Multiple roles for actin in secretory and endocytic pathways

Abstract

Actin filaments play multiple roles in the secretory pathway and endosome dynamics in mammals, including maintenance of Golgi structure, release of membrane cargo from the trans-Golgi network (TGN), endocytosis, and endosomal sorting dynamics. In addition, both TGN carrier transport and endocytosis occur by multiple mechanisms in mammals. Actin likely plays a role in at least four mammalian endocytic pathways (clathrin-mediated, CLIC/GEEC, FEME, and IL2R endocytosis), five pathways for membrane release from the TGN (clathrin-mediated, Rab6-mediated, GOLPH3-mediated, sphingomyelin-enriched, Arf1-mediated), and three processes on endosomes (regulated PM recycling, transport to the trans-Golgi network, and transition from early to late endosomes). Finally, the mammalian Golgi structure is highly dynamic, and actin likely plays roles in these dynamics. One challenge for many of these processes is the need to deal with other membrane-associated structures, such as the cortical actin network on the PM or the matrix that surrounds the Golgi. Arp2/3 complex is a major actin assembly factor in most of the processes discussed, but roles for formins and tandem WH2 motif-containing assembly factors are being elucidated and are anticipated to expand with further study. The specific role for actin is not defined for most of these processes, but is likely to be in generating force for membrane dynamics, either by actin polymerization itself or by myosin motor activity. Defining these processes mechanistically is necessary for understanding membrane dynamics in general, as well as pathways utilizing these processes, such as autophagy.

Introduction

At first glance, the actin cytoskeleton of a typical cultured mammalian cell appears a rather unorganized jumble. Upon closer inspection, one discerns several large structures: stress fibers running long distances along the ventral surface, a lamellipodium in a tight band at the leading edge, and perhaps some finger-like filopodia jutting out of the plasma membrane (PM). Beneath these seemingly dominant structures, however, exists a myriad of smaller, more transient pools of actin filaments, assembling and disassembling in seconds to fulfill specific roles (Figure 1). These roles often involve organelle dynamics, with which we concern ourselves here.

Figure 1: Actin around organelles.

A) HeLa cells transfected with mCherry-Rab5 (EE, early endosomes), then fixed and stained with anti-GM130 (cis-Golgi) and rhodamine-phalloidin (actin filaments). Image is a maximum-intensity projection of 10 × 0.4 μm z-slices by spinning disk confocal microscopy. Zooms to right represent actin puncta associated with EE (top) and with Golgi (bottom). Bars, 10 μm (left) and 2 μm (zooms). B) Hypothetical migrating cell with lamellipodium rich in Arp2/3 complex-branched actin filaments at left. Cortical actin (50–200 nm thick) is on the cytoplasmic PM face around the rest of the cell. Four actin-dependent endocytic pathways shown: clathrin-mediated endocytosis (CME), clathrin-independent endocytosis/GEEC (CLIC/GEEC), fast endophilin-mediated endocytosis (FEME) and interleukin 2 receptor (IL2R) endocytosis. Three transport pathways from the early endosome (EE): 1) bulk PM recycling (actin-independent), 2) regulated PM recycling (actin-dependent), and TGN trafficking (actin-dependent). ER is shown contacting both the regulated and TGN carriers, and is likely involved in their scission. Actin might also be involved in EE-derived carrier transport, either driving motility by actin polymerization or myosin activity. Five actin-dependent transport pathways from TGN: clathrin-coated vesicles (CCV), Rab6-mediated transport, GOLPH3-mediated transport, sphingomyelin-rich carriers (SMRC), and Arf1-mediated transport. In addition, actin might be involved in dynamic membrane inter-connections between Golgi cisternae and mini-stacks. Also shown are actin acting during mitochondrial fission, although these are not discussed in detail here. While we do not depict it here, actin plays multiple roles in autophagosome assembly.

In many cases, the role of actin filaments is to produce force, and it can do so in two ways (Figure 2). Actin polymerization itself can be used to push objects forward, with examples being lamellipodial and filopodial protrusion, as well as the movement of intracellular pathogenic microbes in the cytoplasm¹. Alternately, actin filaments can serve as tracks for myosin-based force production, which comes in several flavors. Bi-polar myosin II filaments produce contractile forces on actin filaments having opposite polarities². Alternately, monomeric or dimeric myosins, such as members of the myosin 5, 6, and 19 classes, can translocate cargo such as organelles along actin filaments³. Another variation is exemplified by class 1 myosins, who can sense and/or create tension on membranes without translocation⁴. Actin-based and myosin-based forces can work together, such as in lamellipodia where actin polymerization provides protrusive force at the front of the structure, while myosin II at the back provides contractile force for retrograde flow¹.

Figure 2: Actin polymerization and actin-based force generation.

Gray arrows show direction of actin- or myosin-based force. A) Three mechanisms of initiating actin filaments: Arp2/3 complex-mediated branched actin nucleation (Arp2/3 complex is green balls at branch points); formin-mediated nucleation (blue donut), and tandem WH2 motif protein-mediated nucleation (green semi-circle). Arp2/3 complex and tandem WH2 motif proteins stay at the filament pointed end after nucleation, whereas formins stay at the barbed end. Barbed end elongation toward membranes provides a pushing force against the membrane. B) myosin II-mediated force generation. Bi-polar myosin II filaments use their motor heads (green ovals) to move toward the barbed ends of oppositely-oriented actin filaments, producing a contractile force. C) Some myosins translocate cargo along actin filaments, such as myosin V (toward barbed end) and myosin VI (toward pointed end). D) Some myosins (such as myosin 1 family members) might not induce movement, but rather resist and/or sense an opposite force (blue arrow).

In order to assemble, disassemble, and convert its dynamics into productive force, actin and myosin interact with many additional proteins⁵. Perhaps chief among these are the ,polymerization factors’ that allow initial filament assembly⁶. Three protein families accelerate nucleation of new actin filaments: Arp2/3 complex (Arp2/3), formins and WH2 motif-containing proteins (Figure 2). The seven-protein Arp2/3 nucleates dendritic networks of branched filaments, such as in lamellipodia. Initially, only one ,type’ of Arp2/3 complex was thought to exist, but heterogeneity in several mammalian subunits is revealing significant differences⁷. There are 15 mammalian formin genes, providing a large bredth of possible function⁸. Major mammalian WH2 motif-containing proteins include Spire1, Spire2, COBL and leiomodin⁹. These families can work together in actin assembly in certain situations^10–12.

