Neuropeptide precursors packaged into SGs are cleaved into active peptides and hormones by SG-resident PCs. The PC family contains nine members: PC1/3, PC2, furin, PC4, PC5/6, PACE4, PC7, SKI-1/S1P, and PCSK9 (Seidah et al., 2013). PC1/3 and PC2 (proprotein convertase subtilisin/kexin type 1/3 and 2) are encoded by PCSK1 and PCSK2 genes and act as basic proprotein convertases that cleave after polybasic clusters (Seidah et al., 1991; Cendron et al., 2023). PC1/3 and PC2 are selectively expressed in endocrine and neuroendocrine cells, suggesting they are important in prohormone processing within SGs (Figure 3) (Halban and Irminger, 1994; Zhou et al., 1999). PC1/3 and PC2 process a plethora of prohormones, including pro-opiomelanocortin (POMC), neuropeptide Y (NPY), agouti-related peptide (AGRP), progrowth-hormone releasing hormone (GHRH), prothyrotropin-releasing hormone (TRH), proinsulin, and proglucagon (Paquet et al., 1996; Nillni, 2010; Seidah, 2011; Cawley et al., 2016). Consequently, defects in PC1/3 or PC2 activity result in several endocrinopathies in both humans and rodents (Furuta et al., 1997; Jackson et al., 1997; Steiner, 1998; Zhu et al., 2002; Scamuffa et al., 2006; Mbikay et al., 2007; Anini et al., 2010; Creemers et al., 2012; Seidah and Prat, 2012).

After endoproteolytic cleavage by PC1/3 and/or PC2, the newly exposed C-terminal basic residues of prohormones are removed by CPE, another resident protein of SGs (Burgoyne and Morgan, 2003). CPE was first identified as enkephalin convertase and subsequently found to cleave the C-terminally extended basic residues from diverse peptide intermediates, including POMC and probrain-derived neurotrophic factor (BDNF) (Fricker and Snyder, 1982; Hook et al., 1982; Fricker, 1988; Lou et al., 2005). CPE is expressed primarily in endocrine tissues and specific areas of the central nervous system (Figure 3) (Fricker, 1988; Cawley et al., 2012; Ji et al., 2017). CPE differs from other carboxypeptidases in that its optimal pH is in the acidic range, consistent with its localization to acidic compartments of the TGN and to the dense core of SGs where prohormone processing occurs (Supattapone et al., 1984; Ji et al., 2017).

Like other proteins in the regulated secretory pathway, CPE is synthesized in the ER as a 476-amino acid precursor containing an N-terminal signal peptide that directs proCPE into the ER before its removal (Song and Fricker, 1995a). ProCPE is transported through the Golgi to SGs where the 17-amino acid “pro” region is cleaved after a penta-arginine sequence to generate mature membrane-bound CPE that is glycosylated at two N-linked glycosylation consensus sites, Asn139 and Asn390 (Song and Fricker, 1995a; Cawley et al., 2012; Ji et al., 2017). Within SGs, the membrane-bound CPE can be further processed by cleavage of its C-terminal cytoplasmic tail to generate a soluble form of CPE that is enzymatically more active (Hook, 1985; Fricker and Devi, 1993).

In addition to its enzymatic function, CPE acts as the quintessential regulated secretory pathway sorting receptor for many prohormones (e.g., POMC, pro-BDNF, proenkephalin, proinsulin, and prophogrin). Under mildly acidic conditions and increasing calcium concentrations, CPE aggregates and binds SG cargo proteins through their prohormone sorting signals, such as those found in POMC (Song and Fricker, 1995b; Rindler, 1998; Zhang et al., 1999). CPE binding enables protein condensation and directs cargo proteins to SGs (Cawley et al., 2012). The C-terminus of CPE forms an amphipathic α-helix under acidic conditions that binds to lipid rafts in the TGN membrane and directs prohormones bound to CPE into nascent granules (Fricker et al., 1990; Dhanvantari and Loh, 2000; Dhanvantari et al., 2002; Zhang et al., 2003; Cawley et al., 2012). In a neutral environment (pH 7.2), CPE’s C-terminus does not exhibit a helical secondary structure, preventing premature association with the membrane (Dhanvantari et al., 2002). Interestingly, CPE may only sort certain prohormones to the regulated secretory pathway, as CPE depletion did not affect sorting of chromogranin A (CHGA) (Cool et al., 1997; Normant and Loh, 1998; Ji et al., 2017).

