Removal of cellular protrusions

Abstract

Cell-cell communications are central to a variety of physiological and pathological processes in multicellular organisms. Cells often rely on cellular protrusions to communicate with one another, which enable highly selective and efficient signaling within complex tissues. Owing to significant improvements in imaging techniques, identification of signaling protrusions has increased in recent years. These protrusions are structurally specialized for signaling and facilitate interactions between cells. Therefore, physical regulation of these structures must be key for the appropriate strength and pattern of signaling outcomes. However, the typical approaches for understanding signaling regulation tend to focus solely on changes in signaling molecules, such as gene expression, protein-protein interaction, and degradation. In this short review, we summarize the studies proposing the removal of different types of signaling protrusions—including cilia, neurites, MT (microtubule based)-nanotubes and microvilli—and discuss their mechanisms and significance in signaling regulation.

Introduction

Signaling protrusions are specialized structures and the communication tools between cells. Therefore, if cells lose such structures, they cannot interact with other cells, just like if we lose our cell phones. Loss of signaling protrusions must greatly impact cells, which implies the possibility that cells may use the removal process to negatively regulate protrusion-mediated signaling. However, the detection of many types of protrusions is often difficult in vivo due to their highly fragile nature, and therefore the impact of structural regulation on signaling outcome is often overlooked by researchers.

Not all protrusions are too fragile to observe. Undoubtedly, the most prominent and robust example of protrusion-mediated signaling occurs in neurons forming synapses with other neurons or non-neuronal cells. Axons and dendrites are the signal sending and receiving structures, respectively, that permit a neuron to pass an electrical or chemical signal. Another example is cilia, which are often referred to as a cell’s “antenna” as they can sense extracellular stimuli or mechanical cues or work as special sites of receptor-ligand interaction (reviewed in [1, 2]). Defective removal of neuronal connections or cilia are both found to be the cause of signaling defects and are implicated in a number of human syndromes [3–7].

Other than neurons and cilia, mounting evidence suggests that protrusion-mediated signaling could be occurring in almost all cell types in a variety of contexts. These include morphogenesis during embryogenesis, immune functions, niche and stem cell interaction, and tumor progression [8–11]. Importantly, signaling protrusions are prevalent tools for cell-cell communication in complex tissue architecture. In the conventional model of paracrine signaling, signal-sending cells secrete ligands that can be received by essentially any cell located nearby. In complex tissue environments with many different cell types, protrusions can reach other cells to make direct “contact” to communicate privately and efficiently [9]. Moreover, protrusions even allow cells to communicate with distant cells contact-dependently [12].

To contribute to the efficiency of cell-cell communication, protrusions often have a specialized molecular composition for signaling. For example, the membrane of a cilia concentrates receptors and signaling molecules to locally initiate the signaling cascade. Thus, bringing the signaling factors in close proximity might be a strategy for efficient signaling activation (see [13] for a detailed review of structure-function relationships of cilia). Axons and dendrites in neurons concentrate ion channels, receptors, lipids, and cytoskeletal components for transport (see [14] for a recent review summarizing diverse structures of neurites). Cytonemes are actin-based signaling filopodia found in Drosophila wing imaginal disc cells. Cytonemes permit interaction of signaling proteins, such as ligands and receptors, between cells. Interestingly, cytonemes are induced/stabilized by the signaling molecules themselves and each cytoneme concentrates specific receptors to communicate with particular ligands [15].

The structural versatility of protrusions suited for signaling indicates that the loss of the protrusion should greatly impact the signaling outcome. Removal of protrusions or portions of protrusions must affect the turn-over time of signaling molecules, and thus must be an essential regulatory component of signaling outcomes. However, much remains unknown regarding the dynamics of these structures and how their structural changes influence signaling outcome.

