MicroRNAs (miRNAs), the tiny regulators of gene expression, can be transferred between neighbouring cells via extracellular vesicles (EVs) to control the expression of genes in both donor and recipient cells. How the EV-derived miRNAs are internalized and become functional in target cells is an unresolved question. We have expressed a liver-specific miRNA, miR-122, in non-hepatic cells for packaging in released EVs. With these EVs, we have followed the trafficking of miR-122 to recipient HeLa cells that otherwise do not express this miRNA. We found that EV-associated miR-122 is primarily single-stranded and, to become functional, is loaded onto the recipient cell argonaute proteins without requiring host Dicer1. Following endocytosis, EV-associated miR-122 is loaded onto the host cell argonaute proteins on the endosomal membrane, where the release of internalized miRNAs occurs in a pH-dependent manner, facilitating the formation of the exogenous miRNP pool in the recipient cells. Endosome maturation defects affect EV-mediated entry of exogeneous miRNAs in mammalian cells.
Intercellular communication in metazoan cells, within the same or different tissues, can be achieved via exchange of membrane-enclosed vesicles that can carry proteins or nucleic acids (Maas et al., 2017). This exchange of materials is considered as a very important physiological phenomenon in animals (Stahl and Raposo, 2019). These extracellular vesicles (EVs), as they are collectively known, either originate from the multivesicular bodies (MVBs) or the plasma membrane and are classified according to their origin and physical status as microvesicles, ectosomes, microparticles or exosomes (van Niel et al., 2018). Exosomes are 40–100 nm-sized vesicles that are positive for the tetraspanin protein CD63 and are formed during membrane invagination of late endosomes to generate MVBs. Following MVB fusion with the cell membrane, exosomes are released into the extracellular space (Colombo et al., 2014). EVs, and exosomes in particular, are known to carry different cargoes, including mRNAs, microRNAs (miRNAs), proteins and macromolecules, to the recipient cell to ensure intercellular communication and exchange of materials.
miRNAs are 20–22-nucleotide-long gene regulatory RNAs, which by base pairing to target messages can trigger translational repression or degradation of target RNAs (Fabian et al., 2010; Filipowicz et al., 2008). miRNAs target important regulatory genes and play crucial roles in disease pathogenesis. miRNAs have been found to be present in the body fluids either as ‘free’ entities (sometimes in complex with the argonaute proteins) or in enclosed vesicles, such as exosomes (Arroyo et al., 2011; Kosaka et al., 2010a).
Transfer of functional miRNAs by EVs to recipient cells has been observed to evoke a physiological response both in cancer and immune cells (Kogure et al., 2011; Mittelbrunn et al., 2011; Montecalvo et al., 2012; Pegtel et al., 2010; Salido-Guadarrama et al., 2014; Valadi et al., 2007). EV-mediated crosstalk between cells having different miRNA expression profiles has also been reported (Basu and Bhattacharyya, 2014). Specific sets of proteins have been observed to selectively sort and package certain miRNAs into MVBs. These MVBs then fuse with the cell membrane to release the intraluminal EVs, containing specific miRNAs, in a context-dependent manner (Cha et al., 2015; Mukherjee et al., 2016; Shurtleff et al., 2016; Villarroya-Beltri et al., 2013).
The internalization of EVs into recipient cells occurs either by endocytosis (Svensson et al., 2013; Tian et al., 2010), macropinocytosis (Tian et al., 2014), phagocytosis (Feng et al., 2010) or engulfment with the help of filopodia (Heusermann et al., 2016). However, despite recent advances in the study of EV-mediated transport of miRNA, the mode of uptake of these miRNAs and the factors that render them functional in recipient cells are entirely unexplored. Using an ectopic miRNA expression system in human HeLa cells, we have observed functional miRNA transfer to the recipient cell by EVs. This exogenous miRNA enters the recipient cell in a single-stranded form. Internalized miRNAs become associated with the recipient cell argonaute proteins in a Dicer1-independent manner. We have observed that the internalized miRNAs utilize the endocytic pathway to reach the endoplasmic reticulum (ER) of target cells to elicit their repressive action. We have substantiated our data in an in vitro assay to show that a decrease in endosomal pH causes the release of the miRNA from the internalized EVs – a prerequisite process for functional loading of internalized miRNAs onto argonaute 2 (Ago2) of the recipient cells.
Exogenously expressed liver-specific mature miR-122 is packaged in the EVs released by non-hepatic HeLa cells
To understand the mechanistic aspects of uptake of EV-enclosed miRNAs, we used HeLa cells ectopically expressing a ‘foreign’ liver-specific miR-122 as donor cells that would produce EVs containing miR-122. This miRNA is expressed abundantly in the liver cells, and is otherwise not detectable in HeLa cells (Landgraf et al., 2007). HeLa cells without the miR-122 expression cassette thus could serve as recipient cells to measure the EV-mediated uptake and functional activity of transferred miR-122. To characterize the miR-122-containing EVs, HeLa cells transfected with miR-122 expression plasmid were subjected to EV isolation by ultracentrifugation of culture supernatants (Fig. S1A). The EVs isolated from either control or miR-122-expressing cells were subjected to nanoparticle tracking analysis (NTA) to obtain an estimate of size and number of the two groups of EVs. No difference in size or concentration was observed for the two EV groups (Fig. S1B,C). Atomic force microscopy (AFM) revealed that there was also no difference in their shapes (Fig. S1D). The EV protein content showed no major change between the control and miR-122-containing EVs. The isolated EVs were found to be positive for CD63, flotillin 1, Alix (also known as PDCD6IP) and HRS (also known as HGS), and were devoid of most cellular markers, including calnexin, Lamp1 and GAPDH. Furthermore, the EVs used for analysis were free from the apoptotic body marker cytochrome c as well as lipoprotein contaminants like ApoE (Fig. S1E).