A major question in this review concerns the polymerization factors used in specific types of organelle dynamics, with Arp2/3 complex being a dominant character. Arp2/3 complex nucleates new filaments from the sides of existing filaments (,mother’ filaments), with Arp2/3 complex itself defining the 70° branch between the two¹. Arp2/3 complex generally requires two activators: 1) mother filament binding, and 2) nucleation promoting factors (NPFs). There are seven NPF families: WASP/N-WASP, WAVE, WASH, WISH, WHAMM, JMY, and DIP/WISH. In addition, cortactin is a ,weak’ NPF alone, but can synergize with other NPFs, and also prolongs Arp2/3 complex-mediated dendritic branches¹³. Each NPF is subject to its own regulation, with key regulators for this review being Cdc42 for WASP/N-WASP, Rac for WAVE, and tyrosine phosphorylation or phosphatidylinositol-3-phosphate for WASH complex. Specific formins can be activated by Rho GTPases and other mechanisms^{1, 6}.

This review discusses the many ways in which actin participates in membrane trafficking, mostly focused on mammals but also referring to other organisms at times. Unfortunately, due to the vast nature of this topic we have had to focus on certain aspects, namely: endocytosis, dynamics of endocytic compartments, and exit from the trans-Golgi network (TGN). We also touch on exciting ways in which actin might contribute to Golgi ultra-structure. Actin’s roles are diverse in membrane dynamics, the current knowledge base varies widely between specific processes. In some cases, such as clathrin-mediated endocytosis, a large amount is known. In others, very little is known. It is our view that understanding this complexity will be achieved by two seemingly contradictory strategies: 1) by not assuming there is a unified mechanism by which actin participates in membrane dynamics, and 2) by looking for similarities between mechanisms where appropriate. We do not cover the extensive new work on actin’s roles in mitochondrial function, and refer to recent reviews^{14, 15}. We also do not cover actin’s roles in autophagy, which have been well-covered recently^{16, 17}. Finally, we do not cover actin’s roles in exocytosis or regulated secretion, which have been covered recently^{18, 19}.

Actin and endocytosis

Endocytosis is critical in regulating antigen presentation, signal transduction through PM receptors, the levels of cell surface proteins, and up-take of substances such as iron²⁰. Appropriate endocytosis and endosomal sorting is vital for correct degradation-versus-recycling decision making for proteins like EGF receptor and CFTR^{21, 22}, as well as for endosomal GPCR signaling²³.

To accommodate the functional diversity of cell surface proteins, there are multiple endocytic pathways that generate endosomes of <500 nm diameter, with clathrin-mediated endocytosis (CME) being the best characterized²⁰. Forces are required at two steps of endocytosis: invagination of the PM, and scission of the invaginated membrane from the PM. Coat proteins, BAR domain proteins, and dynamin GTPases often are involved at these steps²⁰. In addition, dynamic actin filaments are involved in at least four of these endocytic pathways²⁴. We summarize key factors for these pathways in Figure 3 and Table 1. In all cases, Arp2/3 complex is the important actin nucleator. We do not discuss caveolae because, while initially thought to be endocytic structures, their currently understood function is in tension sensing²⁵. We also do not discuss internalization of larger volumes through micropinocytosis or phagocytosis, and direct readers to recent reviews^{26, 27}. Before discussing endocytic mechanisms, we must describe an important but often overlooked structure that lies beneath the PM, the cortical actin network.

Figure 3: Actin in multiple modes of endocytosis.

From left to write are depicted clathrin-mediated endocytosis (CME), CLIC/GEEC endocytosis, fast endophilin-mediated endocytosis (FEME) and IL2 receptor endocytosis (IL2RE). The cortical actin meshwork is in gray, while actin filaments directly involved in endocytosis are in red. Two myosin II minifilaments, components of cortical actin, are shown at their approximate length (300 nm) to emphasize that these are longer than the widths of any of the endocytic pits (100–200 nm). For all mammalian endocytic mechanisms, the cortical actin meshwork must be traversed, but mechanisms for dealing with cortical actin have not been addressed for any system. For CME, two models for actin filament organization are shown: 1) dendritic network extending from cortical actin filaments and barbed ends abutting the planar plasma membrane, neck and nascent vesicle⁴⁶; and 2) dendritic network extending from the nascent vesicle (tethered by Hip1r) and barbed ends abutting the planar plasma membrane⁴⁵. For CLIC/GEEC, FEME and IL2RE, not enough data are available to postulate the orientations or positions of actin filaments although all are Arp2/3 complex-dependent. Positions of the BAR proteins involved in CLIC/GEEC (PICK1 and IRSp53) and FEME (endophilin) are speculative. IL2RE has the additional feature of taking place at the base of plasma membrane protrusions, which are likely actin-dependent but it is not clear to us whether these are filopodia or ruffles.

Table 1:

proteins involved in endocytic mechanisms and cortical actin.

	CME	CLIC/GEEC	FEME	IL2R	Cortical actin
Key molecules	Clathrin, Dynamin, Endophilin*	PICK1, IRSp53	Endophilin, Lamellipodin, Dynamin	Dynamin	NR
Actin polymerizers	Arp2/3	Arp2/3	Arp2/3	Arp2/3	Arp2/3, mDia1
NPF	N-WASP, cortactin	N-WASP	N-WASP, WAVE (?)	N-WASP, WAVE, Cortactin	WAVE, Spin90
Other ABPs	Hip1p, ABP1, coronin, cofilin	IRSp53	Ena	NR	α-actinin, plastins, fascin, filamin, ERM proteins, IQGAP
Myosins	Myosin 1E, Myosin VI, Myosin II (?)	NR	NR	NR	Myosin II

^*facilitates but is not required

NR = not reported

The cortical actin network

This network is a 50–200 nm thick ‘mat’ of actin filaments, crosslinked by several proteins (such as α-actinin, plastins, fascin, and filamin A) and held under tension by myosin II filaments. The mesh size of the network is variable, with a loose average of 100 nm in interphase cells. Cortical actin has two main purposes: 1) to resist the outward osmotic pressure exerted by the high cytoplasmic solute concentration, with defects resulting in membrane ‘blebs’; and 2) to provide cortical tension used in processes such as cell migration and cell division. Excellent reviews of cortical actin have recently appeared^{28, 29}.

Several aspects of cortical actin must be considered when discussing any dynamics at the mammalian PM. First, cortical actin constitutes a major difference between mammals and yeast, which use an external cell wall to resist their significantly higher osmotic pressure³⁰. Therefore, mammalian endocytosis is likely to vary from the yeast process in some respects. Second, some actin binding proteins important for endocytosis may also be important for cortical actin. In particular, Arp2/3 complex is an important nucleator of cortical actin, working with the formin mDia1 and the NPF Spin90³¹. Third, it is unclear how cortical actin varies between regions of the PM. For example, in cultured cell model systems, there may be differences in cortical actin composition and dynamics on the apical versus basal surface. Fourth, it is unclear how cortical actin interfaces with other actin-based structures at the PM, such as lamellipodia and filopodia, although cortical actin does interact with stress fibers³². It is possible that endocytosis at one ‘type’ of PM (for example, in a non-attached cell like a lymphocyte) might be different to another type (eg. on the basal surface of an adherent cell). These points are largely ignored in our subsequent discussion, but must eventually be addressed for full understanding of any mammalian endocytic process. An excellent discussion of cortical actin remodeling during exocytosis is available¹⁸.

