Extracellular vesicles containing ACE2 efficiently prevent infection by SARS‐CoV‐2 Spike protein‐containing virus

Abstract SARS‐CoV‐2 entry is mediated by binding of the spike protein (S) to the surface receptor ACE2 and subsequent priming by host TMPRSS2 allowing membrane fusion. Here, we produced extracellular vesicles (EVs) exposing ACE2 and demonstrate that ACE2‐EVs are efficient decoys for SARS‐CoV‐2 S protein‐containing lentivirus. Reduction of infectivity positively correlates with the level of ACE2, is much more efficient than with soluble ACE2 and further enhanced by the inclusion of TMPRSS2.


INTRODUCTION
SARS-CoV-2 is the infectious causative agent of the COVID-19 pandemic (Zhou et al., 2020). Viral entry into host cells is mediated by the interaction of the spike (S) protein on the surface of SARS-CoV-2 with the surface receptor angiotensin-converting  of  COCOZZA et al. ACE2 is expressed at the surface of pneumocytes and intestinal epithelial cells which are potential target cells for infection (Ziegler et al., 2020). A soluble form of the ACE2 ectodomain can be released after cleavage by ADAM10 or ADAM17 in different physiological conditions (Jia et al., 2009). In addition, TMPRSS2 can compete with the metalloproteases for ACE2 cleavage (Heurich et al., 2014). Soluble recombinant ACE2 neutralizes SARS-CoV-2 by binding the S protein and has been proven to reduce entry of SARS-CoV-2 into Vero-E6 cells and engineered human organoids (Monteil et al., 2020). ACE2, however, is synthesized as a transmembrane protein, like TMPRSS2. We postulate that ACE2 could be present on the surface of extracellular vesicles (EVs), which could result in better efficacy as a decoy to capture SARS-CoV-2. Furthermore, concomitant presence of an active serine-protease TMPRSS2 on the same EVs could further interfere with viral infectivity, by forcing viral fusion on the EV membrane, rather than on target cells.
EVs are lipid bilayer-enclosed structures containing transmembrane proteins, membrane-associated proteins, cytosolic proteins and nucleic acids that are released into the environment by different cell types (Mathieu et al., 2019). Since EVs have the same membrane orientation as cells, they expose at their surface the extracellular domains of transmembrane proteins that can bind to nearby or long-distance targets. By specifically binding to different proteins and protein-containing structures, EVs can act as a decoy for virus (De Carvalho et al., 2014) and bacterial toxins (Keller et al., 2020), thus suggesting a potential role as therapeutic agents.

