RNA-Based Information Transfer via Extracellular Vesicles

RNA-Based Information Transfer via Extracellular Vesicles: History

Please note this is an old version of this entry, which may differ significantly from the current revision.

Subjects: Agriculture, Dairy & Animal Science

Contributor:

Extracellular vesicles (EVs) are 50–1000 nm vesicles secreted by virtually any cell type in the body. They are expected to transfer information from one cell or tissue to another in a short- or long-distance way. RNA naturally present in EVs might be limited in a physiological context.

extracellular vesicles
exosome
RNA
miRNA

1. Introduction

Extracellular vesicles (EVs) are sub-cellular entities delineated by a lipid bilayer, containing biomolecules from parental cells, released either spontaneously or after induction. EVs are subcellular entities that partly reflect the composition of their parental cells, containing a portion of parental cell cytosol (proteins, RNA, and even organelles, etc.) encapsulated by a bilayer membrane with membrane proteins, lipids, etc. EVs contribute to intercellular communication by delivering a variety of bio-molecule cargo like nucleic acids, proteins, and lipids that modify the recipient cells. Their composition depends on the mother cells and on the environmental cues triggering EV secretion. Extracellular vesicles are usually described in three common subtypes: exosomes are 50–200 nm entities produced in multivesicular bodies and secreted after fusion with the plasma membrane, microvesicles are 100–1000 nm vesicles shed directly by the plasma membrane by budding and apoptotic bodies are 50–5000 nm objects secreted specifically during apoptotic cell death.

The hypothesis that EV’s main activity could be due to RNA transfer has been raised in the seminal publication from Valadi et al. [9] in 2007. In this article, the authors convincingly show that EVs purified from murine cell lines contain small size RNAs, and that these RNAs may be found in human cells after exposition to EVs. The translation of these RNAs in protein and its relevance in physiology is more subject to caution as it is only based on a proteomic screening that reported the detection of three murine proteins (not detected in all samples) in lysate from human cells (incubated with murine EVs) that were originally not present in the EV preparation (only 271 protein detected in EVs). Of these three, one was not detected in EV’s RNA. Due to intrinsic limits in the sensibility and specificity of non-targeted proteomics, it is possible that this small signal may be a false positive, and more importantly that this transfer is due to the use of high and non-physiologic concentrations of EVs that do not mirror the in vivo environment (8:1 producing/recipient cell ratio).

2. RNA-Based Mechanism of Action for Native EVs

Many teams reported an (mi)RNA-based mechanism of action for EVs. Most proofs of concept are however in vitro, and in vivo data are relatively scarce or not designed to claim an RNA based mechanism of action delivered via EVs at long distances and/or limited to short-distance transfer (Table 1):

In glioblastoma (GBM), Abels et al. describe short distance communication through EVs from tumor cells to microglia to induce microglia reprogramming. The presence of EVs was detected in 0.3% of microglial cells, and the presence of the miRNA of interest transferred by EVs was detected in these sorted 0.3% of cells. However, no clear target protein silencing was found (only 4 out of 59 validated targets). It may be possible that the miRNA detected would partly be coming from the retention of EVs (and its associated miRNA) in endosome of microglial cells, without intracytosolic delivery [19].
Lucero et al. demonstrate the short distance effect of glioblastoma-derived EVs to induce angiogenesis via miRNAs in vitro, and claim it to be also valid in humans only based on a correlation with a human glioblastoma transcriptomic “fingerprint” [20]. However, no clear demonstration of causality is proposed.
Shen et al. demonstrate the effect of EVs derived from tumors to induce stemness via miRNA in surrounding cells in vitro (at supra physiologic doses) and claim it to be also valid in vivo in tumor-bearing mice. However, they used Rab7 KO tumors as a control to inhibit EV production, a KO that also has a lot of other side effects [21]. It is therefore difficult to know whether this effect is mediated by EVs and by the miRNA inside them.
Ying et al. demonstrate a role for miR-155 transferred by EVs in vitro in glucose tolerance and use an elegant system of bone marrow transplantation to investigate the role of hematopoietic derived miR-155 in a KO mouse. They later claim that the partial rescue of physiologic glucose tolerance is mediated by EVs in vivo although it may also be mediated by other intercellular transfer mechanisms like tunneling nanotubes (TNT), especially to transfer at short distance miRNA from a very macrophage-rich organ like liver to surrounding hepatocytes [22]. The same miRNA-155 has indeed been shown to be able to be transferred through TNT [23].
Chen et al. claimed that miR-375 overexpressing EVs were able to promote bone regeneration but the effect in vivo is not significantly different from the EV control group [5].
Thomou et al. help us to raise other non-trivial questions on vesicular versus non vesicular mediated RNA transfer. He proposed that EVs from adipose tissue would be able to transfer miRNA to liver cells and induce RNA silencing in vivo. The protein expression is reduced by up to ∼95% after injection of serum-derived EVs (from donor mice with brown adipose tissue expressing the miRNA of interest) to miR-KO mice [24]. Strictly speaking, the demonstration proves that a serum factor purified with common EV purification protocols from the donor mice leads to specific miRNA-mediated silencing in mice. It raises the question of whether this effect may be at least partly mediated by an extra-vesicular miRNA in serum co-purified with EVs.
Other teams claimed the demonstration of an efficient transfer of CRE-mRNA via EVs [27,28,29] in vivo. This highly sensitive “on/off” system induces or stops the expression of a particular fluorescent protein upon delivery of the CRE-recombinase protein or its RNA. Although it is very different from a physiologic system, it may still be of interest as a proof of concept. However, this assay has shown limited transduction efficacy even with a high dose of EVs (e.g., in Ilahibaks et al. [30] achieved ∼15% transduction efficacy by ∼8300 EV/cell in vitro, i.e., intra-cytosolic transfer of at least one CRE protein or RNA). More importantly, it may be biased by the transfer of a single CRE recombinase protein (instead of CRE-mRNA) from the donor EVs, although it was not detected in these articles. On the contrary other teams clearly reported the presence of CRE protein in EVs produced from CRE-producing cells [31].

