Fed-EXosome: extracellular vesicles and cell–cell communication in metabolic regulation

In multicellular organisms, where interplay and co-ordination within and between distant organs with distinct functions are required, communication between cells is crucial. In the case of whole-body regulation of metabolism, communication is paramount to co-ordinating the activities of very specialized organs, such as the brain, pancreas, liver, muscle, and adipose tissue (AT). Together, these organs regulate and co-ordinate functions as diverse and fundamental as fuel partitioning and preferential allocation, nutrient uptake, storage and release, among others. Traditionally, non-synaptic cell–cell communication is believed to occur locally via paracrine signalling or over longer distances via endocrine signalling, both involving secretion of signalling molecules such as growth factors, cytokines and hormones. However, cell–cell communication by secretion of extracellular vesicles (EVs) has recently gained recognition as an important enabler of intercellular/organ signalling, as EVs allow ‘custom-made packages’ with specific molecular information to be addressed to defined target cells and delivered in a timely fashion anywhere in the body.

EV is a term which covers different types of vesicles such as exosomes, ectosomes, micro/nanoparticles/vesicles, shedding vesicles and apoptotic bodies. EVs are secreted by many cells of the body and can be found in most body fluids [1,2]. With the exception of exosomes, EVs are generally characterized as being 150–1000 nm in diameter and released from the cell by direct budding of the plasma membrane (PM). Exosomes are vesicles smaller than 150 nm in diameter, generated through inward budding of endosomal membranes [3,4]. This inward endosomal membrane budding leads to the formation of intracellular multivesicular bodies (MVBs) which then fuse with the PM to release the exosomes extracellularly [3,5].

The biogenesis of EVs and exosomes has already been described in detail [2,4]; briefly, direct budding of the PM involves specific changes to lipid and protein characteristics at the site of budding, including an increase in Ca²⁺ levels, phosphatidylserine translocation to the PM extracellular side, lipid raft formation and cytoskeletal protein reorganization [2]. The formation of MVBs requires specific membrane remodelling, membrane budding and vesicle scission that require involvement of endosomal sorting complex required for transport (ESCRT) [4]. ESCRT is also believed to be involved in the sorting of cargo into EVs, as disruption of ESCRT proteins has been found to alter EV cargo composition [6]. However, the specific mechanisms that enable the selection and packing of accurate cargo into the EV is a key logistic problem that is yet unsolved [2,7].

EV cargo is diverse. It includes proteins and nucleic acids found in the lumen of EVs, as well as membrane proteins, phospholipids and lipid species in the EV membrane [4,7]. Several studies have found differences between donor cell and EV composition of proteins, lipids and RNAs (including mRNA, rRNA and miRNA) [7–9], suggesting that EV composition is not merely a representative subset of donor cell composition but a purposely assembled collection of molecules to deliver a specific message. Importantly, EVs have been suggested to play a role also in cellular stress response and removal of unwanted molecules [10]. For instance, active caspase3 may be removed from cells via EVs to protect from cell death [10], harmful cytoplasmic DNA can be removed via exosomes [11], and exosomes have been suggested as a mechanism for the transmission of protein aggregates in neurodegenerative disease [12]. Interestingly, in addition to protecting the cell from accumulation of damaging molecules, the release of EVs during stress may function to induce stress resistance in recipient, non-stressed cells [13]. These findings support the notion that EVs are not simply a random leak of vesicles but may be a sophisticated and highly regulated mechanism of cell–cell communication.

