On the Existence of Endocytosis Driven by Membrane Phase Separations

Abstract

Large endocytic responses can occur rapidly in diverse cell types without dynamins, clathrin, or actin remodeling. Our experiments suggest that membrane phase separations are crucial with more ordered plasma membrane domains being internalized. Not only do these endocytic processes rely on coalescence of membrane domains, they are promoted by participation of membrane proteins in such domains, one important regulatory influence being palmitoylation. Membrane actin cytoskeleton in general resists membrane phase transitions, and its remodeling may play many roles. Besides membrane 'caging' and 'pinching' roles, typically ascribed to clathrin and dynamins, cytoskeleton remodeling may modify local membrane tension and buckling, as well as the presence and location of actin- and tension-free membrane patches. Endocytosis that depends on membrane phase separations becomes activated in metabolic stress and in response to Ca and PI3 kinase signaling. Internalized membrane traffics normally, and the secretory pathway eventually resupplies membrane to the plasmalemma or directs internalized membrane to other locations, including the extracellular space as exosomes. We describe here that endocytosis driven by membrane phase transitions is regulated by the same signaling mechanisms that regulate macropinocytosis, and it may play diverse roles in cells from nutrient assimilation to membrane recycling, cell migration, and the initiation of quiescent or hibernating cell states. Membrane ordering and phase separations have been shown to promote endocytosis in diverse cell types, including fibroblasts, myocytes, glial cells, and immune cells. We propose that clathrin/dynamin-independent endocytosis represents a continuum of related mechanisms with variable but universal dependence on membrane ordering and actin remodeling.

1. Introduction.

The evolution of cellular life must have been coupled closely to evolution of the complex lipid membranes that constitute the boundary of cells to their environment [1-4]. One fundamental property of complex lipid membranes, which has gained increasing attention, is a tendency to develop membrane phase separations (MPSs), thereby generating membrane domains with different compositions and physical properties [5]. Once formed, such domains have a tendency to develop curvature and form buds, either inwardly or outwardly, with a dependence on their sterol content [6]. At least in part, this tendency is dictated by 'line tension' between domains that tends to minimize domain perimeters [7-9]. The sources of additional energies required to excise small vesicles remain a matter of debate. However, numerous studies document that budding can indeed proceed to fission of relatively small vesicles in the absence of classical endocytic proteins or dedicated 'membrane bending' proteins [5, 10-13]. In short, it would be difficult to believe that MPSs did not play a fundamental role in the evolution of membrane trafficking and signaling mechanisms, eventually in collaboration with proteins that mechanically modify membranes.

'L-form' bacteria, which lack a peptidoglycan wall, provide the best living cell model to interrogate potential early steps in the evolution of cellular life [14, 15]. Impressively, these most primitive of all known cells undergo profuse vesiculation, blebbing and tubulation during their growth and proliferation [15, 16]. Although the biophysical mechanisms underlying these membrane remodeling events are not entirely elucidated, an involvement of cytoskeleton-like, membrane-interacting proteins was surprisingly not confirmed [1]. Rather, it was found that remodeling and cell viability have a strong dependence on lipid synthesis per se [17]. It seems at minimum likely that MPSs play a major role, and the existence of MPSs in other prokaryotic cell membranes has also gained attention in recent years [18-20].

The idea that endocytosis could have evolved from mechanisms other than membraneinteracting/binding proteins has been hinted previously, at least implicitly [21-23]. At minimum, it seems reasonable that MPSs could have functioned as a catalyst for the subsequent evolution of dedicated membrane remodeling proteins. The idea that lipid 'rafts' internalize as an endocytic mechanism without the aid of adapters or cytoskeleton is implicit in a great many publications in the last 20 years [22-25]. However, proofs that this is occurring remain few, and principles that underlie this form of endocytosis remain controversial. Furthermore, the existence of constitutive clathrin-independent endocytic mechanisms was recently challenged by experiments in which universal chemical labelling of plasma membrane (PM) proteins was used to identify the sources of endocytic vesicles [26]. Unexpectedly, the results suggested that 95% of constitutive endocytosis in standard cell cultures occurs via clathrin-dependent mechanisms. One major reservation about this conclusion is that the experiments relied on quantitative labeling of surface membrane proteins with a reducible SNAP tag. Certainly in our hands, clathrin/dynamin-independent endocytosis is highly sensitive to perturbation of the outer monolayer, and the presence of ionized groups at the cell surface effectively blocks these forms of endocytosis [27]. In any case, the firm conclusion that clathrin-independent pathways do not contribute significantly to endocytic flux is not well justified by experiments with one cell type (HeLa) employing temperature jump to initiate membrane turnover after long-term cooling and chemical labeling.

The interest of our group in MPSs began with the observation that perfusion of the phosphoinositide, PIP2, into cells via patch clamp caused substantial endocytic responses that did not seem to depend on conventional endocytic mechanisms [28]. Later, we found that in many cell types more than one half of the PM could be rapidly internalized subsequent to a large Ca elevation, and it quickly became a dilemma that we could not confirm an involvement of dynamins, clathrin, or actin cytoskeleton remodeling [29]. In five further articles we described support for the hypothesis that endocytosis relies on MPSs with the more ordered membrane domains being internalized, we described evidence that this type of endocytosis was activated in multiple forms of cell stress, we described that palmitoylation of membrane proteins was crucial to promote these processes, and we described that this type of endocytosis depended strongly on the presence of palmitoylated cargo [27, 30-33]. Perhaps because our key methods are not commonly employed in the traditional cell biology community and/or because we usually employ 'suspension' cells rather than adherent cells, this body of work has not been discussed in recent reviews [34-36] or in articles communicating related experimental work [37-41]. We point out in this connection that the Ca-activated 'delayed' endocytosis described by us appears in some ways to be similar to the clathrin-independent endocytic process, dubbed clathrin-independent carrier (CLIC) endocytosis by Rob Parton and colleagues [21, 37, 38, 42]. However, a major difference is that the endocytic processes described by us do not seem to require actin cytoskeleton remodeling during their progression.

With this background, we use this short review to summarize our work and provide a few new results that support our conclusions. We explain in simple terms our methodologies and rationales, as well as criteria that may help to distinguish a role for MPSs in endocytosis. Finally, we relate this work to that of others, as well as to the physiological and pathological processes where these forms of endocytosis are likely to play important roles.