Chromogranins A and B (CHGB), secretogranins II and III (SCGII and SCGIII), and a few additional related proteins together comprise the granin family of water-soluble acidic glycoproteins (Bartolomucci et al., 2011). These granin proteins serve essential roles in the regulated secretory pathway, with the chromogranins comprising much of the SG matrix (O'Connor and Frigon, 1984; Borges et al., 2010), and, accordingly, are predominantly expressed in endocrine and neuroendocrine cells (Figure 3) (Day and Gorr, 2003; Dominguez et al., 2018). Like CPE and other SG resident proteins, granins are also synthesized at the rough ER, inserted into the ER cisternae via a signal peptide located at their N termini, and trafficked to the TGN via transport vesicles (Bartolomucci et al., 2011). Several biochemical properties that are critical for the function of granins include an acidic isoelectric point, Ca²⁺ binding, and thermostability (Yoo and Albanesi, 1991; Taupenot et al., 2003; Bartolomucci et al., 2011). Additionally, granins aggregate in an acidic environment (pH 5.5) with a millimolar concentration of calcium ions and, by doing so, induce granule formation (Chanat and Huttner, 1991; Parmer et al., 1993; Koshimizu et al., 2010). Granins are negatively charged, which may prevent premature aggregation, but a surplus of calcium ions and protons in the SG lumen may help neutralize the repulsive forces among the granin proteins to allow aggregation (Glombik and Gerdes, 2000). An alternative explanation proposed is that the pH gradient prompts interactions between negative and positive charges of cargo proteins, and then divalent ions may generate a chelate bridge between two negatively charged granins to permit aggregation (Ma et al., 2008; Zhang et al., 2010).

CHGA, the most well-studied granin family member, is a prohormone and a granulogenic factor in neuroendocrine tissues (Laguerre et al., 2020). While CHGA is mostly hydrophilic, its C- and N-termini contain hydrophobic and cell-specific evolutionarily conserved sequences necessary for sorting and granulogenesis (Yoo and Lewis, 1993; Cowley et al., 2000; Montero-Hadjadje et al., 2009; Elias et al., 2010). Cysteine residues within the amino terminus form an intramolecular disulfide loop to interact with the membrane (Parmer et al., 1993; Yoo, 1994; Kang and Yoo, 1997). The primary structure of CHGA contains several glutamic acid stretches that can interact with Ca²⁺, leading to aggregation in specific environments, such as the TGN and SG (Parmer et al., 1993). CHGA may associate with membrane either directly by binding to specific lipids (e.g., phosphatidic acids enriched in TGN and SG membranes) (Carmon et al., 2020; Tanguy et al., 2020) or indirectly by interacting with SCGIII, which in turn binds to cholesterol-rich membranes and targets proteins to the regulated secretory pathway (Hosaka et al., 2004; Han et al., 2008).

Besides their granulogenic function, granins contribute to calcium homeostasis and many are also precursors of bioactive peptides that, upon further processing in ISGs, modulate different physiological processes, including pain pathways, inflammatory responses, metabolic and mood disorders, and blood pressure (Montero-Hadjadje et al., 2008). As an example, the CHGA-derived peptide pancreastatin, which was the first granin-derived peptide discovered, strongly inhibits glucose-induced insulin release (Tatemoto et al., 1986; Bartolomucci et al., 2011).

Even though the granin family is ubiquitously expressed in neuroendocrine tissues, individual proteins exhibit tissue specificity and redundancy (Figure 3). For example, the ablation of Chga and Chgb reduced the size and number of SGs in adrenal chromaffin cells, while SGs in mouse hippocampal neurons were unaffected (Dominguez et al., 2018). In addition, ablation of Chga upregulated the expression of Chgb and secretogranins II-VI in the adrenal medulla and other endocrine glands (Hendy et al., 2006). These data indicate that chromogranin-mediated dense core formation of SG is tissue-specific and that other granin family members may perform a similar function.

MPRs facilitate the transport of soluble acid hydrolases from the TGN to the lysosome by binding to mannose-6-phosphate (M6P) modifications on the enzymes. There are two different MPRs, the larger cation-independent receptor (CI-MPR) and the smaller cation-dependent receptor (CD-MPR) that binds M6P more efficiently in the presence of divalent cations (Gary-Bobo et al., 2007). Besides binding to phosphomannosyl residues, the extracellular region of CI-MPR, also known as IGF2R, binds insulin-like growth factor (IGF)-II to facilitate its endocytosis and clearance by lysosomal degradation (Oka et al., 1985; Oshima et al., 1988; Schmidt et al., 1995; Brown et al., 2008). CI-M6PR also binds other ligands, such as retinoic acid, granzyme B, latent TGF-β, urokinase-type plasminogen activator receptor, and leukemia inhibitory factor, impacting a variety of biological pathways (Purchio et al., 1988; Kang et al., 1997; Blanchard et al., 1999; Godar et al., 1999; Veugelers et al., 2006).

