Mannose-6-phosphate receptors (MPRs) transport lysosomal hydrolases from the trans Golgi network (TGN) to endosomes. Recently, the multi-ligand receptor sortilin has also been implicated in this transport, but the transport carriers involved herein have not been identified. By quantitative immuno-electron microscopy, we localized endogenous sortilin of HepG2 cells predominantly to the TGN and endosomes. In the TGN, sortilin colocalized with MPRs in the same clathrin-coated vesicles. In endosomes, sortilin and MPRs concentrated in sorting nexin 1 (SNX1)-positive buds and vesicles. SNX1 depletion by small interfering RNA resulted in decreased pools of sortilin in the TGN and an increase in lysosomal degradation. These data indicate that sortilin and MPRs recycle to the TGN in SNX1-dependent carriers, which we named endosome-to-TGN transport carriers (ETCs). Notably, ETCs emerge from early endosomes (EE), lack recycling plasma membrane proteins and by three-dimensional electron tomography exhibit unique structural features. Hence, ETCs are distinct from hitherto described EE-derived membranes involved in recycling. Our data emphasize an important role of EEs in recycling to the TGN and indicate that different, specialized exit events occur on the same EE vacuole.
In mammalian cells, most soluble lysosomal hydrolases in the Golgi complex are equipped with a mannose-6-phosphate moiety. In the trans Golgi network (TGN), these phosphorylated enzymes are recognized by the cation-dependent (CD)- and cation-independent (CI)-mannose-6-phosphate receptors (MPRs) and efficiently sorted to endosomes. In contrast, yeast Saccharomyces cerevisiae vacuolar hydrolases do not contain mannose-6-phosphate and are mostly sorted by Vps10p (1), a protein receptor unrelated to MPRs. Vps10p recognizes specific, non-conserved amino acid sequences in the hydrolases (1,2) and continuously cycles between the Golgi complex and endosomes. Recently, a receptor family sharing homology to the Vps10p luminal domain has been identified in mammalian cells, which currently comprises sortilin, SorLAs and SorCSs. The multi-ligand receptor sortilin has been implicated in a variety of functions (3–5), all related to events taking place at the plasma membrane. The major pool of sortilin, however, resides inside the cell. There, it functions in different trafficking routes (3,6,7), including the delivery of sphingolipid activator proteins (SAPs) and the acid sphingomyelinase to lysosomes (8,9). Thus, in addition to MPRs, sortilin could play a role in TGN-to-endosome trafficking of proteins destined for lysosomes.
Anterograde transport of MPRs from the TGN to endosomes requires Golgi-localized, gamma-ear-containing ADP ribosylation factor (ARF)-binding proteins (GGAs), the adaptor protein (AP)-1, the GTPase Arf-1 and clathrin (10–12), resulting in clathrin-coated transport vesicles (CCVs). After releasing their cargo in the endosomes, MPRs return to the TGN for another round of transport. Multiple retrograde transport machineries have been implicated in MPR retrieval from endosomes. These include AP-1 (13), Rab9 and the tail-interacting protein of 47 kD (TIP47) (14), phosphofurin acidic cluster-sorting protein-1 (PACS-1) (15), the t-SNARE syntaxin 16 (16), the clathrin and AP-1-interacting protein EpsinR (17) and the retromer complex (18,19). Presumably, these players act in more than one retrograde pathway emerging at different places in the endocytic system.
The retromer complex was initially identified in yeast where it is involved in Vps10p recycling from endosomal compartments back to the Golgi (20). The yeast retromer consists of five subunits organized into two subcomplexes (21). The first is formed by Vps35p, Vps26p and Vps29p and is implicated in cargo recognition and assembly of subcomplexes. The second comprises Vps17p and Vps5p and is thought to drive vesicle formation through its ability to assemble onto endosomal membranes. The retromer subunits are conserved in mammalian cells (22,23). The mammalian orthologue of Vps5p is sorting nexin-1 (SNX1), a member of the SNX protein family that is characterized by a phosphoinositides-binding phox homology (PX) domain (24). The SNX1-PX domain can bind to phosphatidylinositol-3-phosphate and phosphatidylinositol-3,5-bisphosphate (25), which are lipids enriched in early endosomes (EEs) and late endosomes (LEs), respectively (26).
In contrast to the MPRs, the intracellular transport routes used by sortilin have remained largely mysterious. A chimeric protein consisting of the luminal part and the transmembrane domain of CI-MPR and the cytoplasmic C-terminus of sortilin can restore transport of mannose-6-phosphate (M6P)-lysosomal enzymes in MPR-deficient cells, indicating that the sortilin tail can restore the MPR phenotype (27). Within the sortilin C-terminus, an amino acid motif [DDSD, (27)] is present that is similar to the MPRs in which it mediates binding to AP-1 (28). In addition, the sortilin tail contains a GGA-binding motif, essential for sortilin trafficking and correct sorting of SAPs (8,27). Together, these studies suggest that sortilin could exit the TGN through CCVs, however direct evidence thereof is lacking. It remains unknown whether MPRs and sortilin can share the same carrier vesicles and if and how sortilin recycles back from endosomes to TGN.
