Life in the lumen: The multivesicular endosome

Abstract The late endosomes/endo‐lysosomes of vertebrates contain an atypical phospholipid, lysobisphosphatidic acid (LBPA) (also termed bis[monoacylglycero]phosphate [BMP]), which is not detected elsewhere in the cell. LBPA is abundant in the membrane system present in the lumen of this compartment, including intralumenal vesicles (ILVs). In this review, the current knowledge on LBPA and LBPA‐containing membranes will be summarized, and their role in the control of endosomal cholesterol will be outlined. Some speculations will also be made on how this system may be overwhelmed in the cholesterol storage disorder Niemann‐Pick C. Then, the roles of intralumenal membranes in endo‐lysosomal dynamics and functions will be discussed in broader terms. Likewise, the mechanisms that drive the biogenesis of intralumenal membranes, including ESCRTs, will also be discussed, as well as their diverse composition and fate, including degradation in lysosomes and secretion as exosomes. This review will also discuss how intralumenal membranes are hijacked by pathogenic agents during intoxication and infection, and what is the biochemical composition and function of the intra‐endosomal lumenal milieu. Finally, this review will allude to the size limitations imposed on intralumenal vesicle functions and speculate on the possible role of LBPA as calcium chelator in the acidic calcium stores of endo‐lysosomes.

a result, in this network, the net distinction between late endosomes, endo-lysosomes and lysosomes is often blurred. 2 Late endosomes also function at a crossroad with the autophagy pathway, which, in addition to endocytosis and TGN-derived traffic, provides an additional entry point in the endocytic pathway for the degradation of cytoplasmic material, including organelles. [7][8][9] In addition, endosomes engage in physical contacts with other organelles, including in particular the endoplasmic reticulum, via membrane contact sites that play a key role in lipid movement, calcium exchange and endosome dynamics. [10][11][12][13][14] Endosomes and lysosomes can also acquire the capacity to fuse with the plasma membrane as secretory endo-lysosomes-a process reminiscent of the regulated exocytosis of lysosome-related organelles in specialized cell types. [15][16][17] As a consequence, ILVs not only mediate protein and lipid transport to lysosomes for degradation, but can also be released extracellularly as exosomes, which package cellular molecules that, upon delivery to target cells, regulate a wide range of functions at a distance from the exosome-secreting cell. [18][19][20][21] ILVs may also meet additional fates in specialized cell types, 22 and contribute to the biogenesis of melanosomes in melanocytes, 23,24 or harbor MHC class II molecules loaded with peptides for presentation at the plasma membrane in antigen-presenting cells. [25][26][27] They may also undergo backfusion with the endosome limiting membrane [28][29][30] -as do exosomes after endocytosis by the target cell. 31 Well-integrated with the above functions, late endosomes serve as key sensing/signaling platforms that inform the cell about the cell nutrient situation. 2,32,33 2 | AN ATYPICAL LIPID WITH TWO NAMES Lysobisphosphatidic acid (LBPA) was discovered as a structural isomer of phosphatidyl glycerol (PG) in 1967 by Body and Gray 34 (Figure 2), close to a decade after PG, 35 and a century after the description of the first phospholipid (lecithin or phosphatidyl choline). 36,37 Soon after its discovery, it was found that LBPA accumulates in the lysosomal Membrane dynamics in the endosomal system F I G U R E 1 Outline of the endocytic pathway. Organization of the endosomal pathway in mammalian cells, but not in yeast or plant cells. 