 RESULTS
In order to explore the hypothesis that EVs can be used as SARS-CoV-2 decoy agents, we first assessed whether ACE2 can be present in EVs from cell lines derived from tissues expressing ACE2. As cell lines endogenously expressing ACE2, we used the human lung epithelial cell line Calu-3 and the epithelial colorectal cell line Caco-2 which are known targets for SARS-CoV-2 infection (Hoffmann et al., 2020). In a pilot experiment, Calu-3 and Caco-2 were cultured in medium without fetal bovine serum (FBS) for 24 h and EVs were isolated from the cell conditioned medium (CCM) by size-exclusion chromatography (SEC). This technique allows the separation of EVs from soluble proteins ( Figure 1a). We analysed side-by-side fractions enriched in EVs (pooled fractions 7-11), in soluble factors (pooled fractions 17-21), and the fractions in-between (pooled fraction 12-16) ( Figure 1a). Particle quantification by nanoparticle tracking analysis (NTA) confirmed that the majority of particles released by Calu-3 and Caco-2 cells were isolated in EV-containing fractions (Supplementary Figure A). In this experiment, Caco-2 cells released less EVs but of similar mode size than Calu-3 ( Figure 1b). These EVs contained ACE2 protein as well as known EV markers (CD63, CD81 ADAM10) ( Figure 1c). In addition, although Caco-2 and Calu-3 expressed TMPRSS2, this protease was not released in EVs (Figure 1c). To obtain EVs with high amounts of ACE2 and TMPRSS2 to be tested as decoy agents, we switched to 293FT cells that could be easily genetically-edited. We transduced 293FT cells with lentivirus containing ACE2 alone (293FT-ACE2) or in combination with TMPRSS2 (293FT-ACE2-TMPRSS2). 293FT cells transduced with lentivirus containing empty plasmids were used as a control (293FT-mock). The three 293FT cell lines were cultured in FBS-containing EV-depleted medium and EVs were isolated from concentrated CCM by SEC. We observed a high count of particles with comparable mode sizes in EV fractions from all 293FT cell lines ( 293FT-mock and 293FT-ACE2 cells expressed TMPRSS2, but the predicted full-length 54 kDa form was not detected in their EVs: a 30 kDa N-terminal fragment devoid of the serine protease domain (Zmora et al., 2015) was mainly detected, together with a low level of the glycosylated full-length 70 kDa form. Importantly, EVs from 293FT cells overexpressing TMPRSS2 contained higher levels of the full-length TMPRSS2 protein and less of a cleaved form than EVs from 293FT-mock or 293FT-ACE2 cells ( , 2001). Since the ACE2 ectodomain can be released after cleavage (Jia et al., 2009), we evaluated whether cells overexpressing ACE2 release soluble ACE2 in addition to EV-associated ACE2. To do this, we analysed the presence of ACE2, the EV marker CD81 and the non-EVs component AChE (Liao et al., 2019) by WB on EV fractions (1 × 10 9 particles) and the intermediate and soluble fractions obtained from the same amount of CCM. We detected some particles in intermediate and soluble fractions from the three 293FT cell lines, that probably came from the depleted medium since they did not contain EV markers by WB (Supplementary Figure A, D, E). AChE was enriched in soluble fractions whereas CD81 was mainly found in EV fractions validating our isolation protocol for EVs and soluble components ( Figure 1f). Importantly, ACE2 was found as a full-length transmembrane form in the EV fractions, as two shorter cleaved forms in the soluble fractions, and a mixture of all forms in the intermediate fractions ( Figure 1f). Intermediate fractions thus represent a mixture of EVs and soluble proteins and for this reason were not further analysed. Low levels of ACE2 were found in the soluble SEC fractions of 293FT-ACE2 cells when compared to the levels present on the EV-containing fraction (Figure 1f, 1g). By contrast, 293FT-ACE2-TMPRSS2 cells released similar amounts of EV-associated as of soluble ACE2 (Figure 1f,1g), suggesting that the overexpression of the protease cleaves ACE2 as previously described (Heurich et al., 2014), favouring its secretion to the extracellular space as a soluble protein instead of associated to EVs. We then analysed the capacity of ACE2-and ACE2-TMPRSS2-containing EVs to reduce the infection of target cells by a lentivirus containing SARS-CoV-2-S protein. The virus used in this study was produced using an HIV packaging vector pseudotyped with SARS-CoV-2 Spike protein and including a GFP-coding sequence, which is expressed in infected cells. A variant of SARS-CoV-2 Spike protein bearing a region of the tail of the HIV gp41, instead of the Spike tail, previously shown to enhance infection of pseudotyped virus (Moore et al., 2004) was used. First, we determined the infectivity of the target cells Caco-2, Calu-3 and 293FT-ACE2 by SARS-CoV-2-S-pseudotyped lentivirus and observed that each of these cell lines is infected similarly in a concentration-dependent manner (Figure 2a). To assess the ability of ACE2-containing EVs to decrease virus infectivity in vitro, we infected target cells with SARS-CoV-2-S-pseudotyped virus in the presence or the absence of EVs isolated from 293FTmock (mock-EVs) or 293FT-ACE2 (ACE2-EVs) or 293FT-ACE2-TMPRSS2 cells (ACE2-TMPRSS2-EVs) (Figure 2b). Infection of 293FT-ACE2 cells in the presence of ACE2-EVs and ACE2-TMPRSS2-EVs was reduced while infection remained unaffected by mock-EVs (Figure 2c and quantification in 2d). Importantly, this inhibition was dependent on the dose of EVs (Figure 2d). Caco-2 infection was also reduced in the presence of ACE2-EVs and ACE2-TMPRSS2-EVs (Figure 2e). We then quantified by enzymelinked immunosorbent assay (ELISA) the amount of ACE2 released by these cell lines. As previously documented by western blot (Figure 1f and 1g), we observed that 293FT-ACE2 cells released high levels of ACE2 mainly associated with EVs while 293FT-ACE2-TMPRSS2 cells released lower ACE2 levels that were equally distributed between EV and soluble fractions (Figure 2f). Strikingly, ACE2 in the soluble fractions from these latter cells could not inhibit SARS-CoV-2-S-pseudotyped virus infection as compared to a comparable amount of ACE2 associated with EVs ( Figure 2g). When re-plotting the results obtained in Figure 2d as a function of the absolute amount of ACE2 measured for these same samples by ELISA (Figure 2f), we observed that EVs from cells overexpressing the full length TMPRSS2 together with ACE2 were more efficient at inhibiting SARS-CoV-2-S-pseudotyped viral infection than those from cells overexpressing ACE2 alone (Figure 2h). Moreover, to achieve similar levels of inhibition of lentiviral infection as those observed with ACE2-or ACE2-TMPRSS2-EVs, 500 to 1500 times more of the soluble recombinant human ACE2 had to be used (Figure 2h) in accordance with previous publications (Monteil et al., 2020). Altogether, these findings highlight the increased efficiency of EVs containing full-length ACE2 to inhibit SARS-CoV-2-S-pseudotyped viral entry when compared to the soluble protein alone. The enhanced efficiency of EVs from cells overexpressing TMPRSS2 could be due to the presence of TMPRSS2 together with ACE2, leading to fusion of the virus with the EV thus impairing their capacity to infect cells, and/or to other modifications of the EV composition induced by overexpression of TMPRSS2 in the EV-secreting cells.