To identify the transport carriers and cellular pathways involved in sortilin transport, we analyzed the precise subcellular distribution of endogenous sortilin in direct comparison with that of CD- and CI-MPR and molecular machinery proteins. We demonstrate that the unrelated sortilin and MPRs have identical intracellular trafficking pathways and share identical CCVs to exit the TGN. Moreover, we characterize a novel, SNX1-dependent transport carrier – the endosome-to-TGN transport carrier (ETC) – specific for retrograde transport of lysosomal protein receptors, which is different from the previously described tubular-sorting endosome (TSE) (29) or tubular endosomal network (TEN) (30) that also emerges from EEs. Together, our findings emphasize an important role for EEs in recycling to the TGN and demonstrate that distinct specialized carriers can emerge from the same endosomal vacuole, which adds to our understanding of the functional architecture of EEs.
HepG2 cells express endogenous sortilin
In order to investigate sortilin transport in direct relation to the MPR trafficking pathways, we needed to identify a cell system in which these three proteins could be monitored at endogenous levels. MPRs are ubiquitously expressed, whereas sortilin is mostly restricted to brain, skeletal muscle, spinal cord and testis (28). In addition, sortilin is expressed in cancers in which growth is activated by neurotensin, including fibrolamellar hepatocarcinoma (31,32). The human hepatocarcinoma cell line HepG2, in which we previously established CI- and CD-MPR distributions (33,34), belongs to this family. To determine whether sortilin is present in HepG2 cells, we performed a Western blot using an antibody against the luminal (Figure S1A) or cytoplasmic (data not shown) portion of sortilin. A single band with a molecular weight of 100 kD was detected with the same weight as full-length sortilin overexpressed in HeLa cells (Figure S1A). This result demonstrates that HepG2 cells express significant levels of sortilin as well as CD- and CI-MPR, providing us a suitable model system for our studies.
Sortilin and MPRs have identical subcellular distributions in HepG2 cells
To investigate the sortilin transport routes in relation to MPRs, we first monitored the overall cellular distributions of these proteins by immunofluorescence (IF). For MPR localizations, we focused our studies on CI-MPR, which in HepG2 cells has a largely overlapping distribution with CD-MPR (33,34). By IF, CI-MPR extensively colocalized with sortilin in the perinuclear region and in punctate structures distributed throughout the cytoplasm (Figure S1B), indicating a similar overall distribution pattern. Because IF does not provide the necessary resolution to determine whether proteins are present in the same membrane within a given organelle, we prepared ultrathin cryosections that were double-immunogold labelled for sortilin and CI-MPR [immuno-electron microscopy (immuno-EM)] to determine the transport carriers, organelles and organellar subdomains involved in MPR and sortilin transport. The immuno-EM localization of sortilin and CI-MPR revealed a strikingly similar distribution at the subcellular level: the two proteins localized to the same intracellular compartments (Figures 1 and 2). Moreover, by counting the distribution of gold particles over the distinct compartments (quantitative immuno-EM), we established that the two receptors had almost identical, relative subcellular distributions (Table 1). A first major pool of both receptors (∼35%) was found in the TGN, defined as the tubular membranous network within a distance of less than 400 nm from the trans Golgi cisterna (35). The second major pool (∼55%) was present on endosomes, defined as endosomal vacuoles plus associated vesicles and tubules (Table 1, columns 1 and 2). We conclude from these data that the two unrelated receptors sortilin and CI-MPR are similarly distributed over the same subcellular compartments.
Table 1. Relative distribution of sortilin and CI-MPR in HepG2 cells: effects of SNX1 depletiona
Number of gold particles per cell profile
Cryosections of control and SNX1-depleted HepG2 cells were double-immunogold labelled for sortilin (10 nm gold) and CI-MPR (15 nm gold). Numbers represent the percentage of total gold ± SEM on the listed compartments. The percentage of total sortilin present in the TGN decreased from 37 to 24% (p = 0.0038) upon depletion of SNX1. Conversely, the amount of sortilin in ETCs slightly increased from 37 to 41% (p = 0.006). Moreover, there was a slight but significant (p = 0.034) increase of sortilin at the plasma membrane in SNX1-suppressed cells.
