New directions for the clathrin adaptor AP-1 in cell biology and human disease

Abstract

The clathrin adaptor protein complex-1 (AP-1) is a central player in cell physiology and human health. It is best known for its role in linking clathrin to its cargo at the trans-Golgi network and endosomes. It participates in traffic important for the correct function of a large number of organelles, including the trans-Golgi network, endosomes, lysosomes, lysosome-related organelles, and plasma membrane. Although it was one of the first clathrin adaptors identified, new discoveries about cargo and pathways that depend on AP-1 continue to emerge. This review summarizes new research into AP-1 that further illuminates its roles in the traffic of plasma membrane proteins, in maintaining TGN content, and in human disease.

Introduction

The clathrin adaptor protein complex-1 (AP-1) was among the first adaptors identified [1]. It is a soluble cytosolic complex that links transmembrane cargo to clathrin through direct physical interaction with both cargo and clathrin [2]. AP-1 also recruits additional accessory proteins that perform diverse functions necessary for clathrin-mediated traffic, including binding to additional cargo, deforming the membrane, assembly of the clathrin lattice, and motor activity. Additionally, AP-1 may recruit factors important for vesicle fusion to a downstream organelle [3]. Thus, AP-1 defines where and when clathrin coats assemble, the cargo that will be trafficked, and the next destination for that cargo. Amazingly, despite decades of research, the pathways that depend on AP-1 and its cargo are still being discovered.

AP-1 complexes are evolutionarily conserved adaptor complexes that localize to the trans-Golgi network (TGN) and endosomes. Several excellent resources are available for a detailed description of the subunit structures, regulation, and physical interactions of AP-1, and the reader is referred to those for a background on the subject [2,4-7]. An AP-1 complex is composed of four subunits termed β, γ, μ, and σ. In most organisms, each subunit is encoded by multiple alternate genes. Thus, most organisms express a family of AP-1 complexes composed of different combinations of the alternate gene products. For example, humans potentially express 12 different AP-1 complexes based on different combinations of the single β-subunit (β1) with two alternate γ-subunits (γ1 and γ2), two alternate μ-subunits (μ1A and μ1B), and three alternate σ-subunits (σ1A, σ1B, and σ1C). However, it is not yet known whether all of these complexes form in vivo, much less whether they perform redundant or distinct functions.

AP-1 works at many locations in the cell and directs traffic to many different organelles. Previous research established that AP-1 acts in recycling from the endosome to the TGN, exit from the TGN, sorting to the dendritic domain in neurons, enabling the formation of lysosome-related organelles, and basolateral sorting in epithelial cells [5,6]. This review highlights several studies from the past two years that have expanded or clarified the repertoire of pathways and cargo that depend on AP-1 as well as the significance of AP-1 complexes to human health.

AP-1 in polarized protein traffic

Several new studies have revealed unexpected roles for AP-1 in polarized protein localization. Although AP-1 is best known for its roles in traffic between the TGN and endosomes, it is well-established that it also functions in traffic to the plasma membrane in epithelial cells. An epithelial-specific μ-subunit (μ1B) mediates traffic to the basolateral membrane [6,8,9]. Initially, this was thought to be a unique function limited to epithelial cells and limited to traffic to the basolateral membrane. However, nearly a decade ago, work in C. elegans began to suggest a role for AP-1 in apical protein localization [10-12].

It is now becoming clear that a role for AP-1 in polarized protein localization is a broadly conserved function for AP-1. In Arabidopsis, AP-1 is important for the apical secretion of the seed coat [13]. Moreover, AP-1 can be seen on vesicles emerging from the TGN that contain secretory cargo [14]. In Candida albicans, AP-1 is essential for filamentous growth, a process that requires sustained polarized secretion, [15]. Similarly, in Aspergillus nidulans, AP-1 localizes to secretory vesicles and is required for polarized secretion [16]. However, in budding yeast, this function seems to have been co-opted by other pathways [15].