2. The PM is poised to phase separate when taunted by diverse perturbations.

The first 'case report' of massive endocytosis (MEND) was delivered by Lazaro Mandel and colleagues, describing that epithelial cells can internalize nearly all their Na/K pumps and E-cadherins during ATP depletion [43]. This form of MEND occurred in polarized epithelial cells with a time course that paralleled depolymerization of the cortical actin network and dissolution of the actin ring. In subsequent studies, the membrane structures formed by internalized PM were found to be highly heterogeneous and generally inconsistent with conventional clathrin-dependent endocytosis [44]. While this MEND process was found to be poorly reversible in some cells [44], it has been exploited extensively to study Na/K pump trafficking that requires reversibility and maintained cell viability [45, 46]

The first really decisive example of 'lipid-driven' MEND came from the Fred Maxfield group. Fred and colleagues discovered that the treatment of cells with bacterial sphingomyelinases caused reversible MEND as a result of extensive cleavage of sphingomyelin to form ceramide [47]. Remarkably, the vesicles formed in this protocol clearly integrated into the secretory pathway, and over the course of 30 to 60 min cells were able to resynthesize sphingomyelin and reestablish a normal PM without activating cell death programs. That the MEND occurring in sphingomyelinase-treated cells indeed could be a direct result of sphingomyelin cleavage, followed by the internalization of ceramide rich membrane domains, was supported almost in parallel by studies of complex giant vesicles containing cholesterol and sphingomyelin [11]. Entirely consistent with the Maxwell experiments, treatment of giant liposomes with bacterial sphingomyelinase causes small vesicles to shed profusely into the lumen of the liposomes. Using patch clamp to monitor cell area as PM capacitance (Cm), as described in Fig. 1A, we verified the occurrence of MEND within a few seconds during treatment of cells with bacterial sphingomylinases [29]. Besides documenting the loss of >50% of the PM in seconds, we verified that this form of MEND could 'reverse' over 10 min via ATP-dependent processes, presumably reflecting normal function of the secretory pathway together with resynthesis of sphingomyelin. While we did not verify the hypothesis that native sphingomyelinase activities cause physiological endocytosis, as proposed by the Norma Andrews group [48], the results encouraged us to investigate the role of MPSs in Ca-activated MEND that occurs in the wake of large Ca transients.

Figure 1.

Massive endocytosis (MEND) induced by perturbing the outer monolayer of BHK cells by detergents at submicellar concentrations. Modified from published work [27, 29] with permission by Rockefeller University Press. A. Methods. Experiments are performed by the patch clamp method using BHK cells that have been removed from dishes by trypsin treatment [29]. The patch pipette tip provides access to the cytoplasm so that ions exchange with the pipette solution within less than 20 s. The PM capacitance, a measure of charge moved to change membrane voltage, is monitored on-line via the continuous application of square wave voltage pulses, usually 20 mV in magnitude and 2 ms in duration. The extracellular solutions contain 0.5 mM EGTA and no added Ca. The endocytic responses described here were unaffected by chelating all cytoplasmic Ca with 10 mM EGTA [27], B. Induction of MEND by Triton X-100 (TX100). FM 4-64 dye (5 μM) was applied and removed as indicated below the capacitance record, calibrated as percent of maximal PM membrane area. Initially, FM dye labels the cell rapidly and can be washed off completely within a few seconds. After applying 0.12 mM TX100 together with the dye, 75% of FM dye remains associated with the cell after dye wash-out. In the same time frame (<30s), membrane area has decreased by 70%. Accordingly, the majority of the PM was internalized, and this occurred in the absence of any detectable membrane leaks. Essentially identical results were obtained in the presence of 5 μM latrunculin with 2 mM nonhydrolyzable ATP, AMP-PNP, in the cytoplasmic solution [27]. C. Repeated cycles of TX100-induced MEND. As indicated below the capacitance (PM Area) record, TX100 (0.12 mM) was applied five times for 15 s. MEND amounted to >50% of the PM in each case. PM area recovered with a time constant of 12 to 25 min. Recovery was dependent on the presence of cytoplasmic ATP and was blocked by low concentrations of NEM [27]. D. The hydrophobic cation dodecyltriphenylphosphonium (C10-TPPP) induces strong MEND at a concentration of 40 μM, >20 times less than its CMC. Inward membrane current, plotted upwardly, reflects C10-TPP permeation into the BHK cell. The current decreases by only 20% during the time over which 70% of the PM is internalized. The peak current on second application of C10-TPP is reduced by only 17% in relation to the first application, although membrane area is reduced by 70%. Accordingly, the PM that remains at the cell surface is substantially more permeable to C10-TPP than PM that is internalized. We conclude that detergents result in the internalization of primarily more ordered PM and therefore that these MEND responses are dependent on the generation of MPSs.

Next we considered other interventions that cause MPSs to form and vesiculate in giant liposomes. We were impressed by studies showing that some detergents at low concentrations, especially cholesterol-phobic detergents such as TritonX-100 (TX100), could cause complex membranes to phase separate [10]. When tested in a wide variety of cells, we found that rapid extracellular application of TX100, as well as several other detergents at sub-CMC concentrations, caused the internalization of more than 50% of PM within a few seconds [27]. In contrast to results with giant liposomes, the vesiculations occurring in cells occurred in a preferential inward direction, rather than to cause membrane shedding. At least from our perspective, it appeared that MPSs induced by detergents could cause MEND just as profoundly as the activity of sphingomyelinases. That membrane reordering, rather than phospholipid extraction, was essential became evident from the fact that detergents with ionized head groups did not cause MEND during their application. Rather, MEND occurred upon washing off ionic detergents, such as SDS, as expected if their presence induced MPSs with life times of a few seconds, but inhibited the final excision of vesicles (i.e. the fusion of the outer monolayer with itself) [27].

Fig. 1B illustrates an experiment using both optical and electrical methods to determine membrane movements. The cell capacitance is plotted as percent of initial PM area (red curve), and cellular fluorescence is plotted as percent of maximal fluorescence upon applying the rapidly reversible membrane dye, FM4-64 (5 μM). Before applying TX100, the FM dye can be applied and washed off with fluorescence returning to baseline within a few seconds. When TX100 is applied rapidly together with dye, membrane area decreases by nearly 70%, and thereafter the membrane fluorescence decreases only by about 25% upon washing off dye. This indicates that dye has been locked into vesicles below the outer rim of PM. We mention that FM4-64 is a particularly good dye for these experiments because it binds rather evenly to the cell membrane that is internalized and membrane that remains at the cell surface. This is not the case for most other membrane probes that we have employed in optical and electrical measurements. Rather, most amphipathic probes bind substantially less well to the membrane that internalizes [27, 33], Related to this, our analysis of electrical signals associated with TX100 binding suggested that TX100 binds primarily to membrane domains that remain at the cell surface, a conclusion that helps to explain why the vesicles formed in these experiments are not toxic. Rather, they immediately couple to and take part in the physiological trafficking mechanisms of cells. As illustrated in Fig. 1C, PM area reestablishes itself after MEND over the course of ~15 min. Thereafter, MEND responses can be repeated, and sequences of PM loss and replenishment can be repeated several times in stable recordings. That the membrane trafficking mechanisms mediating this turnover are 'physiological' is supported by the findings that PM expansion requires ATP and is blocked by sulfhydryl reagents as well as by oxidative stress [27].