We then further confirmed the presence of mature miR-122 in the EVs isolated from miR-122-expressing donor HeLa cells compared to EVs from the control HeLa cells. The levels of mature miR-122 detected in EVs from miR-122-expressing donor cells remained unchanged upon RNase-free DNaseI treatment. This data thus ruled out the possible presence of miR-122-encoding DNA fragment as a contaminant in the EVs (Fig. S1F). Upon RT-qPCR threshold cycle (Ct) value analysis, we detected a marginal level of pre-miR-122 in EVs that was resistant to DNaseI (Fig. S1G). Previous reports suggest that EV-associated RNAs are usually localized inside the EVs (Del Pozo-Acebo et al., 2021; Fabbiano et al., 2020; Shurtleff et al., 2017; Valadi et al., 2007). To confirm the miR-122 localization within EVs, we treated the EVs with RNase A. We found protection of the majority of miR-122 from RNase A unless the EVs were subjected to sonication before the RNase A treatment. Sonication disrupts membranous structures (Alvarez-Erviti et al., 2011; Kojima et al., 2018), and we found decreased miR-122 content in EVs that were sonicated and treated with RNase A (Fig. S1H,I). Thus, the majority of EV-associated miRNAs are present inside the vesicles and protected from RNase A.
Ct value comparison revealed that the antisense strand of miR-122 (miR-122*) was present in a negligible amount, as reflected by the very high Ct value associated with it (Fig. S1J, left panel). This signifies strand-specific packing of mature miR-122 and not miR-122* into EVs. Before proceeding further, we also quantified the copy number of mature miR-122 present in the HeLa EVs either isolated from miR-122-transfected or untransfected cells and found that EVs obtained from miR-122-transfected cells contain ∼2 copies (1.7 copies on average) of the miRNA (Fig. S1J, right panel). This is consistent with a previous report where a similar copy number of miRNAs in EVs has been reported (Alexander et al., 2015). Finally, a TUNEL assay showed that miR-122 transfection did not induce apoptosis in HeLa cells (Fig. S1K).
EVs can transfer functional miRNAs between mammalian cells
Having established that the mature miR-122 is localized inside the EVs, we wanted to see the transfer and functionality of the internalized miRNA in recipient cells devoid of miR-122. To visualize the transfer process kinetics, we transfected our donor cells with a plasmid expressing GFP-tagged tetraspanin marker CD63. The EVs were isolated from these donor cells and added to recipient cells (Fig. 1A, left panel). Confocal microscopy analysis suggested internalization of CD63–GFP EVs in recipient cells compared to control cells. Three-dimensional (3D) reconstitution from several z-plane confocal slices further demonstrated the internalization of GFP-tagged CD63-containing EVs (Fig. 1A, right panels). The CD63–GFP EVs, used for the purpose of microscopy, were isolated using a commercial exosome isolation reagent. Thus, we compared them with the EVs obtained from ultracentrifugation and found no difference in size or concentration (Fig. S2A,B,D,E). This was consistent with a previous report (Cheng et al., 2019). However, upon western blot analysis, protein levels were found to be higher in the CD63–GFP-positive EVs than in the control set, suggesting an increase in protein content upon expression of CD63–GFP in donor cells (Fig. S2C). Upon incubating the recipient cells with CD63–GFP positive EVs for 0, 4 and 16 h, we found that the number of internalized GFP-positive EVs per cell increased at 16 h (Fig. 1B and Fig. 1C, left panel). Consistent with this, we measured the levels of EV-derived miR-122 in the recipient cells over time upon treatment with miR-122-positive EVs isolated by ultracentrifugation and found an increase in the internalized miRNA content at 16 h (Fig. 1C, right panel).
The miR-122 antisense strand (miR-122*) was not detected in recipient cells incubated with miR-122-positive EVs (Fig. 1D, left panel), suggesting exclusive transfer of mature miR-122 to the recipient cells. We found ∼2800 copies of miR-122 after 16 h of treatment of the recipient cells with miR-122-positive EVs (Fig. 1D, right panel). We also quantified the levels of precursor miR-122 in the recipient cells and found no change in the precursor levels in both control and miR-122 EV-treated cells (Fig. 1E), which were 500-fold less compared to those in hepatic origin Huh7 cells (data not shown).
In order to investigate the functionality of the transferred miRNA, luciferase assays were performed using a Renilla luciferase (RL)-based reporter plasmid containing a perfect binding site for miR-122 (RL-Perf-miR-122; Fig. 1F). This plasmid was transfected into recipient cells that were either treated or untreated with miR-122-containing EVs. The firefly luciferase-normalized expression levels of RL-Perf-miR-122 were found to be low in the recipient cells incubated with miR-122-positive EVs (Fig. 1G, left panel). Upon transfer of miR-122 via EVs, we found decreased expression of miR-122 target mRNAs, including cationic amino acid transporter 1 (CAT1, also known as SLC7A1) and aldolase A (ALDOA), in EV-treated cells, which signifies functional transfer of EV-derived miR-122 across donor and recipient cell boundaries (Fig. 1G, right panel). Similar to miR-122, ectopic expression of miR-146a in HeLa cells also ensured its EV-mediated transfer to recipient HeLa cells that otherwise do not express miR-146a (Fig. 1H). This result signifies the universality of the heterogeneous miRNA transfer mechanism. We investigated whether the transferred miR-122 can affect the levels and activity of endogenous miRNAs and found no change in endogenous let-7a and miR-16 levels upon miR-122 transfer. Let-7a repressive activity also remained unchanged upon miR-122 uptake for both the reporter mRNA and mRNA levels of the endogenous let-7a target K-RAS (Fig. 1I–K).
The internalized miRNAs associate with Ago2 in recipient cells in a Dicer1-independent manner
One possible mechanism that ensures miRNA homeostasis in animal cells is the EV-mediated transfer of miRNAs between neighbouring cells. This may also happen between cells in contact co-culture (Basu and Bhattacharyya, 2014). In order to revalidate the uptake of EV-derived miRNA, we adopted a contact co-culture technique to follow the miRNA transfer across cellular boundaries (Fig. 2A). In order to check for transferred miRNAs, the donor cells expressing miR-122 were co-cultured with recipient cells expressing FLAG–HA–Ago2. Immunoprecipitation of FLAG was performed to quantify the miR-122 transferred to the recipient cells (Fig. 2B). By co-culturing two different pools of HeLa cells, one expressing miR-122 and the other with a miR-122 reporter, we tested the activity of the transferred miRNA across cell boundaries. We noted transfer of miR-122 repressive activity to cells having no expression of miR-122, as expected, and this transfer had no effect on endogenous let-7a miRNA activity (Fig. 2C,D). The functionality of the transferred miR-122 was also validated by a decrease in CAT1 levels in cells co-cultured with miR-122-expressing cells (Fig. 2E).