Clathrin mediated endocytosis (CME)

CME is the process by which PM is internalized into a vesicle containing a clathrin coat on its cytoplasmic surface. In mammals, binding of the AP2 adaptor complex to the endocytic site is an early step, with clathrin recruitment following. About 50 different accessory proteins are involved, most being conserved from yeast to mammals³³.

CME is extremely well-studied in both budding and fission yeast, including a complete parts list of proteins required³⁴, allowing detailed modeling of force production^{35, 36}. Arp2/3 complex-mediated actin polymerization is an important driver of membrane invagination, with several NPFs being involved³³. After coat assembly on the PM, polymerization of the dendritic actin network around the coat drives membrane invagination. A lattice of the clathrin adaptors Sla2p and Ent1 mediates actin network attachment to the invaginating membrane³⁷. Myosin 1 motors on the PM confine and anchor the network to focus the invagination force^{38, 39}. As the endocytic pit elongates, a BAR-domain protein (Rvs167p) is recruited to the bud neck. At a relatively constant invagination length, membrane scission occurs.

CME in mammals is understood in much less detail. There have also been conflicting results on the necessity of actin for mammalian CME, but this has been resolved with the finding that actin increases in importance with increasing membrane tension^{40, 41}. Another variable might be the extent and nature of the cortical actin network in specific cellular situations.

Actin is polymerized during mammalian CME by N-WASP-activated Arp2/3 complex^42–44, with approximately 200 Arp2/3 complex molecules localizing to the internalization site⁴⁵. Platinum replica electron microscopy shows a dense dendritic network of short (50–100 nm) actin filaments surrounding the endocytic pit at an early stage, and between the endocytic pit and the PM at a later stage⁴⁶. Other actin binding proteins recruited to the endocytic site include Hip1R (a Sla2p homologue), cortactin, Abp1, cofilin and coronin 1B⁴⁶. The latter two proteins arrive late in the process, and likely mediate dendritic network disassembly. In addition to a role in force generation, dynamic actin filaments are important for dynamin2 recruitment to the endocytic site⁴⁷.

A large number of questions remain concerning actin’s role in mammalian CME. First, in which direction does the dendritic network exert force? In budding yeast, a variety of evidence suggests that barbed ends are oriented towards the planar PM, with pointed ends interacting with the coat lattice³³. Fission yeast have a second population of Arp2/3 complex-generated filaments at the tip of the invagination, that also contribute to invagination force³⁴. Platinum replica EM evidence in mammals suggests that barbed ends orient toward the neck of the pit, with pointed ends anchored as branches from other cellular filaments⁴⁶. Super-resolution microscopy studies suggest a more variable orientation of the network, dependent on membrane tension⁴¹, and EM tomography shows longer filaments that are bent, as if under tension⁴⁵. Given that force is exerted at the barbed ends of a polymerizing dendritic network¹, its orientation dictates the direction of force, and is thus an important parameter to establish.

A second question concerns roles for myosins in CME. Deletion or suppression of non-muscle myosin IIA and B inhibit CME, and coated pits reside adjacent to ventral stress fibers⁴⁸. Both myosin 1E and myosin VI are recruited to CME sites⁴⁶. Myosin 1E might be required for N-WASP recruitment⁴⁹, and its suppression inhibits transferrin endocytosis and prolongs clathrin and dynamin lifetimes on the PM^{49, 50}. Myosin VI interacts with clathrin in a competitive manner with Hip1R⁵¹. A number of mechanistic questions remain. A) Is the myosin II role actually on cortical actin or stress fibers? B) Could myosin 1E be important for dendritic network assembly, as in yeasts? C) Does myosin VI act after Hip1R?

Clathrin-independent endocytosis (CLIC)

Compared to CME, the mechanisms driving the actin-dependent steps in clathrin-independent carrier (CLIC) endocyic pathways are not elucidated in detail, for several reasons. Most or all of these pathways are not found in yeast. Also, some of these pathways are extremely rapid. We cover CLIC/GEEC endocytosis, FEME, and IL2R endocytosis.

CLIC/GEEC endocytosis is a clathrin-independent, dynamin-independent pathway for internalization of tubular membranes that are enriched in GPI-anchored proteins (GEEC, GPI-anchored protein enriched endosomal compartment). CLIC/GEEC can have impressive throughput, turning over PM contents in 12 min in some cell types^{52, 53}. This pathway can also account for a major portion of fluid intake but is not micropinocytosis. Of note, CLIC/GEEC appears absent from some cell types, such as HeLa²⁰, and much of the mechanistic work has been conducted in mouse fibroblasts.

CLIC/GEEC is initiated by Arf1 accumulation and activation at the PM, followed by recruitment of Cdc42 and two BAR domain proteins, PICK1 and IRSp53^{20, 54}. Actin polymerization, Arp2/3 complex, and N-WASP are required for CLIC/GEEC^{54, 55}. N-WASP and Arp2/3 complex are recruited relatively early in the process, but actin filaments only appear towards the end⁵⁴, suggesting that actin polymerization contributes to late internalization stages or scission. PICK1 might contribute to early Arp2/3 complex inhibition⁵⁴. It is not clear where these actin filaments polymerize at the endocytic site, or what their specific function is.

FEME (fast endophilin-mediated endocytosis) is an inducible endocytic pathway triggered by membrane receptor engagement⁵⁶. Many receptors can also use other endocytic pathways, with FEME used perhaps when rapid clearance is necessary. A key protein at all steps is endophilin. Prior to FEME activation, endophilin rapidly cycles on and off the PM, through interaction with lamellipodin at PI-3,4,5-P₃ enriched regions. Ligand-bound receptor interaction with endophilin initiates the pathway, resulting in a tubulovesicular carrier that remains endophilin-bound after scission. Dynamin and endophilin are required for scission, after which carriers move along microtubules by dynein. The process is rapid upon cargo binding (< 10 sec). Actin and Arp2/3 complex are necessary for FEME⁵⁷. However, details on the stage at which Arp2/3 complex is recruited, and its activation pathway, are not known. Both Cdc42 and Rac1 are also required, suggesting WASP/N-WASP and WAVE family NPFs. In addition, endophilin can interact with N-WASP⁵⁸. Lamellipodin interacts with Ena/VASP proteins, which control actin filament elongation in other structures^{59, 60}. Indeed, the Ena/VASP protein Mena appears to act in FEME⁶¹. It is not clear where these actin filaments polymerize at the endocytic site, or what their specific function is.