The conclusion that PM, which is internalized in these experiments, is more ordered than membrane that remains at the PM is supported by the analysis of more than 20 amphipathic membrane probes that could be employed in relevant electrical and optical experiments [33], Fig. 1D illustrates use of the hydrophobic cation, dodecyltriphenylphosphonium (C10-TPP), to induce MEND [27]. Delocalization of the positive charge of this cationic detergent allows it to permeate membranes and generate an electrical current upon doing so. As illustrated in this experiment, a BHK cell develops about 50 pA of inward current when C10-TPP is applied at a concentration of 40 μM. Similar to results for TX100, the PM area decreases by nearly 70% within 10 s. During the MEND response, the current generated by the hydrophobic cation decreases very little. When C10-TPP is removed and reapplied, the peak current at the second application is decreased by only 17% in relation to the first application. This, in spite of membrane area being decreased by 70%. Therefore, it can be concluded that the affinity of C10-TPP for PM that remains at the cell surface is about 3.5-fold higher than for membrane that internalizes. In this connection, we point out that phospholipid affinities for one another and for cholesterol in ordered versus disordered PM domains likely differ only by a few fold [49-51]. Similar to this result, we found that many other hydrophobic probes were internalized poorly, indicative of more ordered membrane being internalized [33]. In summary, large fractions of the PM of many cell types can be rapidly internalized when challenged by amphipathic compounds that cause it to develop MPSs. The concentrations of detergents required to induce MEND are about two orders of magnitude less than the concentrations used to isolate so-called 'detergent insoluble membrane fractions'.

3. Actin cytoskeleton remodeling may prevent or activate MPS-dependent endocytosis.

Given the evident tendency of complex membranes to phase separate, one must ask how cells normally prevent the PM from developing MPSs to an excessive extent. As background to this issue, the vacuolar membrane of yeast, for example, is physiologically phase separated into large domains [52] , while the PM is clearly not. Multiple types of experiments suggest that the membrane actin cytoskeleton plays an important role in preventing MPSs [53]. Perhaps most persuasive, bleb membranes lacking cytoskeleton readily undergo MPSs in a temperature-dependent manner, while live cells with well-developed membrane cytoskeleton do not [54, 55]. This corresponds reasonably to our observation that delayed Ca-activated MEND is often more pronounced in cells in which actin cytoskeleton is perturbed by latrunculin [29]. This is illustrated in Fig.2 using a BHK cell that is exposed to 3 μM latrunculin via both the cytoplasmic and extracellular solutions employed. In these experiments, cytoplasmic Ca elevations are generated by constitutively expressed NCX1 Na/Ca exchangers. To do so, 40 mM Na is included in the pipette and a Na-free extracellular solution is employed with n-methyl-d-glucamine (NMDG) as the major monovalent cation. Upon applying 3 mM Ca to the extracellular side, outward membrane current reflects 3-to-1 exchange of extracellular Ca for intracellular Na. Initially PM area increases by 30%, reflecting a powerful Ca-dependent PM expansion mechanism [56] . Then, after a delay of several seconds, PM area begins to fall and continues to fall almost linearly over 4 minutes when more than one-half of the PM has been internalized. At that point, a second Ca elevation in this experiment reestablishes the original PM area, as presumably a large population of cytoplasmic membrane vesicles fuse with the PM. We stress that the expansion response likely does not reflect the same physical membrane that is internalized during MEND, since FM dye taken up during MEND (not shown) is not quantitatively released during the second expansion response. The delayed Ca activated MEND is obviously not dependent on actin remodeling that can be affected by latrunculin, although it was recently shown that all endocytic processes are eventually disrupted by suppressing actin expression [41]. Whether other forms of actin remodeling might be occurring during MEND, e.g. gelsolin-mediated F-actin cleavages [57], remains uncertain. In any case, the control of membrane tension by membrane cytoskeleton is likely to be a major regulatory factor. Remodeling of membrane cytoskeleton in discrete zones during the Ca elevation might facilitate MPSs in those same zones, in particular if cytoskeletal constrictions would reduce local membrane tension. We point out that the 'CLIC' endocytosis pathway has been described to be highly mechanosensitive, being upregulated by a sudden reduction in membrane tension [37]. Similar to that report, we also often observe that cells internalize large amounts of PM after being removed from dishes by trypsin treatment. To what extent this occurs by CLICs or MEND is an open question. Should it turn out that CLIC and MEND are in fact not different endocytic mechanisms, this will of course become a mute question.

Figure 2.

Ca-induced delayed MEND responses in BHK cells are not inhibited by a high concentration (3 μM) of the F-actin disrupting agent, latrunculin. Standard experimental conditions [29]. BHK cells express constitutively cardiac NCX1 Na/Ca exchangers, and the cytoplasmic solution contains 40 mM Na with 0.5 mM EGTA. A large cytoplasmic Ca elevation is induced by applying 3 mM Ca on the extracellular side. Ca influx by 3 Na/1 Ca exchange generates an outward membrane current, shown in blue below the PM Area record. Electrically defined PM area increases by nearly 30% during Ca influx, and delayed MEND causes internalization of 63% of the PM over 8 min. Reapplication of extracellular Ca then causes the PM to expand again to its previous peak value, presumably representing a massive exocytic response. Insensitivity of the MEND response to latrunculin presumably indicates that actin remodeling is not required for MEND.