2 Endocytosed components are delivered to a common early endosome, from where some proteins and lipids are recycled back to the plasma membrane, or routed by retrograde transport to the trans-Golgi network. Molecules destined for late endosomes are sorted into ILVs forming on early endosomal membranes, giving rise to multivesicular endosomes. These detach (or mature) from early endosomes and transports cargoes toward late endosomes and lysosomes. Eventually, some ILVs are delivered to lysosomes where they are degraded together with their protein cargo. Late endosomes and lysosomes exchange membrane components and solutes, forming a transient hybrid endo-lysosome, which is then reconverted into secondary lysosomes, where hydrolases are stored. Endosomes and lysosomes can also undergo fusion with the plasma membrane as secretory endo-lysosomes, and ILVs can also be released extracellularly as exosomes. The endosomal pathway also serves as an input or output for other membrane trafficking pathways, as indicated. In particular, endosomes and lysosomes also function at a crossroad with the autophagy pathway, and engage in physical contacts via membrane contact sites with other organelles, including the endoplasmic reticulum storage disease Niemann-Pick at a time when a precise diagnosis of this lipidosis was uncertain, 38 and later that the lipid is enriched in rat liver lysosomes. 39  LBPA seems to be ubiquitously distributed in all mammalian cells and tissues of high eukaryotic cells. However, with the possible exception of Dictyostelium, 41 the lipid has not been detected in lower eukaryotes, including yeast. Prokaryotes 42 and perhaps plants, 43 however, contain the related lipid, acyl-PG. Immunofluorescence and immunogold labeling of cryosections using a monoclonal antibody against LBPA revealed that the lipid is present exclusively in multivesicular regions of late endosomes and abundant in intralumenal membranes (Figure 3), a distribution further confirmed by subcellular fractionation. [44][45][46] This distribution is consistent with the original finding-before endosomes had been characterized 47 -that LBPA is present in lysosomes. 39

| Stereo-configuration and biosynthesis
LBPA is an unconventional phospholipid not only because of its restricted distribution, but also because it exhibits a unique sn-1glycerophosphate-sn-1 0 -glycerol (sn-1:sn-1 0 ) stereo-configuration [52][53][54] ( Figure 2). LBPA is thus a poor substrate for most phospholipases, 46,55 and a perfect candidate to reside in the degradative environment of late endocytic compartments. However, despite its unusual headgroup and acyl chain organization, LBPA does not act like a detergent and has properties similar to other phospholipids. 56 Both the unconventional stereo-configuration and sub-cellular dis- While LBPA is present in the exoplasmic leaflet of the bilayer, translocation across the bilayer to the cytoplasmic leaflet must occur because the lipid also interacts with the cytosolic ESCRT-protein ALIX. 74,75 So far, no LBPA flippase has been identified. However, like other negatively-charged phospholipids, LBPA may rapidly flip across the membrane if the charge were neutralized at low pH. 76 The close proximity of the headgroups because of LBPA self-assembly or clustering 46 may cause partial protonation of proximal LBPA phosphate groups and transbilayer redistribution of the protonated form. 77,78 In turn, this may drive membrane shape changes, consistent with the capacity of LBPA to deform the bilayer in a pH-dependent fashion 74 keeping in mind that the redistribution of a very small fraction of phospholipids (< 0.1%) can induce significant shape changes. 78 The unique

| Acyl chain composition
In several cell-types, LBPA is predominantly present as dioleoyl isoform (50%-80%), 46,80 but the acyl chain composition of LBPA in rat liver and brain is more complex, including long polyunsaturated acyl chains. 81 In vivo, acyl chains are predominantly present on the 2 and LDL endocytosed by the LDL receptor. 94 Once in late endosomes, cholesteryl esters are de-esterified and free cholesterol is rapidly incorporated into nearby membranes, 95 including LBPA-containing membrane. Cholesterol then reaches the endosome limiting membrane and becomes available for further export to the endoplasmic reticulum for cholesterol-sensing, 96 and to other organelles including the plasma membrane. 50,97 LBPA-membranes also regulate the flux of cholesterol through endosomes during lipid droplet biogenesis induced by Wnt. 98,99 Cholesterol transfer from endosomes to the endoplasmic reticulum may be direct 12,13 or indirect via the plasma membrane, 100,101 and likely involves nonvesicular transport routes at membrane contact sites. 