In order to function as specific messengers, EVs must be able to modulate the function of target cells; this can be achieved through delivery of EV cargo to the cell by EV internalization or through interaction of EV surface molecules with the PM of the target cell [14]. Whether these two approaches have different functional implications is unclear. In vitro studies have shown that adhesion molecules such as integrins and intercellular adhesion molecule 1 (ICAM1) can facilitate EV capture by target cells [15], though it is unclear how the fate of the EV, be it cargo release or EV degradation through the normal endosomal pathway, is decided thereafter. A large body of evidence, discussed in detail in Mulcahy et al. (2014) [14], suggests that internalization of EVs is an energy-dependent process that occurs largely through active endocytosis, and several mechanisms of endocytosis have been described for EVs. These include clathrin-mediated endocytosis, macropinocytosis, caveolin-dependent endocytosis, phagocytosis or lipid raft-mediated endocytosis [14]. Despite considerable efforts to understand EV internalization, there is still no clear consensus in the literature as to which mechanism of endocytosis is most prevalent, how these internalization mechanisms are regulated or how the cargo of actively endocytosed EVs is released into the cytosol.

The presence of RNA species in EVs has sparked particular interest. While many types of RNA have been found in EVs, including mRNAs, tRNAs, viral RNAs and long non-coding RNAs [16], the ability of miRNAs to directly regulate post-transcriptional gene expression has encouraged research into the contribution of these molecules to EV signalling. As a reflection of this emerging field, this review will focus mainly on studies describing miRNA-mediated EV signalling. miRNAs alter protein production by inhibiting mRNA translation, mainly through destabilization of the mRNA target [17]. However, understanding protein output regulation by miRNAs is highly complicated, as one miRNA can have several mRNA targets, and mRNAs can be targeted by several different miRNAs [18]. In the circulation, miRNAs are protected from nucleases by encapsulation in EVs or by forming complexes with Argonaute 2 (Ago2), a component of the RNA-induced silencing complex (RISC) through which miRNAs exert their functions on mRNA [19–21]. Ago2-bound miRNAs are thought to be mainly by-products of cell death with no specific signalling function [19,22], whereas functional miRNA transfer through EVs has been demonstrated in vitro from cells infected with human Epstein–Barr virus [23], from circulating AT exosomes [24], as well as ex vivo from AT macrophages (ATMs) [25].

Interestingly, miRNAs may be packaged into EVs along with proteins required for their processing or function [26,27]. miRNA is first synthesized in the cell as pri-miRNA and then sequentially processed by endonuclease complexes containing Drosha and Dicer to generate pre-miRNA and mature miRNA respectively [28]. Melo et al. (2014) [27] demonstrated the presence of miRNA processing proteins, including Dicer, in cancer exosomes and showed that pre-miRNAs can be processed to mature miRNAs within exosomes in a cell-independent manner. In another study, Ago2 was found to play a protective role for EV miRNAs by protecting them from lysosomal degradation in the target cell [26]. Likely, Ago2 facilitates packaging of its associated miRNA into EVs and later directs the function of the miRNA in the target cell without the need for endogenous Ago2 [26].

These studies demonstrate how packaging of a specific subset of operationally related molecules into EVs may confer functional advantages such as ensuring cargo molecule integrity akin to how retail products may be packaged in protective bubble wrap to ensure their integrity during transport. In addition, miRNAs delivered in the same EV might be able to act together to inhibit multiple targets simultaneously, allowing the generation of a co-ordinated, multifaceted signalling response in the target cell. Of note, also exosomal mRNAs are functional and can be translated in the target cell [9], thus providing a direct method of modulating recipient cell protein production. Observations in glioblastoma [29] and mesenchymal stem cells [30] support this notion.

Evidence suggests [31,32] that EVs interact with target cells in a cell type-specific manner, and thus the idea has emerged that EVs could potentially be used as shuttle vectors for delivery of drugs or other therapeutically relevant molecules to specific cellular ‘addresses’ in vivo. However, the question of how EVs are targeted to particular cells in vivo remains unanswered.