4. More extensive and faster than dynamin/clathrin-dependent endocytosis.

That clathrin is not critical for delayed Ca-activated MEND was originally supported by findings that this form of MEND is not blocked by protein domains that block clathrin-dependent endocytosis and does not require potassium [29]. That dynamins are not involved in this form of MEND was supported by experiments with dynamin inhibitors and expression of dominant negative dynamins [29]. To test these conclusions in a different way, we recently examined delayed Ca-dependent MEND in mouse embryonic fibroblasts (MEFs) in which all three dynamin isoforms can be knocked out in an inducible fashion [40]. Experiments described employ standard experimental solutions [58] with 6 mM cytoplasmic ATP. As shown in the left panel of Fig. 3, Ca influx caused by applying ionomycin (5 μM) together with 3 mM Ca, causes the PM area of un-induced control MEF cells to expand within seconds by more than two-fold. We mention that this expansion appears to be coupled to scrambling of phospholipids between monolayers by the TMEM16F protein [56]. After Ca influx is terminated by removal of both extracellular Ca and ionomycin, PM area is relatively stable for a minute. Then, the 'delayed' Ca-dependent MEND activates and reduces PM area back to approximately the control value. Although PM area returns to baseline, our equivalent experiments using membrane tracer dyes do not support the idea that the membrane internalized is physically the same membrane enters the cell surface during membrane expansion [21]. As shown in the right panel of Fig. 3, the PM expands to only a small extent in the triple KO cells when Ca and ionomycin are applied, possibly because basal PM area is already highly expanded in the KO cells. In any case, the subsequent delayed MEND responses in the KO cells are just as large and as fast as in WT cells.

Figure 3.

Ca-induced delayed MEND responses in MEF fibrobasts are similar in 'un-induced' control cells and cells in which deletion of all three dynamin isoforms has been induced. Standard experimental conditions [29]. The left panel shows an example record together with composite results for application of ionomycin (5 μM) together with 3 mM Ca. PM area increases by nearly three-fold within a few seconds during Ca influx. After terminating Ca influx, PM area is stable for about 1 min, MEND then activates rapidly, and PM area returns to baseline within 2 further minutes. The right panel shows an example record together with composite data for MEF cells on the third day after induction. Baseline PM area is increased in relation to control cells, and the expansion of PM area that occurs with ionomycin/Ca application us less than 20%. MEND reduces PM area by >50%, very similar to the equivalent responses in control cells.

5. The lipidic basis of MPS-driven endocytosis.

The cartoon in Fig. 4 illustrates how MEND responses may occur from the standpoint of PM lipid organization [27]. It is expected in the unstimulated basal state of cell function, i.e. when cells are growing without any major signaling perturbations, that the PM is heterogenous only on a nanoscale. Ordered domains exist as collections of a few proteins that tend to attract a lipid shell [59] that can be more ordered than the surrounding lipid, wherby lipid shells generally would exchange to their surrounding lipids in microseconds (1). Events that change the lipid compositions substantially (e.g. high sphingomyelinase activities) and/or the application of amphipathic compounds (e.g. detergents) can promote ordered domains to grow or coalesce. Amphipathic compounds will in general cause cholesterol to be displaced from disordered domains and then associate with sphingomyelin in ordered domains (2). As domains grow the PM will buckle on a submicroscopic scale, somewhat like the ruffling that occurs on a much larger scale during micropinocytosis [60] (3). Line tension between domains will promote the development of membrane budding, potentially outwardly as well as inwardly. One factor that may determine whether membrane is shed as ectosomes or internalized as endocytosis may be the degree to which the outer monolayer tends to fuse to itself to promote endocytosis, versus the tendency of the inner monolayer to fuse to itself to shed membrane to the extracellular space. More disordered membrane in the outer monolayer of the closing neck of the budding membrane would naturally support sealing off of a vesicle containing on average more ordered membrane. In addition, the architecture of the membrane will be influenced by the presence of lipids on the cytoplasmic side that may be attracted to interfacial regions where bending is greatest. Diacylglycerols, for example, are known to have strong influence on the shapes of organelles, such as endoplasmic reticulum [61], and may support the formation of vesicles with small diameters. To some degree, diacylglycerols—especially unsaturated diacylglycerols—may affect membranes in a manner similar to ceramides [62]. In addition, local membrane tension will with good certainty play a major role in determining if a small vesicle can be excised or not. Finally, we stress that the interplay of membrane cytoskeleton and domain growth is a key area for future research, not only in the relation to endocytosis but in the formation and regulation of membrane protein interactions.

Figure 4.

Hypothetical lipidic basis of endocytosis driven by MPSs. Modified from published figure [27] with permission by Rockefeller University Press. 1. The outer monolayer consists of ordered (Lo) and disordered ( Ld) domains on a nanometer scale. 2. Amphipathic molecules bind to and tend to expand Ld domains, thereby promoting membrane buckling on a submicroscopic scale. 3. Line tension between domains enhances membrane buckling, and a vesicle is eventually excised inwardly in dependence on local membrane tension, the presence of fusogenic lipids in the extracellular monolayer, and the presence of inverted 'v'-shaped lipids that will tend to accumulate in areas of high curvature and promote further curvature. 4. Vesicles formed by MPS-dependent endocytosis likely traffic similarly to descriptions of trafficking that occurs after CLIC endocytosis [42].

Vesicles formed without the aid of clathrin may in general be larger than those formed with clathrin for the simple reason that more energy is required to make higher curvature. How much larger are they? From patch clamp analysis we can conclude that vesicles must be smaller than about 250 nm, which is roughly our limit to detect clearly discrete endocytic events. That is of course not a great restriction. As already noted, the endocytosis form that seems most like delayed MEND is clathrin-independent carrier (CLIC) endocytosis [21, 37, 38, 42]. The vesicular structures formed during CLIC endocytosis, very distinct from caveolar membranes, are tubular with diameters of 50 to 80 nm. They can extend for lengths of several hundred nanometers, endowing them with an area that could approach 1 μSm², equivalent to a capacitance of μ10 fF. While this is still out of range to resolve as capacitance steps in the quite large cells which we typically employ, steps of this size should be readily resolved in smaller cells. In efforts to visualize membrane changes occurring during MEND, we observed that vacuoles often form profusely within cells close to the cell surface. Vacuoles could be readily identified both with confocal microscopy [29] and via electronmicroscopic observation of HRP uptake [27]. This rapid appearance of vacuoles during MEND is very reminiscent of vacuolation of skeletal muscle that takes place during tetanus contractions and evidently involves reversible vesiculation of t-tubules [63].