12,13,50,95 Within endosomes, cholesterol transfer to the limiting membrane depends on the proteins Niemann-Pick type C1 and C2, and loss-offunction mutations in either of these proteins result in a cholesterol storage disease. 102,103 NPC1 is a multi-spanning protein of the limiting membrane and NPC2 a globular protein present in the lumen, 104,105 and both proteins bind cholesterol. 106,107 Structural and mutagenesis evidence indicate that cholesterol is transferred from NPC2 to NPC1, thereby facilitating export from endosomes, [108][109][110][111][112][113][114] and atomistic simulations indicate that LBPA is required for NPC2-membrane interactions. 115 Recent studies showed that NPC2 interacts directly with LBPA and that these interactions are necessary for cholesterol trafficking from endo-lysosomes. 116

| LBPA in NPC cells
Elevated levels of LBPA have been found in NPC 38 and other lysosomal storage diseases. [120][121][122] This increase may reflect some specific need for LBPA, for example in sphingolipid degradation. 89  confirming the notion that a metabolic relationship exists between LBPA and sLBPA. 126 One may thus speculate that such changes reflect some additional adjustment in LBPA-membrane chemical and physical properties to better accommodate the changes caused by cholesterol accumulation. [127][128][129] There is no approved treatment against NPC except for Miglustat, which delays but does not arrest the progression of the disease. 130 Cyclodextrins clear cholesterol storage and restore cholesterol feedback regulation in NPC mice, [131][132][133][134][135] improve symptoms and survival in NPC animal models, 136,137 and decrease the neurological progression of the disease in phase 1-2 trials in NPC patients, 138 suggesting that cyclodextrins may emerge as therapeutical strategy. However, the mechanism of action is being debated. 139,140 Recent studies indicate that hydroxypropyl-cyclodextrin acts by promoting the secretion of the endo-lysosome content, including LBPA, via a mechanism that requires the lysosomal cation channel mucolipin-1 (MCOLN1 or TRPML1) 141 (see Figure 6), which is itself responsible for the lysosome storage disease (LSD) mucolipidosis type 4 when mutated. 142 Interestingly, endo-lysosome secretion elicited by cyclodextrin in NPC cells decreases endosomal cholesterol but not total cell cholesterol, indicating that the secreted cholesterol is presumably incorporated into the plasma membrane or released and recaptured by cells, and eventually redistributed intracellularly. 141 On the whole, these data fit nicely with observations that secretory endosomes or lysosomes 15 mediate the secretion of storage material in lysosome storage disorders via activation of TFEB-family transcription factors, 143 The current view is that ESCRT-0 initiates the process by binding both PtdIns3P on the membrane and ubiquitin conjugated to cargo molecules, and recruits ESCRT-I, which in turn recruits ESCRT-II as nucleator for ESCRT-III filaments 151,152 (Figure 4). In addition to ESCRT-0, -I and -II, the filaments of ESCRT-III can also be nucleated by other factors, including the LBPA partner ALIX, 30,153,154 and perhaps HD-PTP, which shares a Bro-1 domain with ALIX. [155][156][157] ALIX mediates the ESCRT sorting of the GPCRs PAR1 and P2Y1, [158][159][160] while HD-PTP is required for the downregulation of the EGF receptor, 155 PDGF receptor, 161 α5β1 integrin, 162 and virally ubiquitinated MHC class I. 156 In vivo and in vitro observations show that ESCRT-III filaments drive the membrane deformation process that leads to ILV formation, [163][164][165] presumably in conjunction with the triple A ATPase VPS4. 166,167 LBPA itself may also play a direct role in this process. 74 In addition, ESCRT-III drives other membrane deformation processes that share the same topology, including cytokinetic abscission, viral budding, nuclear envelope reformation, [168][169][170][171] as well as plasma membrane 172,173 and endo-lysosome membrane repair. 153,154,174 Hence, ESCRT-III functions as the general membrane deformation and fission machinery with an orientation opposite to endocytosis, away from the cytoplasm.