1. Protrusions are sites of release of extracellular vesicles (EVs)

Apart from the increasing examples of signaling protrusions, the biology of extracellular vesicles (EVs) is rapidly growing as a field of interest. EVs are the generic term for the nanoscale membranous structures that are secreted or shed from many different types of cells and are abundant in body fluids. EVs found in body fluids are highly diverse and can be categorized based on their size and mode of biogenesis. Two types of EVs—exosomes (small EVs) and ectosomes (large EVs)—appear to be formed by distinct mechanisms [16–22]. Exosomes are produced in the endosomal pathway and are secreted by essentially any cell type [23, 24], ranging in size from 30–150 nm [25–28]. Ectosomes are formed by budding or scission of plasma membrane extensions and are generally larger than exosomes, ranging in size from 100–1000 nm [22, 25–28]. Ectosomes are membranous vesicles shed directly from portions of plasma membranes, a process often seen at the membrane of protrusions. This involves shedding of the entire lengths or portions of the protrusions and their release into the extracellular space as vesicles.

EVs can also be categorized based on their functions, as bioactive or non-bioactive vesicles. Increasing evidences of bioactive EVs have drawn researchers’ attention. Bioactive EVs participate in a diverse range of cellular events, and are considered to be a new mode of intercellular communication (reviewed in [29–31]). Moreover, EVs are studied not only for their biological functions, but also for their potential as clinical biomarkers and as cargo vehicles for the delivery of drugs or other components to specific cells for the treatment of diseases [32, 33]. Interestingly, structurally and functionally distinct types of cellular extensions such as cilia, axons, and other membrane protrusions are often found to release vesicles [34, 35], indicating the possibility that mechanisms of EV biogenesis, at least in some cases, could be related to the removal of protrusions or their components. Moreover, the requirement of common molecules, for example, tetraspanins [36], both for protrusion formation and EV biogenesis, suggests a strong link between the protrusion and EV biogenesis [36].

There are several classical examples of EV release from protrusions. Platelets are derived by shedding of plasma membrane protrusions of megakaryocytes [37]. Melanocytes release vesicles containing melanin (melanosomes) from filopodia to deliver melanin to keratinocytes [38, 39]. Cilia are also often found to produce EVs. One clear example of this was found in an unicellular organism, the green alga Chlamydomonas [40]. Ciliary EVs were shown to contain the peptide amidating enzyme peptidylglycine α-amidating monooxygenase (PAM), which converts prohormone into an active form. This EV transfer occurs between two mating algal types and is indispensable for successful mating [40]. Cilia-derived EVs have been also reported in mammals. Primary cilia in mammals are present in essentially all cell types, and are found to release EVs in addition to receiving EVs (reviewed in [41]), suggesting that ciliary EVs are a highly conserved mechanism and may be broadly utilized for different purposes.

How can protrusions do both jobs apparently independent each other, such as signaling and bioactive EV production, within a same structure? A single cilium seems to be able to release more than one type of EV. For example, EVs derived from the same cilia were shown to contain both bioactive and non-bioactive cargos. In mammalian cells, small EVs released from the base region of cilia activate Wnt signaling in target cells, whereas large EVs derived from a different portion (tips) of the same cilia, do not show any activity [42]. Similarly, in Caenorhabditis elegans (C. elegans), ciliated sensory neurons can shed EVs containing different components produced from distinct locations of the cilia (the tip or base) in response to a mating partner [43]. A more recent study in C. elegans suggests that EVs released specifically from the tip of the cilia are likely purposed for the “disposal” of ciliary materials [44]. In this case, EVs released from the ciliary tip are phagocytosed by the associated glial cells and a defect in this process results in local accumulation of ciliary proteins [44, 45]. These studies indicate that cargo sorting and the release of EVs are tightly regulated even within a protrusion. Cilia in different organisms (mammal and C. elegans) show a similar pattern of EV release, such that large ectosomes released from the tip are for disposal and small EVs (exosomes) are released from the base are for bioactivity. This may reflect a universal pattern of EV production in cilia or other protrusions as well (Figure 1). It is still unclear why these two positions of Cilia are used for different types of EV production.

Figure 1. Vesicles derived from different portions of cilia.

Large EVs (ectosomes) are typically shed from the tip region of cilia and phagocytosed, while small EVs (exosomes) are released from the base of cilia. Small EVs often exhibit bioactivity.

Certainly, not all EVs are bioactive. Indeed, EVs were initially described as a mean of disposal of unnecessary components from the cell [46]. However, because most studies of EVs have focused on their bioactivity, the impact of EV release on EV producing cells is less understood. Given that EVs are often released from protrusions utilized for signaling, it is worth revisiting the idea of disposal for signaling regulation.