Although the in vivo relevance of RNA-mediated information transfer via EVs lacks clear demonstration in physio-pathological settings, plenty of proofs-of-concept are described in vitro with efficient silencing efficacies [20]. This may be partly explained by the fact that stoichiometric analysis to use physiologically relevant EV concentrations is usually not considered, leading to a typical in vitro EV/cell ratio of about >150,000 EVs/cell [20]. Secondly, for both in vitro and in vivo data, most reports are using cell lines KO for a specific miRNA as a control, but this control might be questionable. Indeed, miRNA usually have about 90–300 targets [34,35], i.e., the comparative effect may be mediated by a lot of other non RNA-mediated effects due to cell physiology dysregulation. Another important factor to consider is the kinetic effect of EVs, which has rarely been investigated. In the few studies that we found, EVs mediated their effect within minutes (less than 60 min, peak effect at less than 20 min [36,37]), a kinetic that is not the one expected by RNA-mediated information transfer.

3. Physiological Effect of RNA Cargo in EVs: A Natural RNA Vector?

3.1. Stochiometric Evaluation of RNA Loading in EVs

Sverdlov claimed that it is very unlikely that naturally circulating EVs transfer a significant part of information through RNA in vivo at long distances in physiological states [14]. He argued that the best candidates for information transfer would be self-amplifying (e.g., mRNA) and/or have a regulatory function (e.g., a transcription factor, a miRNA). At the time, he made the hypothesis that RNA inside EVs was not subject to strong selection. Baglio et al. [38] and other groups found that most RNA in various types of EVs (from tumor, MSCs, immune cells and serum, isolated by various methods (ultra)-centrifugation or affinity column) were small <400 nucleotides (nt) long RNA [39,40,41,42]. Among them, most are tRNAs (that can hardly be expected to have an effect) and miRNA only constituted ∼0.9% [43] of RNA reads. Although the miRNA are relatively enriched (∼10 fold compared to cell RNA⁴), enrichment may largely be due to the nonspecific size selection biased to the smaller sizes such as tRNAs. As an example, 16 S RNA (1,6 kB), a typical medium-size RNA has a hydrodynamic diameter of ∼30 nm [44], whereas miRNA (20–83 nts) have a cylinder shape with a 2 nm diameter and a 7–20 nm length. mRNA encapsulation inside EVs also depends on their local concentration around EV formation sites, as well as mRNA interaction with membrane lipids and proteins [45]. Before being functional, miRNA are getting through the pri- and pre-miRNA state. To be potentially active if they get to the target cell cytosol, miRNA needs either (i) to be not yet associated with Ago2 to form the RISC complex but still able to bind to it (i.e., being pri- or pre-miRNA) and therefore they would be able to bind it later on in the recipient cell cytosol or (ii) to already be associated with the RISC complex as a miRNA, a state in which they can exert their silencing activity directly. Importantly, association of miRNA to the RISC complex allows them to be much more stable than if left alone where it can be rapidly degraded by nuclease, in particular in the context of EV travel through endosomes (containing nucleases) in the target cell.