While it is generally believed that EVs selectively bind target cells through interaction of specific surface receptors/ligands, the study of these interactions is complicated by the difficulty in visualizing nanoscale EVs using standard confocal microscopy, as well as the lack of robust methods for isolating and identifying specific types of EVs, reviewed by others [3,33]. New methods have been developed to visualize the transfer of EV cargo to target cells in vitro and in vivo through modification of cargo proteins or mRNAs to generate reporter molecules, including Cre-recombinase-mediated LacZ expression [34], membrane-fused fluorescent proteins [35] and luciferase-fused biotin [36]. Although these methods do not allow distinction between different EV species, further development of EV-reporter molecules could enable identification and better understanding of EV–target cell interactions.

Despite incomplete understanding of the regulatory mechanisms of EV uptake and target cell interactions, the role of EVs in many physiological processes and pathologies, including stem cell maintenance [37], cancer [2,38,39], kidney injury [40], cardiovascular disease [41], wound healing [42], neurological diseases [43,44], immune suppression [45] and metabolism (see below), is becoming progressively more apparent. These findings have potentially important pathogenic and therapeutic implications. For instance, breast cancer cell lines resistant to chemotherapy drugs are able to transmit this chemoresistance to non-resistant cell lines via exosomal miRNAs in vitro [38].

Moreover, as EVs carry highly specific cargo molecules, EVs may have clinical relevance as biomarkers of disease, facilitating the diagnosis of obesity [46], metastatic cancer [47] and Alzheimer’s disease [48]. The relevance of EVs in metabolic disease was shown by Karolina et al. (2012) [49]: dysregulation of specific circulating and exosomal miRNAs with roles in pathways of metabolic regulation, such as lipid metabolism and glucose homoeostasis, could be found in patients with the metabolic syndrome or individual metabolic disorders such as hypertension, type 2 diabetes and hypercholesterolaemia.

Thus, one may think of EV-mediated communication as the courier service of the cell, where each EV is a letter or a parcel with a unique message of cargo molecules, delivered to target cells through an ‘address’ composed of a unique combination of surface molecules (Figure 1). Any one cell may send out a range of messages to distinct targets, and the messages and targets may vary depending on the metabolic state of the cell. Therefore, mechanisms must be in place to sense and select cargo and package this cargo into EVs equipped with the appropriate address. While much is left to be understood about these mechanisms, the importance of EVs as a means of cell communication is becoming increasingly clear.

The maintenance of whole organism metabolic homoeostasis requires interplay and communication between key metabolic organs. A growing amount of evidence suggests a role of EVs in this interorgan communication and this role has been investigated extensively in situations of metabolic dysfunction occurring during disorders such as obesity, diabetes, lipodystrophy and non-alcoholic fatty liver disease (NAFLD). Since obesity and associated metabolic disorders are increasing health problems, understanding the role that EVs play in metabolic communication, and how EVs may be used as biomarkers of metabolic disease, may be of high value for improved diagnostics/prognostics as well as the development and administration of drugs without side effects. We will focus on literature describing EVs from AT, liver and skeletal muscle as these are the major insulin-sensitive tissues. The role of pancreatic EVs in metabolism has been reviewed in detail by others [50,51]. Furthermore, we have briefly discussed placental EVs as this organ is pivotal to metabolic regulation during pregnancy.

In addition to lipid storage and release, AT contributes to the regulation of metabolism through the secretion of molecules such as cytokines and hormones (termed ‘adipokines’), including leptin and adiponectin. However, the presence of AT exosome-like vesicles has been documented in vitro and in vivo [8,52,53], representing another potential, and likely more selective and specific, mechanism for AT to communicate with other organs.

A characterization of EVs released from 3T3-L1 cells during adipocyte differentiation showed stage-specific changes with respect to lipid and protein content, as well as EV number and size distribution, during differentiation [8], suggesting different signalling functions of preadipocyte- and mature adipocyte-EVs. The present study also supports the notion that EV composition is not necessarily representative of the donor cell, as differences were found between cell and EV composition of proteins and lipids, including the adipokine fatty acid binding protein 4 (FABP4 or adipocyte protein 2 (aP2)), a major lipid transport protein. Also miRNA content has been found to differ between adipocyte donor cells and EVs [46], indicating specific mechanisms of cargo selection and a specific signalling role for adipocyte EVs.