Interleukin-2 receptor (IL2R) endocytosis is an intriguing clathrin-independent endocytic sub-type concerning the constitutive endocytosis of IL2 receptors containing the beta or gamma chains, which are expressed in a sub-set of hematopoietic cells and some epithelial cells. IL2R endocytosis occurs at about half the rate of CME and does not require IL2 binding^{62, 63}. The resulting carriers are small (95 nm diameter) and largely spherical⁶⁴.

While dynamin 2 is necessary^{63, 65}, there is no current evidence for BAR domain protein, Rab or Arf GTPase involvement. Endocytosis is triggered by PI-3 kinase interaction with IL2R, stimulating recruitment of the Rho-GEF Vav2⁶⁶ which can act on both Rac1 and Cdc42⁶⁷. Arp2/3 complex activity is required for IL2R endocytosis, and both WAVE complex and N-WASP appear to be required NPFs^{64, 68}. IL2 receptor might bind WAVE complex directly⁶⁴. Cortactin also plays a role in IL2R endocytosis⁶⁵.

Two intriguing features about IL2R endocytosis are: 1) it appears to take place at the base of cell surface protrusions, akin to filopodia or ruffles; and 2) there are two waves of Arp2/3 complex activity, one that is WAVE-activated early in the process and the other is N-WASP-activated late⁶⁴. Clearly, there are many questions here. What are the protrusions (filopodia or ruffles), and are they present before or formed during endocytosis? What are the roles of the two distinct Arp2/3 complex-requiring steps? As for all endocytic processes, how is cortical actin dealt with?

We also point out that this process might bear resemblance to retroviral up-take, in which internalization occurs at the base of filopodia after viral transport down the filopodium to the cell body^{69, 70}. Many viruses likely hijack existing endocytosis pathways⁷¹.

Actin and Endosome Dynamics

Current thinking is that, regardless of the specific internalization mechanism used, all endocytosed membranes enter into the same downstream pathway⁷². A series of membrane fusion reactions coincide with sorting processes to form the early endosome (EE, also called ‘sorting endosomes’) rapidly after endocytosis^73–77. Production of phosphatidylinositol-3-phosphate (PI3P) by the Rab5-recruited PI-3 kinase Vps34 is an important recruitment signal for EE proteins. Major players in fusion include Rab5, the Rab5 effector EEA1, and SNARES. Major players in sorting include the large family of sorting nexins (SNX). The EE is morphologically heterogeneous, with multiple ‘tubular’ domains extending off ‘vacuolar’ domains. Luminal acidification through vacuolar ATPase (V-ATPase) acidifies the vacuolar domain to ~pH 6. The vacuolar domain also starts the process of internalizing portions of its membrane (termed the ‘limiting membrane’) to make intra-luminal vesicles (ILVs), enriched in ubiquitinated cell surface receptors destined for degradation. A key player in ILV formation is ESCRT complex.

Transition of EEs to late endosomes (LEs) is characterized by: loss of Rab5, gain of Rab7, loss of most tubular domains, change in membrane fusion characteristics, increased acidification (to ~pH 5), increased luminal calcium, conversion of some PI3P to PI-3,5-P2, increased degradative enzyme import, increased ILV formation, and microtubule-based movement from the cell periphery to a peri-nuclear region. LAMP1 and LAMP2 form a protective glycocalyx on the limiting membrane to inhibit its degradation. LEs can also fuse with autophagosomes. The transition from LE to lysosome is blurry, but lysosomes are more acidic (pH 4.5), more spherical (no tubular domains) and denser. Escape routes from LEs include: release of tubular domains, release of ILVs at the PM (exosomes), transport of specific molecules (such as cholesterol) to ER through contact sites, and possibly ILV fusion with the limiting membrane for luminal content release into the cytoplasm. Microtubule motors play a major role in membrane transport at all endosomal stages⁷⁸. ER contact also is important in endosome dynamics^79–81.

Transport from the EE comes in three flavors: 1) rapid ‘bulk’ recycling back to the PM of proteins like transferrin receptor (TfR), 2) slower ‘regulated’ recycling back to the PM of proteins such as the beta2 adrenergic receptor (b2AR), and 3) transit of proteins to the TGN such as cation-independent mannose-6-phosphate receptor (CI-MPR)^{75, 76}. Proteins that stay in the vacuolar domain are lysosomally degraded. The mechanism for bulk recycling is unclear, but regulated recycling and TGN transit seem to be related, with key proteins being the heterotrimeric retromer or retriever complexes⁷⁷. Regulated recycling passes through an intermediate compartment, the recycling endosome (RE), which is Rab11-positive and near the MTOC⁷⁵. EEs themselves might also undergo dynein-mediated transport to the MTOC and mature into REs without the need of tubular carriers⁷⁵. Scission mechanisms for all of these pathways are poorly characterized but might involve dynamins or EHD proteins⁸².

There are three separate roles for actin in endosomal dynamics. Actin plays a major role in assembly and/or sorting for regulated tubular carriers. In addition, actin assembly on EEs might act in EE-to-LE conversion. Finally, actin-based motility might drive transport of endocytic carriers after fission from EEs/LEs. We summarize key players in these pathways in Figure 4 and Table 2.

Figure 4: Actin in endosomal dynamics.

Schematic of early endosome showing cargo enrichment in a tubular domains prior to their scission from the vacuolar domain to produced regulated carriers or TGN-directed carriers. An Arp2/3 complex-dependent actin network accumulates at the base of the tubular domain, activated by WASH complex in two ways: 1) by un-capping of the actin-like filament in the dynactin complex through the CPI domain of the Fam21 subunit, which creates the necessary ‘mother filament’ for Arp2/3 complex binding and activation; and 2) by direct binding of Arp2/3 complex through the WCA domain of the WASH1 subunit, similar to other NPFs⁹³. Subsequent endoplasmic reticulum interaction with the tubular domain occurs just prior to scission, which might negatively regulate WASH complex. A bulk recycling is also shown, with no cargo enrichment and no actin filaments at the base. A vesicle is shown near that tubule, which is engaging in actin-based motility after scission, but it is not clear whether this motility is employed by preferentially by bulk carriers, regulated carriers, TGN-directed carriers, or all three.

Table 2:

proteins involved in endosome-associated processes.

