The Small GTPase Ral orchestrates MVB biogenesis and exosome secretion

ABSTRACT

Extracellular vesicles are novel mediators of cell-cell communication. They are present in all species and involved in physiological and pathological processes. One class of extracellular vesicles, the exosomes, originate from an endosomal compartment, the MultiVesicular Body (MVB), and are released from the cell upon fusion of the MVB with the plasma membrane. Although different molecular mechanisms have been associated with MVB biogenesis and exosome secretion, how they coordinate remains poorly documented. We recently found that the small GTPase Ral contributes to exosome release in nematodes and mammalian tumor cells. More specifically, we found that C. elegans RAL-1 is required for the biogenesis of MVBs, and later for MVB fusion with the plasma membrane. Here, we discuss our results in relationship with other factors involved in extracellular vesicle production such as the ESCRT complex and Phospholipase 1D. We propose models to explain Ral function in exosome secretion, its conservation in animals, and its possible role in tumor progression.

Over the past 20 years, extracellular vesicles have emerged as major mediators of cell-cell communication. They convey proteins, RNAs (coding and non-coding) and lipids from a secreting cell to a receiving cell, in various physiologic and pathologic contexts.¹ Examples of their physiological roles range from morphogen transport during development,^2,3 to the regulation of inflammation,⁴ coagulation⁵ or sexual behavior.^6,7 Exchange of extracellular vesicles occurs between different cells of an organism,⁸ between different species in host-parasite communication⁹ and is even suspected to take place outside of organisms.¹⁰ In addition, their function has been investigated in the mediation of cell-cell communication in pathological conditions. In particular, they have been extensively studied in the context of tumor progression. Extracellular vesicles released by tumor cells have been shown to modify the local environment surrounding the tumor by impacting the composition of the extracellular matrix, by modulating of the immune response or by promoting angiogenesis (reviewed in ref. 11). In addition, tumor extracellular vesicles can act at distance of the primary tumor to induce the formation of a pre-metastatic niche, favoring the subsequent seeding of metastatic cells.^12-14 Besides, as they contain molecular information from the tumor, circulating cancer extracellular vesicles constitute promising diagnostic targets from liquid biopsies.¹⁵ However, and despite their rising importance, the molecular mechanisms regulating their biogenesis are far from fully understood.

Two main categories of extracellular vesicles, whose cellular origin differs, have been described: microvesicles and exosomes. Microvesicles constitute a heterogeneous population of vesicles (100-1000 nm diameter) budding from the plasma membrane through mechanisms involving small GTPases.¹⁶ By contrast, exosomes (below 150 nm diameter) originate form an endosomal compartment, the MultiVesicular Body (MVB), through a variety of possible molecular mechanisms (reviewed in ref. 17 and discussed thereafter). MVBs contain small intraluminal vesicles (ILVs), which, upon secretion outside of cells, are named exosomes. Formation of ILVs includes the concentration of cargos (proteins and RNAs) and lipids in subregions of the endosomal membrane, the inward budding of this region toward the lumen of the endosome and the pinching of this vesicle that carries cytoplasm (Fig. 1A). MVBs can either fuse with lysosomes inducing the degradation of their content, or be directed to the plasma membrane, where they fuse with it and secrete their content outside of the cell (Fig. 1A). It is now clear that different populations of MVBs co-exist within the same cell^18,19 and that different subpopulations of exosomes^20-23 can be secreted by one cell type. These different populations of exosomes could arise from different MVBs. Alternatively a single MVB could contain distinct populations of ILVs, which could thus lead to the secretion of different exosomes. Several molecules, including a number of small GTPases, have been shown to affect exosome secretion by acting at different levels. We recently found that the small GTPase RAL-1 modulates exosome secretion by controlling ILV formation and the attachment and fusion of MVBs to the plasma membrane.²⁴

Figure 1.

Models of Ral function in exosome biogenesis and secretion. (A) MVBs can either enter a degradation pathway by maturing into lysosomes or induce the secretion of the heterogenous populations of intraluminal vesicles (ILV) they contain by fusing with the plasma membrane and liberating exosomes in the extracellular space. (B) In total absence of the GTPase RAL-1 in C. elegans, the formation of ILVs is reduced and MVBs are enlarged (one extreme example of ral-1 phenotype is presented). Ral could orchestrate the inward budding of ILVs by controlling the localization or the activation of PLD, directly or via Arf6 activation, which in turn affects the local enrichment of the lipid Phosphatidic Acid (PA). PA induces negative membrane curvature and directly interacts with syntenin (Synt), which promotes ILV budding trough its interaction with Alix (Alx) and syndecan (Synd). (C) In addition, partial depletion of RAL-1 in C. elegans leads to an accumulation of MVBs docked to the plasma membrane. Ral could control the fusion between MVB and the plasma membrane, by promoting the localization of the t-SNARE Syntaxin 5 (Syx5), which would interact with an unknown v-SNARE present at the surface of the MVB to promote fusion and exosome release.