One of the signalling roles of AT EVs, summarized in Figure 2, may be to communicate with other cells of the AT, including fibroblasts, preadipocytes, endothelial cells and immune cells, to co-ordinate the tissue plasticity response to varying fuel availability, such as AT expansion [54]. In particular, ATMs have been found to change significantly during obesity in terms of number, location and inflammatory status [54]. When incubated with EVs from human adipocyte cell lines and AT explants, monocytes differentiated into ATM-like macrophages, and conditioned medium from these macrophages inhibited adipocyte insulin signalling in vitro [52]. These findings suggest a role for AT EVs in the communication between adipocytes and macrophages, and in the development of AT dysfunction and insulin resistance typically observed in obesity. In line with this, EVs from M1-like proinflammatory macrophages were found to reduce insulin signalling in human adipocytes, possibly mediated by nuclear factor κB (NF-κB) activation, while the opposite effect was seen with M2-like derived EVs [55]. These findings support previous in vivo studies demonstrating that AT EVs from obese mice can induce macrophage activation and insulin resistance [53], though the specific molecular mechanisms driving these changes remain unclear. In a recent study [56] it was shown that exosomes from adipose-derived stem cells (ADSCs) improved insulin sensitivity and hepatic steatosis and decreased obesity when injected into obese mice. These findings were associated with increased browning of inguinal white AT (WAT) and increased M2-like macrophage polarization in WAT [56]. Furthermore, Ying et al. (2017) [25] demonstrated that ATM exosomes from obese mice were able to cause systemic insulin resistance and glucose intolerance in lean mice, while these factors were improved in obese mice when treated with ATM exosomes from lean mice. The ATM exosomes were taken up by AT, liver and skeletal muscle in vivo, and in vitro ATM exosomes were shown to decrease insulin signalling in these cell types [25]. miR-155, a repressor of the adipogenic transcription factor peroxisome proliferator-activated receptor γ (PPARγ), was suggested as a possible key mediator of the effects of ATM exosomes on insulin resistance [25]. Together, these studies highlight the potential importance of exosome-mediated cross-talk between key metabolic tissues in the regulation of metabolism both in physiological and pathophysiological situations.

During obesity, AT dysfunction is the result of adipocyte stress characterized by hypertrophy and hypoxia. The notion that characteristics of AT EVs change during adipocyte stress is supported by several studies [46,49,57,58]. In plasma of obese subjects Eguchi et al. (2016) [46] observed an increase in EV number that correlated with insulin resistance, as well as an increase in EV PLIN 1A, a lipid droplet surface protein. Of relevance, distinct protein, lipid and miRNA profiles were found between control and stressed adipocyte EVs in vitro [46]. In another study, hypoxia was found to influence the composition of adipocyte exosomal cargo by increasing the levels of proteins related to metabolic processes, in particular enzymes related to de novo lipogenesis [57]. These exosomes were found to increase lipid accumulation in recipient adipocytes [57]. Similarly, a previous study demonstrated that CD73-containing microvesicles released from large rat adipocytes in vitro can stimulate lipid synthesis in recipient small adipocytes [59]. The transfer of lipogenic machinery may represent an attempt for hypertrophic adipocytes to shift the burden of lipid storage to recipient, non-stressed adipocytes. Additionally, Wnt/β-catenin and transforming growth factor-β (TGF-β) pathways, which play roles in fibrosis and chronic inflammation, have been found to be targeted by AT exosomal miRNAs from obese adolescents [58], suggesting that AT exosomes may play a role in the development of the systemic metabolic disturbances observed in obesity.