	Regulated Recycling	EE-to-LE transition	Motility (actin-based)	Motility (myosin V-based)	Motility (myosin VI-based)
Key molecules	Retromer or Retriever, Sorting nexins, EHD proteins, dynamins	Annexin A2, HOPS complex	Protein kinase C	Rab11	APPL
Actin polymerizers	Arp2/3	NR	Arp2/3	Spir, FMN formins	*
NPF	WASH	Cortactin	N-WASP	NR	*
Other ABPs	NHERF/ABP50	Moesin	NR	NR	*
myosins	NR	NR	NR	Myosin V	Myosin VI

NR = not reported

^*Myosin VI translocates along cortical actin filaments.

Actin and tubular domain dynamics

The two recycling pathways (bulk flow and regulated) have very different characteristics. Bulk flow tubules are transient (<30 sec lifetime) while regulated tubules are longer lived⁸³. This longer duration is crucial for enrichment of G protein-coupled receptors (GPCRs) like b2AR, whose diffusion rate is 5-fold slower than for the bulk flow-recycled TfR. In addition, b2AR is enriched in regulated tubules over bulk endosome, suggesting mechanisms to concentrate it there⁸³. Microtubule-based motor activity is also an important player in tubulation dynamics⁸⁴.

Arp2/3 complex activation by the five-protein WASH complex, containing the NPF WASH1, is required for correct sorting to the TGN⁸⁵ and regulated recycling to the PM^{83, 86}. Loss of WASH1 also causes collapse of the endo/lysosomal network, suggesting a fundamental role in maintaining the network⁸⁷. Increased PI3P levels lead to WASH complex recruitment through several mechanisms⁸⁸, which might be further activated by tyrosine phosphorylation⁸⁹. Cortactin is also important^{83, 90}, and might cooperate with WASH1 in Arp2/3 complex activation. WASH1, cortactin, Arp2/3 complex and actin filaments enrich in foci at the base of tubular membranes^{85, 86, 91} that are enriched in b2AR⁸³, and actin filaments turnover in seconds in these foci⁸³. Key to this enrichment is the Fam21 subunit of the WASH complex, that contains multiple motifs for interaction with SNXs and retromer/retriever complexes^{85, 92}. A fascinating recent work suggests that WASH1 activates Arp2/3 complex in a unique manner with two components: 1) direct Arp2/3 complex binding in a manner similar to other NPFs, and 2) by Fam21-mediated un-capping the dynactin complex, which contains a barbed end capped minifilament consisting mostly of Arp1⁹³. This second component allows actin filament elongation from the dynactin complex, creating an Arp2/3 complex-activating mother filament. It is tempting to speculate that actin polymerization and dynein activity could be sequentially coordinated in this process.

The mechanistic role of actin in this process is not entirely clear and could involve cargo sorting and/or tubule scission. WASH1 suppression causes an increase in tubular endosomal protrusions^{85, 86, 91}, suggesting a scission defect. Dynamin proteins might also act in scission^{80, 86}, and WASH complex co-immunoprecipitates with dynamin 2⁸⁶. However, the scission mechanism is still unclear. An interesting finding is that contact between ER and actin-enriched foci at tubulation sites occurs within 10 sec of scission⁸⁰ and appears important for scission⁷⁹. ER recruitment is not dependent on actin polymerization but does require enrichment of the actin-binding protein coronin 1C to the tubule base^{79, 80}. Interestingly, ER-endosome interaction promotes a reduction of endosomal phosphatidylinositol-4-phosphate (PI4P), which contribute to WASH activation⁹⁴. The combination of coronin, which is generally considered to antagonize Arp2/3 complex^{95, 96}, and reduction of WASH activation might therefore terminate actin polymerization prior to scission.

Actin might play a role in receptor sorting, by prolonging the lifetime of regulated tubular membranes⁸³. In addition, sorting of b2AR depends on its PDZ-binding motif, that mediates interaction with the actin filament binding protein NHERF/EBP50⁹⁷, and replacing the PDZ-binding motif with the actin filament binding sequence of ezrin also results in correct sorting⁸³, suggesting that actin binding contributes to enrichment of some proteins. Less direct interactions with actin might also contribute to enrichment⁹⁸.

Actin and EE-to-LE conversion

Annexin A2 (AnxA2)-dependent actin polymerization appears to act in the conversion of EEs to LEs⁹⁹ in a WASH-independent manner¹⁰⁰. This pathway also involves cortactin¹⁰⁰ and moesin, an ERM family member that links actin filaments to membranes. Results from cell-free systems suggest that Arp2/3 complex is involved¹⁰⁰, but its activation pathway is unclear. Also, the role of these actin filaments in EE/LE conversion is unclear but might involve intralumenal invagination of the limiting membrane. Another study links AnxA2 to endosome membrane dynamics, but in a function related more to tubular membrane protrusions¹⁰¹.

Actin and carrier translocation

Actin has been linked in multiple ways with transport of endosomal carriers. Studies in Xenopus oocytes and in HeLa-derived endosomes show that phorbol ester stimulation causes actin comet tails to assemble on acidified EEs and LEs in a Cdc42-dependent manner, with N-WASP enriching in these tails¹⁰². These results imply an Arp2/3 complex-dependent dendritic nucleation motility mechanism similar to Listeria and other intracellular pathogens¹⁰³. Perhaps in separate pathways, both myosin VI^{104, 105} and myosin Vb^{106, 107} have been reported to move EE-derived carriers. Myosin VI translocates an APPL-bound sub-set of Rab5+ endosomes to the PM along cortical actin filaments¹⁰⁵. Myosin Vb moves Rab11-positive vesicles long distances¹⁰⁸ and plays an important role in nuclear positioning in mouse oocytes¹⁰⁹. Interestingly, both myosin Va and Vb interact directly with Rab11 as well as the tandem WH2 motif Spir¹¹⁰. Since Spir works with Fmn formins to assemble actin networks^{12, 111}, coordination between actin filament assembly and myosin V is feasible. Finally, actin might play a role in docking endosome-released carriers at their destinations. Cortactin is necessary for MVB docking at the PM prior to exosome release, but this effect appears to be more on cortical actin than on MVB-associated actin¹¹². The WASH complex subunit Fam21 retains association with TGN-bound carriers after scission and mediates TGN association¹¹³, although it is not known whether this is an actin-dependent role.

Actin and ER structure/function

In addition to its myriad cellular roles (including protein synthesis, folding and transport, lipid synthesis, and steroid and carbohydrate metabolism, and cellular homeostasis), the ER contacts all other cellular organelles, and these contacts are of increasingly appreciated consequence¹¹⁴. Structurally, the ER comprises a single continuous membrane from the nuclear envelope to the peripheral ER. The peripheral ER consists of tubules and sheets that can inter-convert, with the luminal diameter for both being 60–100 nm for mammals¹¹⁵. Tubules are highly curved through the actions of reticulons, and undergo homotypic fusion using atlastin¹¹⁶. It is not clear how ER tubules undergo fission, but ER fragments during mitosis^117–119. In mammals, ER tubule movement is generally toward the PM and is microtubule-dependent through multiple mechanisms¹²⁰. Sheets are flattened structures containing a high density of bound ribosomes, and are major sites of synthesis and folding for secreted and integral membrane proteins¹¹⁵. While sheets are often thought of as continuous membrane surfaces, fenestrations have been observed^121–123. More recently, high resolution imaging showed that peripheral ER sheets in fact appear to be tightly packed arrays of ER tubules¹²⁴.