Sverdlov proposed a rough approximation of the maximal amount of RNA per EV if they are densely packed in EV of 100 nm diameter: ∼1600 RNA/EV for 1000-nt RNA and ∼6700 RNAs/EVs for 200 nts RNA. However, when measured by total RNA quantification [48], the number of RNA per EV was less than one in serum-derived EVs. Another team reported the presence of ∼7 µg of RNA per 10¹⁰ EVs dosed by bulk representing ∼6500 RNA molecules per EV [43], but the presence, as discussed by the authors, of contaminating surrounding extra-vesicular RNA may artificially enhance this number. As an example, once extra-vesicular RNA is removed from serum-derived EV preparations (using differential centrifugation and size-exclusion chromatography) only ∼2.5% of total miRNA remains in the serum-derived EV fraction [25,26,49,50,51,52]. Most of the time, purification strategies used are not allowing complete extra-vesicular RNA removal (in particular in serum where it represents a large fraction of RNA), therefore attribution of a particular effect to intra-vesicular EVs may be difficult. Quantitative results on the amount of miRNA per EVs estimates that most represented miRNA can hardly be found in 1 out of 100 exosomes (the range varies for each miRNA from one copy per 9 exosomes to one copy per 47,162 exosomes, mean of 1 copy per 121 exosomes using digital PCR, a reliable and sensitive quantitative method) [25]. Knowing that they detected 131 miRNA in total, the estimated miRNA per EV should be considered to be ∼1 per EV.

3.2. Navigating the Bloodstream and Getting to the Target?

Once loaded with RNA, in order to exert an effect at a significant distance, EVs have to get inside the bloodstream to reach other tissues. If not produced directly inside the bloodstream, they are secreted in interstitial fluid to later on get to the bloodstream. The main physiological barrier to the long-distance travel of EVs from organs is probably their limited ability to get to the bloodstream. There is a probably significant recapture by surrounding cells in the interstitial fluid (ISF) before getting to the bloodstream. Indeed, the amount of interstitial fluid is expected to represent about 16% of the body weight (11 L in a 70 kg adult), and the concentration of EVs in ISF is on overall ∼12 times more concentrated than in plasma [55]. The ISF flow from interstitial fluid to plasma was calculated to be about 2.9 L/24 h [56] meaning that EVs may spend a significant time in ISF. Although the half-life of EVs in blood is relatively well estimated (∼3–15 min), it is not known in ISF, but it is probably in the same order of magnitude (10 min). Therefore, if about 1/4 of ISF goes to the bloodstream every day, the turnover of ISF (∼96 h) is far longer than the typical EV half-life in ISF, meaning that most EVs (96 h × 60/10 = 576 EV half-lives) are recaptured inside the tissue before reaching the bloodstream.

A simple model taking into consideration key EV pharmacokinetic parameters helps to get an idea of what happens in physiological conditions for EVs navigating in the bloodstream. Indeed, the amount of EVs received per cell from blood per 24 h can be estimated by the following Equation (1):

where Ctot(EV) is the total concentration of EVs in the blood (in EV/L), Vol(plasma) is the volume of plasma in the organism (3 L for humans), f(EV subtype) is the fraction of a particular EV subtype of interest,

τ ½

is the EV half-life in the blood stream (in minute), t is the number of minutes per day (1440), f(target tissue) is the fraction of EVs that target a particular tissue of interest and Nb Cell(tissue) is the number of cell in the tissue of interest. Of note, this model does not take into account the excretion of EVs in urine and other biofluids as it is considered to be negligible in biodistribution studies [57]. Furthermore, this model is only valid for steady states with EV generation and recapture balance being relatively stable and therefore does not apply to bolus injections of EVs. An estimation of relevant parameters to consider is provided in Table 2, some of them are estimated from mouse data (e.g., half-life), the model suffers a lot of approximations but yet gives an interesting approximation in terms of order of magnitude.

Using this equation, the mean total amount of EVs received from blood from all kinds of parental cells by all cells of the organism is estimated to be ∼4.3 × 10¹⁴/day, the mean number of EVs received per cell is ∼11.5 although it largely varies from an organ to another, e.g., is ∼1069/day per liver cell and 0.7/day per brain cell. On another side, significant variations occur depending on the subtype of interest considered: contrary to a quite common vision, most (99.8%) of EVs in the blood come from hematopoietic cells whereas only 0.2% of them come from non-hematopoietic tissues, most of them (81%) coming from adipose tissue.