The formation of vacuoles presumably reflects fusion of vesicles formed during MEND both with themselves and with endosomes present within cells. We stress two observations relevant to the fate of membrane internalized during MEND. First, using FM dyes to label the PM and monitor MEND, we have never observed that dye taken up during MEND can be quantitatively excreted back into the extracellular space, neither in a constitutive fashion nor when Ca influx is activated to stimulate fusion events. Rather, the internalized dye moves over time deeper into cells and becomes dispersed. In other words, the average vesicle integrates into the endosomal membrane network. Second, we have observed that HRP taken up via MEND can appear within a few minutes in multi-laminar structures that are presumably multivesicular bodies [27]. Thus, membrane internalized via MEND may eventually be released from cells as exosomes [64].

6. The role of proteins: Regulation via protein cargo and palmitoylation.

Knowing that detergents and certain lipases can powerfully induce MPS-dependent MEND responses, we expected to find that MEND could be triggered physiologically by lipases which would generate lipids that could effectively substitute for detergents to induce MPSs. While this may still be the case, our initial efforts were unsuccessful. Rather, we identified two key factors that relate to PM proteins themselves: First, we found that the palmitoylation state of PM proteins was critical for the occurrence of MEND [30-32], presumably by promoting the coalescence of proteins into domains. Second, we found that the expression level of some membrane proteins, most definitively NCX1 Na/Ca exchangers, strongly influences the extent of MEND that occurs. Importantly, this 'cargo' effect could be demonstrated using protocols that did not involve Ca (i.e. MEND activated by GTPγS in the presence of 10 mM cytoplasmic EGTA) [30]. The NCX1 Na/Ca exchanger has a very large cytoplasmic domain whose function remains enigmatic in many respects [65]. We have speculated in this connection that proteins with large cytoplasmic domains may promote membrane bending as they associate into domains in a manner reminiscent of dedicated adapters [66]. In the case of NCX1, the presence of a single palmitoylation can strongly influence the rate and extent of MEND that occurs [30]. Given that NCX1 has 10 transmembrane helices, it is impressive that a single palmitoylation has such a large effect. Closely related to these observations, it has recently been described that dopamine transporters (DATs) can be internalized in dependence on their clustering and oligomerization without the involvement of classical endocytic mechanisms [67], very similar to our suggestions for Na/K pumps and NCX1 Na/Ca exchangers [66].

Figure 5 illustrates experimental support for the notion that palmitoylations can promote delayed Ca-activated MEND and that the pathway is likely regulated by classical signaling mechanisms . The upper panel of Fig. 5 shows that MEND is reduced by 65% when two major acyl transferases that are active at the cell surface, DHHC5[68-70] and DHHC2[71, 72], are knocked out by CRISPR technology in TreX293 cells expressing NCX1. We point out in connection with these experiments that NCX1 palmitoylation does not affect NCX1 localization in cells or NCX1 activity at the cell surface, at least not when the protein is overexpressed under standard conditions [30]. Accordingly, outward currents generated by NCX1 activity in protocols like those of Fig. 2 were not significantly different in WT cells and cells lacking both DHHC2 and DHHC5. Evidently, the single palmitoylation of NCX1 promotes exchangers to selfassociate and thereby supports the internalization of transporters by MPS-dependent endocytosis in multiple types of cell stress.

Figure 5.

MEND depends on acyl transferases that palmitoylate PM proteins and is promoted by accumulation of PIP3. A. Dual knockout of DHHC2 and DHHC5 acyl transferases. Method. Cells were transfected via Lipofectamine 3000 with a LentiCRISPRv2 (Addgene #52961) construct containing sgRNA for either DHHC5 or DHHC2. Primers were annealed and ligated with BsmBI digested plasmid prior to transformation into Stbl3 chemically competent E. coli (Invitrogen, C737303) as described [109, 110]. DHHC5-FP, CACCGTGAATAACTGTATTGGTCG; DHHC5-RP, AAACGCGACCAATACA GTTATTCAC; DHHC2-FP, CACCGCGTAGTAGGACCAGCCGAGC; DHHC2-RP, AAACGCTCGGCTGGTCCTACTACGC. A hU6F primer (GAGGGCCTATTTCCCATGATT) was used for sequencing verification of the sgRNA/Cas9 containing LentiCRISPR V2 constructs. 48 hrs after transfection, puromycin (10 ug/ml) was added to the cells and cultured for one week prior to single cell isolation of puromycin resistant clones and subsequent western blot analysis. For simultaneous knockout of DHHC2 and DHHC5, cells were co-transfected with LentiCRISPRv2-DHHC5 and LentiCRISPRv2-DHHC2, followed by the same selection process. Cells were lysed in RipA buffer. 25 μg of total protein was loaded per lane and subjected to SDS-PAGE. Primary antibodies (1:2000 DHHC5 and DHHC2, 1:1000 for actin). The Western blot, documenting dual knockout, uses anti-ZDHHC5 antibody produced in rabbit (HPA014670, SIGMA) and anti-ZDHHC2 antibody produced in mouse, Clone 1035A ( gift from Masaki Fukata,Japan). B. MEND responses using the same protocol as in Fig. 2 amounted to a loss of 38% of PM area in WT TreX293 cells. MEND responses were reduced to 17% of PM area in equivalent cells lacking both DHHC2 and DHHC5. Na/Ca exchange current magnitudes were not significantly affected by DHHC deletion. C. MEND responses in BHK cells, grown to only 50% of confluence, amounted to only 17% of total PM area. Acute treatment with the PTEN inhibitor, boV(HOpic) (5 μM) increased MEND responses in underconfluent cells to 38% of PM area, and the PI3 kinase inhibitor wortmannin (1 μM) reduced MEND responses in the presence of bpV(HOpic) to 7% of PM area. Na/Ca exchange currents of not significantly different in the three sets of experiments.

In initial work implicating palmitoylation in the occurrence of delayed Ca-activated MEND, we proposed that large Ca elevations induce a wave of acyl CoA in the cytoplasm of cells caused by the release of CoA and acyl CoA from mitochondria [30]. However, acyl CoAs do not usually cause MEND when perfused into cells in the absence of Ca elevations. In other words, Ca elevations must promote palmitoylations by one or more mechanisms besides increasing acyl CoAs. In the meantime, we have found that MEND occurs in multiple circumstances without Ca elevations. For example, MEND occurs in cardiac tissue during episodes of anoxia/reoxygenation when oxygen is reintroduced [73]. A routine observation using BHK cells is that MEND responses are larger and faster when cells are grown to complete confluence. In testing numerous possible signaling pathways, we found that manipulations of PIP₃, and potentially also tyrosine phosphorylations, strongly influence the magnitudes and speed of MEND. As shown in Fig. 5C, the PTEN inhibitor, bpV(HOpic) (5 μM) [74], strongly enhances delayed Ca-dependent MEND in BHK cells that are 'subconfluent'. Furthermore, MEND responses are largely suppressed by the PI3 kinase inhibitor, wortmannin [75]. First, these results verify that key cell signaling pathways likely control the occurrence and function of this form of endocytosis. Second, the control of MEND by PI3 kinase signaling, if correct, would be similar to well-defined signaling pathways that regulate macropinocytosis [76-79]. In other words, the suspicion is growing that MEND and CLIC endocytosis, both powerful micropinocytosis mechanisms, may be employed in similar cell circumstances as macropinocytosis with all three mechanisms having overlapping functions.