In mammals, ER has not generally been thought of as an actin-associated organelle. Only one ER-associated actin-binding protein has been identified, a splice variant of the formin INF2, but its suppression does not lead to noticeable ER changes¹²⁵. However, INF2-polymerized actin filaments increase ER-mitochondrial interaction, leading to calcium transfer between ER and mitochondria and mitochondrial fission^{15, 126, 127}. Actin appears to play a different role in ER association with PM during store operated calcium entry, although this might be through cortical actin rather than ER-bound actin^128–131. Finally, actin filaments localize to regions of fenestrated ER sheets, and inhibiting actin assembly disrupts sheet organization¹³². The mechanism by which these filaments exert their effect is unclear but may involve myosin 1c.

In mammals, ER-to-Golgi transport occurs in two phases: 1) movement from ER exit sites (ERES) to the ER-Golgi intermediate compartment (ERGIC), and 2) movement from ERGIC to cis-Golgi^{133, 134}. This process requires the COPII coat in mammals, and has generally been thought to occur through COPII-coated vesicles¹³⁴. A number of lines of evidence suggest the existence of variations in the COPII-mediated transport mechanism, including direct membrane connections between ER and ERES^{135, 136}. Microtubules are involved in ER-Golgi transit^{137, 138}, but are not required in non-polarized culture cells¹³⁹, which might be due to the close apposition of the fragmented Golgi to ERES that occurs upon microtubule depolymerization.

The role of actin in ER-to-Golgi transport is unclear. The NPF WHAMM can bind microtubules and membranes and enriches at ERGIC and cis-Golgi¹⁴⁰, but recent studies have linked WHAMM more with autophagy^{141, 142}. A recent study implicated a specific population of tropomyosin 4.2-coated actin filaments as being important for ER-to-Golgi trafficking in an Arp2/3 complex independent manner, but requiring myosin II¹⁴³. More work is needed to define actin polymerization factors and roles in transport here.

Actin and Golgi

It is humbling that many features of the mammalian Golgi are still unclear after over 100 years of study. This lack of clarity is mainly due to the large number of nanometer-scale membrane convolutions that are densely packed in this organelle¹⁴⁴. An ‘average’ Golgi in a culture cell might occupy ~3 × 5 μm, but within that is a stack of 4–11 cisternae, flattened sheets 20–40 nm thick. Furthermore, cisternal stacks are divided laterally into roughly 100 units called “mini-stacks”. Lateral membrane tethers between mini stacks dynamically assemble to create the Golgi “ribbon”, which can dissociate into stacks during Golgi “fragmentation”. Other fascinating but incompletely understood Golgi features are longitudinal inter-connections between cisternae, as well as fenestration of cisternae. These features have been elucidated largely using elegant EM tomography^145–147, with future studies by FIB-SEM promising to expand and perhaps modify our understanding.

Golgi cisternae are divided into ‘cis’, ‘medial’ and ‘trans’, in order of decreasing proximity to ERES. Cis-Golgi receives ER-derived vesicles from the ERGIC. Cargo make their way through cisternae toward the trans-Golgi while receiving modifications. The means by which cargo pass through Golgi, however, is not clear, with two possible methods: 1) vesicular traffic between Golgi cisternae of fixed identity, and 2) continuous maturation of cisternae that themselves move through the apparatus en bloc^{148, 149}. Exciting new tools to track individual cargo now show that different routes can be taken through the same Golgi¹⁵⁰. In addition, recent work shows extensive cargo and resident Golgi protein segregation laterally within an individual cisterna¹⁵¹, suggesting specialized functional zones.

The trans-most compartment is called the trans-Golgi network (TGN), and is the exit site for delivery to multiple destinations¹⁵². In addition, it is increasingly clear that the TGN interacts directly with the ER, which is a site of non-vesicular lipid transfer¹⁵³. The TGN is somewhat different from the rest of the Golgi in its extensive tubulation, and some consider it a separate organelle. Certain cargo can leave Golgi prior to the TGN¹⁵⁴.

One special Golgi feature is the existence of a proteinaceous ‘matrix’ surrounding all cisternae to a thickness of 100–150 nm, excluding ribosomes and other organelles¹⁵⁵. Major components of the matrix include golgins, which are extended coiled-coil proteins; and GRASPs, that mediate cisternal pairing both laterally and longitudinally¹⁵⁶. Specific GRASPs and/or golgins are found at specific cisternae. The matrix plays a major role in Golgi dynamics. In addition, most movement to and from cisternae must negotiate this matrix, with matrix components often specifying/facilitating these processes¹⁵⁷. It is unclear to what extent the matrix covers the region of the TGN from which membrane cargos release.

Microtubules attach to both the cis-Golgi and TGN by their minus ends^{158, 159}. These microtubules are important trafficking conduits, as well as playing roles in Golgi positioning and structure^{160, 161}. Due to the minus end microtubule attachment, essentially all MT-mediated transport toward Golgi is through dynein and all transport away from Golgi is through conventional kinesins.

Actin filaments enrich on the Golgi in specific situations, and actin is clearly involved in two aspects of Golgi structure and function: 1) membrane release from TGN, and 2) dynamics within/between Golgi stacks. We summarize key players in these pathways in Figure 5 and Table 3.

Figure 5: Actin and Golgi dynamics.

Actin is involved in multiple carrier assembly mechanisms from the TGN, with five mechanisms shown here: clathrin-coated vesicles (CCV), Rab6-dependent transport, GOLPH3-dependent transport, sphingomyelin-enriched carriers (SM), and Arf1-dependent transport. CCV engages largely in transport to endosomes, whereas the others are PM-directed. Mechanistic details for actin assembly and organization in all pathways are unclear. In addition, actin may be involved in dynamic membrane connections between cisternae in the Golgi interior. The Golgi matrix, which extends approximately 100–150 nm from the Golgi, is also shown. We do not show the matrix extending beyond the TGN, although we are not certain about this detail.

Table 3:

Proteins involved in TGN carrier assembly and Golgi dynamics.