36.8 ± 2.24
56 ± 3.31
19.4 ± 1.8
36.6 ± 1.30
7.4 ± 0.91
27.7 ± 2.68
33.7 ± 1.10
55.8 ± 1.85
24.0 ± 1.04
31.8 ± 2.97
10.5 ± 0.82
19.6 ± 1.79
SNX1 RNA interference
23.6 ± 2.75
65.1 ± 1.98
24.3 ± 3.95
40.8 ± 2.58
11.3 ± 1.36
11.6 ± 1.38
13.6 ± 2.63
61.8 ± 2.78
27.4 ± 0.61
34.4 ± 1.23
25.2 ± 1.21
9.32 ± 0.77
Sortilin and MPRs are incorporated into the same TGN-derived CCVs
To investigate if sortilin and MPRs use the same transport carriers to leave the TGN, we performed a series of double-immunogold labelling with sortilin, MPRs and the various transport machinery proteins. Others and we have previously found that export of CI-MPR from the TGN requires AP-1, GGA and clathrin (11,12,36,37). Within the TGN area, sortilin and CI-MPR colocalized in vesicles displaying the characteristic clathrin coat (Figure 1A) (34). The identity of this coat was confirmed with anti-clathrin antibodies (data not shown). Quantification of the sortilin–CI-MPR double-labelling showed a high degree of colocalization: ∼40% of the sortilin-positive, TGN-associated CCVs also contained CI-MPR. Very similar data were obtained by double-labelling for sortilin and CD-MPR (data not shown). Importantly, the sortilin-containing CCVs at the TGN were also positive for GGA3 (Figure 1B) and AP-1 (Figure 1C), a feature also shared by MPR-positive CCVs (34). All together, these data provide strong evidence for a CCV-mediated exit of sortilin from the TGN and show that sortilin and MPR can share a single CCV.
Sortilin is evenly distributed over EEs and LEs
Because sortilin and CI-MPR exit the TGN in the same carriers, they likely enter the endocytic system at the same point. For MPRs, it is well established that upon entering the endosomes, they release their ligand and then recycle back to the TGN (38,39). For sortilin, this has not been addressed yet. As mentioned above, more than half of sortilin and MPRs are present in endosomes (Table 1), and we found that about 20% of both sortilin and CI-MPRs decorated the same endosomal vacuoles (Table 1, column 2a; Figure 2b). Within these endosomes, we also found sortilin together with its known ligand, prosaposin (a SAP protein family) (40) (data not shown). To determine the endosomal distribution of sortilin in more detail, we developed a method to distinguish EEs and LEs by a combination of morphological, kinetic and molecular criteria. By definition, EEs are accessible to transferrin (Tf) and reached by endocytic tracer after a short incubation time BSA-gold internalized for 10 min (data not shown). LEs are inaccessible to endocytosed Tf (Figure S2A) and reached by internalized BSA-gold only after 15 min (41,42). It has also been shown that an increasing number of intraluminal vesicles reflects the endosomal maturation process (43). With these notions in mind, we correlated the number of intraluminal vesicles (i.e. within the endosomal vacuole) with the occurrence of internalized Tf (Figure S2A) and BSA-gold (data not shown). This analysis showed that in HepG2 cells, Tf-positive EE vacuoles on average contained one to eight intraluminal vesicles, whereas endosomal vacuoles with nine or more intraluminal vesicles were mostly devoid of Tf and therefore designated as LE (Figure S2A). To further investigate whether the number of intraluminal vesicles could be used as a reliable criterion to distinguish between EEs and LEs, we subsequently performed a similar analysis with various marker proteins of the EE (Figure S2A). We found that all early endosomal marker proteins are exclusively or predominantly localized in endosomes with up to eight intraluminal vesicles. Based on this analysis, we used in our further studies the number of internal vesicles as a rapid method to morphologically distinguish EEs from LEs.
By counting the number of sortilin gold particles over endosomal vacuoles classified in this manner, we found that sortilin was evenly distributed over EE and LE vacuoles (10.9 ± 2.03% and 8.5 ± 1.85%, respectively). In contrast, lysosomes, identified by an extensive labelling for CD63 (Figure 2D, inset) and lysosome associated membrane protein 1 (LAMP1) (data not shown), were devoid of sortilin. These observations precisely mirrored previous data obtained for the MPRs (34). Taken together, these results show that sortilin and CI-MPR are evenly distributed over EEs and LEs in which they also colocalized.
ETCs recycle sortilin and MPRs to the TGN in a SNX1-dependent way
The largest pool of endosome-associated sortilin was present in a characteristic population of electron-dense tubules and vesicles surrounding or in connection with the endosomal vacuoles. We will further refer to these membranes as endosome-to-TGN transport carriers (ETCs) (Table 1, column 2b). MPRs were also enriched in ETCs (Table 1; Figure 2B, C and E), again displaying a high level of colocalization with sortilin: ∼40% of the sortilin-positive ETCs also contained CI-MPR. However, ETCs were devoid of lysosomal proteins, such as CD63 (Figure 2D), LAMP1 and LAMP2 (data not shown). Notably, also prosaposin and cathepsin D, typical sortilin and MPR ligands, respectively, were absent from ETCs (data not shown), which we took as a first indication that ETCs are involved in retrograde rather than anterograde transport to endosomes.
ETCs are highly reminiscent to the so-called ‘endosome-associated vesicles’ previously described to contain SNX1 (44) and SNX2 (45), two components of the mammalian retromer complex involved in the endosome-to-TGN recycling of CI-MPR. Because of the intimate sortilin and CI-MPR colocalization, we hypothesized that ETCs might also mediate retrograde transport of sortilin from the endosomes back to the TGN. To address this question, we first performed double-immunogold labelling for sortilin and SNX1. In favour of our hypothesis, the two proteins extensively and almost exclusively colocalized to ETCs (Figure 3A,B). Similar results were obtained with SNX2 (Figure 3C).