In addition to ESCRT-dependent mechanisms, ILVs may also form via ESCRT-independent pathways. 175 In melanocytes, the melanosomal protein PMEL is sorted into ILVs in an ESCRT-independent 176 but CD63-dependent manner 177 (Figure 4). Similarly, different ILV populations may be formed in a Hrs-or CD63-dependent manner in HeLa cells. 178 It should be noted that EGF, which triggers EGF receptor endocytosis and sorting into ILVs, also increases multivesicular endosomes biogenesis and ILV formation 179 in an ESCRTdependent manner. 175 However, the mechanism driving the increase in ILV formation is not known, perhaps dependent on annexin 1 179 and SCAMP3. 180 In addition, stress exposure triggers the ligandindependent internalization of EGF receptor via a route that diverts from the canonical pathway and that depends on WASH and Tsg101-ALIX, leading to EGF receptor accumulation in a subset of LBPA-rich multivesicular endosomes. 181

| Microautophagy and exosome biogenesis
In a process clearly reminiscent of ILV biogenesis, cytosolic components can be engulfed within the lumen of nascent ILVs via microautophagy, and then delivered to lysosomes. 7,182 Microautophagy may be mediated via more than one pathway, dependent or not on autophagy-related (ATG) genes. In budding yeast, the NPC orthologs, Ncr1p and Ncr2p, promote microautophagy presumably by increasing sterol in the vacuole limiting membrane. 183 In fission yeast, Nbr1 was identified as autophagy receptor for the ESCRT-dependent targeting of soluble cargos to the vacuole. 184 Accumulating evidence support the notion that the ESCRT machinery is required for microautophagy. [185][186][187][188][189][190] In addition, evidence also suggests that proteins encoded by ATG genes have pleiotropic effects on exosome biogenesis and release. 9 In particular, the ATG3-ATG12 conjugate was reported to interact with ALIX in order to promote autophagy and exosome biogenesis. 191 Exosomes correspond to a sub-population of extracellular vesicles that originate from ILVs and are released outside cells upon endosome fusion with the plasma membrane 31,192,193 (Figure 1). Consistently, exosome biogenesis depends on ESCRT-III, 194 and ALIX 71,195,196 -although exosomes secreted by oligodendrocytes may form in a ceramide-dependent but ALIX-and ESCRT-independent manner 197 ( Figure 4). In addition, LBPA is present in exosomes 198 and ALIX is considered as one of the best-established exosome markers, 31,199,200 which is surprising given the fact that ESCRTs remain cytosolic and are typically excluded from ILVs. 201,202 Essentially nothing is known about the mechanisms that control the alternative fates of ILVs-degradation in lysosomes, back-fusion or secretion as exosomes. Neither is anything known about the principles responsible for the lysosomal targeting of ILV cargoes or retrieval to other destinations, including exosomes.

| Biochemically-distinct populations of ILVs
The sub-cellular distribution of LBPA clearly demonstrates that biochemically-distinct populations of ILVs co-exist within endosomes.
Indeed, the lipid cannot be detected in early endosomes, 44 where ILV biogenesis begins. 148 Neither is the lipid detected in canonical multivesicular endosomes/bodies, which serve as intermediate between early and late endosomes (Figure 1). LBPA is found, and thus likely synthesized, in late endosomes or endo-lysosomes, 44 which are filled with internal membranes of various origins, including exosomes in the F I G U R E 4 Multivesicular endosome biogenesis. The figure outlines the proposed mechanisms driving the formation of ILVs and exosomes in most cell types (green), exosomes in oligodendrocytes (brown) and melanosomes in melanocytes (blue). In most cell types, sorting into ILVs is mediated by ESCRT-0, -I and -II, HD-PTP or ALIX, as is presumably the nucleation of ESCRT-III filaments, which drive the membrane deformation process. However, ILVs may also be formed in a CD63-dependent and ESCRT-independent manner-a process presumably akin to the biogenesis of melanosomes in melanocytes. ILVs formed in early endosomes presumably lack LBPA, because the lipid is only found in late endosomes. The biogenesis of exosomes may require ALIX and ESCRTs, as well as syntenin presumably, but not in oligodendrocytes where the process seems to depend on ceramides and to be ALIX-and ESCRT-independent. Once formed, ILVs and exosomes follow different pathways. ILVs can be targeted to lysosomes for degradation, or undergo back-fusion with the limiting membrane. Exosomes by contrast are secreted upon endosome fusion with the plasma membrane. The relationship between ILVs and exosomes are not clear. Neither are the mechanisms that discriminate their selective fates. The factors that have been reported to control each process are indicated. Membranes shown in the black color imply that it is not known whether the corresponding processes involve LBPA-containing membranes making, ILVs destined for lysosomes, as well as remnants of organelles delivered by autophagy (see tomogram of late endosomes in Cos cells-Movie S1). LBPA itself seems to be enriched in one subpopulation of these intralumenal membranes. 46 Consistent with this notion, PtdIns3P and LBPA localize to different ILV populations within endosomes. 203 The notion than more than one population of ILVs co-exist in endosomes 204 is clearly further supported by observations that, in addition to ESCRT-dependent mechanisms, ILVs may also form via ESCRT-independent pathways, as discussed above. One of the future challenges will be to establish what are the overlapping vs unique mechanisms, dependent or not on ESCRT subunits or ESCRTassociated proteins, which may drive the biogenesis of functionallydistinct populations of ILVs, microautophagosomes or exosomes.