We wanted to check the involvement of recipient cell Ago2 in the uptake process. Using isolated miR-122-containing EVs as described in the experimental scheme shown in Fig. 1A, we observed a similar time-dependent increase in the levels of Ago2-associated miR-122 upon EV treatment at 16 h (Fig. 2F). Transfer of EV-derived miRNAs to argonaute proteins was not found to be exclusive to Ago2; Ago3, upon contact co-culture, was also found to be loaded with internalized miRNAs, compared to negative control NHA–LacZ protein, when expressed in recipient HeLa cells (Fig. 2G).
In the canonical pathway, Ago2 loading of miRNA is dependent on its interaction with Dicer1. After processing of pre-miRNA by Dicer1 to generate the double-stranded miRNAs, the miRNA sense strand is then loaded onto argonaute proteins in a Dicer1-dependent manner (Song and Rossi, 2017). To test whether the internalization and argonaute-association of internalized single-stranded miRNA is dependent on the prime components of the miRNA biogenesis pathway in the recipient cell, Ago2 and Dicer1 proteins of the recipient cell were targeted using specific siRNAs. When the recipient cells were knocked down for Ago2, the internalization of EV-transferred miR-122 decreased in the recipient cell, suggesting a requirement of recipient cell Ago2 for the internalization of the EV-transferred miR-122 (Fig. 2H).
When the recipient cells were knocked down for Dicer1 and incubated with miR-122-containing EVs, the levels of internalized miR-122 remained unchanged, as compared to levels in the control (Fig. 2I). Ago2-associated miR-122 levels in cells treated with control non-targeting siRNA (siControl) or siRNA targeting Dicer1 (siDicer1) also remained unchanged in co-culture (Fig. 2J, left panel). However, in HeLa cells co-transfected with pre-miR-122 expression plasmid and siDicer1, the levels of mature miR-122 were reduced significantly along with a similar decrease, as expected, for endogenous let-7a levels (Fig. 2J, right panel). Therefore the data on internalization of EV-derived miRNA and its argonaute-loading suggest that, independent of the presence of Dicer1 in recipient cells, the transfer of miR-122 via EVs involves the single-stranded and mature form of the miRNA. If the precursor miRNA (pre-miR-122) were transferred, then knockdown of Dicer1 would be expected to affect miRNA incorporation onto recipient cell Ago2, as Dicer1 is required for pre-miRNA processing. Dicer1-independent non-canonical argonaute-loading of EV-derived mature miRNA is possibly required to elicit a quick and effective response in the recipient cells.
The internalization of miRNAs is dynamin2-dependent, and ER stress can hinder the uptake process
To understand how the EV-derived miRNA enters the recipient cell and where it is finally localized, the involvement of specific proteins known to control endocytic processes were targeted. Dynamin2 is needed for pinching-off of vesicles into the interior of the cell from cell membrane in certain endocytic pathways (De Camilli et al., 1995; Feng et al., 2010). The dependence of EV internalization on dynamin2 was confirmed by microscopic uptake analysis of CD63–GFP-tagged vesicles in recipient HeLa cells where dynamin2 expression was reduced using siRNA (Fig. S3A,B, left panel). Upon knockdown of dynamin2 in the recipient cell, the internalization of miRNA was significantly reduced, as was the functionality (Fig. S3B, right panel, C). This suggests that dynamin2, a protein prerequisite for endocytosis, is also needed for the internalization of EV-containing miRNAs.
Having checked the effect of dynamin2 on miR-122 uptake, we used inhibitors for different endocytosis pathways, as used in previous reports (Costa Verdera et al., 2017), to see the uptake of miR-122 via EVs. We have seen that upon treatment of recipient cells with EIPA [5-(N-ethyl-N-isopropyl)amirolide], a macropinocytosis inhibitor, there was a decrease in the internalization of miRNA content. Similarly, upon treatment with genistein (a clathrin-independent pathway inhibitor) as well as chlorpromazine (a clathrin-dependent pathway inhibitor), we found a decrease in the uptake of miR-122, suggesting that macropinocytosis, clathrin-dependent and clathrin-independent pathways can facilitate the EV internalization process of miR-122 in HeLa cells (Fig. S3D). Some clathrin-independent pathways (like CDC42- and ARF6-regulated pathways) and the macropinocytosis pathways do not use dynamin2 for the pinching-off of internal vesicles (Doherty and McMahon, 2009; Mayor and Pagano, 2007). Thus, in our case, it is possible that the EV-associated miRNAs are either internalized by a clathrin-dependent pathway or by dynamin-dependent but clathrin-independent pathways (for instance caveolae-dependent pathways).
Because it has been reported that the ER serves as the nucleation site for the miRNA machinery (Stalder et al., 2013), we wanted to see what happened to the internalization of EV-derived miRNA upon ER stress, which is known to affect miRNA machineries (Mukherjee et al., 2016). Thapsigargin is an ER-stress inducer, and treatment of recipient cells with thapsigargin followed by incubation with miR-122-containing EVs resulted in low internalization and activity of EV-derived miRNA-122 in treated cells (Fig. S3E–G). Unlike the Dicer1-dependent canonical pathway, where availability of the target messages has a positive influence on cognate miRNA formation (Bose and Bhattacharyya, 2016), we did not detect an increase but rather a decrease of EV-delivered miR-122 content in the presence of its target message (RL-3×Bulge-miR-122). No change in Ago2-associated miRNA level in co-cultured recipient cells was detected in the presence of target message as well (Fig. S3H,I). This data further supports a non-canonical miRNA loading pathway prevalent for EV-internalized miRNAs.
EV-derived miRNAs enter recipient cells via the endosomal pathway
We wanted to assess the subcellular localization of the internalized miRNA in the recipient cells. We initially quantified the EV-derived miR-122 levels in the cytosolic and membrane fractions of the recipient cells by digitonin fractionation (Fig. S3J). Digitonin treatment allows permeabilization of cells. This detergent makes the cell membrane porous by solubilizing high-cholesterol cell membranes and thus separating the cytosolic part as soluble fraction and the detergent-insoluble membranes as the pellet fraction (Barman and Bhattacharyya, 2015). We observed that in cells that received EVs containing miR-122, the internalized miRNA was found in the membrane fractions (Fig. S3K,L). This was similarly observed when miR-122 was delivered via liposomes (Fig. S3K). To further dissect the subcellular localization of these internalized miRNAs, cellular lysates of recipient cells post 16 h EV treatment were collected in isotonic solution and fractionated using Optiprep density gradient ultracentrifugation to separate the organelles. Western blot analysis was performed using the fractions to demarcate different subcellular compartments (Fig. 3A,B). The different subcellular fractions were collected and pooled together depending on the presence of marker proteins for different organelles (Fig. 3C), and RNA was isolated from the different pooled fractions. Fractions 2 and 3 were enriched for early endosomes, while fractions 4–6 were enriched for lysosomes and fractions 7–9 had enrichment for ER marker proteins. In a steady-state context, the majority of the internalized EV-derived miRNAs (total and Ago2-associated) were found to be associated with the ER (Fig. 3D) and possibly with ER-attached polysomes. From the relative Ct value analysis, there was also an increase in polysomal miR-122 content upon treatment of recipient cells with miR-122 containing EVs (Fig. 3E). This was accompanied by a concomitant decrease in the target CAT1 mRNA level associated with the polysomes (Fig. 3F).