In mammalian cells, several studies have now demonstrated roles for AP-1 in apical protein localization in addition to its roles in basolateral protein localization. In a recent study using MDCK cells, AP-1 was found to be important for the apical localization of the apical proteins megalin and syntaxin 3 [17]. In its absence, a fraction of each protein was mislocalized to the basolateral membrane. Notably, this study explored the contributions of the two alternate γ-subunits and revealed unique localization and functions for each subunit. γ1 localized to the TGN, common recycling endosome, and apical recycling endosome and it controlled apical recycling of megalin. In contrast, γ2 localized to only the TGN and common recycling endosome and controlled the localization of syntaxin 3, but this effect was only seen when γ1 was also disrupted. A second study focused on the effect of combined loss of the two alternate μ-subunits. This study found that in addition to syntaxin 3, GPI-anchored proteins were mislocalized to the basolateral surface [18]. These results highlight the role of AP-1 in the apical protein localization of select cargo in MDCK cells.

A third study used a comparative cell surface biotinylation approach to probe the apical and basolateral proteomes in MDCK cells lacking μ1A, μ1B, or both [19]. This study revealed many additional apical and basolateral proteins that depend on AP-1 and found that the loss of μ1A disrupted the localization of some basolateral proteins. This is contrary to prior work that suggested μ1B could compensate for the loss of μ1A [20]. However, the prior work looked at a more limited number of basolateral proteins, which may explain why the unique basolateral functions of μ1A were previously missed. The impacts of AP-1 on the plasma membrane proteome are not limited to polarized cells. A recent study explored the effect of AP-1 on the cell surface proteome in HELA cells and reported over 900 proteins were depleted from the cell surface in cells lacking the γ1-subunit, indicating a role for AP-1 in plasma membrane protein localization in unpolarized cell types [21]. Finally, work in human pluripotent stem cells revealed defects in the localization of the apical protein podocalyxin when the γ1-subunit was disrupted [22]. Curiously, the mislocalization of podocalyxin was only observed in cells grown as 3-D cysts, suggesting a differential requirement for AP-1 depending on cellular topography. This exciting finding has precedence from research on Rab35, a small GTPase that regulates a fast recycling pathway [23]. Rab35 controls podocalyxin localization in MDCK cells grown in either 2-D or 3-D conditions. However, different Rab35 effectors and upstream activators are required in 2-D and 3-D conditions [24,25]. Together with the differential requirement for AP-1 in 2-D versus 3-D, these findings indicate that cells grown in 3-D have surprisingly different trafficking wiring than those grown in 2-D. Altogether, the recent work on AP-1 in mammalian epithelia firmly establishes a role for AP-1 in apical protein localization in multiple mammalian cell types.

It is still unclear how AP-1 controls apical protein localization in epithelial cells. It could function directly in the traffic of apical proteins by sorting the cargo into clathrin-coated vesicles at the TGN or at the recycling endosome. At the TGN, AP-1 could direct cargo to the plasma membrane or into an endosome where plasma membrane sorting occurs (Figure 1 arrow 1). A function for AP-1 in apical protein sorting at the recycling endosome is particularly appealing because AP-1 localizes to the recycling endosome and the recycling endosome is particularly important for the localization of many apical proteins [17,22,26-29]. In this case, AP-1 could mediate traffic directly to the plasma membrane from the recycling endosome (Figure 1 arrow 2) or move cargo from an early endosome to a second endosome where traffic to the plasma membrane occurs via an additional AP-1 independent mechanism (Figure 1 arrow 3). However, apical proteins generally lack the types of sorting motifs known to be required for direct interactions with AP-1. Indeed, GPI-anchored proteins lack any cytosolic domain to interact with AP-1 or any other component of the coat. This raises a question as to how AP-1 could mediate traffic of apical proteins at any location within the cell.

Figure 1.

Schematic for the potential roles of AP-1 in apical protein localization. On the left, models for how AP-1, if it interacts directly with apical proteins, could contribute to apical protein localization. A. AP-1 (yellow cage and magenta arrow #1) binds apical protein cargo at the TGN and packages them into vesicles directed to the apical plasma membrane or to an endosome from where another pathway (black) sorts apical proteins to the plasma membrane. AP-1 binds apical proteins at the endosome and packages them into vesicles directed to the apical plasma membrane (magenta arrow #2). AP-1 binds apical proteins at the endosome and packages them into vesicles directed to another organelle where another pathway sorts apical proteins to the plasma membrane (magenta arrow #3). On the right are models for how AP-1 could control apical protein localization without direct binding to apical proteins. AP-1 controls the positioning of an endosome that is important for apical protein localization (grey arrow #4). Due to the mispositioning of this endosome, the apical proteins are mislocalized. AP-1 contributes to endosome identity by sorting material out (magenta arrow #5) of or into (not shown) an endosome. This sorting process allows the endosome to mature into one competent for apical protein sorting. AP-1 delivers proteins to the endosome that are important for apical protein sorting (magenta arrow #6).