Special attention should be paid to the short estimated EV half-life compared to most hormones, proteins and drug delivery systems. Indeed, apart from being an interesting parameter for modelization, it is also an important driver to control the potential specific targeting of a tissue/cell type by EVs. Indeed, the whole blood volume circulates ∼3 times/min [67], and therefore may get through the cerebral circulation and potentially interact with brain cells only ∼7 (half-life) × 3 × 0.15 = 3.15 times (cerebral blood flow is ∼700 mL/min [68], representing ∼15% of blood circulation).

Altogether, these assumptions give a quantitative insight on how difficult the long-distance trip of a particular EV may be to a particular cell of interest in a distant organ. Once in the bloodstream, long-distance travel of EVs is mostly limited by their non specific capture and subsequent elimination by RES, and this has proved true for all kinds of EVs tested from hematopoietic origin or not [57,65]. However, this estimation may be different in pathological conditions in which injured organs, inflammation sites or tumors might be a significantly enhanced source of EVs but also a sink for circulating EVs.

3.3. A Very Interesting Intra-Cytosolic RNA Delivery (Endosomal Escape)

Once EVs reach an acceptor cell, the way they are internalized and the mechanisms by which they deliver their cargo are still not fully clarified. De Jong et al. reported with an interesting CRISPR-based system that delivery was partly dependent on micropinocytosis (in particular Pak1, Rak1) or endocytosis (in particular Cav1 and RhoA) [13]. It is now well accepted that intra-cytosolic delivery depends on acidic endosomal escape [71]. Although data are very scarce, the reported EVs endosomal escape efficacies vary from 10% after 2 h to 24.5% after 12 h in Joshi et al. [72] and about 20–30% in Bonsergent et al. [71] Although direct fusion with the plasma membrane is possible, it is expected to represent a much smaller fraction of intra-cytosolic delivery when looking at the delivery kinetic. These numbers are to be compared with the natural endosomal escape that is reported to be about 2–7% depending on the cell type [73], about 0.1–2% for synthetic vectors [74] and about 40–50% for viral vectors like AAVs [75].

3.4. Is the Physiologic RNA in EVs Dose Sufficient to Achieve an Effect?

The RNA-based mechanism of action (MOA) for effects mediated by non-modified native EVs in therapeutic conditions has previously been challenged by comparing it to data obtained from siRNA experiments. In most preclinical studies, EV doses usually range from ∼1 to 200 µg per mouse [81], corresponding to about 10¹⁰ to 10¹² EV/mouse depending on EV preparation and dosage methods. If we consider ∼1 miRNA per EV, this dose represents ∼10¹⁰ to 10¹² miRNA per dose, corresponding to about ∼0.2–20 ng of miRNA/mouse or ∼0.016–1.6 pmol/mouse. siRNA doses reported to be efficient in vivo in systemic injections are rather in the microgram range (27 to 750 µg/mouse [82,83]). One explanation is that the observed therapeutic effect of native EVs is not mediated by their naturally loaded (mi)RNAs. Indeed, this ∼10³–10⁴ fold difference was though too big to be explained by a very high difference in delivery efficacy [43,80]. However, this may be now discussed in view of recent results comparing engineered EVs to synthetic RNA nanovectors.

Indeed, recently reported delivery efficacy of EVs obtained in vivo show a ∼10–300 fold improvement in favor of EVs [84] compared to lipid nanoparticles (although the authors discuss the estimation of miRNA concentration with their method may favor EV reported efficacy by ∼10 fold [85,86]). The authors used the natural ability of pre-miR-451 to be enriched preferentially in EVs and used it as a backbone to couple with an siRNA of interest in order to target it inside EVs [84]. They then used these engineered EVs to target the liver, intestine or kidney glomeruli and achieve various target knockdown. Interestingly, this ∼10 to-300 fold improvement in terms of RNA cytosolic delivery in favor of EV in vivo is fully consistent with independent data on delivery efficacies reported for synthetic vectors: EVs reach a ∼20% endosomal escape rate [71] compared to 0.1 to 2% for synthetic vectors [87], which leads to a ∼50 fold increased cytosolic delivery. Even higher differences (up to 10⁴) were reported in the delivery efficacy in favor of EVs in vitro [32]. Importantly, such a fold change also takes into account the very different endocytosis rate that favors EVs compared and synthetic vectors in vitro but not in vivo [88].

Altogether, these quantitative estimates show (Figure 1) that distant communication by EVs via RNAs probably has limited efficacy in physiological conditions, although it may be a bit different in pathological conditions and in the therapeutic use of EVs that are engineered to load large amounts of specific RNA.