We explored the effect of depletion of specific endosomal maturation pathways proteins in mammalian cells on EV-mediated miRNA entry and function. The endosomal pathway comprises distinct membrane-bound compartments, which internalize molecules from the plasma membrane and recycle them back to the surface or sort them for lysosomal degradation. Three proteins of the endosomal pathway, Rab5a (an early endosomal GTPase required for early to late endosomal maturation), Rab7a (a late endosomal GTPase required for endosome maturation and lysosome interaction) and RILP (Rab-interacting lysosomal protein, which facilitates the interaction of late endosomes with lysosomes) were targeted specifically to dissect which steps of endocytosis are required for functional miRNA transfer to ER attached polysomes (Fig. 3G). In a steady state, endosome numbers were not found to significantly alter on knockdown of the endosomal proteins individually (Fig. S4A–C). Endosomes were visualized in cells expressing an endosomal targeting variant of YFP protein (Endo-YFP). Measuring the effect of knockdown of these factors, we found a stronger effect of RILP knockdown on internalized miRNA content (Fig. 3H). However, the knockdown of RILP marginally altered the levels of endogenous let-7a and exogenously expressed miR-122 in recipient HeLa cells (Fig. 3I). Interestingly, Rab5a depletion had a similar effect on both EV-internalized miR-122 and endogenous miRNAs, and in each case an increase in overall miRNA content was detected. Surprisingly, the increase in miR-122 content was not reflected in its repressive activity. A reduction in miRNA activity was noted after knockdown of Rab5a or Rab7a and more significantly upon RILP depletion (Fig. 3J). Using the co-culture model of miRNA transfer (Fig. 3K), we did observe an increase in Ago2 incorporation of internalized miRNAs upon depletion of the Rabs or RILP but, conversely, we also observed a decrease in repressive activity (Fig. 3L,M). We performed microscopic analysis of the number of CD63–GFP-positive EVs internalized into recipient cells either knocked down for Rab5a, Rab7a or RILP. Contrary to our finding that the levels of miRNA increase upon knockdown of RILP, we found an overall decrease in the number of EVs internalized in the recipient cells knocked down for either Rab5a or RILP (Fig. S4D,E). As the endocytic pathway is hampered upon siRNA transfections, the uptake of EVs is probably blocked but the internalized miRNA is stored elsewhere away from its target mRNA. The mRNA–miRNA interaction, which leads to degradation of the miRNA (Bitetti et al., 2018; de la Mata et al., 2015; Ghini et al., 2018), being absent causes accumulation of EV-derived miR-122 in Rab7a- or RILP-depleted cells. However, this accumulation and Ago2 association appeared to be highest in the case of RILP depletion due to low degradation of EV-derived miR-122, which is consistent with a previous observation that RILP knockdown delays the trafficking of EGFR from early to late endosomes and prevents their degradation in lysosomes (Progida et al., 2007). Thus, our data suggest that the newly formed miRNP is stable and does not reach its target, perhaps because of an altered localization when endosomal pathways are hampered.
Bafilomycin interferes with acidification of the endosomal compartments thus leading to hindered EV-derived miRNA compartmentalization
The endosomal pathway consists of different compartments that carry cargo within organelles. Each component of the endosomal pathway has a specific pH, culminating in the lysosomal compartment having a pH of less than 4 (Fig. 4A). From the analysis described in the above experiments, it seems that majority of the internalized miRNAs are localized with ER-attached polysomes at steady state upon treatment with miR-122-loaded EVs. With depletion of endogenous Rab7a protein and RILP individually, we observed an increase in miRNA–Ago2 association specifically in the endosome fraction, which is possibly a result of the miRNA remaining blocked in the endosomes due to a faulty endosomal pathway (Fig. 4B). Thus, our data suggest that the EV-derived miRNA is released in the host cell at the level of endosomes, and a simple mechanism for this would be the fusion of the endocytosed EV membrane with the surrounding endosomal membrane.
Bafilomycin is a proton pump blocker that acts on the endosomal maturation pathway by preventing the acidification of endosomes, which is a prerequisite for fusion with lysosomes (Johnson et al., 1993). The effect of bafilomycin has also been observed on the transfer of cargo from early endosomes to the late endosomes (Bayer et al., 1998). Importantly, it has also been reported that low pH promotes the fusion of exosomal membrane with cell membranes (Parolini et al., 2009). Thus, it can be hypothesized that upon lowering the pH of endocytic vesicles, fusion between the endosomal membrane and internalized EV membrane should occur. Since the luminal pH lowers as endosomes mature, endosomal maturation should be an effective mechanism to trigger the release of EV-derived miRNAs in the proximity of endosomes, and its loading on host Ago2. To test this hypothesis, we blocked endosomal acidification using bafilomycin. We detected an increase in EV-derived miRNA content upon treatment of recipient cells with bafilomycin, concomitant with a decrease in the level of repression by internalized miRNAs (Fig. 4C–F). Digitonin fractionation revealed an increase in membrane association of miR-122 in bafilomycin-treated cells compared to that in control cells, whereas the levels of miR-122 remained unchanged in the cytosolic fractions (Fig. 4G). We also revalidated the subcellular fractionation in recipient cells receiving miR-122 EVs either treated with DMSO or bafilomycin and found a consistent accumulation of miR-122 in the endosomes compared to the ER (Fig. 4H). This was accompanied by a relative reduction in Ago2-bound internalized miR-122 levels compared to those in RILP-knockdown cells, where RILP depletion also increased the total internalized miRNA content with a concomitant increase in Ago2 association (Fig. 4I). Thus, our data suggest that bafilomycin, like knockdown of RILP, can block the EV-derived miRNAs in the endosomal compartment. However, bafilomycin interferes with proper acidification of the endosomes, which thereby hinders the endosomal Ago2 association of the internalized miRNAs.