CI-MPR is one of the most thoroughly studied proteins that is removed from ISGs and then recycled back to the TGN (Figures 1, 2, 4, 5) (Bonnemaison et al., 2013; Seaman, 2018). As SGs mature, the concentration of MPRs declines by about 90% (Klumperman et al., 1998). MPRs are sorted from ISGs by binding to adaptor protein 1 (AP-1) and the Golgi-localized, γ-ear containing, ARF-binding (GGA) family of proteins in clathrin and syntaxin 6–positive vesicles that are delivered to endosomes (Figures 1, 2) (Klumperman et al., 1998; Coutinho et al., 2012). Then, the multiprotein retromer complex mediates the endosome-to-TGN retrieval of CI-MPR (Bonnemaison et al., 2013; Seaman, 2018). Retromer is critical for the recycling and reuse of MPRs in a new cycle of acid hydrolase transportation.

In the following sections, we discuss sorting into the regulated secretory pathway, as well as the retromer-dependent recycling of CI-MPR, potential retrograde trafficking targeting motifs, and finally, the tissue-specific role of MAGEL2 in the retromer-dependent recycling of mature and immature SG components.

For secreted proteins, there are two proposed sorting models that are not mutually exclusive, as supporting evidence suggests both models may occur in cells (Figure 4) (Dikeakos and Reudelhuber, 2007; Goronzy and Weyand, 2009). In the “sorting-at-entry” model, constitutively secreted proteins are segregated from regulated proteins by binding to specific receptors clustered in the TGN before granule formation (Figures 4, 5) (Chidgey, 1993). Initial aggregation of cargo proteins (e.g., granins and prohormones) in a mildly acidic pH (6.0–6.5) and cation-dependent manner excludes certain constitutive proteins. Then, the aggregates can bind to sorting receptors at the TGN membrane. One of the first sorting receptors proposed was CPE, which sorts POMC and proBDNF (Cool et al., 1997; Lou et al., 2005). SCGIII can also act as a sorting receptor, as SCGIII associates with cholesterol-sphingolipid-rich membrane microdomains (i.e., lipid rafts) in the TGN membrane and, by doing so, serves as a sorting receptor for CHGA in pituitary and pancreatic cells (Tooze, 1998; Hosaka et al., 2004; Park and Loh, 2008). Receptors are then recycled through vesicles budding off ISGs (Tooze, 1998).

In the “sorting by retention” model, both regulated and non-regulated proteins enter the nascent granule with the latter proteins excluded from the ISG by budding of clathrin-coated vesicles (Chidgey, 1993). In support of this model, lysosomal proteins have been found in immature SGs, which are then most likely excluded through binding to MPRs and budding of a constitutive-like vesicle. All evidence points to the conclusion that the best-fitting sorting model is reliant on the specific protein, its affinity for aggregation, the relative speed of its synthesis, and tissue or cell specificity (Tooze, 1998). As an example of cell specificity, the sorting domain in the C-terminus of PC2 is essential for sorting in Neuro2A cells (Assadi et al., 2004) but is not required in corticotrophic AtT-20 cells (Taylor et al., 1998; Lacombe et al., 2005). Similarly, the SG sorting of CHGA and CHGB requires an N-terminal domain in neuroendocrine PC12 cells but not in endocrine GH4C1 cells (Chanat et al., 1993); rather, a C-terminal region of CHGA was required for proper sorting in GH4C1 cells (Cowley et al., 2000). Different sorting domains on the same protein offer insights not only into targeting efficiency but also into the possibility of targeting signal redundancy, which may offer protection from mutations and other damage.

Mechanistically, several processes contribute to sorting RSPs into the regulated secretory pathway. An intrinsic propensity to aggregate combined with Ca²⁺-binding motifs underly one sorting mechanism–selective aggregation of a subset of soluble proteins in the TGN, excluding soluble non-aggregating proteins and giving rise to SGs (Figure 5). In this aggregation mechanism, only one or perhaps a small number of proteins need to interact with the membrane, as the other regulated proteins are targeted to the SG through protein-protein interactions mediated by helper proteins, like the granins. Accordingly, overexpression of CHGB led to more effective sorting into granules in cells (Natori and Huttner, 1996; Huh et al., 2003). Expression of granulogenic proteins (e.g., CHGA) in regulated secretory-deficient cells was sufficient to induce the formation of vesicles that resembled SGs (Kim et al., 2001; Huh et al., 2003; Beuret et al., 2004). However, aggregation alone is insufficient to target all RSPs to SGs (Quinn et al., 1991; Jutras et al., 2000), suggesting additional sorting mechanisms (Lacombe et al., 2005).