2. Examples of protrusion removal via shedding

Shedding of portions or even the entirety of protrusions is observed in a variety of developmental and physiological processes, and many protrusions including neurons and cilia are removed from the cell rather than retracted or reabsorbed into the cell body. Often, removal occurs in an actively regulated manner via extrinsic or intrinsic factors. Moreover, the vesicles for disposal are often taken up or engulfed by neighboring cells for recycling of their contents, or by professional phagocytes for degradation (see Section 3 for more details). These processes are either “phagocytosis” or a mechanism called “trogocytosis” in which phagocytes directly contact a protrusion to ingest or “bite” small parts of cells, which are also released into extracellular spaces as EVs. Table1 summarizes examples of removal processes of different types of protrusions.

Table 1.

Example of removal of protrusions or their contents for the purpose of signaling regulation

Type of protrusions	Cell type	Cargo	Fate of cargo	References
Cilia	Mammal, Photoreceptor	Opsins in outer membrane	Opsins taken up by RPE cells are recycled	[47–50]
Cilia	Mammal	G protein-coupled receptors (GPCRs)	Unknown	[51]
Cilia	Chlamydomonas reinhardtii	Vegetative lytic enzyme (VLE)	Release mitotic daughter cell from mother cell wall	[52]
Cilia	Chlamydomonas reinhardtii	peptidylglycine α-amidating monooxygen ase	Taken up by the other mating-type	[40]
Cilia	Chlamydomonas reinhardtii	membrane polypeptide, SAG1-C65	Removed	[53]
Cilia	C. elegans, sensory neurons	Cargo required for mating behavior	Phagocytosed by the associated glial cells	[54] [45]
Filopodia	Mammal, melanocyte	Melanin in melanosomes	Taken up by keratinocytes	[55]
Microtubule-based (MT)-nanotubes	Drosophila, germline stem cells (GSCs)	Tkv receptor	Taken up by niche cells and degraded	[56]
Microvilli	Mammal, T-cells	T-cell receptor	Trapped on the surface of antigen presenting cells	[57]
Dendrites	Mammal, neurons	Neurites/synapses	Engulfed by microglia and degraded	[58]
Axons	Mammal, neurons	Neurites/synapses	Engulfed by astrocytes or microglia and degraded	[59] [60]

2.1. Removal of portions or entire cilia

Cilia are specialized cellular projections found in a broad range of organisms, from unicellular organisms to humans. The structures of cilia are highly conserved, containing a nine-fold doublet microtubule bundle, called the axoneme, that extends from a mother centriole-derived basal body wrapped by a ciliary membrane. The primary cilium is non-motile and appears in most mammalian cell types. The primary cilium often functions as a specialized signaling center of cells; thus it has been deemed the cell’s antenna. Cilia detect various signals such as hedgehog, noncanonical Wnt, and GPCR-mediated signals including odorants and light [61, 62]. To execute their functions, the composition of cilia is specifically regulated apart from other regions of the cell. The ciliary membrane is concentrated with specialized lipids and proteins and separated from other cellular membranes via a diffusion barrier called the transition zone [63].

Given the function of cilia as the antennae, removal of the entirety or parts of cilia or contents should have a significant impact on rest of the cell. Shedding, ectocytosis or decapitation are the pathways for removing parts of cilia which have mainly been considered as a form of cellular disposal [64]. As a consequence of shedding or decapitation, the removed part of the cilium is released into the extracellular space [64]. Ciliary-derived vesicles have compositions different from the cilia from which they were derived. A subset of ciliary proteins including proteases, ESCRT proteins, small GTPases, and ubiquitinated proteins have been found to be enriched in cilia-derived EVs [65], indicating that cells sort components to be removed. The removal of ciliary components by EVs often regulates appropriate turnover time of proteins within cilia. For example, in the green alga Chlamydomonas, ectocytosis is required for the timely removal of ciliary components [52, 53]. Similarly, ectocytosis was reported to prevent excess accumulation of ciliary cargo in C. elegans sensory neurons [45]. Moreover, a study used real-time imaging of mammalian cells and showed that the GPCRs that localize on primary cilia are shed into vesicles from the tips of cilia upon signal activation [51] (Figure 2B). GPCRs are enriched in ectosomes and their removal is required for signal attenuation, indicating that signal-dependent ectocytosis regulates ciliary signaling [51] (Figure 2B). These examples suggest the universality of removal of ciliary components, which is strictly regulated in response to extrinsic cues.