Alternatively, several indirect mechanisms are possible. For example, AP-1 could control apical protein localization by correctly positioning the recycling endosomes (Figure 1 arrow 4). AP-1 controls endosome positioning by linking endosomal vesicles to kinesins [30,31]. Indeed, consistent with a defect in recycling endosomes, depletion of AP-1 in worms disrupts the localization of RAB-11 to endosomes, suggesting that endosome position or identity depends on AP-1 [11,27]. However, in human pluripotent stem cells lacking γ1, Rab11 endosomes appear to be positioned correctly, indicating that endosome positioning may not explain the apical protein localization defect seen in these cells [22].

Additionally, AP-1 could be required to maintain or establish endosome identity (Figure 1 arrow 5). It is now clear that some compartments of the endolysosomal system undergo a process of maturation. Membrane traffic can contribute to this maturation process by removing transmembrane proteins and lipids that define the earlier stage endosome or delivering the proteins and lipids that define the later stage endosome [32,33]. Because AP-1 is known to mediate traffic to and from the endosomal system, it could participate in this maturation process.

Finally, AP-1 could disrupt the function of the endosome by disrupting the localization of proteins required for traffic into or out of the recycling endosome (Figure 1 arrow 6). Appealing candidates for this model are SNARE proteins, which are important for the fusion of incoming vesicles. Several SNARE proteins are the direct cargo of AP-1 binding proteins from the EpsinR family. Thus, EpsinR mediated SNARE traffic provides a potential mechanism by which AP-1 could indirectly disrupt traffic [34-37]. Indeed, one study proposed mislocalization of syntaxin 3 as the explanation for the defect in GPI anchor protein localization seen in MDCK cells [18]. However, another study revealed that apical protein localization was disrupted in cells without defects in syntaxin 3 localization, which suggests this may not be the main way AP-1 controls apical protein localization [17]. Together the recent studies provide ample evidence for a role for AP-1 in apical protein sorting in many systems. However, it is equally clear from these studies that much future work is required to determine which, if any, of the proposed models explain how.

AP-1 in dynamic recycling at the TGN in yeast

AP-1 complexes localize to the TGN in many, if not all, systems. However, there remains much uncertainty about its function at the TGN. This uncertainty emerges from the ability of AP-1 cargos to retain their normal distribution in the absence of AP-1 by recycling or rerouting via different pathways, the redundancy of AP-1 with other clathrin adaptors at the TGN, and, in yeast, the dynamic and multifunctional nature of the TGN [38-42]. New research in yeast has overcome some of these problems to illuminate the role of AP-1 in retaining proteins within the dynamic TGN.

In the yeast Saccharomyces cerevisiae, disruption of AP-1 causes no effect on cell growth under standard laboratory conditions. However, several previous studies pointed to a role for AP-1 in retaining proteins in the TGN. Initial evidence came from studies using synthetic cargo targeted to the lysosome. In these studies, AP-1 appeared to slow its delivery from the TGN to the late endosome, suggesting a potential role in retaining proteins at the TGN [43,44]. Additionally, when AP-1 is disrupted, some TGN proteins traffic to the plasma membrane [45-47]. These proteins are frequently rapidly endocytosed and delivered back to the TGN, so the defects are only observed when endocytosis or other trafficking pathways are disrupted. Together these findings formed the basis for a model that AP-1 retains proteins at the TGN. Recent studies further strengthen and clarify this retention model.