Association with lipid rafts is another sorting-at-entry mechanism used by several RSPs (Figure 5). Through insertion into lipid rafts, the transmembrane domains of several prohormone-processing enzymes, like PAM, mediate their own sorting into the regulated secretory pathway (Bell-Parikh et al., 2001). In addition, several sorting receptors in the TGN recruit and fasten SG-destined proteins to membrane sites where a nascent vesicle will bud (Park and Loh, 2008). For example, SCGIII, CPE, PC1/3, and PC2 have all been reported to associate with cholesterol- and sphingolipid-rich lipid rafts, which is crucial for their own targeting to the regulated secretory pathway (Dhanvantari and Loh, 2000; Jutras et al., 2000; Arnaoutova et al., 2003; Assadi et al., 2004; Hosaka et al., 2004; Dikeakos et al., 2009).

Besides interacting with sorting receptors, various cargo adaptors, such as adaptor protein (AP) complexes and monomeric GGA proteins, can help direct proteins into their appropriate transport carriers (Figure 5) (Robinson, 2004). The AP family includes five cytosolic heterotetrameric complexes, AP-1 to AP-5, that mediate sorting of transmembrane proteins on defined intracellular routes (Bonnemaison et al., 2013). AP-1, AP-3, and AP-4 are associated with the TGN, and AP-1 also removes material from ISGs (Bonnemaison et al., 2013; Guardia et al., 2018). AP-2 regulates clathrin-mediated endocytosis at the plasma membrane, and AP-5 facilitates the retrograde transport of proteins from endosomes to the TGN (Bonnemaison et al., 2013; Hirst et al., 2013; Guardia et al., 2018). The subunits of these AP complexes have several isoforms in mammals, suggesting evolutionary adaptation to finetune the process of regulated secretion, particularly in neurons that uniquely express two AP-3 subunits (Dell'Angelica et al., 1997; Boehm and Bonifacino, 2002; Bonifacino, 2014; Li et al., 2016; Guardia et al., 2018). GGA proteins act as monomeric clathrin adaptors (Bonifacino, 2004). Arf small G proteins in an active GTP-bound, membrane-associated state mediate the membrane recruitment of AP complexes and GGAs (Traub et al., 1993; Austin et al., 2000; Collins et al., 2003; Ren et al., 2013). Then, AP complexes can bind to the cytoplasmic tails of cargo proteins and recruit coat proteins (i.e., clathrin) and accessory proteins to drive vesicle formation (Bonnemaison et al., 2013; Tan and Gleeson, 2019). Upon the release of secretory vesicles from the TGN membrane, coat proteins are dissociated and recycled for additional rounds of vesicle formation (Tan and Gleeson, 2019).

The motifs and domains on RSPs that are responsible for their proper sorting into SGs remain enigmatic and not uniform. In general, sorting signals facilitate the aggregation of cargo proteins or piggyback anchoring through association with membrane domains and other proteins (Figure 5).

Within the TGN lumen, proteins targeted to SGs tend to aggregate, not only promoting formation of the dense core of the SGs but also enabling their own sorting to the regulated secretion pathway (Figure 5) (Burgess and Kelly, 1987). Several motifs and domains within RSPs promote aggregation and sorting to SGs. In the case of CHGA, several glutamic acid repeats interact with Ca²⁺ and promote aggregation at the TGN and in SGs (Parmer et al., 1993). Small disulfide (CC) loops also act as SG sorting signals by promoting self-aggregation at the TGN (Reck et al., 2022). These CC loops are present frequently at the very N- or C-terminus of proteins or close to processing sites and thus potentially exposed (Reck et al., 2022). For example, in POMC, an N-terminal 13-residue CC loop is necessary and sufficient for granule sorting (Tam et al., 1993; Cool et al., 1995; Loh et al., 2002). Longer disulfide loops in CHGA and CHGB are also involved in their sorting to SGs (Kang and Yoo, 1997; Kromer et al., 1998; Glombik et al., 1999; Taupenot et al., 2002). In addition, di-basic processing sites and acidic motifs in prohormones were shown to promote granule sorting, likely via interaction with PCs or CPE (Brechler et al., 1996; Lou et al., 2005). The sorting signal motif for CPE recognition was first identified as two acidic residues and two aliphatic hydrophobic residues in POMC (Cool et al., 1995), and similar sorting motifs were subsequently found in proinsulin (Dhanvantari et al., 2003), BDNF (Lou et al., 2005), and proenkephalin (Normant and Loh, 1998; Cawley et al., 2016).