Brown AT (BAT), a regulator of metabolism and energy expenditure through non-shivering thermogenesis, represents a promising target organ for the development of therapeutics against obesity and associated metabolic disorders. The role of miRNAs in brown adipocyte (BA) biology, including development and functional regulation, has previously been described in detail [60], and recently two studies have investigated the role of exosomal miRNAs in BAT [24,61]. Chen et al. (2017) [61] demonstrated that the release of miRNA-containing exosomes from BAs was increased in vivo by BAT activation through cold stimulation. Specifically, the authors identified the miRNA, miR-92a in BA exosomes, and the exosomal level of this miRNA was decreased following BA activation in vitro and in vivo [61]. Furthermore, the level of miR-92a in serum exosomes was found to correlate inversely with BAT activity in humans [61]. Based on these findings, exosomal miR-92a could function as a possible biomarker of BAT activity. To shed light on the role of miRNAs derived from AT, Thomou et al. (2017) [24] used a previously generated mouse model with an AT-specific knockout of Dicer (ADicerKO mice) [62]. These mice showed loss of WAT, whitening of BAT, insulin resistance and lipodystrophy [63]. Circulating exosomal miRNA levels were decreased in ADicerKO mice, however these levels were restored upon engraftment of wild-type BAT, but not WAT, which also caused an improvement of glucose tolerance [24]. Moreover, transplantation of wild-type BAT was able to reduce the increased levels of circulating fibroblast growth factor 21 (FGF21) and hepatic Fgf21 mRNA found in ADicerKO mice, possibly mediated by the BAT exosomal miRNA, miR-99b [24]. FGF21 is known to influence hepatic metabolism, and thus the present study demonstrates a functional role of BAT exosomal miRNAs in the regulation of key metabolic tissues.

During obesity, an increase in intrahepatic triglyceride content, termed as steatosis, is characteristically observed in the liver [64]. This disease, known as NAFLD, is strongly associated with metabolic dysfunction such as type 2 diabetes, hypertension and dyslipidaemia, although the cause–effect relationship between NAFLD and these metabolic dysfunctions is unclear [64]. As the diagnosis of NAFLD currently requires invasive biopsy, better non-invasive biomarkers of the disease are warranted, though as NAFLD often occurs along with obesity and metabolic dysregulation it is difficult to selectively isolate the effects of NAFLD on circulating EV characteristics.

Nonetheless, analysis of serum EVs from patients with NAFLD showed increased levels of monocyte/macrophage- and invariant natural killer T cell-associated surface markers, allowing distinction of NAFLD patients from healthy controls or those with chronic hepatitis [65]. Chronic hepatitis C could also be distinguished from healthy liver and NAFLD by analysis of serum EV miRNAs, and classification of NAFLD through miRNA expression patterns may be possible with further methodological and analytical development [66]. In a mouse model of NAFLD, the disease was associated with increased EV number in plasma and liver, along with increased levels of two miRNAs, miR-122 and miR-192, both known to be associated with NAFLD [67]. A unique proteomic profile was also found in exosomes from this model, further supporting the possibility of using circulatory EVs as diagnostic/prognostic biomarkers for NAFLD [68]. Furthermore, EVs from obese individuals caused dysregulation of TGF-β signalling in hepatocytes [69], and these findings are supported by later studies [58]. The TGF-β signalling pathway is implicated in the pathogenesis of NAFLD and thus these findings may provide some insight into the pathogenic role of EVs in NAFLD.

As the main site of insulin-stimulated glucose consumption, insulin resistance of skeletal muscle is a key factor in the metabolic dysregulation observed during obesity. Although EV release from skeletal muscle has been observed [70,71], little is known about the contribution of skeletal muscle EVs to physiological effects of exercise or during metabolic disease.