A key development that allowed the clarification of the retention model came from the discovery that the TGN is a dynamically maturing compartment that receives endocytosed material in yeast [39,48,49]. The identification of the TGN as an endosomal compartment simplifies earlier models that needed to incorporate roles for AP-1 at distinct TGN and early endosome compartments. New work using a synthetic secretory cargo has now revealed that AP-1 plays a dynamic traffic-based role in retention. This synthetic cargo allowed kinetic measurements of a pulse of fluorescent cargo to show that AP-1 mediates active recycling of material from a late-stage TGN back to an earlier stage [50]. This function likely involves direct interactions between AP-1 and the cytosolic tails of TGN-resident proteins that bind the synthetic cargo. Finally, subsequent work has identified the clathrin adaptor Ent5 as performing a partially redundant role in this retention function (Figure 2) [51]. This has allowed the identification of additional TGN resident proteins that are recycled within the TGN via AP-1. Moreover, the redundancy between Ent5 and AP-1 may explain why phenotypes for AP-1 mutants are difficult to observe in Saccharomyces cerevisiae.

Figure 2.

Schematic for the role of AP-1 and Ent5 in recycling to maintain Golgi and TGN identity. In yeast, Golgi form de novo when e coalescence of ER-derived vesicles fuse with vesicles containing Golgi resident enzymes (not shown). The newly formed Golgi then mature into a TGN by a process of vesicle budding and fusion. AP-1 and Ent5 participate in two rounds of recycling in the TGN. In recycling Step a, the late Golgi matures into the early TGN, AP-1, and Ent5 package Golgi resident enzymes into recycling vesicles that are delivered to a different newly maturing Golgi. In recycling Step b, as an early TGN matures into a late TGN AP-1 and Ent5 package TGN resident enzymes and delivers them to a different newly maturing TGN. These recycling reactions maintain the proper distribution of Golgi and TGN enzymes despite continued de novo generation of new Golgi, their maturation into the TGN, and ultimate dissipation of the organelle TGN through the process of secretion and recycling.

The role of AP-1 in retaining proteins within a dynamically maturing TGN in budding yeast likely reflects a role analogous to its functions in mammalian cells in endosome to TGN traffic. Unlike yeast, in mammalian cells, the early endosome is clearly distinct from the TGN, however, a large number of proteins cycle between the TGN and endosome [52]. These include the mannose-6-phosphate receptors which engage in repeated cycles of traffic between the TGN and endosome. In cells derived from a mouse knock-out line, loss of AP-1 resulted in redistribution of mannose-6-phosphate receptors from the TGN to the endosomes and impaired its recycling [53]. More recently, studies using an acute inhibition system have verified a role for AP-1 in endosome to TGN traffic for mannose-6-phosphate receptors and additional proteins [38,54]. Thus, despite significant differences in the organelle identity between yeast and mammalian cells, the function of AP-1 in maintaining TGN content appears conserved.

The evidence supporting the role of AP-1 in recycling mannose-6-phosphate receptors to the TGN is notable considering the historical view of the role of AP-1 in lysosomal enzyme traffic. AP-1 was initially thought to mediate the exit of lysosomal enzymes from the TGN in both yeast and mammalian cells. This idea emerged from early studies showing that mammalian AP-1 interacted with mannose-6-phosphate receptors, a receptor for many lysosomal enzymes [55,56]. This idea was called into question when subsequent work showed that AP-1 did not bind the key residues that are needed mannose-6-phosphate receptor to exit the TGN [57]. Subsequent work shed light on how mannose-6-phosphate receptor exited the TGN by identifying a second class of clathrin adaptors the Golgi-localized, γ-ear–containing, ARF-binding proteins (GGAs) that bound to these residues [58]. In mouse L-cells, GGAs are required for mannose-6-phosphate receptor incorporation into clathrin-coated vesicles [59]. Notably, GGAs themselves were not found in clathrin-coated vesicles in L-cells. This observation together with colocalization of GGA and AP-1 on budding structures at the TGN observed by immuno-gold labeling led to the model that Gga2 concentrates cargo and hands it off to AP-1 which assembles the clathrin-coated vesicle [59]. More recent work in HeLa cells using acute inhibition supports this model [38]. In this system, Gga2 was found in clathrin-coated vesicles and was important for mannose-6-phosphate and lysosomal enzyme incorporation into the vesicle. However, Gga2 incorporation into CCV depended on AP-1, supporting the idea that Gga2 is a cargo selective adaptor that recruits cargo to vesicles formed by AP-1. However, some questions remain, most significantly, why mannose-6-phosphate receptors mislocalize to the periphery after acute or prolonged inhibition of AP-1 [53,60,61]. This argues that the mannose-6-phosphate receptors can exit the TGN without AP-1, although this may be a non-physiological effect of blocking its normal exit route.