Amphipathic helices enable interaction with the granule membrane and are important for the incorporation of RSPs and enzymes, like PCs and CPE, into SGs. Besides a disulfide-bonded hydrophobic loop that interacts with SG membrane (Kang and Yoo, 1997), the N-terminal region of CHGA also contains an amphipathic α-helix that may be necessary for sorting CHGA/hormone aggregates into SGs (Thiele and Huttner, 1998a; Taupenot et al., 2002; Laguerre et al., 2020). Similarly, proglucagon is sorted by α-helices present in the mature hormone domains rather than in the typical prohormone domain (McGirr et al., 2013; Guizzetti et al., 2014). Tissue-specific processing of proglucagon by PCs yields glucagon in pancreatic α cells (Furuta et al., 2001) or glucagon-like peptide 1 (GLP-1) and GLP-2 in intestinal L cells and hypothalamic neurons (Dhanvantari et al., 1996; Damholt et al., 1999; Holt et al., 2019). Although each of these products contains α-helices, only the non-amphipathic, dipolar α-helices on glucagon and GLP-1 efficiently target them to the regulated secretory pathway (Guizzetti et al., 2014). In contrast, GLP-2 has a more uniform negative charge distribution along the length of its α-helix (Guizzetti et al., 2014). PC1/3, PC2, and PC5/6A also contain α-helices in their C-termini that form electrostatic interactions and help sort proteins into the regulated secretion pathway (Assadi et al., 2004; Dikeakos et al., 2007b; Dikeakos et al., 2009). There seems to be no correlation between the helix length or the isoelectric point and sorting efficiency, but helices with a positive or negative charge and a hydrophobic segment seem the most effective in sorting (Dikeakos et al., 2007a), suggesting electrostatic interactions also play a role in aggregation, either inter- or intra-molecularly (Zhang et al., 2010). Indeed, Ma et al. (2008) identified four to five residues, two of which are charged, as the elementary sorting unit for protachykinin targeting to SGs. Clustering of these charged elementary units improved aggregation, leading to an additive and graduated effect that also improved sorting (Ma et al., 2008). In summary, charged amphipathic helices or non-amphipathic helices that have segregated charges and a hydrophobic patch are sufficient for targeting to secretory granules (Dikeakos et al., 2007a).

On the cytosolic side, specific motifs in the cytoplasmic domains of SG membrane proteins enable binding to adaptor proteins (Figure 5). The acidic-cluster-dileucine motif (DXXLL, where X is any amino acid) found in MPRs and sortilin, for example, is recognized by GGA proteins (Johnson and Kornfeld, 1992a; b; Chen et al., 1997; Nielsen et al., 2001; Puertollano et al., 2001). Serine and threonine residues embedded within cytosolic acidic clusters serve as substrates for casein kinase II (CKII) phosphorylation that enhances the sorting of proteins, like PAM, CI-MPR, and furin (Meresse et al., 1990; Jones et al., 1995; Chen et al., 1997; Steveson et al., 2001). In addition, VAMP4, phogrin, MPRs, and lysosomal proteins contain tyrosine-based (YXXΦ, where Φ is a bulky hydrophobic residue) and dileucine [(D/E)XXXL(L/I)] sorting motifs that are recognized by AP complexes (Glickman et al., 1989; Honing et al., 1997; Peden et al., 2001; Ghosh and Kornfeld, 2004; Torii et al., 2005; Wasmeier et al., 2005; Braulke and Bonifacino, 2009). These interactions between sorting motifs and adaptor proteins directs MPRs and their cargo proteins into TGN-derived, clathrin-coated vesicles that fuse with endosomes (Doray et al., 2002; Waguri et al., 2003; Braulke and Bonifacino, 2009).

In summary, several years of research uncovered how cells control the secretion of proteins and proposed two major models, sorting-at-entry and sorting by retention, to describe how proteins are sorted into the proper pathway at the TGN. Both models incorporate several sorting mechanisms that promote protein aggregation and binding to unique membrane lipids, leading to SG formation. After budding from the TGN, both constitutive and regulated secretory vesicles are transported to secretion sites at the plasma membrane via microtubule-based transport systems (Park and Loh, 2008).

Kinesin, dynein, and myosin are molecular motors that transport SeVs and SGs along microtubule or actin tracks. In general, SeVs and SGs use the same type of microtubule motor, such as kinesin, for anterograde transport to the secretion sites, and cytoplasmic dynein for retrograde transport back to the cell body (van den Berg and Hoogenraad, 2012). Tight regulation of the transport machinery is critically important to ensure that proteins are picked up and delivered to the right place at the right time. At the end of microtubule-based transport, SGs are transferred to the actin cortex close to the plasma membrane with the help of myosin V and the F-actin motor proteins (Rose et al., 2003; Rudolf et al., 2003). Actin and myosin have established themselves as key players in regulated secretion by providing tracks to target SGs to fusion sites, actively squeezing cargoes from fused vesicles, and following fusion, retrieving excess membrane to maintain cell surface area and recycle several SG-resident proteins (Rudolf et al., 2003; Rojo Pulido et al., 2011; Li et al., 2018; Miklavc and Frick, 2020). Besides its role as the transport platform for myosin motors, F-actin acts as a physical barrier for SG exocytosis and is also involved in the regulation of sorting at the TGN (Park and Loh, 2008; Gutierrez and Villanueva, 2018). During SG biogenesis, F-actin is recruited to the budding granule by actin-related protein-2/3 (ARP2/3) and myosin 1b (Delestre-Delacour et al., 2017). ARP2/3 complex binds to actin and exerts an active role in SG formation through its nucleation and branching activities that provide a structural or force-generating scaffold (Goley and Welch, 2006). The nucleation core activity of ARP2/3 is activated by nucleation promoting factors, such as members of the Wiskott-Aldrich syndrome family (WASP, N-WASP, WAVE, and WASH proteins) (Welch and Mullins, 2002; Alekhina et al., 2017).