Biogenesis of ILVs

Extracellular vesicles are named exosomes only once ILVs contained within MVBs are secreted in the extracellular space after fusion of MVBs with the plasma membrane. Therefore, several mechanisms affecting ILV formation that could impact exosome content, maturation and secretion have been described. Among them, the ESCRT complex has been extensively studied and its sequential action in ILV biogenesis has been well characterized in the context of endocytosis (for a review see ref. 25). In parallel, ESCRT-independent mechanisms relying on particular lipids, the ceramids,^26,27 or on a family of transmembrane proteins, the tetraspanins, which are enriched in exosomes,²⁸ have been described. Finally, a mechanism involving the ESCRT associated molecule, Alix, the type I membrane protein syndecan and the PDZ protein syntenin has been dissected in mammalian cells in the past years.^29-31 Syntenin interacts with both syndecan and Alix to induce ILV budding in MVBs and thereby promotes the formation of a subpopulation of exosomes containing these 3 proteins. Interestingly, the laboratory of Pascale Zimmermann has recently identified 2 new actors involved in ILV budding through the syntenin-dependent pathway: the GTPase Arf6 and its downstream effector, the phospholipase D2 (PLD2), an enzyme metabolizing phosphatidylcholine into phosphatidic acid (PA).³¹ PA is a cone-shaped lipid, which binds to syntenin³¹ and induces negative membrane curvature which might favor endosomal intraluminal budding.³² Whether these distinct mechanisms are independent or whether they can co-exist within the same cell is unknown. However, it is likely that other molecules are involved in these processes, at least to coordinate them.

In a RNAi based screen in the nematode C. elegans, we recently identified 73 genes, which could potentially be involved in exosome biogenesis or secretion.²⁴ Among them, we decided to focus on the small GTPase RAL-1 (Ras-Like), the homolog of mammalian RalA and RalB, which had previously been shown to be involved in different types of secretion and in secretion independent mechanisms.^33-35 The cycling of Ral GTPases between an active GTP bound state and an inactive GDP bound state is controlled by Ral GEFs (Guanine nucleotide exchange factors, such as RalGDS, RGL1-4 and RalGPS1-2) and GAPs (GTPase Activating Proteins such as RalGAPα1-2 and TSC1-2), which themselves function downstream of the Ras oncoprotein. Active Ral-GTP interacts with a number of downstream effectors, including components of the exocyst, a complex tethering secretory vesicles to the plasma membrane, or the proteins RalBP1, phospholipase D (PLD) or Filamin. Interestingly, mammalian RalA and its effector PLD1 have been localized to late endosomal/lysosomal compartments in macrophages.^36,37

Using a newly described tag visible in electron microscopy, APEX2,³⁸ we found that RAL-1 localizes at the surface of MVBs, suggesting it could be involved in ILV budding and exosome secretion. To understand more precisely the function of RAL-1, we conducted a systematic quantitative analysis of MVBs morphology and position in the secreting epidermal cells using electron microscopy of full and partial ral-1 loss of function individuals. Of note, this analysis allowed us to describe precisely the morphological characteristics of MVBs in wild type animals (Average diameter: MVB 386,5 ± 9,2 nm; number of ILV /MVB: 22,6 ± 1 ,3; Average ILV diameter: 56,3 ± 0 ,33 nm), based on a significant numbers of MVBs (n = 214). We observed that in a total absence of RAL-1, MVBs are 25% larger and contain 33% less ILVs than in control animals (Fig. 1B). This phenotype, together with RAL-1 subcellular localization at the surface of MVBs, suggested that RAL-1 could function in the inward budding of ILVs. The most likely hypothesis is that in the absence of RAL-1, ILV budding is impaired leading to a reduction of the number of ILVs visible in the lumen of MVBs. As a consequence, there is an excess of membrane at the surface of MVBs, leading to an increase of their diameter. How RAL-1 precisely functions is however not known. One possibility could be that it functions through the PLD-Syntenin-Alix axis.³⁹ Indeed, mammalian Rals have been shown to act upstream of PLDs and to positively control their function.^35,36,40-42 More precisely, RalA and Arf6 act synergistically to stimulate PLD1.^35,43 In addition, both RalA and RalB have recently been shown to activate Arf6 in normal and cancer cells.⁴⁴ Therefore, it is tempting to speculate that Ral GTPases activate or localize PLD at the MVB membrane and thereby affect the local levels of PA, which in turn favors ILV budding (Fig. 1B). Ral GTPases could control PLD activation in parallel to Arf6, in synergy with Arf6, or directly by activating Arf6. These different hypotheses could be tested by comparing the phenotypes of ILV budding of single Ral and Arf6 mutants to the phenotype of a Ral-Arf6 double mutant.