The cooperation of AP-1 and GGAs in lysosomal enzyme traffic in mammalian cells stands in contrast to work in budding yeast which suggests that the two do not cooperate. In yeast, AP-1 and Gga2 are recruited to the TGN independently of one another [41] and exhibit different kinetics at the TGN, suggesting they participate in different pathways [40]. New work further establishes that AP-1 has little or no role in traffic from the TGN to the lysosomal compartment, called the vacuole in yeast. Instead, the GGAs are responsible for most if not all clathrin-mediated traffic of proteins from the TGN towards the vacuole. Early studies in budding yeast, showed that although clathrin was important for several acid hydrolases to arrive at the vacuole, AP-1 was not [62-64]. Even though GGAs were later identified to impair acid hydrolase sorting at the TGN [65-67], the idea that AP-1 was involved in acid hydrolase sorting remained prevalent because mutations in AP-1 enhanced the sorting defect of a GGA mutant [42]. One recent study explored the requirements for sorting of acid hydrolase receptor Vps10, using a synthetic cargo. This study found that GGAs but not AP-1 were required for sorting to the vacuole [68]. Moreover, the timing of GGAs arrival at the TGN coincided with when lysosomal targeted cargo began to exit the TGN whereas AP-1 was not recruited until most of this cargo has left the TGN. This is consistent with earlier work that showed AP-1 was dispensable for Vps10 incorporation into clathrin-coated vesicles in yeast and was dispensable for an in vitro TGN to vacuole assay [69,70]. The earlier observation of enhanced defects in double mutants lacking both AP-1 and GGAs can be explained by a model where AP-1 retains vacuolar targeted proteins within the TGN where GGA sorts them for vacuolar delivery.

A second study came to a similar conclusion about the relative roles of AP-1 and GGAs from looking at the traffic of endocytosed proteins. This study followed plasma membrane proteins targeted to the vacuole after endocytosis [49]. These proteins transit through the TGN. However, in cells lacking Gga2 the delivery of the endocytosed proteins to the vacuole was impaired, consistent with several prior reports [71-74]. Notably, deletion of a gene encoding an AP-1 subunit did not impair vacuolar delivery on its own, however, it subtly enhanced the defect caused by deletion of GGA2 resulting in less protein delivered to the vacuole, less accumulated at the TGN, and more accumulated at the plasma membrane [49]. These results further support a role for AP-1 in retaining proteins at the dynamic TGN whereas GGA proteins sort proteins towards the vacuole. It is currently unclear whether a similar separation of function occurs for AP-1 and GGAs in other organisms either at the TGN or in other locations in the cell.

AP-1 in human disease

In recent years, multiple human disorders have been mapped to defects in AP-1 or proteins important for AP-1 functions. These disorders highlight the broad physiology that depends on AP-1. The first disorders identified were due to defects in the various σ-subunit genes [75,76]. Interestingly, mutations in the three different σ-subunit genes have largely unrelated clinical manifestations. This may reflect functional differences between the alternate σ-subunits or their differential expression. Mutations in AP1S1, which encodes σ1A, cause a multisystem disorder that impacts the skin, intestine, and nervous system, causing congenital chronic diarrhea, congenital hyperkeratosis, ichthyosis, intellectual disabilities, and hearing loss [77,78]. In contrast, mutations in AP1S2, which encodes σ1B, cause X-linked intellectual disabilities with no reported involvement of the skin or intestine [79]. Finally, mutations in AP1S3, which encodes σ1A and is primarily expressed in a subset of skin cells, cause a form of psoriasis [80].