Late endosomal maturation and low pH are required for release and Ago2 loading of endocytosed EV-derived miRNAs
To prove that low pH favours release of miRNA and miRNP formation with endosomal Ago2, we first isolated endosomes from HeLa cells using ultracentrifugation. The cell lysates were first subjected to subcellular fractionation followed by further purification of endosome-positive fractions by ultracentrifugation (Fig. 5A). These endosomes were analysed by western blotting to check for characteristic proteins and other subcellular organelle contaminations (Fig. 5B).
Bafilomycin prevents the lowering of endosomal pH, and thus endosomes isolated from bafilomycin-treated cells should have a lower capacity to release endocytosed EV-derived miRNAs, and this capacity should be restored by lowering endosomal pH. We thus purified endosomes after treatment of cells with bafilomycin and miR-122-containing EVs, then incubated them in vitro in a buffer at pH 5 with or without FCCP, a protonophore that can equalize the pH across biological membranes. We then followed the release of EV-derived miRNAs from endosomes after the in vitro reaction by separating the endosomes as a pellet from the supernatant by ultracentrifugation (Fig. 5C). It was found that the internalized miR-122 content in the supernatant was increased in the presence of FCCP, and this increase occurred in a time-dependent manner (Fig. 5D, left panel, E). We also quantified the levels of endogenous let-7a purified with the endosomes, and found a release of let-7a in the supernatant of the endosomes treated with pH5 and FCCP (Fig. 5D, right panel). The endosomal pellets and supernatants were analysed for the endosomal protein content, which was unchanged upon treatment with FCCP (Fig. 5G). The endosomes were analysed using AFM and NTA, which showed that FCCP did not cause damage or a change in size and number of the endosomes (Fig. S5A–C).
These data thus substantiate the idea of pH-dependent release of miRNA content from endosomes for subsequent reloading to membrane-attached Ago2. This was further supported by an in vitro loading assay performed using endosome-associated Ago2. Incubation of single-stranded miR-122 with endosomes resulted in a time-dependent miRNA loading of endosomal Ago2 (Fig. 5F). With isolated endosomal fractions from an Optiprep gradient, we found that the Ago2 present on endosomes was sensitive to Proteinase K even in the absence of any detergent, such as Triton X-100. ApoE is a lipoprotein that is endocytosed after binding with low-density lipoprotein receptors, and hence should be present within endosomes (Dekroon and Armati, 2002). ApoE was found to be resistant to increasing concentrations of Proteinase K treatment but was degraded when the endosomes were treated with both Proteinase K and Triton X-100. Therefore, a large fraction of endosome-localized Ago2 is present on the outer side of the endosomes and thus is sensitive to Proteinase K, as was observed for other endosomal membrane proteins including EEA1 and HRS (Fig. 5H). Since there are contact points between endosomes and the ER (Friedman et al., 2013), it is expected that endosomal Ago2 freshly loaded with EV-derived miRNAs can easily access ER-located target mRNAs for repression.
There is ample evidence in support of a role for EVs in carrying genetic material and other biologically important cargoes to other target cells to elicit a response there (Au Yeung et al., 2016; Cossetti et al., 2014; Thery et al., 2002; Valadi et al., 2007). However, a major limitation in the EV and miRNA field is a deficit in the understanding of the mechanistic aspect of functional internalization of EV-carried miRNAs into recipient cells. Our findings shed light on this mechanism of transfer of miRNA between mammalian cells.
In this study, we have elucidated how a functional miRNA can be transferred from donor to recipient cells via EVs (Fig. 5I). The transferred miRNA does not affect the pre-existing endogenous miRNA levels or their activities. We have observed that the EV-derived miRNA is transferred primarily in a Dicer1-independent manner in its mature single-stranded form. EV sorting of Ago2 has been reported to stabilize miRNAs in exosomes or microvesicles (Beltrami et al., 2015; Lv et al., 2014; McKenzie et al., 2016). Despite a previous claim that Ago2 is present in EVs, a recent publication has questioned the presence of protein components of miRNP machinery in mammalian cell-derived EVs (Jeppesen et al., 2019). Therefore, it is most likely that single-stranded miRNAs may travel from one cell to another via EVs in an Ago2-unbound form. We have observed that the EV-derived miRNA gets bound to the recipient cell Ago2 that drives the sustained function of the transferred miRNA in mammalian cells.
We have demonstrated that the uptake of the EV-containing miRNA is dependent on dynamin2, a GTP binding protein involved in the endocytosis pathway. The involvement of the endocytic pathway in the uptake of EVs has been discussed in previous studies (Bonsergent and Lavieu, 2019; Joshi et al., 2020). We also conclude that miRNAs utilize the endosomal pathway to reach the ER of the recipient cell. The compartmentalization of these miRNAs to the ER suggests adaptability of the EV-derived miRNAs to elicit a repressive activity in the recipient cell. We have found that knockdown of the endosomal pathway effector protein RILP leads to accumulation of the internalized miRNA in the recipient cell and defects in repression of target genes. Subcellular localization assays revealed a differential localization and Ago2 association of the transferred miRNAs upon knockdown of RILP. These results explain why the internalized miRNA cannot be functional; it fails to reach the ER for its target mRNA interaction and repressive activity in endosome maturation-defective cells. The increase in miRNP content could be due to lowered target-dependent miRNA turnover (Bitetti et al., 2018; de la Mata et al., 2015; Ghini et al., 2018) in RILP-depleted cells as a result of possible reduced miRNP–target RNA interaction.
The pH of the endosomal vesicles plays a role in release of cargo molecules to the late endosomal and lysosome-attached cytoplasmic components. It has been observed that endosomal pH, and hence maturation, plays a role in uncoating of viral particles for their infectivity (Li et al., 2014). Furthermore, low pH can influence the release and uptake of exosomes by fusion in cancer cells (Parolini et al., 2009). Hence, an in vitro assay was performed to validate the hypothesis that the change in pH during endosomal maturation may play a role in release of the late endosomal contents into the ER via contact points. The observed loading of single-stranded miRNA to Ago2 attached to endosome membranes strongly supports the concept of endosomal miRNP formation involving membrane-bound Ago2 and internalized miRNAs.
miRNA uptake via EVs ensures sensing of extracellular status and its coupling with gene repression machineries of the respective cells. By responding to the cellular miRNA levels, the miRNA uptake machineries may balance the cellular miRNA content to stabilize gene repression. Therefore, this may be considered as a salvage pathway for de novo miRNA formation in mammalian cells to buffer the cellular miRNA content.
