Fisher Scientific) according to the manufacturer's protocols.

Briefly, RNA or lipofectamine was diluted with opti‐MEM medium, respectively, and then, diluted RNA solution was mixed with lipofectamine solution and incubated for 5 min at room temperature. Finally, the mixture was plated onto 70% confluent monolayers of Vero E6 cells that had been prepared 24 h earlier in 48‐well plates. Cells were further incubated and harvested in indicated time point for total RNA extraction and following ZIKV genome quantification. GAPDH was analyzed as internal control.

Size distribution and CD9 labeling of UC‐EVs by flow NanoAnalyzer

For determination of particle size distribution, a silica nanosphere cocktail (S16M‐Exo) was used as the size standards, with the correlation of particle size and the side scattering intensities, a calibration curve was generated and the size of EV particles could be calculated based on the corresponding side scatting intensity.

For surface protein profiling, the EV sample was resuspended in 50 μl of PBS, and Alexa Fluor 647‐conjugated anti‐CD9 (NanoFCM, NHA009‐A647‐100 T) was added. The mixture was incubated at 37°C for 30 min and washed twice with PBS via ultracentrifugation at 100,000 g for 30 min at 4°C. The final pellet was resuspended in 100 μl of PBS for analysis with a Flow NanoAnalyzer model type U30E (NanoFCM Inc., Xiamen, China) equipped with a laser (638 nm).

Real‐time PCR analysis

To measure ZIKA RNA copies in culture medium or purified EVs, quantitative RT‐PCR was performed. First, total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer's protocol. ZIKV RNA levels were measured by a one‐step quantitative realtime reverse transcription polymerase chain reaction assay (qRT‐PCR, QuantStudio 6 Flex, ABI) using One‐Step RT‐PCR Master Mix (Toyobo).

Electron microscopy (EM)

To characterize EVs' morphology, negative staining was performed. Five microliters of EV sample was applied to glow‐discharged, carbon‐coated formvar grids (230‐mesh, Zhongjingkeyi Technology Co., Ltd.) and incubated for 10 min. Grids were then washed twice with PBS and fixed using 2.5% glutaraldehyde for 5 min. After following washes with distilled water, grids were stained by using 2% UA in water for 1 min. After dried by air, images were captured by FEI F20 TEM at 200 Kv.

To further observe EVs in native condition, cryo‐electron microscopy (Cryo‐EM) which allows the visualization of biomaterial in their native state was performed. Isolated EVs were directly adsorbed onto glow‐discharged holey carbon grids (QUANTIFOIL, Germany). Grids were blotted at 95% humidity and rapidly plunged into liquid ethane with the aid of VITROBOT (Maastricht Instruments BV, The Netherlands). Then, grids were observed using a 200 kV electron microscope (Tecnai G2 F20 Sphera, FEI).

For detection surface distribution of ZIKV envelope protein on EVs, we chose immuno‐EM. Briefly, 30 μl of EV suspension was dropped on parafilm, and a carbon‐coated nickel grid was positioned with the coating side facing the EV droplet for 30–60 min. The grids were washed three times with PBS. Later, the grids were fixed in 2% paraformaldehyde for 10 min. After three washes with PBS, the grids were incubated with primary antibody 4G2 against ZIKV‐E (anti‐Flavivirus Group Antigen Antibody, Merck Millipore) (dilution ratio 1:50) for 40 min. After three washes, the grid was incubated with the goat antimouse 10 nm‐gold antibody (dilution ratio 1:50) for 40 min, postfixed with 2.5% glutaraldehyde for 10 min and washed three times with PBS. The sample was treated with 2% uranylacetate and examined by FEI F20 TEM at 200 kV.

ADE of ZIKV or DENV infection in K562 cells

The capability of neutralizing monoclonal antibodies (2B10/Z3L1) or human sera from DENV patients to mediate ADE of infection was measured by a flow cytometry using K562 cells as previously described. Briefly, mAbs or heat‐inactivated plasma samples were serially diluted in RPMI medium and preincubated with EV‐free ZIKV (ZL1, 1.5 × 10⁵ PFU) for 1 h at 37°C. This virus‐antibody complex was mixed with indicated amounts of ZIKV‐infected cells derived IC‐EVs (10 μg, 2.5 μg, 0.6 μg, 0 μg) in a total volume of 200 μl and added to 1.5 × 10⁵ K562 cells. After incubation for 2 h at 37°C, cells were then washed twice, resuspended in RPMI‐1640 medium supplemented with 2% fetal bovine serum and cultured at 37°C for an additional 2 days. Cells were then washed with PBS, fixed with 2% paraformaldehyde (PFA), permeablized, and stained with 4G2 (anti‐Flavivirus Group Antigen Antibody, Merck Millipore). Infection levels were read by the use of flow cytometer (Celesta, BD).

ZIKV infection of ifnαr1^−/− C57BL/6 mice

Ifnαr1^−/− C57BL/6 mice were maintained under specific‐pathogen‐free conditions at the animal facility of Institute Pasteur of Shanghai (IPS, China). Experimental protocols related to ZIKV infection in mice were approved by the Institutional Animal Care Committee and Bio‐Safety Committee of IPS (approval number: P2020039) and were performed according to the guidelines of Animal Bio‐Safety Laboratory of IPS. For in vivo ZIKV ADE assays (Bardina et al, 2017; Shim et al, 2019), 6‐week‐old mice were intraperitoneally injected with neutralizing monoclonal antibody 2B10 (0.1 μg per mouse) diluted in 100 μl PBS. These mice were infected 1 h later by subcutaneous back injection of 2 × 105 PFU of EV‐free ZIKV with or without E‐coated EVs and were monitored daily for body mass, survival, and clinical symptoms.

Methods availability

We have submitted our methods to the EV‐TRACK knowledgebase (EV‐TRACK ID: EV210141) (Van Deun et al, 2017).

Statistical analysis

All data were analyzed using Prism6 software (GraphPad). Statistical evaluation was performed using t test and one‐way ANOVA as appropriate. Data were presented as means ± SD. P < 0.05 was considered statistically significant (*P < 0.05; **P < 0.01 and ***P < 0.001).

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix S1

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

The EMBO Journal (2023) 42: e112096

Data availability

Mass spectrometry raw data were submitted to ProteomeXchange via the PRIDE database (Project accession: PXD039152). Project Name: Mass spectrometry analysis of Extracellular vesicles derived from naive HUVEc cells or ZIKV‐infected HUVEc cells (https://www.ebi.ac.uk/pride/archive/projects/PXD039152).