To directly assess a role for SNX1 in sortilin recycling, we suppressed SNX1 expression by using specific small interfering RNAs (siRNAs). If SNX1 is involved in sortilin recycling to the TGN, it is anticipated that SNX1 depletion leads to retention of sortilin in endosomes, resulting in an increased degradation in lysosomes. A similar phenotype was previously reported for CI-MPR (44). Accordingly, we tested sortilin levels in mock- and SNX1-depleted HepG2 cells by Western blot. We found that the total amount of sortilin was decreased by 50% in SNX1-depleted cells when compared with the control (Figure 4A). When we blocked protein synthesis with cycloheximide, sortilin levels in the SNX1-suppressed cells were only slightly more decreased than in the absence of this inhibitor (Figure 4A and B), indicating that SNX1 depletion does not cause a decrease in sortilin synthesis. Importantly, treatment with lysosomal protease inhibitors restored sortilin levels in SNX1-depleted cells, unequivocally demonstrating that its reduced amount is because of mistargeting and degradation in lysosomes. These data are in agreement with a role for SNX1 in recycling of sortilin from endosomes.
Sortilin mistargeting after SNX1 depletion was also seen by IF. In mock-treated cells, sortilin was concentrated in the perinuclear area (probably representing the TGN pool) and in puncta dispersed throughout the cytoplasm (Figure 4C). SNX1 depletion resulted in a dramatic redistribution towards the dispersed puncta, whereas the perinuclear labelling was decreased (Figure 4C,D). CI-MPR was affected in the same manner. Remarkably, however, despite this massive redistribution, the two proteins continued to display a high degree of colocalization (Figure 4D, inset). Notably, the localization pattern of the TGN marker TGN46 was normal in SNX1-depleted cells (Figure 4D), indicating that the effects on sortilin and CI-MPR distribution were not because of TGN dispersion but a specific failure of bringing back these receptors to the TGN.
Finally, we monitored the effects of SNX1 depletion on sortilin and CI-MPR localization by quantitative immuno-EM. As a general observation, and in agreement with our Western blots, we noticed a significant decrease in sortilin and CI-MPR labelling levels in SNX1-depleted cells: the overall numbers of sortilin and CI-MPR gold particles were decreased by ∼42 and ∼47%, respectively, in the SNX1-depleted cells when compared with that of the control (Table 1). Of the remaining labelling, the percentage of total sortilin present in the TGN had decreased significantly (Table 1, column 1), which is in agreement with a decreased recycling from endosomes. In accordance herewith, we found a slight increase in ETCs (Table 1, column 2b) and at the plasma membrane (Table 1, column 3). Notably, whereas the number of endosomal vacuoles in SNX1-depleted cells was basically unmodified, the number of ETCs was reduced by almost half (Table S1, column 2) and the remaining ETCs were seen about twice as much in a direct connection with endosomal vacuoles (Table S1, column 3; Figure S3). Taken together, these data show that SNX1 depletion causes an impairment of ETC formation at the endosomal-limiting membrane as well as a probable impairment of their processing, resulting in a decreased endosome-to-TGN recycling of sortilin and MPRs.
ETCs are a new class of recycling compartments distinct from the TSEs
Recent studies have shown that EEs can give rise to an extensive TEN or TSE in which proteins destined for recycling to distinct cellular destinations, i.e. plasma membrane, lysosomes, TGN and specialized storage vesicles (29,30), are sorted to multiple exits. These TSE/TEN membranes amongst others give rise to the well-characterized recycling endosomes, mediating recycling to the plasma membrane. Because TSE/TEN also contains low levels of CI- and CD-MPR, we needed to establish the relationship between ETCs and the TSE/TEN.
First, we determined if ETCs are preferentially associated with either EEs or LEs. Hereto, we randomly counted the number of ETCs present in close proximity (≤100 nm) of EE and LE vacuoles (characterized by their number of internal vesicles, see above). These countings showed that ETCs are most often found around EE vacuoles (Figure 5A). The number of ETCs increased proportionally and significantly with the number of intraluminal vesicles, reaching a maximal density around EE vacuoles containing seven internal vesicles. Interestingly, when analyzing SNX1 and sortilin distribution, we observed that both proteins were mostly concentrated in the same EE vacuoles (four to seven internal vesicles) (Figure 5B). Thus, the sortilin-positive ETCs exit the endocytic pathway primarily from EE vacuoles, just before their maturation into LE vacuoles.