7. MPS-driven endocytosis and macropinocytosis may be regulated similarly.

To summarize our impressions up to now, MEND and 'CLIC' would traditionally be classified as different forms of micropinocytosis. However, the usual experiments with actin reagents reveal no actin-dependence of delayed MEND (Fig. 2), whereas CLIC endocytosis [38] and macropinocytosis [80] are strongly or entirely dependent on actin cytoskeleton remodeling. Vesicles formed in MEND and CLIC endoctyosis are small compared to macropinosomes. Lipid ordering plays key roles in MEND, whereas no critical dependence on membrane ordering has been suggested for micropinocytosis (although cholesterol can be essential [81]). Nevertheless, micropinocytosis and macropinocytosis often occur in parallel and often appear to have similar or complementary functions. As might be guessed, we find that the same cell signaling pathways can regulate all three mechanisms. To illustrate this, experiments are shown in Fig. 6 using identical experimental conditions in three cell lines: TreX293 and BHK cells, which show pronounced micropincytosis, and A431 cells which show profuse macropinocytosis [81, 82]. Fig. 6 illustrates that PIP₃ and/or tyrosine kinase-mediated phosphorylation events can initiate MEND in the absence of Ca elevations. The requirements, as expected for the involvement of palmitoylations, are that CoA is present in the cytoplasm (15 μM) and that the extracellular solution contains albumin (50 μM) complexed with palmitate (150 μM). This corresponds roughly to half-saturation of albumin by fatty acid. To promote the generation of PIP₃ and tyrosine phosphorylations, the cytoplasmic solution contains the phosophatase inhibitor, pervanadate at a low concentration (6 μM). Otherwise, the experiments employ our standard experimental conditions with 0.5 mM EGTA [27]. As illustrated in Fig. 6A, we routinely photograph the cells employed to determine cell dimensions.

Figure 6.

Comparison of MEND induced by cytoplasmic CoA (15 μM) and pervanadate (6 μM) in TreX293 cells, A431 cells, and BHK cells. A. Cells were patch clamped using standard solutions [29], including CoA and pervanadate in the pipette solution, as indicated, and including albumin and palmitate in the extracellular solutions, as indicated. PM area was then monitored with no further intervention for 3 to 5 minutes at 37°C. B. TreX293 cell area decreased on average by 28% over 3 min, and this response was suppressed to 15% in cells lacking DHHC2 and DHHC5. C. In equivalent experiments, A431 cells internalize ~40% of their PM within 5 min. Loss of PM occurs with large fluctuations, apparent in the derivative signal (PM area change per 5 s) shown below the PM area record. Fluctuations of PM area are indicative of the internalization of PM as large vacuoles (>2 microns in diameter), consistent with micropinocytosis. After the MEND response, total PM amounts to twice that of a sphere with the diameter of the cell. When equivalent experiments were performed in the absence of CoA and presence of 3 μM triascin c to inhibit acyl CoA synthesis, the average PM area increased by 5% (n=5). D. In equivalent experiments, BHK cells internalize 33% of their cell surface within 2.5 min. However, MEND in BHK cells proceeds without large fluctuations, consistent with endocytosis occurring as a form of micropinocytosis, rather than macropincytosis. As illustrated in this example, PM area can decrease to nearly the area of a smooth sphere with the diameter of the cell. Omission of CoA and addition of triascin c reduce the MEND responses by 56%. Together, the results suggest that signaling mechanisms activated by low concentrations of pervandate regulate both macro- and micropinocytosis.

Fig. 6B shows composite results for TreX293 cells. Within 5 min after rupturing the surface membrane sealed into the pipette tip, thereby establishing diffusion from the cytoplasm into the tip , TreX293 cells took in 28% of their surface area, and this was reduced by one-half in cells in which the DHHC2 and DHHC5 transferases had been deleted. Fig. 6C shows results for A431 cells, a human cancer cell line routinely employed in studies of macropinocytosis [81, 82]. As shown in the recording in Fig. 6B, endocytosis begins profusely in these experimental conditions when cells are warmed from room temperature to 35°C. Capacitance records during the decline of PM Area are remarkably noisy, indicating that large endocytic events are occurring, as expected for macropinocytosis. Capacitance fluctuation, defined as capacitance change over 5 s, is shown below the record. Downward fluctuations on the magnitude of 1 to 2 pF are indicative of the excision of vacuoles with diameters of 2 to 5 microns from the PM. As indicated in bar charts in the figure, the average endocytic responses amounted to 36% of the PM over 5 min, and the endocytosis was entirely blocked when CoA was omitted and the acyl CoA synthetase inhibitor, triascin C (3 μM) [83], was included in solutions. In other words, the macropinocytotic processes activated in this protocol seem to be dependent on acyl CoA signaling, similar to results for delayed MEND. It will therefore be of great interest to determine if macropinocytosis, like delayed MEND, requires palmitoylations. Fig. 6D shows equivalent experiments for BHK cells. The endocytic responses are similar in magnitude to those in A431 cells, and they occur at a similar speed. However, capacitance fluctuations during endocytosis are negligible in BHK cells compared to those in A431 cells. In other words, the endocytosis occurring in BHK cells is clearly a form of micropinocytosis, not macropinocytosis. As shown in bar charts, the MEND responses in BHK cells were reduced by 56% when CoA was omitted from the pipette and triscin c was included in the experimental solutions. The results together raise the possibility that related signaling mechanisms activate macropinocytosis as well as micropinocytosis, including perhaps both MEND and CLIC endocytosis. It will be of great interest in this connection to determine when and if MEND responses might become actin-dependent and therefore might become equivalents of CLIC endocytosis. Certainly, membrane cytoskeleton in these experiments will be severely affected by pervanadate-dependent signaling processes [84].