 DISCUSSION
Our data demonstrate that EVs containing ACE2, alone or in combination with TMPRSS2, block SARS-CoV-2 Spike-dependent infection in a much more efficient manner than soluble ACE2. Thus, ACE2-EVs represent a potential versatile therapeutic tool to block not only SARS-CoV-2 infection but also other coronavirus infections that use the ACE2 receptor for host cell entry, such as SARS-CoV (Li et al., 2003) and NL63 (H et al., 2005). Further studies to determine the efficacy of the ACE2/TMPRSS2-EVs in experimental models of SARS-CoV-2 virus need to be conducted to validate their therapeutic use for COVID-19, but also the lack of side-effects. The use of engineered EVs as therapeutic agents has been proposed several years ago and is currently being explored in humans (Wiklander et al., 2019), suggesting that well-designed EV therapeutics against COVID-19 may be feasible to prevent initial infection or further internal dissemination of the virus, and thus reducing the virus burden and disease severity. However, as recently highlighted by the International Societies for EV (ISEV) and for Cellular Therapies (ISCT) (Börger et al., 2020), despite the urgency induced by the current pandemic, EV-based therapeutic developments for COVID-19 will have to meet as strong criteria of manufacturing processes, quality controls and compliance to safety regulation as any other therapies, before they can be implemented in human subjects. Note: a speculative article discussing the idea that we demonstrate experimentally here was published while we were preparing this article, thus showing concomitant emergence of similar scientific ideas (Inal, 2020).

. Preparation of EV-depleted medium
EV-depleted medium was obtained by overnight ultracentrifugation of DMEM supplemented with 20% FBS at 100,000 x g in a Type 45 Ti rotor (Beckman Coulter, K-factor 1042.2). After ultracentrifugation, EV-depleted supernatant was carefully pipetted from the top and leaving 7 ml in the bottom. Supernatant was filtered through a 0.22 µm bottle filter (Millipore) and additional DMEM and antibiotics were added to prepare complete medium (10% EV-depleted FBS medium).

. EV isolation by size-exclusion chromatography
239FT-mock, 293FT-ACE2 and 293FT-ACE2-TMPRSS2 cells were cultured in FBS EV-depleted medium for 24 h. Caco-2 and Calu-3 cells were cultured in FBS-free DMEM for 24 h. Cell conditioned medium (CCM) was harvested by pelleting cells at 350xg for 5 min at 4 • C three times. Supernatant was centrifuged at 2,000xg for 20 min at 4 • C to discard 2K pellet and concentrated on a MiIlipore Filter (MWCO = 10 kDa, UCF701008) to obtain concentrated CCM. Medium was concentrated to 1 ml from 12 to 41 ml for Caco-2 and Calu-3 and from 75 ml for 293FT cells and overlaid on a 70 nm qEV size-exclusion column (Izon, SP1). 0.5 ml fractions were collected and EVs were recovered in fractions 7 to 11 following manufacturer's instructions. We additionally

. Infectivity assay
20,000 Caco-2 or Calu-3 cells or 10,000 293FT-ACE2 cells were seeded in a 96 well plate 6 h before infection with SARS-CoV-2 Spseudotyped virus. Infection was performed in the absence or in the presence of different amount of EVs or human recombinant ACE2 (Abcam, 151852) by spinoculation at 1200 x g for 1 h 30 min at 25 • C. 48 h after infection, cells were washed, trypsinized and fixed. Infection was measured by analysing eGFP expression using a CytoFLEX LX cytometer. Data were analysed using FlowJo software.

. ACE enzyme-linked immunosorbent assay
Quantification of the amount of human ACE2 in the different EV preparations and other SEC fractions was done using the human ACE2 ELISA kit (Abcam, ab235649) following manufacturer's instructions.