To further confirm that ETCs derive predominantly from EE vacuoles, we performed double-immunogold labelling between sortilin and various EE marker proteins. This showed that sortilin-positive endosomal vacuoles with associated ETCs were also positive for the specific EE markers early endosome antigen 1 (EEA1), Hrs, Rab11 and Rab4 (Figure S2B). Remarkably, the ETCs per sé were not labelled by any of these EE marker proteins (data not shown). We took this as a first indication that ETCs are different from the TSE/TEN, to which Rab4 and Rab11 have been localized (29,30).
The TSE/TEN consists of elongated, branched, partially clathrin-coated, tubular membranes. To establish the architecture of ETCs, we performed three-dimensional (3D) electron tomography and made reconstructions of endosomal vacuoles exhibiting ETCs. This revealed that ETCs are non-branched tubules (on average 170–230 nm in length and 20–50 nm in diameter) and vesicles (with a 30–50 nm diameter) (Figure 6). Remarkably, a given EE vacuole can display multiple ETCs (Figure S4). These observations indicate that ETCs are small, electron-dense, non-branched tubules. Like the long, tubular-branched TSE/TEN, they originate directly from EE vacuoles, but they are of an entirely different shape and size.
To further investigate the relationship between ETCs and TSE/TEN, we next investigated whether ETCs contain proteins known to be recycled to the plasma membrane. Both the asialoglycoprotein receptor (ASGPR) and transferrin receptor (TfR) use this pathway in HepG2 cells (46). We therefore performed a series of immunogold double-labelling for sortilin, ASGPR and internalized Tf. These studies revealed that sortilin and Tf can occasionally be found at the limiting membrane of the same EE vacuoles. However, Tf prevailed in vacuole-associated tubules of the TSE/TEN that were morphologically and qualitatively different from the sortilin-positive ETCs (Figure 7A). We found similar results for sortilin and ASGPR (Figure 7B). Only 5.1% of the endosomal ASGPR was present in the ETCs compared with 49.8% of sortilin (Table S2). These results demonstrate that during their itinerary, Tf, ASGPR and sortilin can pass through the same EE vacuoles but exit this organelle through different recycling tubules, i.e. the TSE/TEN and ETCs, respectively.
All together, our analyses prove that ETCs are a novel tubular/vesicular carrier mediating the specific retrograde transport of lysosomal protein-sorting receptors from EEs. In addition, they reveal that segregation and sorting for different exits from the EE vacuole can occur at the limiting membrane of the same EE vacuole, implying the existence of specialized sorting subdomains.
In this paper, we provide the first detailed intracellular localization of sortilin, a receptor that has been implicated in the transport of a set of lysosomal hydrolases and their cofactors from the TGN to endosomes and lysosomes. We provide evidence that sortilin is a recycling receptor that together with the MPRs travels from TGN to endosomes through CCVs and from endosomes back to the TGN through a novel class of recycling carriers, the ETCs. We show that ETCs form in a SNX1-dependent manner on EE vacuoles and to a much lesser extent on LE vacuoles. Moreover, we show that ETCs bud from the same endosomal vacuoles that give rise to the previously characterized TSE/TEN, indicating the existence of a hitherto unknown sorting event at the EE-limiting membrane (See Figure 8 for summarizing model).
By both IF and immuno-electron microscopy, we found that sortilin, CI- and CD-MPR colocalized throughout the cell, i.e. in TGN, endosomes and ETCs. These observations confirm and extend previous fluorescence data (28,47). The colocalization was very tight: we did not observe compartments that exclusively hosted sortilin or MPR. Moreover, quantification of the immunogold labelling showed that also the relative distributions of sortilin and CI-MPR were similar, suggesting that sortilin and MPR not only share the same transport pathways but also do so with similar transport kinetics. This is surprising because although both sortilin and MPRs are implicated in the transport of lysosomal proteins, they are from unrelated protein families.
At the TGN, sortilin and CI-MPR were present in the same AP-1- and GGA-positive CCVs. We only show colocalization of sortilin with GGA3 because the endogenous levels of GGA1 and GGA2 were too low for IEM detection. However, all three GGAs are known to co-operate and colocalize to the same AP-1-positive CCVs at the TGN (36,48). Moreover, GGA1 and GGA2 interact with sortilin in vitro(27,49). The presence of a putative AP-1-binding motif in the cytosolic tail of sortilin (28,50) as well as the localization of sortilin to AP-1 positive CCVs (this study) are both in agreement with a role of AP-1 in sortilin sorting. However, sorting of sortilin in AP-1-positive CCVs could also be mediated through the GGAs, whereas recruitment of AP-1 occurs through MPR together with GGA (10,51). Thus, although the exact sorting process of sortilin at the TGN has still to be resolved, our present data indicate that sortilin does exit the TGN through the same CCV as the MPRs.