In mice on a palm oil-enriched diet, suggested to cause skeletal muscle insulin resistance, ex vivo skeletal muscle exosome secretion was increased, and this finding was mimicked in vitro [72]. Although insulin resistance was not found to be transferred between muscle cells through exosomes, their myotube phenotype and myoblast proliferation were affected, suggesting that the deleterious effects of a palm oil-enriched diet can be communicated between skeletal muscle cells through exosomes [72]. Furthermore, when injected in vivo, fluorescently labelled EVs from skeletal muscle cells [72] or from purified skeletal muscle from palm oil-fed mice were found to be taken up by a number of tissues, including the spleen, liver and pancreas [73]. In vitro, these EVs were able to modulate gene expression and proliferation of β cells, providing a possible link between skeletal muscle and the pancreas during situations of insulin resistance [73].

The placenta is one of the main determinants of pregnancy success in mammals and controls maternal–foetal exchange of nutrients and oxygen, as well as regulating maternal metabolism to optimize resource allocation during pregnancy [74–76]. Maternal obesity is associated with a high risk of developing obstetric complications such as gestational diabetes mellitus (GDM), which can cause metabolic complications in both mother and offspring later in life [77]. During pregnancy, circulating levels of placental exosomes are significantly increased [78], and placental exosome release and composition is affected by circulating glucose and oxygen levels [79,80]. During GDM, the release of placental exosomes is further exacerbated [81], and differential expression of lipids [82] and proteins [83,84] in circulating placental exosomes has been associated with preeclampsia and recurrent miscarriages. Though little is known about the functional role of placental exosomes in pregnancy regulation, circulating placental exosome levels and composition may be used as biomarkers of placental dysfunction.

A growing body of evidence is suggesting that EVs contribute to cell–cell and interorgan communication, and the transfer of miRNAs has been particularly highlighted in EV-associated signalling. EVs are strongly implicated in the communication between key metabolic organs during metabolic homoeostasis and dysregulation. In particular, EVs seem to be important for the cross-talk between adipocytes and ATMs and the regulation of insulin resistance. However, additional research is required before EV-mediated communication may be utilized for therapeutic purposes. On a mechanistic level, the recruitment and packing of EV cargo are poorly understood, as are the processes involved in addressing the EV to a specific target cell. Some of the questions that remain to be resolved include whether EV message and address vary in response to different metabolic states, how these processes are regulated, and in what way characteristics of EVs differ between different cell types and organs. In order for these mechanisms to be elucidated, techniques for isolation and visualization of different EV species must be improved.

A number of reports have described alterations in circulating EV number or composition during situations of metabolic dysregulation and obesity, suggesting that although EV signalling may not be understood on the molecular level, EVs may be used as clinical biomarkers for diagnostic and prognostic purposes in the near future. With further understanding of EV biology, this novel signalling mechanism could one day open new avenues for the development of therapeutics against obesity and associated metabolic disorders.

• EVs mediate cell signalling through the transfer of cargo molecules from the donor cell to the target cell and/or through interaction between EV and target cell surface molecules.

• Evidence suggests that circulating EV characteristics, including miRNA content, can be used as biomarkers for a number of pathologies such as metabolic dysfunction.

• EVs released from AT may regulate the metabolic function of distant key metabolic tissues.

• Cross-talk between distinct cells of the AT may be mediated by EVs.

The authors gratefully acknowledge support from the Medical Research Council (PO 4050281695), the British Heart Foundation (RG/12/13/29853) and Wellcome (109153/Z/15/Z).

The authors declare that there are no competing interests associated with the manuscript.

ADicerKO mice

mouse model with an AT-specific knockout of Dicer

Ago2

argonaute 2

AT

adipose tissue

ATM

AT macrophage

BA

brown adipocyte

BAT

brown AT

ESCRT

endosomal sorting complex required for transport

EV

extracellular vesicle

FGF21

fibroblast growth factor 21

GDM

gestational diabetes mellitus

MVB

multivesicular body

NAFLD

non-alcoholic fatty liver disease

PM

plasma membrane

TGF-β

transforming growth factor-β

WAT

white AT