Figure 2. Shedding of cilia regulates turnover of signaling molecules.

A) Shedding of outer segments of photoreceptors (purple cell). Outer segment discs contain rhodopsin, which belongs to the GPCR superfamily. Rhodopsins undergo both a conformational change (becoming opsins) and activation of the G protein upon receiving a photon, and need to be renewed/replaced daily. Part of the outer segment disk containing used opsins is shed, phagocytosed by neighboring retinal pigment epithelium (RPE) cells, transformed back to rhodopsins, and then transported into photoreceptor cells.

B) GPCRs (receptors) localize to cilia and accumulates in the tip portion. Upon stimulation by the hedgehog ligand, GPCRs that fail to be retrieved from the tip are removed by ectocytosis.

The “disposed” portion of cilia can also be recycled. A unique example of this is observed in the photoreceptor cells of mammalian retina (Figure 2A). Photoreceptor cells form specialized cilia (outer segment) that contain rhodopsin (the substance for photo-sensing), and the tip of each outer segment is released approximately 6 times per day to remove used opsins with oxidized retinols (reviewed in [50]). Released segments are then taken up by the neighboring retinal pigmented epithelium where the retinol is reduced and transported back to the photoreceptor cells to regenerate functional rhodopsin (Figure 2A).

In proliferating cells, extension and retraction of cilia occur in a cell cycle-dependent manner. Typically, the basal body derived from the mother centriole extends cilia in post-mitotic phases (G1 or G0 phases) of the cell cycle, then cilia dissociate prior to mitotic entry [66]. The exact timing of the ciliary cycle varies by cell type [66]. Two modes have been suggested for how cilia disappear during the cell cycle: resorption, in which the axoneme is depolymerized and ciliary contents are incorporated into the cell, and deciliation, in which the axoneme is excised near its base and the entire cilium is released into the extracellular space [67]. It is still unclear which proposed mechanism is true, or if it varies in different cases. A recent study used live imaging of cilia during serum-induced ciliary assembly and disassembly to determine the predominant mechanism of ciliary loss, finding that deciliation is more prevalent than resorption in mammalian cells [68].

A recent study demonstrated the decapitation of the ciliary tip appears to induce subsequent entire ciliary retraction in mouse embryonic fibroblasts. In this case, decapitation removes the tip region of ciliary components including the intraflagellar transport-B (IFT-B) complex which delivers α and β tubulin subunits towards the ciliary tip for elongation. Unexpectedly, this study showed that the loss of cilia is not just the consequence of routinely occurring cell cycle process, instead cilia removal and cell cycle regulation are mutually dependent processes [69].

2.2. Receptor removal from microtubule-based nanotubes

Another type of protrusion known to undergo the process of removal of its contents or portions is the MT-nanotube, which is spefically present on Drosophila germline stem cells (GSCs). In the Drosophila testis, 8–10 GSCs attach to the cluster of somatic cells, a major niche component called the hub (Figure 3) [70]. Hub cells secrete niche ligands, including Decapentaplegic (Dpp) (Figure 3). Dpp functions in the BMP signaling pathway and is required for GSC maintenance [71]. The receptor for Dpp, Thickveins (Tkv), localizes to the surface of MT-nanotubes, where Dpp-Tkv interaction and Tkv activation both occur. Therefore, MT-nanotubes promote Dpp-Tkv interaction and signal activation. However, at the same time, Dpp signal attenuation also occurs on MT-nanotubes [56].

Figure 3. Tkv removal from MT-nanotubes is required for signal adjustment.