We found about 20% of the cell’s sortilin and CI-MPR in ETCs that form at the later stages of EE vacuoles and to a lesser extent at the early stages of LE vacuoles. Based on several observations, we propose that ETCs represent recycling structures that specifically recruit sortilin and MPRs. First, SNX1 and SNX2, two subunits of the retromer (44,45,52), specifically localize to ETCs, a finding in agreement with a previous study where SNX1 was shown to be involved in MPR recycling from endosomes to the TGN (44). Moreover, it was recently shown that CI-MPR and sortilin share the T-L-M tail-motif required for interaction with the retromer (53). Second, sortilin and MPRs are specifically enriched in ETCs, but their cargo proteins (cathepsin D and prosaposin) are not present in these carriers. Third, suppression of SNX1 by siRNA induced a shift in the sortilin distribution from the TGN to endosomal compartments and an ∼50% reduction of the overall sortilin levels in the cell. This reduction was prevented by lysosomal protease inhibitors and largely independent of cycloheximide treatment, indicating that SNX1 depletion causes impaired recycling of sortilin from endosomes resulting in an increased lysosomal degradation (44). Fourth, ETCs did not concentrate plasma membrane receptors or lysosomal membrane proteins. Together, these data indicate that ETCs specifically mediate retrograde transport of sortilin and MPRs from EEs to the TGN in a SNX1-dependent manner. However, it remains to be established whether this transport involves a direct fusion of ETCs with TGN membranes or whether an as yet unidentified intermediate compartment is involved (44,54,55).
To the best of our knowledge, this is the first report demonstrating the involvement of a retromer component in the recycling of endogenous sortilin. Our data are in agreement with previous studies showing that depletion of Vps26, another component of the retromer, affects endosome-to-TGN recycling of an ectopically expressed CD8/sortilin–C-terminus chimaera (54). However, although both SNX1 and Vps26p are involved in CI-MPR retrieval (44,54), it remains a question whether the entire retromer participates in the ETC pathway because SNX1 can also perform sorting functions that are independent of Vps26 (56), and it remains as yet uncertain whether the Vps26 and SNX1 localizations completely overlap ((45) and our preliminary observations). Moreover, depletion of Vps26 results in fragmentation of the Golgi and TGN (54), while we did not observe such an effect upon SNX1 depletion. Recently, both SNX1 and Vps26 have been implicated in the endosome-to-TGN transport of an endocytosed protein; the Shiga toxin B subunit (55). By whole-mount immuno-EM, Vps26 was colocalized with clathrin on Shiga toxin B-positive endosomal tubules, which is in accordance with previous studies implicating a role for clathrin in this pathway. Because ETCs do not bear a clathrin coat, the Vps26/clathrin exit differs from ETCs. Taken together, further studies are required to establish whether Vps26 and other retromer components are functionally involved in ETC formation and if and under which conditions retromer components can localize to different endosomal exits.
In SNX1-depleted cells, the reduced number of ETCs still contained sortilin, while the number of ETCs connected to EE vacuoles had doubled. These data indicate that in the absence of SNX1, CI-MPR and sortilin can still be concentrated in putative ETC-forming domains at the endosomal vacuole but that growth and fission of ETC are impaired or slowed down. In yeast, Vps35p together with Vps29p and Vps10p drives cargo selection, whereas Vps17p together with the SNX1 homologue Vps5p drive vesicle budding. Likewise, in mammalian cells, in vitro studies showed that Vps35 can bind the cytoplasmic tail of CI-MPR (18), whereas SNX1 is not involved in cargo binding but by its C-terminal BAR domain binds to high-curvature membrane profiles and induces membrane curvature (44,57). These data are consistent with our observations that in SNX1-depleted cells, sortilin and CI-MPR can still concentrate at specific subdomains of the endosomal vacuoles but that the completion of ETC formation is strongly reduced.
EEs are long known for their involvement in the recycling of plasma membrane receptors through the formation of recycling tubules that bud from the EE vacuole to form the pericentriolar recycling endosome that mediates receptor recycling to the plasma membrane (reviewed by 43). Recently, we found that proteins as different as the plasma membrane receptors ASGPR and TfR, the lysosomal membrane proteins LAMP1, LAMP2 and CD63 and to a lower extent, MPRs are all present within the same EE-associated tubules (29). Moreover, these tubules display many different adaptor and coat proteins (AP-1, AP-3 and clathrin), suggesting that from these tubules, proteins are sorted to distinct destinations of the cell (29,30,58,59). Hence, they were proposed to be renamed as tubular-sorting endosome (TSE) (29) or tubular endosomal network (TEN) (30). Segregation of proteins for these different destinations may eventually lead to the formation of differentiated subdomains, like the pericentriolar recycling compartment (30).
ETCs are undoubtedly not part of the TSE/TEN but constitute an entirely different population of recycling carriers (Figure 8). ETCs – with the exception of the MPRs – are devoid of any of the above-mentioned proteins of TSE/TEN. The only machinery proteins present on ETC were SNX1 and SNX2. This is in agreement with previous fluorescence microscopy and EM data showing that SNX1 and ASGPR do not colocalize with TfR and CD-MPR, respectively (34,60). Moreover, we did not encounter formation of ETCs from TSN/TEN. Finally, our 3D tomographic analyses clearly illustrate that ETCs are morphologically different from the TSE/TEN as they are non-branched, short tubules and vesicles not organized in a reticulum.