MATERIALS AND METHODS
Cell culture and inhibitor treatments
Human HeLa, HEK293 (from ATCC) and FLAG–HA–Ago2-expressing HEK293 [described in Bose and Bhattacharyya (2016)] cells were cultured in Dulbecco's Modified Eagle's medium (DMEM; Gibco-BRL) supplemented with 2 mM L-glutamine and 10% heat-inactivated fetal bovine serum (FBS). EV-depleted growth medium was made with DMEM and 10% EV-depleted FBS. For depletion of EVs from FBS, the FBS was ultracentrifuged at 100,000 g for 4 h and then added to the DMEM. Characterization of EV depletion was performed as reported previously (Mukherjee et al., 2016). Plasmid transfections were conducted using Lipofectamine 2000 (Invitrogen) following the manufacturer's instruction. All plasmid constructs used have been reported previously (Ghosh et al., 2015), and details are provided in Table S1.
For induction of ER stress, the recipient cells were pre-treated for 1 h with 2.5 µM thapsigargin (Sigma) before treatment with miR-122-containing EVs for 16 h. For uptake studies using different endocytic pathway inhibitors, the recipient cells were pre-treated with 50 µM EIPA [5-(N-ethyl-N-isopropyl) amirolide; Sigma], a macropinocytosis inhibitor; 50 µM genistein (a clathrin-independent pathway inhibitor; Sigma); or 10 µM chlorpromazine (a clathrin-dependent pathway inhibitor; Sigma) for 30 min prior to EV treatment for 16 h.
siRNA and plasmid transfections
siRNAs against different endocytic pathway proteins (dynamin2, Rab5a, Rab7a and RILP), Dicer1 and Ago2, as well as control non-targeting siRNA, were purchased from Dharmacon. siRNA transfection was performed using RNAi Max (Invitrogen) according to the manufacturer's instructions. HeLa cells were transfected with 15 pmoles of siRNA per well of a 24-well plate. The siRNA-transfected cells were split 48–72 h post transfection for proper knockdown. Details of the siRNAs used are shown in Table S5. For EV isolation, donor HeLa cells were transfected with 1 μg pmiR-122 per well of a 6-well plate. pmiR122 plasmid, a plasmid encoding the precursor miR-122, was described previously (Ghosh et al., 2013)
EV isolation and characterization
HeLa cells transfected with pmiR-122 were split 24 h post transfection into 90 mm plates and incubated for 24 h at 70–80% confluency. These cell culture supernatants were subjected to EV isolation as described previously, with minor modifications (Thery et al., 2006). For EV isolation, cells were grown in EV-depleted medium to prevent any background from exosomes or EVs present normally in FBS. Briefly, the cell supernatant (EV-depleted growth medium that was used to culture the donor miR-122-expressing HeLa cells) was centrifuged at 2000 g for 15 min to remove cell debris. Next, the cell supernatant was collected and centrifuged at 10,000 g for 30 min. The supernatant that was obtained was passed through a 0.22μm filter unit to further clear it. This was followed by ultracentrifugation of the supernatant at 100,000 g for 90 min. After ultracentrifugation, the pellet was resuspended in EV-depleted growth medium and added to recipient cells.
For characterization of EVs by AFM, the EVs were isolated on a 30% sucrose cushion, and the layer of EVs was further washed and pelleted with 1× phosphate-buffered saline (PBS) by ultracentrifugation at 100,000 g for 90 min and resuspended in 1 ml PBS. Then, 5 µl of the EV suspension was placed onto a mica sheet (ASTM V1 Grade Ruby Mica from MICAFAB, Chennai, India) and dried for 15 min. The sample slides were then gently washed with autoclaved MilliQ water to remove molecules that were not firmly attached to the mica before being dried again. AAC mode AFM was performed using a Pico plus 5500 ILM AFM (Agilent Technologies, USA) with a piezoscanner maximum range of 9 µm. Microfabricated silicon cantilevers of 225 µm in length were used (Nano sensors, USA). Images were processed by flatten using Pico view1.1 version software (Agilent Technologies, USA).
For NTA, EVs were resuspended in 1 ml PBS. EVs were diluted 10-fold, and 1 ml of diluted EVs was injected into the sample chamber of a nanoparticle tracker (Nanosight NS300).
To detect whether the EV-associated miRNAs were present on the surface or inside the EVs, the EVs were isolated from miR-122-transfected HeLa cells and either sonicated or not sonicated. These EVs were then treated with RNaseA (10 µg/100 µl) for 10 min at 37°C before the RNA was isolated, and RNA content was subjected to real-time qPCR.
In case of CD63–GFP EVs, which were used for microscopic analysis of EV uptake, the EVs were isolated using commercially available EV isolation reagent, the method of which is described below (see section Microscopic analysis of EV entry into recipient cells).
Contact co-culture model
Donor HeLa cells were transfected with pmiR-122, while recipient cells were transfected either with pRL-Con, pRL-Perf-miR-122 or FLAG-HA-Ago2 plasmids depending on whether the luciferase assay or immunoprecipitation (IP) assays were being performed. After transfection, the donor and recipient cells were seeded in equal amounts and allowed to interact for 24 h after which either luciferase assay or IP was carried out. IP was followed by quantification of the amount of miRNA transferred to the recipient Ago2 in the contact co-culture system, whereas the luciferase assay depicted the amount of RL reporter repression in recipient cells in the presence of internalized miRNA from donor cells.
Copy number calculation
In order to calculate the copy number of miR-122 in the EVs isolated from miR-122 untransfected or transfected HeLa cells, we made a standard curve with known concentrations of synthetic miR-122. Following this, the concentration of miR-122 in the EVs was quantified. We used the formula reported previously (Simmonds, 2019):
The yield of RNA was divided by the amount of RNA used for the PCR reaction (200 ng). The copy number of miR-122 in recipient HeLa cells was also calculated in the above manner using the same formula.