The fine-tuned control of actin polymerization on endosomes is fundamental for the retrieval and recycling of several cargoes (Puthenveedu et al., 2010; Simonetti and Cullen, 2019). Retrieval and recycling are orchestrated by several multi-protein complexes, including retromer, commander/CCC/retriever, sorting nexins, and the ARP2/3-activating WASH complex (Simonetti and Cullen, 2019; MacDonald et al., 2020; Placidi and Campa, 2021). Recently, MAGEL2 regulation of the WASH complex and ARP2/3 activation was shown to prevent lysosomal degradation of SG-resident proteins and enable the proper neuroendocrine function of the hypothalamus, suggesting a tissue-specific regulation of compensatory endocytosis in regulated secretion (Figure 6) (Chen et al., 2020). In the next paragraphs, we will briefly summarize retromer-dependent retrograde recycling and its regulation.

Retromer is an evolutionarily conserved complex that regulates the retrograde pathway across all eukaryotes (Koumandou et al., 2011; McGough and Cullen, 2011; Chamberland and Ritter, 2017). In Saccharomyces cerevisiae, where retromer was first discovered, five vacuolar protein sorting (VPS) proteins compose the two retromer subcomplexes: the cargo-selective complex and the tubulation complex, both of which are conserved in higher eukaryotes (Seaman et al., 1998; Neuman et al., 2021). The cargo-selective complex is a trimer of Vps26, Vps29, and Vps35, which recruits cargo via an association between Vps35 and a sorting motif located within the cytoplasmic tail of cargo (Seaman et al., 1998). Retromer is recruited to the endosomal membrane by the sorting nexin (SNX) proteins Vps5 and Vps17 that form the tubulation complex and contain C-terminal Bin/amphysin/Rvs (BAR) domains that promote membrane tubulation and cargo vesicle formation (Horazdovsky et al., 1997; Nothwehr and Hindes, 1997; Seaman et al., 1998; Chamberland and Ritter, 2017).

Compared to the yeast retromer, the mammalian retromer is more complex and allows for more cargo specificity and transport regulation (Chen et al., 2019). In addition, the interactions between the retromer subcomplexes are more transitional and may not always occur in mammalian cells (Harbour and Seaman, 2011; Cullen and Steinberg, 2018). For example, both subcomplexes are needed to facilitate recycling of CI-MPR and other cargo in yeast cells (Yong et al., 2022), whereas the SNX-BAR dimer and not the VPS trimer is required for CI-MPR recycling in mammalian cells (Kvainickas et al., 2017; Simonetti et al., 2017), implying functional segregation of the two subcomplexes (Chamberland and Ritter, 2017). The acquisition of new binding partners in mammals that are not found in yeast provides further evidence of the functional divergence between the subcomplexes (Chamberland and Ritter, 2017). From now on in this manuscript, retromer will refer to the VPS26-VPS29-VPS35 heterotrimer core that is found in mammals (VPS26 has two isoforms in mammals).

The retromer complex serves as a hub for recruiting accessory proteins and complexes, such as receptor-mediated endocytosis-8 (RME-8) (Freeman et al., 2014), Eps15 homology domain-containing protein-1 (EHD1) (Gokool et al., 2007), TBC1D5 (Seaman et al., 2009), and the WASH complex (Kvainickas et al., 2017; Simonetti et al., 2017), that regulate retromer’s role in endosomal trafficking and endosomal tubule dynamics (Harbour et al., 2010). For example, VPS35 binds to SNX3 and the GTPase Rab7, leading to membrane recruitment of retromer (Burd and Cullen, 2014; Harrison et al., 2014). While SNX3 directs retromer to early endosomes by binding to phosphatidylinositol 3-phosphate [PtdIns(3)P] (Harterink et al., 2011), Rab7-GTP recruits retromer to late endosomes (Rojas et al., 2008; Progida et al., 2010; Harrison et al., 2014). TBC1D5, which binds to VPS35 and VPS29, negatively regulates membrane recruitment of retromer by acting as a GTPase-activating protein for Rab7 (Seaman et al., 2009; Jia et al., 2016; Borg Distefano et al., 2018). Another important accessory protein is the WASH complex protein FAM21, which binds directly to VPS35 through its C-terminal repeats of the LFa motif (Gomez and Billadeau, 2009; Harbour et al., 2012; Jia et al., 2012; Helfer et al., 2013; Chen et al., 2019). WASH, which is regulated by MAGEL2 and USP7, then nucleates actin on the membrane (Hao et al., 2013; Hao et al., 2015; Florke Gee et al., 2020). The interaction between WASH and the Prader-Willi protein MAGEL2 will be discussed further in the last section.