The presence of ETCs on, predominantly, EEs highlights an important role for EEs in recycling of lysosomal protein receptors to the TGN (61,62). The predominance of sortilin and MPR in ETCs, however, does not exclude that smaller pools of these proteins recycle through TSE/TEN or LEs. In addition to retromer, CI-MPR recycling from endosomes involves other molecular players, such as syntaxin 16 (16), Rab9/TIP47 (14) and the AP-1-binding proteins PACS-1 (15) and EpsinR (17). However, these proteins are not likely involved in ETC formation. As ETCs are not clathrin or AP-1 or AP-3 positive (data not shown), it is unlikely that they harbour PACS-1 or EpsinR. By contrast, the TSE/TEN exhibits AP-1 and clathrin-coated buds (29), suggesting that PACS-1- and/or EpsinR-mediated sorting of CI-MPR might occur from here. Rab9/TIP47 mediates CI-MPR recycling mostly from LE (14), whereas ETCs principally derive from EEs. We have no evidence regarding a potential implication of syntaxin 16 in the retrieval of sortilin and MPR (16) through ETCs. However, syntaxin 16 has also been implicated in TGN46 recycling, which was absent in ETCs. Our data are in agreement with an additional recycling pathway for sortilin and CI-MPR from LEs, because ETCs are predominantly found on EEs, whereas sortilin and CI-MPR are equally distributed over EEs and LEs. In addition, part of sortilin and MPRs travelling through the LEs could be destined for lysosomes for degradation. However, our data also suggest that the ETCs specifically house the SNX1-mediated retrieving mechanism.
Strikingly, both the TSE/TEN and the ETCs bud from EE vacuoles, whose limiting membrane displays a characteristic bilayered coat itself implicated in yet another protein-sorting event into the internal vesicles of endosomal vacuoles (29,30,63–65). Thus, our data clearly illustrate that protein present in EE vacuoles are sorted at least into three types of specialized sorting domains. They also predict the existence of molecular machineries operating at the non-coated regions of the EE vacuoles that drive proteins to be specifically recruited to either of these two different recycling compartments (i.e. ETC or TSE/TEN). The processes and cellular conditions that mediate sorting into either of these recycling pathways are largely unknown. Future investigations are needed to further unveil the components of these machineries.
Materials and Methods
Antibodies and reagents
Sortilin was detected by monoclonal antibody (mAb) neurotensin receptor 3 (NTR3) against the lumenal part of the protein (BD Biosciences) or a polyclonal antibody (pAb) recognizing the cytosolic tail (66). Both antibodies gave identical labelling patterns and therefore were used in parallel in our studies. pAbs against CI- and CD-MPR were previously described (34,35,44). mAbs against SNX1, SNX2, GGA3, EEA1, LAMP-1 and Rab4 were purchased from BD Biosciences. Anti-SNX1 pAb was a kind gift from Dr M. Seaman (Cambridge IMR, Cambridge, UK). Anti-Alexa® 488 and anti-green fluorescent protein (GFP) mAbs were from Molecular Probes. mAb against TGN46 was purchased from Serotec. The hybridoma mouse cell line H5C6 producing anti-human CD63 mAb is from the Department of Biological Sciences (University of Iowa, Iowa City, IA, USA). The following Abs were generous gifts: anti-Glut1 pAb (Dr G. E. Lienhard, Dartmouth Medical School, Hanover, NH, USA), anti-AP-1 mAb (Dr M. Robinson, University of Cambridge, Cambridge, UK), anti-ASGPR pAb K1 (Dr A. Schwartz, Washington University, St. Louis, MO, USA) and anti-Hrs pAb (Dr S. Urbé, School of Biomedical Sciences, Liverpool, UK). As a bridging step between mouse mAbs and protein A-gold, we used rabbit anti-mouse pAb (DakoCytomation, DAKO). The pCDNA3.1/Zeo-sortilin plasmid was previously described (66). All other reagents were from Sigma.
Cell culture and SDS–PAGE
HepG2 cells (ATTC clone HB-8065) were grown in MEM Hanks’ salts with l-glutamine (Gibco) and 10% foetal calf serum (FCS) (Gibco) for a maximum of 10 passages. HeLa cells were cultured in DMEM high glucose supplemented with l-glutamine and 10% FCS and transiently transfected with Fugene (Roche Molecular Biochemicals) following the manufacturer’s instructions. For SDS–PAGE, cells were cultured in 10-cm Petri dishes before being detached and lysed in PBS buffer containing 1% Triton-X-100 supplemented with the Complete™ protease inhibitor mixture (Roche Molecular Biochemicals). Equal amounts of lysates were analyzed by SDS–PAGE.
siRNA duplexes designed against SNX1 (target: AAGAACAAGACCAAGAGCCAC) or a control duplex (target: AAGACAAGAACCAGAACGCCA) were obtained from Dharmacon. Cells were seeded in 6-well plates at a density of 0.95 × 105 cells per well prior to transfection with the siRNA duplexes using oligofectamine (Invitrogen) and further incubated for 72 h. A small fraction of each sample was lysed and subjected to SDS–PAGE to assess the suppression levels of SNX1 by Western blot.