A pair of recent studies reveal that mutations in AP1B1, the sole AP-1 β1-subunit gene, cause a disorder similar to that caused by mutations in AP1S1 [81,82]. Surprisingly, some patients appear to completely lack β1 protein. Because AP-1 complexes lacking a β subunit are predicted to be non-functional, the survival of people lacking β1 is remarkable. One potential explanation for this observation is that hybrid AP-1 complexes may form using the β2-subunit from the endocytic AP-2 complex to substitute for β1. The β subunits of AP-1 and AP-2 are 84% identical. Moreover, in mice lacking β2, β1 is incorporated into AP-2 complexes indicating that a similar mechanism may occur in the human patients lacking β1 [83].

Finally, a fifth human disorder is caused by mutations in HEATR5B. HEATR5B is a known AP-1 associated protein, although its interactions with AP-1 are thought to be mediated by other binding partners [84,85]. Notably, HEATR5B genes from flies, yeasts, and worms control AP-1 localization suggesting a fundamental and conserved role for the HEATR5B family in AP-1 functions [27,86-89]. Accordingly, mutations in HEATR5B cause a disorder with clinical manifestations similar to those caused by mutations in AP1S2, including intellectual disability, epilepsy, and motor delay [90]. Together, these studies highlight the wide variety of human physiological systems that depend on AP-1.

The mechanistic explanations for the different clinical manifestations caused by mutations in AP-1 are still emerging. Two studies attribute some of the symptoms seen in patients with mutations in AP1S1 to a defect in the localization of copper transporters [81,91]. In addition, the psoriasis symptoms caused by mutations in AP1S3 have been attributed to defects in autophagy [92]. For other clinical manifestations, new research is providing tantalizing clues. Keratinocytes from patients with mutations in AP1B1 provided insight into the hyperkeratosis and ichthyosis symptoms in those patients. In the patient-derived keratinocytes, junctional proteins are mislocalized, including E-cadherin, a known cargo of AP-1 [82,93]. The disruption of the localization of E-cadherin and other junctional proteins may directly contribute to hyperproliferation and skin differentiation symptoms in these patients. Similarly, a pair of in vitro studies have revealed insight into severe and occasionally lethal diarrhea caused by mutations in AP-1 subunit genes. The first study explored the role and significance of AP-1 in the basolateral localization of an innate immune receptor, Toll-like receptor 3 (TLR3) [94]. Here, in intestinal organoids or spheroids lacking μ1B, TLR3 was mislocalized to the apical surface, which is analogous to the intestinal lumen. If such a mislocalization were to happen in vivo, the microbiota of the gut would likely activate TLR3 signaling, which would cause inflammation and diarrhea. A contrasting model emerges from studies of AP1S1 in human colon organoids [95]. Here, loss of AP1S1 resulted in defects in epithelial barrier functions, which could contribute to diarrhea. However, due to the absence of patient intestinal biopsy samples, it remains to be determined whether either or both mechanisms contribute to this clinical feature. This highlights an ongoing challenge for the field in which future advances will require close collaboration between clinical and laboratory research teams.

Summary

In the nearly fifty years since AP-1 was first observed as faint ‘traces’ in a clathrin preparation [96], AP-1 has emerged as a central player in cell physiology and human health. Its functions directly or indirectly impact the composition of a truly stunning number of organelles, including the TGN, endosomes, lysosomes, lysosome-related organelles, and plasma membrane. It is no wonder that mutations in AP-1 affect so many organs and tissues. Yet despite all we know about AP-1, much is yet to be discovered. In yeast, where the function of AP-1 has remained elusive, new research finally provides a mechanistic explanation for its retention functions at the TGN. Recent research highlights the role of AP-1 in apical protein localization in mammalian epithelia, although the molecular mechanism remains unclear. Finally, as the number of human disorders associated with defects in AP-1 continues to expand, there remains a pressing need to tie molecular mechanisms to the clinical observations. The future promises many more discoveries that will surely provide a significant impact on both the laboratory and clinic.

Highlights:

• AP-1 is known to control the protein content of many organelles in the endo-lysosomal system

• Proteomics and other studies reveal widespread effects of AP-1 on plasma membrane content and the polarized localization of many plasma membrane proteins

• Kinetics studies in yeast reveal the involvement of AP-1 in a dynamic retention mechanism at the trans-Golgi network

• Studies from the lab and clinic provide insight into AP-1 dysfunction in human disease