(Left) Microtubule (MT)-nanotube appears on the Drosophila GSC and promotes the reception of Dpp ligand secreted from the niche, a cluster of hub cells. (Right) Shedding of MT-nanotube membrane and potential engulfment by hub cells. Vesicles taken up are fused to lysosomes in hub cells and ultimately degraded, resulting in attenuation of the signal.

Tkv localized on MT-nanotubes is removed from the MT-nanotube and subsequently taken up by the adjacent niche cells where it is degraded. Expression of ubiquitination-defective Tkv, which cannot be removed from MT-nanotubes, overactivates downstream signaling, suggesting that Tkv removal is required for signal attenuation (Figure 3) [56]. MT-nanotubes share their molecular components with cilia, such as IFT-B [70], suggesting a possibility that Tkv removal may utilize a mechanism similar to ciliary ectocytosis. The mechanism of how Tkv is transported into hub cells to be degraded is still unknown. MT-nanotubes are stable structures during the entire cell cycle; however, they seem to disappear during mitosis, similar to cilia. GSCs are a constantly dividing population and the estimated cell cycle time is approximately 12 hours [72]. Therefore, the removal or resorption of the entire length of MT-nanotubes may occur at least once in 12 hours. It is still unclear whether Tkv is transported together with entire MT-nanotubes during the cell-cycle dependent removal process or if Tkv is specifically packaged in a vesicle and removed. Further studies using long-term live observation or ultrastructure analysis will be necessary to identify the mechanism of this phenomenon.

Moreover, it is still unclear why Tkv receptor must be removed in the hub cells, rather than in the GSC itself. When surface receptors are internalized, endosomes containing receptors can still signal into the cells until multivesicular body formation for subsequent lysosomal fusion [73]. Therefore, Tkv receptor release outside of the cell may be a safeguard to protect stem cells from excess activation of the niche signal. It will be interesting to investigate whether similar mechanisms are used for other stem cell systems.

2.3. Removal of microvilli

Microvilli are actin-dependent cellular membrane protrusions often described as thin and finger-like. They appear on a broad range of cell types including lymphocytes [74], intestinal epithelial cells [75], endothelial cells [76, 77], dendrites [78], neurons [79], and oocytes [80]. The function of microvilli varies depending on the cell type; for example, microvilli are well-known to facilitate absorption in the intestinal epithelium, while in lymphocytes they aid in adhesion [81, 82]. In the oocyte, it is suggested that microvilli are important for sperm binding and fusion [83–85]. The best-studied example of role of microvilli on signaling occurs at the immunological synapse formed between T-cells and antigen-presenting cells (APCs). T-cell receptors (TCRs) and co-receptors are highly clustered on the tips of microvilli, suggesting that microvilli play roles on signaling between T-cells and APCs [86–90]. A recent study demonstrated that the TCR containing vesicles are released from the tip of microvilli and localized on the surface of APCs [57], indicating the possibility that vesicle release from microvilli is for the purpose of signaling regulation.

Other than the example of T-cells, the roles of microvilli for cellular signaling are still unclear. However, studies over the past few decades have found that various receptors are enriched on microvilli surfaces, including insulin receptors [91], selectin [92]. Moreover, EVs have been reported to shed from microvilli. In mammalian cell culture, EVs containing the glycoprotein prominin-1 were reported to originate from epithelial cells’ microvilli [93], and the regulation of EV release from these microvilli was found to be dependent on prominin-1 binding to membrane cholesterol [94]. Additionally, it was recently reported in Drosophila that the microvilli of the wing imaginal disc epithelial cells also shed EVs which contain the morphogen Hedgehog [95] required for wing imaginal disc development. These indicate the possibility that EV shedding promotes remodeling of microvilli in order to regulate signaling in other cell types.

2.4. Removal of neuronal connections

Neurons initially extend an excess number of neurites to form multiple synapses from single cells. During development and the early years of life, neurons use a unique way to remove unnecessary neurite connections, which is called neurite pruning (reviewed in [96]). Pruning selectively removes axon or dendrite branches from neurons and is an essential process for establishment of the refined neural circuits [96]. Defective pruning is implicated in a variety of adolescent and adulthood neurological disorders, such as Alzheimer’s disease, multiple sclerosis, schizophrenia, and autism [3–6].