HepG2 cells were transiently transfected with GFP–Rab11 using Saint (Synvolux Therapeutics) following the supplier’s protocol and 24 h later processed for IEM.
Internalization of Alexa 488-conjugated Tf
Alexa Fluor 488-conjugated Tf was purchased from Molecular Probes. HepG2 cells cultured in 6-cm-diameter dishes were preincubated in MEM containing 0.1% BSA in a CO2 incubator maintained at 37°C before adding 25 mg/mL Tf–Alexa Fluor 488 for 30 min. Cells were then rinsed twice with the same buffer at 4°C and fixed overnight with 4% formaldehyde in 0.1 m phosphate buffer, pH 7.4.
HepG2 cells were cultured on microscope coverslip glasses (Marienfeld Laboratory Glassware), fixed for 20 min with 4% formaldehyde in PBS buffer, then quenched for 5 min with 50 mm glycine in the same buffer and finally permeabilized with 0.1% saponin in PBS for 20 min or alternatively with 0.1% Triton-X-100 in PBS for 5 min. Primary antibodies were added for 1 h at room temperature and after incubation with the secondary antibodies conjugated to Alexa fluorophores for 45 min at room temperature and coverslips were mounted in mowiol. The cells were examined using a LSM510 Zeiss confocal microscope with a 32-PMT META detector (Zeiss) or a Leica AOBS confocal microscope. The acquired images were processed using an Adobe Photoshop® v6.0 software.
HepG2 cells grown to 80% confluence were fixed by adding freshly prepared 4% formaldehyde in 0.1 m phosphate buffer (pH 7.4) to an equal volume of culture medium for 10 min, followed by post-fixation in 4% formaldehyde at 4°C overnight. Ultrathin cryosectioning and immunogold labelling were performed as previously described (67,68).
Counting of gold particles, structures and cell profiles was performed by random screening of prefixed areas of sections derived from at least three different grids. The level of colocalization of sortilin and CI-MPR in CCVs was established in sections double-labelled for sortilin (10 nm gold) and CI-MPR (15 nm gold) by randomly selecting 170 CCVs that were within 400 nm of the last Golgi stack. The relative distribution of sortilin and CI-MPR in control and SNX1-depleted HepG2 cells was assessed by randomly analyzing cell profiles in which 948 and 895 gold particles for sortilin and 828 and 870 for CI-MPR were counted, respectively. Gold particles were assigned to a compartment when no further than 25 nm away from its limiting membrane. The relative distribution of sortilin over EEs and LEs was assessed in 300 randomly selected cell profiles. The distributions of the endosomal markers Alexa 488-conjugated Tf, EEA1, Rab4, GFP–Rab11, Hrs and SNX1 were determined by counting 195, 156, 206, 89, 181 and 237 gold particles, respectively. EEs and LEs were defined as described in the Results section and Figure S2. The number of ETCs per endosomal vacuole was established by counting 200 randomly encountered, sortilin-positive ETCs. The relative distributions of sortilin and CI-MPR in control and SNX1-suppressed cells were determined by analyzing 20 profiles for each condition. The numbers of endosomes, of ETCs per cell and of ETCs connected to an endosomal vacuole, were assessed in control and SNX1-suppressed cells by analyzing 20 cell profiles.
Electron tomography, 3D reconstruction and modelling
Three-dimensional models of sortilin- and SNX1-positive ETCs were obtained by electron tomography of 300- to 500-nm thick cryosections as previously described (69).
Protein stability assays
Twenty hours after the second siRNA treatment, control and SNX1-suppressed cells were lysed directly or treated with 40 μg/mL cycloheximide, lysosomal protease inhibitors (1 mg/mL leupeptin and 10 μg/mL E-64) or cycloheximide and protease inhibitors. Cycloheximide and protease inhibitors were replaced every 4 h, and cells were lysed after 12 h. Total lysates were subjected to SDS–PAGE, sortilin detected by Western blot and levels of this protein quantified using an ImageQuant software (Amersham/GE Healthcare).
We thank Viola Oorschot for her help in the EM studies, Marc van Peski and René Scriwanek for excellent photographic assistance and Catherine Rabouille, Fulvio Reggiori and Arjan de Jong for critical reading of the manuscript. We thank the members of the Cell Biology Department for their helpful discussions. We also thank the Medical Research Council for providing an Infrastructure Award (G4500006) to establish the School of Medical Sciences Cell Imaging Facility at the University of Bristol and Mark Jepson and Alan Leard for their assistance. M. M. is supported by an Institute national de sante et de la recherche medicale (INSERM) and an intra-European Marie Curie fellowship. M. V. B. is supported by the Department of Biochemistry, University of Bristol, and PerkinElmer LAS, UK. J. K. is the recipient of VICI grant 918.56.611 of the Netherlands Organization for Scientific research (NWO).