The plasmids pRL-Con (encoding RL control; RL-con) and pRL-Perf-miR-122 (encoding the RL-Perf-miR-122 reporter) were a kind gift from Dr Witold Fillipowicz (Friedrich Meisher Institute Basel, Switzerland). For miRNA repression assays, 30 ng of Renilla luciferase (RL) reporter plasmids (both pRL-Con and pRL-Perf-miR-122) with 300 ng of firefly luciferase (FL) plasmid (pGL3-FF) were co-transfected per well of a 12-well plate. RL and FL activities were measured using a Dual-Luciferase Assay Kit (Promega, Madison, WI) following the supplier's protocol on a VICTOR X3 Plate Reader (PerkinElmer, Waltham, MA). The RL expression levels for reporter and control were normalized against FL levels. These normalized values were then used to calculate fold repression as the ratio of normalized control RL values to normalized reporter RL values.
For IP of Ago2, HeLa cells were transfected with FLAG–HA-tagged Ago2 plasmid. Briefly, Protein G agarose beads (Invitrogen) or anti-FLAG-M2 agarose beads (Sigma) were used for FLAG-tagged Ago2 IP. For HA, beads were blocked with 5% BSA in lysis buffer [20 mM Tris-HCl, pH 7.5, 150 mM KCl, 5 mM MgCl2, 1 mM dithiothreitol (DTT), 0.5% Triton X-100, 0.5% sodium deoxycholate and 1× PMSF (Sigma)] for 1 h followed by antibody incubation (1:50 dilution) for 4 h at 4°C. For IP reactions, HeLa cells were lysed in lysis buffer for 20 min at 4°C. The lysates, clarified by centrifugation at 3000 g for 10 min, were incubated with HA antibody pre-bound Protein G Agarose beads or pre-blocked anti-FLAG M2 beads and rotated overnight at 4°C. Subsequently, the beads were washed thrice with 1× IP buffer (20 mM Tris-HCl, pH 7.5, 150 mM KCl, 5 mM MgCl2 and 1 mM DTT), and separated into two halves for RNA and protein analysis from the bound Ago2 on the beads. All antibody information is in Table S4.
For digitonin fractionation, the recipient cells were pelleted and lysed using digitonin lysis buffer [10 mM Tris-HCl, pH 7.5, 25 mM KCl, 5 mM MgCl2, 1 mM CaCl2, RNase inhibitor (Applied Biosystems), 1 mM DTT, PMSF (Sigma) and 50 µg/ml digitonin (Calbiochem)] for 10 min on ice. This was followed by centrifugation at 2500 g for 5 min, and the supernatant (soluble cytosolic fraction) was collected. The membrane pellet was washed again in the same lysis buffer without digitonin. The fractions were divided for protein analysis and RNA isolation using TriZol LS (Invitrogen).
Optiprep density gradient ultracentrifugation
Optiprep (Sigma-Aldrich, USA) was used to prepare a 3–30% continuous gradient in a buffer containing 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES (pH 7.0) for separation of subcellular organelles. HeLa cells were washed with PBS and homogenized with a Dounce homogenizer in a buffer containing 0.25 M sucrose, 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES (pH 7.0) in addition to 100 μg/ml cycloheximide, 5 mM vanadyl ribonucleoside complex (VRC; Sigma-Aldrich), 0.5 mM DTT and 1× PMSF. The lysate was clarified by centrifugation at 1000 g for 5 min twice, then layered on top of the prepared gradient. The tubes were centrifuged at 133,000 g for 5 h for separation of the gradient, and ten fractions were collected by aspiration from the top. The fractions were pooled accordingly for subsequent analysis of proteins and RNA. For immunoprecipitation of FLAG–HA–Ago2 from the fractions, the pooled fractions were lysed with lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM KCl, 5 mM MgCl2, 1 mM DTT, 0.5% Triton X-100, 0.5% sodium deoxycholate and 1× PMSF) for 20 min at 4°C. The lysate was clarified by centrifugation at 16,000 g for 5 min and incubated with anti-FLAG-M2 agarose beads overnight followed by RNA and protein analysis of the Ago2 associated with the beads.
Microscopic analysis of EV entry into recipient cells
To detect the uptake of EVs by microscopy, the donor HeLa cells were transfected with CD63–GFP expression plasmid encoding a GFP-tagged tetraspanin. The supernatant of these cells was used for isolation of EVs from the cell culture medium by Exosome Isolation Reagent (Thermo Scientific). The supernatant was cleared of cell debris by centrifugation at 2000 g for 30 min, followed by addition of the reagent, which was incubated overnight at 4°C. On the following day, the supernatant was centrifuged at 10,000 g for 60 min to pellet the EVs. These were added to the recipient cells seeded on coverslips in different conditions to observe under the microscope. Cells were fixed using 4% paraformaldehyde for 20 min. For detection of proteins, including β-tubulin, blocking and permeabilization was performed using 1% BSA, 10% goat serum and 0.1% Triton X-100 for 30 mins. The anti-β-tubulin (mouse) was used at 1:1000 dilution. Alexa Fluor 568-conjugated anti-mouse IgG secondary antibody (Molecular Probes) was used at 1:500 dilution. All the microscopic detection was done using a Zeiss LSM800 microscope followed by analysis using Imaris software.
In vitro miRNA release assay
To see the effect of pH on release of miRNA from endocytic vesicles, an in vitro assay was used. For this, the recipient cells were incubated with 25 nM bafilomycin (pre-treated for 1 h) along with miR-122-positive EVs for 16 h. The cell lysates were subjected to Optiprep density gradient ultracentrifugation, and the endosomal marker-positive fractions 2, 3 and 4 were collected and diluted with buffer containing 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES (pH 7.0). This was followed by further purification of the endosomes by ultracentrifugation at 133,000 g for 2.5 h. The endosome pellet was then resuspended in 100 µl buffer (20 mM Tris-HCl, pH 5, and 1 mM ATP) with or without 5 µM FCCP. The endosomal suspension was then incubated for 30 min at 37°C. The reaction was stopped by transferring the mixture to 4°C. This was further diluted with a buffer containing 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES (pH 7.0). Then, the mixture was ultracentrifuged at 133,000 g for 2.5 h to pellet the endosomes, and then the supernatant was used to isolate RNA to see the amount of miRNA released from the vesicles. The remaining supernatant and pellet were used for western blot analysis of marker proteins to rule out the contamination of endosomes in the supernatant.