In past decades, studies in different systems have identified various extrinsic and intrinsic regulatory pathways for the determination of pruning, suggesting that the molecular mechanisms for each pruning event are unique [97]. It has been hypothesized that neural activity, both long-term potentiation (LTP) and long-term depression (LTD) are involved in the selection process to determine which connections are to be pruned [97]. During the dendrite pruning process, “thinning” at the proximal region of neurites is initiated first to trigger the subsequent pruning at distal portions of dendrites [98]. Local endocytosis is required for the dendrite thinning step at the defined region. During a later pruning step, the GTPases Rab5 and Dynamin are required for removal of chopped dendrites [98].

Axons are also subjected to elimination. One study demonstrated that axon pruning in neuromuscular junctions is mediated by shedding of membrane-bound remnant “axosomes”; which was shown with correlating light and serial electron microscopy [99]. Interestingly, axosomes are engulfed and subsequently eliminated by glial cells, reminiscent of the case of outer segment of photoreceptor and MT-nanotubes described above [99]. Microglia engulfment was also shown in presynaptic inputs via pruning of retinal ganglion cell axons [60]. Complement receptor 3(CR3)/C3 serves as the microglia-specific phagocytic signaling pathway and disruption of this pathway resulted in sustained deficits in synaptic connectivity [60].

Developmentally regulated pruning suggests the significance of removal of signaling structures for the purpose of terminating cell-cell interactions. Later, engulfment of neurites or synapses by microglia [58] or astrocytes [59] are both considered subtypes of trogocytosis (see section 3.5), implying the commonality of mechanisms among cilia, neurons, and MT-nanotubes.

3. Mechanisms of removal of protrusions

3.1. The role of the actomyosin cytoskeleton

Structurally, cilia are supported by a microtubule-based axoneme and originally thought to be devoid of F-actin [100]. However, recent studies have demonstrated that actin is required for ectocytosis of cilia, which shortens the cilia [51]. Activated GPCRs can be released from the tip of the cilia via signal-dependent ectocytosis, a process requiring actomyosin activity. Inhibition of actin has been shown to block ectocytosis and result in elongated cilia [51]. Decapitation of the tip of cilia is also induced by ciliary PI(4,5)P2-mediated F-Actin accumulation in ciliary lumen [69]. Consistently, previous work has established that the actin regulators RhoA, Rac, and Cdc42 are required for ectosome release from the plasma membrane [101], a process that is at least in part facilitated by cortical actomyosin contraction [102]. These findings not only demonstrate the crucial role for actomyosin contraction in ectocytosis, but also suggest a consensus in which actomyosin negatively regulates the length of cilia.

3.2. The endosomal sorting complexes required for transport (ESCRT) machinery

Although the exact mechanisms for ectocytosis remain elusive, findings from recent studies are shedding light on the process. The endosomal sorting complexes required for transport (ESCRT) machinery is indispensable for membrane remodeling [103–105]. ESCRT-III and Vps4-dependent shedding from the plasma membrane has been extensively studied in the context of viral budding and the process bears some resemblance to that of ectocytosis. In this mechanism, ubiquitination is a key step that is necessary for ESCRT machinery recruitment to the site of viral budding [106, 107]. Notably, ESCRT-III and Vps4 are recruited to the site of membrane budding just before virion release [108, 109], and these factors, along with the ESCRT-associated factor ALIX, are required for the budding and scission of the packaged virion from the plasma membrane [110]. The ESCRT-III complex forms spiral-like structures around the base of the “neck” of a budding membrane; it is thought that this spiral-like structure, formed by the subunit Snf7, is responsible for deforming the membrane around cargo to form a bud [111]. If ESCRT function is required for protrusion removal via promoting ectosome shedding, loss of ESCRT should cause defect of protrusion removal. Strikingly, a recent study using a combination of proteomic analysis and knockdown of ESCRT-related proteins in Chlamydomonas showed that the ESCRT proteins mediate ectosome release and thereby influence shortening of Cilia [112]. Similar mechanism may be used for removal of other protrusions.