Retromer functions as a coat complex that packages and delivers its cargo via tubular or vesicular structures to the TGN or plasma membrane (Lucas and Hierro, 2017; Wang et al., 2018). Vesicles coated with retromer are defined as retromer-coated endosomal tubular carriers (ETCs). Compared to other protein coats (i.e., clathrin, COPI, and COPII), ETCs are much more heterogeneous with a “loose” assembled coat, possibly an adaptation to different membrane curvatures (Chen et al., 2019). Retromer coat assembly depends on SNX proteins in mammals, as retromer does not possess intrinsic membrane-binding properties (Burd and Cullen, 2014; Chen et al., 2019). Different combinations of SNX proteins and retromer are important for recycling specific proteins (Gallon and Cullen, 2015; Chamberland and Ritter, 2017; Yong et al., 2022).

The SNX protein family expanded from 10 proteins in yeast to 33 in mammals and six of them (SNX1, SNX2, SNX3, SNX5, SNX6, and SNX27) were shown to associate with the retromer complex (Cullen, 2008; Burd and Cullen, 2014; Lucas and Hierro, 2017). SNX1 and SNX2 are Vps5 orthologs, and SNX5 and SNX6 are Vps17 orthologs (McGough and Cullen, 2011). All SNX proteins possess a Phox (PX) domain that binds to phospholipids, in particular PtdIns(3)P, found in endosomes (Bugarcic et al., 2011; Chamberland and Ritter, 2017; Wang et al., 2018). SNX proteins are divided into subfamilies based on their domain architecture (Wang et al., 2018). SNX-PX subfamily members, like SNX3, only possess a PX domain. The SNX-FERM (4.1/ezrin/radixin/moesin) subfamily member SNX27 contains a FERM domain, which can bind to PtdIns(3)P and NPxY motifs, and a PDZ domain, which facilitates protein-protein interactions (Zhang et al., 2018). The SNX-BAR subfamily members contain a BAR domain that is capable of sensing and inducing membrane curvature (Burd and Cullen, 2014). The retromer-interacting proteins from the SNX-BAR subfamily include SNX1, SNX2, SNX5, and SNX6 (Lucas and Hierro, 2017). Although most SNX proteins associate with early endosome-enriched PtdIns(3)P, some also interact with late endosome-enriched phosphatidylinositol 3,5-phosphate [PtdIns(3,5)P]. SNX-PX-retromer and SNX-BAR-retromer mediate retrograde transport to the TGN, and SNX27-retromer mediates recycling to the plasma membrane through its PDZ domain (Temkin et al., 2011; Steinberg et al., 2013; Lucas and Hierro, 2017). Besides binding phosphatidylinositides, SNXs also play a central role in cargo recognition (Wang et al., 2018).

Retromer cargo selection is facilitated directly by VPS35 and VPS26 and/or by adaptor proteins, like SNXs, AP-1, and GGAs (Burd and Cullen, 2014; Cullen and Steinberg, 2018). The cargo proteins CI-MPR, sortilin, Wntless, and polymeric immunoglobulin (pIg) receptor possess a ΦX(L/M/V), where Φ is F/Y/W, consensus sequence that facilitates direct binding to SNX3-retromer (Verges et al., 2004; Seaman, 2007; Canuel et al., 2008; Harterink et al., 2011; Harrison et al., 2014; Lucas and Hierro, 2017; Cui et al., 2019; Yong et al., 2022). In general, at least a hydrophobic (F/W)L(M/V) motif that is commonly present in cargo proteins is needed for retromer-mediated sorting (Seaman, 2007; Cullen and Steinberg, 2018). This direct retromer interaction pathway is likely conserved in all eukaryotes. In contrast, metazoan-specific adaptor-dependent sorting allowed for the evolution of a plethora of sorting signals to finetune retrograde trafficking (Cullen and Steinberg, 2018). For example, a bipartite motif in VPS10, the functional homolog of CI-MPR in yeast, is needed for precise recognition by yeast retromer subunits Vps26 and Vps35 but not Vps17 (Suzuki et al., 2019). However, mammalian SNX5/SNX6 (yeast orthologs of Vps17) may function as the cargo-selecting module that recognizes a bipartite motif in CI-MPR and many other cargo proteins (Kvainickas et al., 2017; Simonetti et al., 2017; Simonetti et al., 2019; Yong et al., 2020). As another example, metazoan-specific SNX27 acts as an adaptor protein to select cargo through its FERM and PDZ domains (Cullen and Steinberg, 2018; Chen et al., 2019). Examples of SNX27 cargo containing a PDZ-binding motif [i.e., (S/T)xΦ] include β2-adrenergic receptor, parathyroid hormone receptor 1 (PTHR), α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor, and the N-methyl-D-aspartate (NMDA) receptor, which are all important in neuroendocrine tissues (Temkin et al., 2011; Steinberg et al., 2013; Lucas and Hierro, 2017). Some of these proteins, particularly their sorting motifs, must be phosphorylated to facilitate high-affinity binding to SNX27, suggesting additional complexity in regulation and finetuning of retrograde transport (Clairfeuille et al., 2016; Lucas and Hierro, 2017).