	Clathrin transport	Rab6 transport	GOLPH3 transport	SM-rich transport	Arf1 transport	Mini-stack dynamics
Key molecules	Clathrin, dynamin	Rab6, Microtubules, Kif5b, Kif20a	PI4P, GOLPH3	Cab45, SPCA1	Arf1, dynamin	GRASP65, microtubules
Actin polymerizers	Arp2/3	NR	NR	NR	Arp2/3	NR
NPF	WAVE, N-WASP (?)	NR	NR	NR	N-WASP, cortactin	NR
Other ABPs	Hip1R, Coronin7	NR	NR	Cofilin	NR	Mena
myosins	NR	Myosin II	Myosin 18A	NR	NR	NR

NR = not reported

Membrane release from the TGN

Proteins and lipids exit the TGN to two general destinations: 1) the PM (for secretion or membrane insertion), or 2) endosomes/lysosomes¹⁵². While endosomally-targeted carriers tend to be small clathrin-coated vesicles, PM-targeted carriers are often non-uniformly shaped tubular-vesicular membranes of varying sizes (0.3–1.7 μm). There are multiple distinct assembly mechanisms for PM-targeted carriers¹⁵², four of which may use actin: 1) Rab6-based, 2) GOLPH3-based, 3) sphingomyelin/Cab45-based, and 4) Arf1-based.

In almost all of these processes, the exact mechanism for actin filament assembly is unknown. One potentially interesting factor is coronin7, which enriches at the TGN and appears to influence anterograde trafficking¹⁶², dependent upon its ubiquitination¹⁶³. Inhibition of Coronin7 Golgi localization results in decreased Golgi-associated actin filaments¹⁶³. However, the link between coronin7 and actin assembly is unclear. Coronins are generally considered to contribute more to depolymerization through effects both on Arp2/3 complex and cofilin^{95, 96}. The biochemical properties of coronin7 on actin have not been studied in detail.

Clathrin-coated vesicles (CCVs) are released from the TGN and often transport to endosomes/lysosomes, but can also go to the basolateral membrane in epithelial cells¹⁶⁴. AP-1, 3 and 4 are the major TGN-associated adaptor complexes, with GGA and epsin-related proteins working either in parallel or together with APs. All of these proteins are likely recruited by Arf1, by direct binding and/or activation of PI-4-kinase¹⁶⁴.

Arp2/3 complex also appears to be recruited to the CCV assembly sites on TGN through the WAVE complex, which itself is recruited by the clathrin coat in an Arf1-dependent manner¹⁶⁵. Rac1 is also recruited, and presumably activates WAVE complex. Curiously, N-WASP also appears important¹⁶⁵. Suppression of these actin-associated factors reduce clathrin-coated carrier assembly from TGN and delivery of cargo to lysosomes¹⁶⁵. The actin binding protein Hip1R, which is important in CME, also localizes to TGN and is important for lysosomal transport¹⁶⁶. Coronin7 may also play roles here through its interaction with Eps15¹⁶³ or AP-1¹⁶².

These findings might suggest that CCV assembly from the TGN resembles clathrin-mediated endocytosis, with Arp2/3 complex-polymerized actin filaments providing the force for budding, and a dynamin providing the scission force. Dynamin2 might serve this role at the TGN^{167, 168}.

We refer to a second mode of TGN carrier assembly as “Rab6/myosin II-based”. Rab6 is an important initiator of TGN carrier assembly, with knock-down causing a decrease in TGN-released carriers and an increase in long tubular TGN protrusions¹⁶⁹. Microtubules and kinesins are also required for Rab6-mediated carrier assembly and transport. Kinesin 1 (Kif5b) takes vesicles from the TGN to the PM, and may even supply some force carrier elongation from the TGN¹⁶⁹, while Kif20a plays additional roles¹⁷⁰.

Actin dynamics occurs transiently during TGN carrier assembly, in only a 2–4 sec pulse at carrier assembly sites prior to carrier release¹⁷⁰. At present, the mechanism for this actin polymerization is unclear. Myosin II also localizes to these sites¹⁶⁹, and myosin II activity appears required for the scission step from the TGN. Small molecule inhibitors of myosin II, or knock-down of myosin IIA or myosin IIB, cause similar tubular TGN phenotypes to Rab6 KD¹⁶⁹. Both Rab6 and myosin II inhibition reduce transport of VSV-G protein to the PM, while GPI-anchored protein exit from TGN appears un-affected. Kif20a appears to anchor Rab6 close to microtubule attachment sites on the TGN, as well as helping to recruit Myosin II through direct interaction¹⁷⁰. These three proteins form “hot spots” on Golgi from which carriers emerge. In HeLa cells, there are ~6 hot spots/Golgi, which release ~2 carriers/min.

So, an interesting cooperation between microtubules and actin is emerging for Rab6-mediated TGN carrier assembly¹⁷⁰. Two outstanding questions are: 1) the mechanism by which myosin II and actin assist fission; and 2) the mechanism for actin assembly at hot spots.

We refer to a third mode of TGN carrier assembly as “GOLPH3/myosin 18A-based”. GOLPH3 is a phosphatidylinositol-4 phosphate (PI4P) binding protein that is highly abundant in the cytosol¹⁷¹, and is recruited to the PI4P-rich TGN. While GOLPH3 participates in retrograde trafficking¹⁷², GOLPH3 KD also causes a decrease in TGN-to-PM transport of VSV-G, a decrease in carrier budding from TGN, and dilation of TGN cisternae¹⁷¹. Myosin 18A (M18A) interacts with GOLPH3, and M18A KD causes similar Golgi transport defects¹⁷¹. In contrast to GOLPH3 KD, M18A KD causes increased TGN tubulation¹⁷³. The resulting model is that GOLPH3 tubulates the Golgi carrier, and M18A-mediated force production on TGN-associated actin filaments mediates scission of this carrier¹⁷³. As with the Rab6 pathway, there is no clear mechanism for actin filament assembly.

One concern with the model is that, while M18A can bind actin filaments, there is strong evidence that M18A does not possess motor activity or other characteristics of bona fide myosin motors^174–177. These results reduce the likelihood that M18A itself can act as a motor. Another possibility is that M18A co-assembles into mini-filaments with myosin II¹⁷⁸. So, M18A could serve as a specific ‘localization’ signal for myosin II mini-filaments through its GOLPH3 interaction, and myosin II might be the actual motor. If myosin II is the motor for GOLPH3/M18A-mediated TGN carrier assembly, one question is whether this constitutes the same or a different mechanism from that of Rab6. Depletion of Rab6 only causes about 50% depletion of myosin II from Golgi¹⁶⁹, possibly suggesting multiple mechanisms for myosin II recruitment.