Interestingly, disruption of the class III PI3-kinase Vps34 in neurons, which is required for both autophagy and ILV formation, triggers the secretion of unique exosomes enriched for undigested lysosomal substrates, specific sphingolipids, and LBPA. 205

| ILVs hijacked by pathogens
Pathogens use all tricks in the book to overcome cellular defenses, and not surprisingly, they also exploit the multivesicular endosome pathway 206 ( Figure 5). The anthrax toxin penetrates the target cell in a process that depends on LBPA, ALIX and other ESCRTs. 71 The ESCRT machinery was also recently shown to play additional roles during bacterial infection, in light with a general role for ESCRTs in repairing endo-lysosome membranes 153,154 and other membranes. 222 Vacuoles containing the intracellular pathogen Coxiella burnetii recruit ESCRTs to maintain an intact vacuole, which presumably provides the bacterium with a replication advantage. 154 Similarly, ESCRTs are required to repair small membrane damage in the vacuole containing Mycobacterium marinum in Dictyostelium discoideum 174 or Mycobacterium tuberculosis in macrophage, 223 presumably to ensure that the pathogen remains contained within intact compartments.
ILVs as exosomes have also been proposed to mediate the spreading of pathogens or pathogenic agents from cell to cell ( Figure 5). In fact, it is being discussed whether viruses and exosomes (or other types of extracellular vesicles) share similarities and may be related. 224 It has been reported that exosomes may mediate the transmission of hepatitis C virus 225 in a process that depends on the ESCRT subunit HRS. 226 Similarly, exosomes have also been proposed to transfer hepatitis C viral RNA. 227,228 as well as nucleic acids from other viruses including HIV. 229,230 The non-enveloped hepatitis A virus was also shown to be released after inclusion within a host-derived exosomallike membrane generated in a process that depends on the ESCRTs, VPS4B and ALIX 231,232 -an observation that blurs the classic distinction between enveloped and non-enveloped viruses. In addition, uropathogenic Escherichia coli (UPEC), which targets lysosomes but avoids degradation by pH neutralization, can be expelled in exosomes by bladder epithelial cells, upon pH sensing via the calcium channel TRPML3 (TRP channel 3 or mucolipin 3) 233 (see Figure 6). Finally, in addition to delivering their toxin cargo to the cytoplasm by backfusion, ILVs containing anthrax toxin may also be released as exosomes so that the toxin can be transmitted to naïve cells. 71 Interestingly, however, anthrax toxin containing ILVs fail to be targeted to lysosomes for degradation. 71 It thus appears that that the machinery controlling ILV formation and dynamics has been hijacked to mediate viral RNA or toxin release to the cytoplasm during infection/intoxication, or secretion to the extracellular medium as exosomes in order to propagate the infection or to spread the toxin to naïve cells.