An essential part of endosomal recycling is membrane remodeling, which is mediated by patches of branched F-actin, and requires membrane localization of the retromer, WASH, and ARP2/3 complexes (Fokin and Gautreau, 2021). Endosomal cargo proteins are recognized by specific combinations of retromer and SNXs (Gallon and Cullen, 2015; Chamberland and Ritter, 2017; Yong et al., 2022). SNX proteins localize the retromer complex to the membrane, and FAM21 binds to retromer’s VPS35 to facilitate membrane localization of the WASH complex (Harbour et al., 2010; Bugarcic et al., 2011; Harbour et al., 2012; Jia et al., 2012; Helfer et al., 2013; Seaman and Freeman, 2014; Chamberland and Ritter, 2017). A recent study indicates that SWIP/KIAA1033 can also recruit the WASH complex to the endosomal membrane independently of the FAM21-VPS35 interaction (Dostal et al., 2023). The endosomal membrane recruitment of WASH complex may also be mediated through interaction with SNX27 (Temkin et al., 2011; Steinberg et al., 2013) or its own direct interaction with endosomal lipids (Derivery et al., 2009).

The WASH complex functions as the major actin nucleation-promoting factor in endosomal recycling and is required for the recycling of several proteins, including CI-MPR, glucose transporter 1 (GLUT1), α5β1 integrin, and major histocompatibility complex II (MHC II) (Gomez and Billadeau, 2009; Zech et al., 2011; Piotrowski et al., 2013; Graham et al., 2014; Dostal et al., 2023). WASH possesses a conserved C-terminal VCA (verprolin homologous or WASP-homology-2, connector, and acidic) domain that binds actin monomers and ARP2/3 to trigger actin filament nucleation (Linardopoulou et al., 2007; Duleh and Welch, 2010). More specifically, binding of ARP2/3 to WASH’s VCA domain induces a conformational change in ARP2/3 that loads the first actin monomer and allows further actin polymerization through rapid ATP hydrolysis on ARP2 (Derivery et al., 2009; Padrick et al., 2011; Espinoza-Sanchez et al., 2018). Prior to actin nucleation, WASH’s VCA domain is autoinhibited by intra- and intermolecular interactions (Hao et al., 2013). Small GTPases and PtdIns(4,5)P are some general regulators that may release VCA motif inhibition, but tissue-specific machinery, like the MUST complex (composed of MAGEL2, TRIM27 and USP7) might have evolved to enable better adaptation to specific environmental challenges (Figure 6) (Fon Tacer and Potts, 2017; Lee and Potts, 2017; Florke Gee et al., 2020).

The WASH complex promotes retromer trafficking by triggering actin polymerization via ARP2/3 complex recruitment and activation. Actin polymerization combined with the action of SNX-BAR proteins and motor proteins ultimately leads to the formation of tubular structures (Fokin and Gautreau, 2021). Subsequent actin polymerization and the activity of the dynein-dynactin complex leads to the fission of tubular vesicles that carry various cargo proteins to their final destinations (Hunt et al., 2013; Wang et al., 2018; Fokin et al., 2021). This pathway is well-established for diverse constitutively secreted receptors; however, recent evidence suggests that essential SG membrane proteins and granins are recycled back to the TGN for nascent granule formation by retromer-dependent retrograde transport, though the mechanistic details and sorting motifs are still more or less unknown (Ferraro et al., 2005; Ma et al., 2020; Neuman et al., 2021). Interestingly, Prader-Willi mouse models with depletion of Magel2 suggested that, in the hypothalamus, SG-resident proteins are the major targets of Magel2-retromer-dependent protein recycling (Figure 6) (Fon Tacer and Potts, 2017; Lee and Potts, 2017; Florke Gee et al., 2020).