8. The experimental dissection of endocytic subtypes: A self-limiting endeavor.

In two concluding sections, we point out key open questions and summarize potential physiological roles of MPS-dependent endocytosis. Delayed MEND is a form of endocytosis that might have a constitutive role, but only becomes undeniably prominent in a variety of cell stresses [32]. As summarized in Table 1, MEND has two unusual properties compared to other clathrin/dynamin-independent mechanisms. First, it has no latrunculin-sensitivity, and second it internalizes more ordered PM domains and therefore likely relies on MPSs. Undeniably, our experiments are unusual in that we typically employ cells after removing them from dishes by minimal trypsin treatment. One key factor that may be altered in such cells is the membrane tension that will resist membrane folding and invagination, as well as formation of membrane buds. As noted already, a major influence of actin remodeling may be to modify membrane tension experienced by PM-dependent processes such as endocytosis , and modification of membrane tension is in fact proposed to explain the variable influence of actin cytoskeleton on clathrin-dependent endocytosis [85]. In this light, it is a question whether the mechanisms studied by us might be actin-dependent when studied in adherent cells, and whether endocytic mechanisms that seem to be primarily actin-dependent may have strong dependence on membrane ordering in a different experimental setting. In short, we now speculate that clathrin/dynamin-independent endocytic mechanisms may form a continuum of related mechanisms that operate with varying dependence on actin remodeling and lipid ordering. One possible contradiction is that the deletion of actin apparently disrupts all forms of endocytosis, at least in synapses [41]. However, these knockdown experiments obviously involve long-term deletion of a protein that regulates many vital cell functions. In this light, it seems prudent to defer conclusions until, at minimum, it is known whether PM area and PM lipid composition are significantly altered.

Table 1.

Listing of endocytic mechanisms and their dependence on actin manipulation with latrunculin and the involvement of PM ordering into domains (MPS-dependence).

	Latrunculin- dependence	MPS- dependence	Comments
Classical clathrin-dependent endocytosis in yeast	YES	Unknown	Clathrin endocytosis in yeast is highly dependent on actin-remodeling with myosin used to anchor actin networks [89].
Classical clathrin-dependent endocytosis in animals	YES/NO	Unknown	Clathrin endocytosis in mammals can function with less direct dependence on actin [85, 111, 112]
Clathrin-independent Dynamin-dependent	YES/NO	Unknown	Clathrin appears to have evolved before dynamin [113], but dynamins can drive select endocytosis w/o clathrin (e.g. [114, 115]).
CLIC: Clathrin-independent Dynamin-independent	YES	Unknown	CLIC endocytosis is a rare example in which external PM monolayer fusion has received serious attention [116]
Neuronal Bulk endocytosis: Clathrin-independent Dynamin-independent	YES	Unknown	Neuronal bulk endocytosis likely reflects multiple mechanisms. It is usually highly latrunculin-sensitive[88], can clearly involve myosin II [90], and can internalize GM1 avidly [117].
Neuronal Ultrafast endocytosis: Clathrin-independent.	YES	Unknown	This form of fast PM retrieval at synapses requires actin-dependent PM invagination and (probably) fission by dynamins [118].
Caveolar endocytosis: Clathrin-independent/Dynamin-dependent	bidirectional	Unknown	Both actin polymerization [119] and depolymerization [22, 120] canpromote caveolar endocytosis.
Delayed MEND: Clathrin/Dynamin-independent	NO	YES	Both delayed MEND and immediate Ca-activated MEND internalize PM that is more ordered than PM that remains at the cell surface (i.e. >10 optical and electrical PM probes are poorly internalized [27, 33]).
Immediate Ca-activated MEND: Clathrin/Dynamin-independent	NO	YES
Macropinocytosis	YES	Unknown	Macropinocytosis can be cholesterol-dependent [81]. As noted in the 3rd column, it is unknown whether more ordered PM may be preferentially internalized.

Turning to the third column of Table 1, while the cholesterol dependence of endocytosis has received substantial attention, only our studies have addressed systematically whhether internalized membrane is more or less ordered than membrane that remains at the cell surface. At least, it would be interesting to know how the amphipathic membrane probes we have employed [33] would behave in other forms of endocytosis. A dependence of endocytosis on the PM cholesterol content (e.g. [86]) may indicate only that cholesterol facilitates membrane bending, and therefore tends to support clathrin-independent endocytosis. The importance this issue for endocytosis is underscored by findings that the budding of COPI-coated vesicles from the Golgi indeed seem to involve lipid sorting [87] that presumably requires MPSs of the Golgi membrane. To summarize Table 1, a large number of endocytic processes are known to operate independent of dynamin and clathrin. These processes may all depend both on membrane ordering and actin remodeling, and it will be of great importance to eventually determine in each case whether actin is acting primarily by modification of membrane tension or via additional mechanisms.

9. Physiological roles and regulation of MPS-dependent endocytosis.

Bulk endocytosis in neurons and glial cells.

The most prevalent form of clathrin-indepenent endocytosis in neurons is called 'bulk endocytosis' [88]. Bulk endocytosis occurs during high frequency and progressively higher intensity stimulation of synapse function. Involvements of clathrin and dynamins have been largely eliminated at this time [40]. Bulk endocytosis is consistently found to be latrunculin-sensitive , as indicated in the Table, and more recently it has been found to be dependent on myosins [89, 90]. Nevertheless, this clear actin dependence does not contradict a possibility that bulk endocytosis may also depend on PM ordering and may be regulated by processes that regulate PM ordering, as described in Figs. 5 and 6 for delayed MEND. Furthermore, it is evident that multiple types of bulk endocytosis may exist. In the rat calyx of Held, for example, a bulk endocytosis mechanism is described to operate with no requirement for ATP but with a clear involvement of sphingomyelinases [91]. Calcium-activated endocytosis that recycles membrane involved in transmitter release from glial cells is both dynamin- and clathrin-independent and has characteristics that seem similar to MEND [92].

The possible involvement of Ca-activated lipases and acyl transferases.

In addition to delayed MEND, we have observed in many cell types the existence of rapid Ca-activated MEND that is unaffected by highl concentrations of latrunculin and that depends on the immediate presence of cytoplasmic Ca [29, 56]. Fig. 7A shows the typical Ca-activated MEND response of a murine cardiac myocyte. The myocyte is patch clamped with standard solutions and Cm is stable for many minutes, as are Na/K pump currents activated by applying extracellular K (7 mM) for a few seconds. When extracellular Ca is applied (3 mM), an outward Na/Ca exchange current is immediately activated, corresponding to Ca influx, and within 3 s Cm starts to fall precipitously, as PM is internalized. The MEND responses can be rapidly terminated by terminating Ca influx if the myocytes do not beat spontaneously (i.e. via Ca-induced Ca release) [29]. In this case, 30% of the PM is internalized in 15 s.