4. Considerations on RNA Based Information Transfer in Therapeutic Settings

4.1. Considerations on the Therapeutic Effect of RNA from Unmodified EVs

RNA delivery has been reported to be the mechanism of action of natural EVs administrated for therapeutic purposes like MSC-EVs [100]. The main difference compared to naturally circulating EV is that the injected concentration in the specific EV of interest will be highly increased compared to their physiological concentration. For example, MSC-EVs are expected to represent much less than 0.1% of serum EVs [60], i.e., <10⁹ out of 10¹². Using the previously described model, if we consider a human therapeutic dose of 10¹² to 10¹³ EVs, one may expect that ∼2.5–25 EVs and ∼0.5–0.5 intra-cytosolic copies of an miRNA of interest would be delivered to each liver cells, or ∼0.75–7.5 EVs and ∼0.015–0.15 intra-cytosolic copies of an miRNA of interest to each cell in the spleen (an organ that contains a lot of immune cells of interest for these therapies). Whether it is sufficient to explain the therapeutic effect of MSC-EVs seems once again unlikely in systemic administrations, apart from a very efficient RNA “cocktail effect”. Local administrations (we approximate that it would touch only 100 g of tissue, 1/650 of the total amount of cells of a human body, i.e., 5.7 × 10¹⁰ cells) of similar doses (10¹²–10¹³) would lead to the delivery of ∼18–180 EVs and ∼0.35–3.5 intra-cytosolic copies of an miRNA of interest to each cell. It seems also difficult therefore that unmodified EVs’ therapeutic effect may be mostly mediated by RNAs.

4.2. Considerations on the Effect of RNA from Engineered EVs

Although unmodified EVs may difficultly have an effect via their RNA cargo, EVs may be very interesting vectors for synthetic delivery of exogenously loaded RNA (mi/si/mRNA). Exogenously EV-loaded synthetic RNA delivery for therapeutic purposes have been reported to treat various pathological conditions [101,102]. The problem of efficient EV loading is still highly debated, in particular in the case of electroporation [103] and was recently reviewed elsewhere by our team and others [11,104]. Briefly, loading of nucleic acid in EVs in EVs have been reported with various techniques ranging from electroporation [103], EV destruction-reformation techniques (slicing, extrusion or sonication [105]), permeabilization via saponin, the use of commercial transfection reagents like lipofectamine [106], heat shock [107], pH gradient [108], etc. Importantly, most of these methods suffer from a low loading efficiency, potential substantial denaturation of EVs and whether these methods lead to intra-vesicular loading or extra-vesicular aggregation on EVs is usually not investigated clearly [11].

6. Conclusions

Quantitative estimates from simplified pharmacokinetic and pharmacodynamic models based on quantitative data reported in the field suggest that information transfer through RNA by naturally circulating EVs may be limited in terms of efficacy at long distances. If it exists, it may rather be at a short distance and/or in the reticulo-endothelial-system [24,84]. Ligand–receptor interactions are more prone to be used by EVs to transfer information (and interact with inflamed tissues or the brain–blood barrier) although the exact effectors (protein, lipids, others?) are probably very varied (and redundant?) depending on each biological process. On the other hand, these potential limitations of RNA transfer by native EV in physiological settings (and potentially in pathological conditions like cancer, inflammation) do not apply to EVs loaded with RNAs of interest for therapeutic purposes. Indeed, engineered EVs may benefit from the very interesting endosomal escape efficacy reported by some studies to deliver RNA molecules with much higher efficiency than synthetic vectors. Of note, to benefit from EVs’ outstanding delivery properties, RNA should probably be encapsulated within the vesicles without EV structure destruction and in sufficient amount to avoid the need for large EV doses that would be cost-prohibitive. Then, challenges faced for the clinical development of EV mediated RNA delivery concern both EV and RNA massive bio-production, as well as efficient encapsulation of RNA in EVs with controlled and scalable processes. Furthermore, EV based delivery may benefit from intrinsic EV properties depending on its parental cell type (angiogenesis [3], immunomodulation [4], increased cell proliferation [5], fibrosis inhibition [6], inflammation resolution [7], etc.) compared to synthetic vectors. Of note, quantitative results on which are based our estimations are usually very limited (if any) and subject to caution. Comparison with literature on other delivery systems is therefore difficult, and the field would highly benefit of a standardized data reporting frame. A particular caution should also be brought to common (and often unknown) purification or engineering artifacts before specifically attributing to EVs some observed effect. More generally, in our point of view more precautions should be taken for data interpretation in the EV domain, a young and interdisciplinary field with limited and difficult to interpret characterization methods.

This entry is adapted from the peer-reviewed paper 10.3390/pharmaceutics13111931

© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.