Proteinase K protection assay
In order to detect whether the Ago2 protein is present on the outer or inner surface of the endosomal membranes, a Proteinase K assay was used. Briefly, HEK293 cells were lysed, then loaded on an Optiprep density gradient and ultracentrifuged as mentioned above. Then, the endosome-enriched fractions 2, 3 and 4 were collected. The endosomal fractions were incubated with 0, 5, 10 or 20 ng/µl Proteinase K without Triton X-100, or with 20 ng/µl Proteinase K with 1% Triton X-100, for 30 min at 37°C. After the reaction, the solutions were subjected to methanol-chloroform protein precipitation. Then, western blotting for Ago2 was performed to see whether the Ago2 was present on the outside or inside of the endosomes.
For total polysome isolation, HeLa cells were lysed in a buffer containing 10 mM HEPES (pH 8.0), 25 mM KCl, 5 mM MgCl2, 1 mM DTT, 5 mM VRC, 1% Triton X-100, 1% sodium deoxycholate and 1× PMSF (Sigma) along with cycloheximide (100 μg/ml; Calbiochem). The lysate was cleared at 3000 g for 10 min followed by another round of pre-clearing at 20,000 g for 10 min at 4°C. The clarified lysate was loaded on a 30% sucrose cushion and ultracentrifuged at 100,000 g for 1 h at 4°C. The sucrose cushion was washed with a buffer (10 mM HEPES, pH 8.0, 25 mM KCl, 5 mM MgCl2 and 1 mM DTT), ultracentrifuged for an additional 30 min, and the polysomal pellet was finally resuspended in polysome buffer (10 mM HEPES, pH 8.0, 25 mM KCl, 5 mM MgCl2, 1 mM DTT, 5 mM VRC and 1× PMSF) for protein and RNA isolation and estimation.
In vitro endosomal Ago2 miRNA loading assay
To analyse the amount of miRNA bound to Ago2 in endocytic vesicles, an in vitro assay was performed. For this, lysate from HEK293 cells expressing FLAG–HA–Ago2 was subjected to Optiprep density gradient ultracentrifugation, as mentioned above, and the endosomal fractions 2, 3 and 4 were collected and diluted with buffer (78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES, pH 7.0). This was followed by further purification of the endosomes by ultracentrifugation for 2.5 h. The endosome pellet was then resuspended in 50 µl buffer containing 0.25 M sucrose, 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA, 50 mM HEPES pH 7.0, 100 μg/ml of cycloheximide, 0.5 mM DTT, RNase Inhibitor (Applied Biosystem) and 1× PMSF. These endosomal suspensions were incubated with 500 fmol synthetic miR-122 for either 0, 15 or 30 min at 37°C. The reaction was stopped by further dilution with the same endosome resuspension buffer. The endosomes were lysed in a lysis buffer comprising 20 mM Tris-HCl, pH 7.5, 150 mM KCl, 5 mM MgCl2, 1 mM DTT, 1% Triton X-100 and 1× PMSF for 20 min at 4°C. The endosomes were then sonicated with three pulses for 10 s each followed by centrifugation at 16,000 g for 5 min at 4°C to clear the lysate. This was then loaded onto anti-FLAG-M2 agarose beads for overnight immunoprecipitation of the cellular Ago2. The beads were washed on the following day and divided for analysis of protein and Ago2-associated RNA.
The samples (cell lysates, membrane fraction and immunoprecipitated proteins) were subjected to SDS–PAGE, transferred to PVDF membranes and probed with specific antibodies for a minimum of 16 h at 4°C. Following overnight incubation with antibody, membranes were washed and incubated at room temperature for 1 h with secondary antibodies conjugated with horseradish peroxidase (1:8000 dilutions: Life Technologies). Imaging of all western blots was performed using an UVP BioImager 600 system equipped with VisionWorks Life Science software (UVP) V6.80. The complete antibody list is provided in Table S4. We observed single CD63 bands in western blots similar to what is reported in many studies (Basu and Bhattacharyya, 2014; Cossetti et al., 2014; Kosaka et al., 2010b; Mukherjee et al., 2016; Sento et al., 2016; Zhu et al., 2014) and unlike the smears expected for a glycosylated protein like CD63.
RNA isolation and real-time qPCR
Total RNA was isolated using TriZol or TriZol LS reagent (Invitrogen) according to the manufacturer's protocol. miRNA assays using real-time qPCR were performed using specific primers shown in Table S3. Real-time analyses by two-step reverse transcription qPCR (RT-qPCR) were performed for quantification of miRNA levels on a Bio-Rad CFX96TM real-time system using an Applied Biosystems Taqman chemistry-based miRNA assay system. Cycles were set according to the manufacturer's protocol. Samples were analysed in triplicates. The comparative Ct method, which includes normalization by the U6 snRNA or 18 s rRNA, was used for relative quantification. Further details of primers and miRNA assays are in Tables S2 and S3, respectively.
All graphs and statistical significances were calculated using GraphPad Prism 5.00 (GraphPad, San Diego). Experiments were performed for a minimum of three times, unless otherwise mentioned. Paired and unpaired Student's t-tests were used for determination of P-values.
We acknowledge Witold Filipowicz for providing plasmid constructs.
Conceptualization: S.N.B.; Methodology: B.G., E.B., S.N.B.; Validation: B.G., S.N.B.; Formal analysis: B.G., S.N.B.; Investigation: B.G.; Resources: E.B., S.N.B.; Data curation: B.G.; Writing - original draft: S.N.B.; Visualization: S.N.B.; Supervision: S.N.B.; Project administration: S.N.B.; Funding acquisition: E.B.; S.N.B.
S.N.B. is supported by The Swarnajayanti Fellowship (DST/SJF/LSA-03/2014-15) from the Department of Science and Technology, Ministry of Science and Technology, India. B.G. acknowledges support from a Council of Scientific and Industrial Research, India, fellowship and an Indo-French Centre for the Promotion of Advanced Research (CEFIPRA) Raman–Charpak travel exchange fellowship. This work was also supported by a High Risk High Reward Grant (HRR/2016/00093) from the Science and Engineering Research Board (SERB), Department of Science and Technology, Ministry of Science and Technology, Govt. of India, India, and by CEFIPRA project grant 6003-J.
Peer review history
The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/DOI/10.1242/jcs.253914
The authors declare no competing or financial interests.