Figure 7.

Rapid Ca-activated MEND in cardiac myocytes that depends on the immediate presence of Ca in the cytoplasm. A. Murine myocyte patch clamped with a low Ca buffer concentration (0.5 mM EGTA) under standard conditions [29] . After stable recording of Na/K pump currents for several minutes, Ca influx via reverse Na/Ca exchange causes internalization of large fractions of the PM within seconds accompanied by no change of cardiac length. B. Human embryonic stem cell (hESC)-derived myocytes are entirely resistant to the induction of MEND in the same protocol, when they have been cultured under standard conditions for several weeks. C. Brief enrichment of hESC-derived myocytes with cholesterol (15 min with 10 mM hydroxypropylcyclodextrin + 0.8 mM cholesterol) enables robust and rapid Ca-activated MEND responses.

While latrunculin has no effect on this type of MEND, whether applied for short or long (1h) times, other experiments are consistent with involvement of membrane ordering. This is illustrated in Figs. 7B and 7C, employing human embryonic stem cell (hESC)-derived cardiac myocytes that were prepared and maintained by established protocols [93]. These myocytes display Na transport and ion channel currents that are comparable in our experience to myocytes isolated from non-rodent animals. As shown in Fig. 7B, however, these myocytes do not undergo MEND in our routine conditions when the outward Na/Ca exchange current is activated by extracellular Ca (>100 observations). Knowing that fast Ca-activated MEND is highly sensitive to the cholesterol content of the PM [29], we enriched the PM with cholesterol by preincubating hESC myocytes with hydroxypropylcyclodextran/cholesterol complexes (10 mM/0.8 mM) for 15 min, as in experiments with other cell types [29]. As shown in Fig. 7C, Ca influx by reverse Na/Ca exchange then activated MEND responses within 2 s that resulted in nearly 50% loss of the PM over 30 s. In our experience this type of Ca-activated endocytosis, requiring the immediate presence of Ca, can occur in a very wide range of cell types, including BHK cells, fibroblasts, Jurkat cells, and myocytes. The Ca sensors that mediate this type of endocytosis are still unknown. They could be membrane proteins that modify the lateral ordering of the PM, they could be lipases that generate lipidic second messengers that promote MPSs, or they could be acyl transferases that act on PM proteins and/or lipids.

Potential pathological importance of palmitoylation-dependent MEND.

Given that fatty acids and CoA are key metabolites in the function of delayed MEND, it is an valid question whether delayed MEND is regulated by the concentrations of these metabolites. It was suggested decades ago that acyl CoA concentrations within cardiac myocytes were sensitive to the prevailing CoA concentration, acyl CoA synthesis being limited by the CoA concentration [94]. Similarly, the generation of acyl CoA may in some circumstances be limited by fatty acid availability. At the next signaling step, it appears likely that excessive palmitoylations mediate at least a component of palmitate toxicity [95], including a role in palmitate-induced insulin resistance [96].

Cell migration.

Surprisingly, the precise roles and function of membrane trafficking in cell migration remain rather enigmatic. If membrane turnover is actively contributing to cell movement, endocytosis would be expected to remove PM from the trailing edge, while exocytosis would be expected to extrude membrane at the leading edge. That PM turnover could in fact contribute to cell migration was impressively verified by a study using FM dyes in dictyostelium in 1999. Specifically, it was demonstrated that the average PM in dictyostelium internalizes every 4 to 10 min, and it was determined that this turnover was only mildly reduced by clathrin deletion [97]. Thus, clathrin-independent pinocytosis must play an major role. A recent study by Tanaka and colleagues in dictyostelium appears to confirm the expected over all membrane flow. Specifically, these authors support a 'fountain-like' model in which membrane on the top, bottom, and sides of the cell would move backwards, relative to the front of the cell, as the cell moves forward [98]. Notably, both macropinocytosis [99] and CLIC endocytosis [42] are taking place primarily at the front end of migrating cells. If Tanaka et al. are correct, other forms of clathrin-independent endocytosis would presumably be taking place at the trailing edge.

Uptake of biologically important particles.

A discussion of virus and particle uptake by cells would vastly exceed the scope of this article. It is well established that pathogens exploit many different entry mechanisms into cells, and biophysical details of how HIV entry is mediated by ordered membrane domains are being rapidly elucidated [100, 101]. It will suffice here to point out two other examples of particle uptake that are of great medical importance and that might involve endocytic processes related to delayed MEND. First, the endocytic mechanisms by which macrophages take up LDL, HDL and/or cholesterol esters per se, leading eventually to foam cell formation, remain a topic of considerable debate [102]. One proposal is that multiple forms of pinocytosis play crucial roles [103]. It is at least noteworthy that the triascin C inhibitor of acyl CoA synthesis, employed in Fig. 6, completely blocks foam cell formation [104]. Besides inhibiting lipid synthesis, depletion of acyl CoAs will inhibit palmitoylations that seem to support some forms of micropinocytosis. Second, the uptake of prions, amyloid peptides and tau protein by neurons are of great neuropathological importance. While prions and tau proteins appear to be taken up by many different endocytic routes [105], the amyloid-beta peptide appears to associate to ordered membrane domains with some specificity and appears to be taken up by mechanisms that are dependent on MPSs [106, 107].

In conclusion, the idea that endocytosis proceeds primarily by clathrin caging of membrane buds followed by dynamin pinching of their necks is comfortable and firmly established. The idea that endocytosis can occur via MPSs is uncomfortable and continues to be controversial. Nevertheless, the number of important examples where MPSs are likely to play important roles, both physiologically and pathologically, continue to grow. In short, it is undeniable that membrane budding and fission can occur by primary lipidic mechanisms and can contribute to the turnover of the surface membranes of cells. MPSs likely become important not only at the PM but during the turnover of complex internal membrane systems, such as the multivesicular body [108].

Highlights.

• Endocytosis can occur without the involvement of dynamins, clathrin, or actin remodeling.

• Clathrin/dynamin-independent endocytosis may occur in different forms with different dependencies on actin remodeling and membrane ordering into domains.

• Clathrin/dynamin-independent endocytosis can be regulated by the same mechanisms that regulate macropinocytosis.

• Clathrin/dynamin-independent endocytosis can be triggered by many forms of cell stress, including calcium signaling and metabolic stresses.