Ral-independent mechanisms of ILV budding also exist, since MVBs containing ILVs are still observed in the absence of RAL-1. Similarly, after co-depletion of 4 ESCRT subunits, ILVs are still visible in multivesicular endosomes.⁴⁵ Interestingly, the remaining ILVs generated by a Ral-independent mechanism are 12% bigger than controls. More precisely, in wild-type animals 44% of ILVs have a diameter larger than 55nm, while this number peaks to 68% in the absence of RAL-1. This could mean that RAL-1 is required for the formation of smaller ILVs and that Ral-independent mechanism of ILV budding specifically generates a population of bigger ILVs. Mechanistically, local RAL-1 dependent activation of PLD could lead to a more efficient budding initiation and to a more rapid budding process and, therefore, to the formation of smaller ILVs. Together, these observations suggest that different mechanisms of ILV budding co-exist within one cell and presumably in the same MVB.

Docking and fusion of MVBs to the plasma membrane

In addition to its role in MVB formation, we found that RAL-1 could also function in the later step of MVB tethering and fusion. Indeed, partial depletion of ral-1 leads to an accumulation of MVBs under the apical plasma membrane of epidermal cells and to an increased proportion of MVBs showing an attachment to the plasma membrane.²⁴ We occasionally observed hemifusion diaphragms between the MVB membrane and the plasma membrane. This suggests that the membrane fusion might have been initiated, but not completed. In our initial screen, we identified other candidates, which could cooperate with Ral to mediate MVB anchoring and fusion. To our surprise, we found that the exocyst complex, which is a well-known Ral effector responsible for anchoring some secretion vesicles to the plasma membrane,⁴⁶ was unlikely to be involved in exosome secretion in C. elegans. Indeed, partial or full deletion of 2 of its components, sec-5 and sec-8, could not fully phenocopy ral-1 defects. While sec-5 mutants displayed a decrease in the average number of ILVs per MVB, sec-5 and sec-8 mutants showed no defects in MVB density in the cytoplasm, no accumulation in proximity of or tethering to the plasma membrane. This result suggests that RAL-1 mediates exosome secretion independently of the exocyst. The exocyst seems to control more conventional secretion, as suggested by the accumulation of small secretory vesicles under the plasma membrane observed after exocyst component depletion.

In our initial screen, we identified genes belonging to 2 families linked to exosome secretion downstream of MVB formation: RAB GTPases and SNAREs. RAB GTPases are known to control multiple aspects of vesicular movement and fusion.⁴⁷ RAB27A and B, which are so far the best-described RAB GTPases in exosome release; they control their secretion in various cell types and species by targeting MVBs to the cell periphery and docking them at the plasma membrane.⁴⁸ Besides, RAB35 has been shown to function in docking/tethering of MVBs to the PM.⁴⁹ Several other RABs have been found to affect exosome secretion levels (reviewed in ref. 17), but how they function remains unknown. Several proteins of the soluble NSF-attachment protein receptor (SNARE) family have been suggested to control the fusion between MVBs and the plasma membrane.⁵⁰ The V-SNARE VAMP7 has been shown to be required for extracellular vesicle secretion by human K562 cells,⁵¹ while the V-SNARE YKT6 has been linked to exosome secretion in human and Drosophila cells.^3,52