3.3. Bardet-Biedl syndrome (BBS) proteins

Bardet-Biedl syndrome (BBS) are caused by the mutation of ciliopathy genes encoding BBS proteins [113]. BBS is characterized by many symptoms including male infertility, retinal degeneration, renal failure, and diabetes caused by the dysfunction of primary cilia [114, 115]. BBS components assemble a coat structure around vesicles [116], similar to COPI, COPII, and clathrin-coating, in an octameric complex referred to as the BBSome [117, 118]. The BBSome has been implicated in ciliary trafficking and is suggested to be required for transport of signaling receptors to ciliary membranes [116, 119, 120], and also regulate cilia disassembly [121, 122]. Studies have shown that BBS components negatively regulate ectocytosis in mammals and C. elegans [42] [123] where BBS proteins seem to regulate ciliary cargo retrieval from ciliary tip to cell body, indicating that BBS can act as a key factor that determines whether cargos will be removed into EV or retrieved into cells.

3.4. Tetraspanins

Tetraspanins form a transmembrane protein superfamily. They interact with a large variety of transmembrane and cytosolic signaling proteins to create membrane microdomains. Because they are abundantly found in the membrane of EVs, tetraspanins have been widely used as EV markers. Tetraspanin genes have been shown to regulate cargo sorting for EVs, and their function on ESCRT-independent endosomal sorting is essential for melanosome biogenesis and transfer [124, 125]. Intriguingly, tetraspanins are also enriched at various types of protrusions, including intestinal microvilli [126], oocyte microvilli [80, 127, 128], microprotrusions of platelets [129, 130], photoreceptor sensory cilia [131], and synapses [132], implicating the role of these protrusions as the platform of EV release. Moreover, studies have demonstrated that tetraspanins regulate a variety of signaling pathways, such as integrins, EGFR, TNF-α, c-Met, c-Kit and TGF-β [133]. This suggests that tetraspanins may be a key factor to determine the fate of signaling molecules by deciding whether to sort them into EVs for disposal. Although a number of signaling molecules have been identified to interact with tetraspanins, the mechanisms by which tetraspanins control signaling still remains largely elusive. Future studies, especially detailed structural analysis, will help us understand these mechanisms.

3.5. Final processes (eat-me signal and phagocytosis/trogocytosis)

What triggers ESCRT-III assembly to initiate ectocytosis from protrusions? The TAT-5 phospholipid flippase maintains the location of phosphatidylethanolamine (PE) on the inner leaflet of the plasma membrane. Absence of TAT-5 results in overproduction of EVs in early embryonic divisions of C. elegans [134, 135]. Flippases also translocate phosphatidylserine (PtdSer) from outer to inner leaflets of the plasma membrane. Opposing enzyme scramblases of the TMEM16 family of proteins function to shuffle PtdSer asymmetrically in a Ca²⁺- and caspase-dependent manner and often show a polarized localization in other cell types [136, 137]. These reports suggest that PE or PtdSer externalization may be upstream of ESCRT machinery. The engulfment of outer segments of retinal pigment epithelium occurs via recognizing PtdSer presented on the surface of outer segments by flippase, P4-ATPase, and Atp8a2 [138]. Surprisingly, the cargo opsin itself has also been reported to possess ATP-independent lipid flippase activity [139], indicating that certain cargos can promote their disposal or transport, and externalization of PtdSer was initially found in apoptotic cells as an “eat-me” signal for phagocytes [140]. This suggests the possibility that PtdSer externalization could be recognized by the same or similar phagocytic receptors present on phagocytes for clearance. It would be tempting to examine if PtdSer externalization is utilized for EV disposal. The membrane-flipping activity is known to be triggered by calcium influx [137]. However, whether PtdSer externalization is the common mechanism that triggers EV shedding across different cell types remains to be determined, and the entire mechanism of how PtdSer externalization is locally regulated is not fully understood (Figure 4).

Figure 4. Phagocytes can bite off protrusions (Trogocytosis).

(Left) PtdSer normally localizes to the inner leaflet of the plasma membranes via the function of flippase. Scramblase is activated by Ca2+ and scrambles PtdSer localization. (Middle) Externally localized PtdSer is recognized by phagocytic receptors, and these receptors recruit actomyosin to the phagocytic cups. Phagocytic cups generate constriction forces to “bite” up the tip of the protrusion (Right). Protrusions are cell-autonomously shed via activating ESCRT. These mechanisms were both reported in the engulfment of photoreceptor outer segment, and thus may occur simultaneously (see text).