We must also point out that one study reports an inability to detect M18A on Golgi, as well as lack of effect of M18A KD or KO on another reported phenotype of M18A KD in¹⁷¹, Golgi compaction¹⁷⁷. This study did not examine effects on Golgi-to-PM trafficking or TGN ultra-structure. In summary, while quite compelling evidence exists for M18A function in TGN carrier assembly, a number of issues are un-resolved.

Yet another distinct mode of membrane carrier assembly, which we refer to as “sphingomylelin-rich”, occurs on regions of the TGN enriched in the lipid sphingomyelin (SM). The ATP-dependent calcium pump SPCA1 concentrates in SM-enriched regions, and the resulting increase in luminal calcium activates cargo sorting by the luminal calcium-binding protein Cab45¹⁷⁹. Only a sub-set of cargo is transported by this pathway.

Actin filaments and the actin-binding protein cofilin are required for this pathway. Depletion of cofilin results in secretion defects for a sub-set of proteins, which is rescued by treatment with actin depolymerizing drugs, suggesting that cofilin’s depolymerization activity is necessary¹⁸⁰. Cofilin co-immunoprecipitates with SPCA1 in a manner dependent on its actin binding, is required for optimal SPCA1 calcium transport activity, and enriches on Golgi in a manner dependent on SPCA1 binding^{181, 182}. The overall model is that cofilin/actin filament binding to SPCA1 stimulates increased luminal calcium, triggering cargo sorting and transport through Cab45.

A number of questions remain for this fascinating mechanism. How do cofilin and actin activate SPCA1? Perhaps through clustering? Does actin play any other roles in membrane tubulation or fission? Why would actin depolymerization be particularly important? Is the force of depolymerization used for membrane dynamics? Finally, what assembles the actin filaments?

Finally, another mode of TGN carrier assembly uses the Arf1 small GTPase, which is long known to trigger COP1-mediated retrograde transport from the cis-Golgi to ER, as well as transport between Golgi cisternae and CCV budding from the TGN^{164, 183}. Interestingly, Arf1 is also found in tubular membrane extensions from both the cis-Golgi and TGN¹⁸⁴. Even more fascinating is the fact that these tubules appear to be hot spots of clathrin-containing puncta on the TGN, and COP1-containing puncta on the cis-Golgi¹⁸⁴. GTP-bound Arf1 alone can tubulate membranes¹⁸⁵, but also can recruit the COPI coat, the AP and GGA families of clathrin adaptors and PI4K¹⁶⁴. Interestingly, Arf1 and PI4K also play roles in mitochondrial interactions with Golgi-derived vesicles, leading to mitochondrial division¹⁸⁶.

Arf1 is also important for actin and Arp2/3 complex recruitment to Golgi^{187, 188}. One mechanism might be through cortactin, which enriches on Golgi in an Arf1- and actin-dependent manner¹⁸⁹. Cortactin’s SH3 domain binds directly to dynamin2, and disruption of this interaction causes VSV-G accumulation in the Golgi¹⁸⁹. Arf1 also recruits Cdc42 to Golgi, which in turn recruits N-WASP, Arp2/3 complex, and a Cdc42 GAP to the Golgi^{188, 190, 191}. However, Cdc42 also has microtubule-dependent effects on Golgi¹⁶⁰.

Overall, a large number of mechanistic questions remain about Arf1-recruited actin and Golgi dynamics. How might Arf1 recruit actin polymerization factors? What role does actin play in tubulation? Given Arf1’s ability to recruit PI4K, could GOLPH3/M18A play a role here?

Golgi stack dynamics

The lateral membrane connections between mini-stacks that form the Golgi ribbon represents a truly spectacular bit of architecture. These connections are clearly dynamic, and might depend on both microtubules and actin filaments¹⁵⁶. Actin and myosin II play roles in the overall response of Golgi to force¹⁹². One of actin’s roles may be to stabilize interactions between GRASPs on adjacent cisternae, through an interaction between GRASP65 and the Ena/VASP protein Mena¹⁹³. The equilibrium of these dynamics is an important factor in the Golgi’s role in autophagy¹⁹⁴. Actin can also regulate the vacuolar H⁺-ATPase (V-ATPase), whose progressively greater activity from cis-Golgi to TGN causes a pH gradient across the Golgi, from pH ~ 6.7 to < 6.0¹⁹⁵. Actin depolymerization reduces V-ATPase activity¹⁹⁶. Clearly, many questions remain here, including filament assembly mechanisms and regulation thereof.

A complication in deciphering this process is that the typical read-out for defective lateral stack interaction is Golgi ‘fragmentation’. While the effect of microtubule depolymerization is consistently Golgi fragmentation¹²⁰, the effect of actin depolymerization is less consistent, with some studies showing compaction^197–199 and others showing fragmentation¹⁹³. These differences may reflect multiple roles in Golgi structure, accentuated by variations in experimental system. In this respect, it is interesting that the formin protein FHDC1, which binds tightly to Golgi-derived microtubules, affects Golgi distribution in an actin-and microtubule-dependent manner²⁰⁰. In addition, the FMNL formins enrich at Golgi, and their suppression results in Golgi fragmentation as well as trafficking defects^{11, 201}.

In addition, Golgi structure is acutely sensitive to changes in entry/exit rates of membrane carriers^{202, 203}, thus can be affected indirectly by changes to ER or to other membrane sources/sinks. A possible example of this effect is suppression of cortactin, which causes large changes in Golgi morphology, but the mechanism for this could be endosome-to-TGN transport⁹⁰. However, cortactin can directly interact with Golgi¹⁸⁹, so the situation is complicated. Another example is the positive role of intersectin-1 on Golgi fragmentation, through interaction with the TGN golgin GCC88, which appears to require actin, but also participates in Golgi dispersal induced by retromer inhibition²⁰⁴. Finally, the effects of FMNL formins on Golgi morphology^{11, 201} could be secondary to roles in trafficking. Resolution of these issues will require novel assays and approaches.

Common Themes and Future Directions

One common theme is that Arp2/3 complex plays a role in most actin-dependent processes in secretory and endocytic pathways. Especially in the cases of membrane carrier assembly (endocytosis and TGN carrier assembly), it is tempting to speculate that the function of Arp2/3 complex is somewhat similar: to provide a force-generating dendritic network that drives membrane invagination. However, for retromer-mediated endosomal carrier assembly, it is possible that the role of actin might be more for cargo sorting. For this reason, mechanistic roles must be determined on a case-by-case basis. For a number of processes discussed, the polymerization mechanism is not known, and might involve Arp2/3 complex, formins, or WH2 motif-containing proteins. In addition, it is possible that formins proteins might play roles in processes where Arp2/3 complex is already known to be required, as occurs in a number of actin-based structures. Finally, the roles of actin in the multiple transport pathways that assemble autophagosomes^{16, 17, 205} are certainly far from understood.