While we do not know which adhesion complexes are responsible for anchoring MVBs at the plasma membrane, we identified a protein that could control their fusion. In our initial RNAi screen, we identified the T-SNARE protein SYX-5 (closest mammalian homolog, Syntaxin-5) and found that it colocalizes with an active mutant form of RAL-1. In the absence of SYX-5, MVBs accumulate in close proximity of the plasma membrane. Therefore we think that RAL-1 is required to activate a SNARE complex formed by SYX-5 and a yet unknown V-SNARE at the site of MVB attachment to the plasma membrane (Fig. 1C). One interesting candidate is YKT6, as it has been shown to form a complex with SYX-5 in the Golgi apparatus of mammalian cells,^53,54 and to mediate exosome secretion.^3,52 However, YKT6 did not come out of our screen in C. elegans (unpublished results). Therefore, understanding precisely how RAL-1 controls SNARES to mediate MVB fusion with the plasma membrane deserves further studies. Another protein, the V0 ATPase subunit VHA-5, could also contribute to the last step of exosome release. Indeed, we previously found that the V0 complex of the H⁺-vacuolar ATPase, in particular VHA-5, contributes to exosome secretion, likely by controlling the final steps of MVB docking and fusion.⁵⁵ Besides, V0 ATPases were shown to mediate membrane fusion with SNAREs.⁵⁶ Furthermore, the docking of MVBs to the plasma membrane is likely to require additional uncharacterized adhesion proteins.

Conservation of ral function

We found that Ral function in exosome secretion is conserved between nematodes and mammals. Indeed, depletion of either RalA or RalB, the 2 mouse homologues of C. elegans RAL-1, in 4T1 mammary tumor cells leads to a decrease in exosomes secretion, as attested by EM analysis of secreted vesicles and markers analysis by western-blot.²⁴ Interestingly, the KRAS oncogene, a well-known upstream regulator of Ral,⁵⁷ has recently been shown to regulate the miRNA and protein content of exosomes in mammalian cells.^58-60 Cells expressing constitutively active KRAS(G13D) mutant have been shown to secrete more exosomes than control cells.⁵⁹ However, an earlier study reported that exosome secretion levels were not affected in KRAS mutants.⁶⁰ Whether Ral also affects the content of the secreted exosomes, in addition to their levels, is not known. These findings are interesting considering that tumor extracellular vesicles are mostly promoting tumor progression¹¹ and that RalA or B are overexpressed in several human cancer types and can favor tumor growth in mice.⁵⁷ For instance, pancreatic tumor cells depleted of RalA have a reduced growth in vitro, form smaller primary tumors and less metastasis in vivo.⁶¹ Meanwhile, mammary carcinoma or melanoma cells secreting decreased levels of exosomes (after rab27A depletion) have reduced tumor growth and metastasis capacities.^12,62,63 In addition, tumor exosomes were recently shown to modulate the microenvironment at distance of the primary tumor, inducing the formation of a premetastatic niche that can favor the implantation of metastatic cells.^12-14 Therefore, dissecting the contribution of exosomes secreted in a Ral-dependent manner in tumor progression appears as an interesting avenue to explore. This would however require to precisely separate Ral function in exosome secretion from its other known cellular activities. Indeed, like all other genes known to date to control exosome release, Ral is also involved in many exosome independent cellular processes,⁵⁷ which can also impact on tumor progression (such as exocytosis, cytokinesis, cell migration, autophagy etc.).

The identification of Ral proteins, and more generally of other small GTPases, in exosome secretion opens several questions. How are the different molecular mechanisms of ILV biogenesis and exosome secretion coordinated within a cell? How are they related to the different MVB populations present in a cell and to the different secreted exosomes? Answering those questions will be possible by combining genetic co-depletion of parallel pathways, careful electron microscopy analysis of secreting MVBs (morphologically and by immunodetection of specific cargoes) and precise characterization of exosome subpopulations (in term of number of vesicles, and protein and RNA content).

Abbreviations

EM

Electron Microscopy

ESCRT

Endosomal Sorting Complex Required for Transport

ILV

IntraLuminal Vesicle

MVB

MultiVesicular Body

PA

Phosphatidic Acid

PLD

PhosphoLipase D

Acknowledgments

We thank the editor and reviewers of this manuscript for their critical input.

Funding

This work was supported by a fellowship from the Fondation ARC pour la Recherche sur le Cancer (V. Hyenne); grants from the Institut National du Cancer (to M. Labouesse), the Agence Nationale de la Recherche (ANR-10-LABX-0030-INRT, a French State fund managed by the Agence Nationale de la Recherche under the frame program Investissements d'Avenir labeled ANR-10-IDEX-0002-02 to the IGB MC), and the Institut National du Cancer and Roche (to J.G. Goetz); and institutional funds from the Center National de la Recherche Scientifique, Institut National de la Santé et de la Recherche Médicale, and University of Strasbourg.