The activation of phagocytic receptors in the phagocytic retinal epithelia leads to myosin recruitment in the phagocytic cup and is required for successful engulfment [141]. This “bite” mechanism is closely related to the mechanism referred to as “trogocytosis” (Figure 4). Trogocytosis has been actively studied in immune cell interactions at immunological synapses [142]. Trogocytosis is similar to phagocytosis but involves only eating parts of cells. Although this process can ultimately kill the target cells [143], in many cases the removal occurs only partially and, interestingly, cellular protrusions are often targeted by this ingestion [143]. A study using time-lapse imaging revealed that trogocytosis is a rapid process, immediately triggered when phagocytes meet filopodia [58]. Increasing evidences suggest that trogocytosis may be broadly utilized even outside of immune responses [144]. Defining further molecular mechanism of this process is a fascinating future study.

4. Unsolved questions

The reason why EV release preferentially occurs from protrusions is not clear. Protrusions may act to compartmentalize cargos away from their cell body for further sorting into EVs. Or they may be structurally suitable to sense extracellular stimuli which drive release or shedding of vesicles. Another possibility is that protrusion formation itself utilizes similar molecular components for exocytosis or ectocytosis, and thus cells have evolved to use the same components for bioactive EV production. The usage of abundant surface area and curvature of plasma membranes on protrusions should be suitable to produce EVs in an energy efficient manner. It is currently unknown if there is any qualitative/quantitative difference between protrusion-derived EV and cellular membrane-derived EV.

Another intriguing question is how protrusions and their contents are determined to be removed or not. During development, different cells may stop or start their interaction or change their interacting partner. Therefore, the remodeling of protrusions needs to be regulated developmentally via extrinsic and/or intrinsic factors. During the remodeling of neural networks, the neuronal activity itself seems to be a primary factor for determination of the fate of connections [97]. It is tempting to speculate whether there is a universal determination mechanism for the formation or removal of signaling protrusions.

In contrast to occasional removal, daily turnover of protrusion contents occurs through constant removal, which appears to be essential for adjusting signaling at the physiological level. Several examples suggest or support this. SAG1-containing cargo moves unidirectionally on the cilia and never enters the cell body [53]. MT-nanotubes carry Tkv receptor into niche cells and Tkv is never observed in the cell body unless trafficking is suppressed [56]. In this way, niche signal reception is tightly limited to stem cell populations by limiting the Tkv receptor to be inherited into differentiating daughter cells during cell division [56]. Removal of protein at a certain rate and location may contribute to the specificity and selectivity of protrusion-mediated signaling.

Do protrusions use any common mechanism for disposal processes? Indeed, many mechanistic and molecular commonalities have been found in different types of protrusions. For example, removal of signaling components is dependent on signal activation itself [51]. Ubiquitinated proteins are often targeted for the removal process, and the cargo sorting processes share similar molecular machinery. More and more studies using proteomics approach aiming to determine composition of EVs and protrusions have identified interesting similarity among different systems. Ultimately, an “eat-me” signal exposed on the surface of protrusions may be a common mechanism for either shedding or targeted by “biting” through trogocytosis. Further studies will be necessary to identify how protrusions or their components are targeted by these common mechanisms to be disposed from cells.

Conclusion and remarks

Signaling protrusions contribute to special temporal regulation of cell-cell communication over a variety of cell types. Studies described in this review suggest that the removal of protrusions occurs constantly in order to maintain physiological turnover of proteins or in response to developmental or environmental signals, and dysregulation of the removal process greatly impacts signaling outcome. Studying regulation of protrusions has been challenging due to technical limitations. Future studies utilizing new techniques such as super-resolution live imaging, cryo-correlative light and electron microscopy (Cryo-CLEM) or high sensitive proteomics approaches may greatly contribute to this field. Identifying common molecular pathways and comprehensively examining the impact of manipulation of these pathways on various cellular protrusions may greatly facilitate our understanding of signaling regulation in vitro and in vivo.