| Protons, anions and cations
In the late endosome lumen, where LBPA is found, ILVs and other intralumenal membranes are packed within a highly crowded environment (Movie S1). Beyond the diversity of membranes already discussed above, relatively little is known about the biochemical and biophysical properties of the lumenal milieu, 234 although much progress has been made in the characterization of endo-lysosomal ion channels and in the description of the ionic situation within the endolysosomal milieu (for recent reviews, see [235][236][237][238][239]. It is well-established that endo-lysosome acidification depends on the V-ATPase, with early endosomes having a mildly acidic pH ≈ 6.2 and late endosomes/lysosomes a more acidic pH ≈ 5.0 234,240,241 ( Figure 6). Numerous physiological processes, including ligand-receptor uncoupling, lysosomal enzyme activity and membrane traffic are controlled by the acidification properties of endo-lysosomes. The low endo-lysosomal pH is also used by enveloped viruses to trigger fusion of the viral envelope with the endosomal membrane and by some toxins to cross the endolysosomal membrane so that the viral nucleic acid or the toxin can reach the host-cell cytoplasm. 206,242 In addition to protons, cations and anions also play important roles in the regulation of the endolysosomal lumenal environment. Chloride controls ion homeostasis F I G U R E 5 Viruses, toxin and ILV-membrane dynamics. The left side of the figure (penetration) outlines the pathways used by some endocytosed pathogenic agents that enter the host-cell cytoplasm through endosomes, in a process that depends on proteins/lipids involved in ILV membrane dynamics. VSV, Lassa virus, LCMV, and Flaviviruses may penetrate cells in a two-step process. First, the viral enveloped undergoes fusion with the ILV membrane (eg, in early endosomes) so that the capsid be delivered into the protected environment of the ILV lumen. Then, the capsid is released into the host-cell cytoplasm upon fusion of the ILV membrane with the late endosome limiting membrane (so-called backfusion). Similarly, the anthrax toxin is first translocated across the ILV membrane and then delivered to the cytoplasm upon ILV back-fusion. Other endocytosed viruses may penetrate cells upon direct fusion of the viral envelope with the late endosome membrane. 206  Much like in the ER, 252 the free Ca +2 in endosomes is estimated to 0.4-0.6 mM. 249,253 In the ER, most Ca +2 is buffered by abundant lumenal Ca +2 -binding proteins. 254,255 However, these proteins or F I G U R E 6 Ions, channels and transporters. The figure outlines the major ion channels and transporters present in endo-lysosome, as well as the estimated ion concentration in the lumen of endo-lysosomes and in the cytoplasm. The intralumenal concentration of Cl − was estimated using a DNA-based, fluorescent chloride reporter 271 and see also. 272 The lumenal concentration of Na is estimated to be around 140-150 mM. 245 Li and collaborators recently proposed that ΔΨ of resting lysosomes is around 0 (±20 mV). 235 Essentially nothing is known about the ionic situation within ILVs or exosomes, except for the observation that ILVs remain neutral until at least 20 minutes after formation. 266 At ER-lysosome membrane contact sites, the ER may sequester lysosomal Ca 2+ , 273 and ER Ca 2+ may refill lysosomal Ca 2+ stores. 274 Ca 2+ is released from ER stores via Ins(1,4,5)P 3 receptor (IP 3 R) and calcium refilling of the endosomes may be driven by the proton gradient via a vertebrate Ca 2+ /H + exchanger (CAX), 275 or depend directly on the ER in a pH-independent fashion. 276 Membranes shown in the black color imply that it is not known whether the corresponding processes involve LBPA-containing membranes. V-ATPase: the vacuolar ATPase 240 ; CLC-3, -6, -7: the 2Cl -/H +exchangers CLC-3, -6, -7 (chloride channels) that distribute in endo-lysosomes 238 ; CAX, a putative endo-lysosomal Ca 2+ /H + exchanger involved in Ca 2* uptake into endo-lysosomes 275 ; P2X 4 , purinergic P2X receptor subtype 4; TPC, two-pore channels; TRPMLs, transient receptor potential channels; BK, big conductance Ca 2+ -activated potassium channel 274 ; TMEM175: K + -selective channel 235 their functional homologs are not found in endosomes and lysosomes, and the nature of the Ca +2 -binding molecules that play similar roles in the acidic calcium stores is unknown. Yet, it can be estimated that ≈99.9% of Ca +2 in acidic stores is chelated, supporting the notion that buffer molecules or matrix must exist. 256 It is appealing to propose that the abundant, negatively-charged lipid LBPA serves as calcium buffer in the lumen of late endosome/endo-lysosomes. Indeed, the capacity of calcium to bind negatively-charged lipids is a universal principle, which is best illustrated by the active translocation of the negatively-charged lipid PS from the outer leaflet of the plasma membrane (high calcium environment of the blood) to the inner leaflet (low calcium environment of the cell). 257 Moreover, calcium exhibits a substantial capacity to bind membrane phospholipids [258][259][260][261] and to alter the properties of the bilayer. 262 In fact, accumulation of the divalent cation Zn +2 in the LBPA-containing late endosomes of cells expressing the ZnT2 zinc transporter caused cholesterol accumulation much like in NPC cells. 64 It can be anticipated that calcium association to LBPArich membranes in the endosome lumen may not only control the fate and dynamics of ILVs, but may also play a key-role in the late endosome/endo-lysosome capacity to modulate calcium-dependent processes, including in lysosomal signaling. 246

| The lumen in the lumen: Size matters
In mammalian cells, typical ILVs form one or more fairly homogenous populations of vesicles with a mean diameter around 50 nm, 263,264 while ILVs in yeast are smaller with a diameter of ≈ 25 nm. 265  Thus, a highly efficient mechanism must exist to produce, sort and package miRNAs into exosomes, and to target these exosomes to the recipient cells, for such a transfer mechanism to operate in a physiologically-relevant manner-miRNAs and RNAs associated to extracellular vesicles are reported to be enriched in certain sorting motifs. 224 Using an assay that measures the biogenesis of ILVs into late endosomes in vitro, the ILV lumenal pH was found to be neutral for a relatively long time, up to 20 minutes after ILV formation. 266 However, given the fact that an ATP-dependent mechanism is unlikely to maintain the pH gradient across the ILV membrane inside endosomes, it is not known whether the pH gradient persists until digestion in the lysosomes, or whether proton permeation across the bilayer eventually acidify the lumen, prior to degradation. In any case, the asymmetry across the ILV membrane driven by pH and ion gradients, as well as the asymmetric protein and lipid composition of the ILV bilayer likely contribute to regulate the fate of ILVs.

| CONCLUSION
Late endosomes/endo-lysosomes are unique organelles of the vertebrate vacuolar apparatus in that they contain membrane vesicles within their lumenal environment, which is topologically equivalent to the extracellular space. These vesicles are highly specialized, in particular because some are rich in LBPA-an atypical lipid that is not found elsewhere in the cell. LBPA not only has an unconventional biosynthetic pathway and stereochemistry, but also has a unique shape and acyl chain migration capacity, likely to influence its impact on membrane organization and dynamics.
A fully unanswered and outstanding question is the nature of the mechanism that drive the sorting of ILVs toward one of their possible fates-degradation in lysosomes, secretion as exosomes, or recycling to the limiting membrane via back-fusion. The privileged and secluded environment of ILVs, bathed into the late endosome/endo-lysosome lumen, is fully disconnected from all cytosolic machineries that drive signaling or protein and lipid sorting, and therefore the fate of ILVs cannot rely on these established mechanisms. Future work will be needed to address this issue. However, some speculations are already possible.
LBPA-rich membranes are involved in the regulation of several features of the endo-lysosome intralumenal membrane system, including cholesterol transport, sphingolipid degradation, and membrane dynamics, as well as perhaps endosomal Ca +2 . LBPA also exhibits a rare capacity for adaptive shape changes, via acyl chain remodeling, because of its unique structure. It is therefore attractive to believe that LBPA-rich membranes play a crucial role in modulating trafficking within the endosome and the fate and dynamics of intralumenal membranes. In particular, given the fact that the LBPA partner ALIX is involved in the biogenesis of at least some exosome populations and is itself found in exosome, LBPA-rich endosomal membrane domains may ultimately control the biogenesis of exosomes.