When ectopically expressed, green fluorescent protein (GFP)-tagged SNX3 showed a characteristic punctate distribution, and colocalized with EEA1, an effector of the small GTPase Rab5, and to some extent with the transferrin receptor, but not with late endocytic markers, including lysobisphosphatidic acid (LBPA) (Figure S1
A) or Lamp1 (C and F). Consistently, GFP-SNX3 cofractionated with Rab5, but not with LBPA (Figure S1
B). GFP-SNX3 was thus primarily present on early endosomes, much like the endogenous protein [40
], and this association depended on PtdIns3P and required an intact PtdIns3P-binding domain PX (Figure S1
C), as expected [40
To investigate the possible role of SNX3, we first tested whether overexpression affected transport along the endocytic pathway. SNX3 overexpression had no effect on EGF receptor endocytosis, since internalized EGF colocalized with both GFP-SNX3 and EEA1 in early endosomes within 10 min of incubation at 37 °C (Figure S2
A), much like in control cells (unpublished data, see [33
]). However, upon longer incubation times at 37 °C, EGF receptor degradation was delayed by SNX3 overexpression (A; quantification in B), in agreement with previous findings [40
]. This delay was caused by defective transport, since the bulk of endocytosed EGF failed to reach late endosomes containing Lamp1 (C and S3
A) or LBPA (unpublished data) after 50 min in cells overexpressing SNX3. Then, EGF remained in early endosomes containing GFP-SNX3 and EEA1 (D and S3
B; quantification in E), whereas it was mostly transported beyond early endosomes in control cells (Figure S4
A). The inhibitory effect of SNX3 was specific, since overexpression of other SNX family members, including SNX1, SNX2, and SNX16, did not interfere with EGF receptor transport (Figure S5
), as expected [41
Much like with EGF receptor, endocytosis of the bulk phase marker rhodamine-dextran was not affected by SNX3 overexpression, and the tracer accumulated in early endosomes containing GFP-SNX3 and EEA1 within 10 min at 37 °C (Figure S2
B). However, after a subsequent 40-min chase at 37 °C, SNX3 overexpression markedly impaired dextran transport to late endosomes. The bulk of the tracer then remained in early endosomes and failed to reach Lamp1-positive late endocytic compartments (F and S3
C), as was observed with the EGF receptor (C and S3
A) and in contrast to control cells (Figure S4
B). These inhibitory effects of excess SNX3 were specific, since SNX3 overexpression did not affect the transport of endocytosed Shiga toxin B-subunit from early endosome (Figure S2
C) to the trans-Golgi network (G and S3
D; compare with controls in Figure S4
C), which requires SNX1 [43
]. Altogether these observations indicate that SNX3 overexpression caused the selective retention of signaling receptors and bulk tracers in early endosomes containing Rab5 and its effector EEA1, and thereby inhibited more-distal steps of early-to-late endosome transport.
To further investigate the role of SNX3 in early-to-late endosome transport, we made use of vesicular stomatitis virus (VSV), which infects cells from the endocytic pathway. Indeed, endocytosed virions must be transported beyond early endosomes for efficient infection to occur [44
]. Overexpression of SNX3 had no effect on endocytosis of VSV (A), as observed with EGF and dextran (Figure S2
A and S2
B), but significantly reduced VSV infection, which was monitored by the synthesis of the viral glycoprotein G (B). This inhibition did not result from some indirect effects of SNX3 on the G-protein biosynthetic pathway, since replication of the viral genome, quantified by real time-PCR (RT-PCR), was also similarly reduced (C), indicating that excess SNX3 prevented efficient release of the viral nucleocapsid into the cytosol.
SNX3 in VSV Infection and MVB Biogenesis
The viral nucleocapsid is released into the cytoplasm after low pH-triggered fusion of the viral envelope with endosomal membranes. To monitor viral fusion events, VSV was labeled with self-quenching amounts of Dil, a fluorescent long-chain dialkylcarbocyanine dye, and bound to the cell surface at 4 °C [44
]. After endocytosis at 37 °C, the fusion of individual virions was revealed in the light microscope by the appearance of fluorescent spots in endosomes, due to Dil dequenching [44
]. Strikingly, overexpression of SNX3 markedly reduced VSV fusion (D). Since VSV fusion normally occurs beyond early endosomes [44
], these observations indicate that virions, much like EGF receptor and dextran () remained trapped in early endosomes in cells overexpressing SNX3.
The effects of SNX3 overexpression on the transport of viral particles, EGF receptor, and fluid phase markers led us to investigate by electron microscopy whether endosome morphology was then affected. SNX3 overexpression caused a dramatic accumulation of multivesicular structures without causing a general expansion of endosome vesicular regions (Figure S6
; quantification in E), perhaps suggesting that only specialized regions of the early endosome are competent to become MVBs. Multivesicular elements after SNX3 overexpression were frequently clustered in groups of five to ten individuals, as revealed in low (Figure S6
) and high (F) magnification views, perhaps connected to each other (F). Immunogold labeling of cryosections showed that GFP-SNX3 itself was abundant on the limiting membrane of these multivesicular structures. These SNX3-positive structures all exhibited a similar multivesicular and spherical appearance (diameter ≈ 0.4 μm), which closely resembles the morphology of MVBs or ECVs that mediate early-to-late endosome transport [46
]—herein referred to as ECV/MVBs. However, in marked contrast to free ECV/MVBs during transport towards late endosomes, these SNX3-positive structures exhibited the characteristic features of early endosomes, including all early endosomal markers that were tested ( and see ). Moreover, the transferrin receptor, which is restricted to early and recycling endosomes, was found closely associated with these SNX3-positive ECV/MVB-like structures (arrowheads in H and I), consistent with data from us () and others [40
]. The transferrin receptor was often present in tubulo-cisternal elements connected to multivesicular elements (I), presumably corresponding to forming recycling endosomes. It thus appears that SNX3 overexpression causes the accumulation of multivesicular regions on early endosomal membranes. These ECV/MVB-like structures fail to detach, or to mature, from early endosomes. Such a frustrated process of ECV/MVB formation accounts nicely for our observations that the transport of EGF receptor, VSV and bulk markers beyond early endosomes is then inhibited (see model, Figure S10
SNX3 Colocalizes with Hrs, Ubiquitinated Proteins, and Clathrin on Rab5Q79L Enlarged Endosomes.
In our electron microscopy analysis, SNX3 was often observed on, or close to, membrane regions containing electron-dense materials on the cytoplasmic membrane face (e.g., G), which resembled the early endosomal Hrs-clathrin coat that mediates ubiquitinated receptor sorting into lumenal vesicles during ECV/MVB formation on early endosomes [9
]. To further characterize SNX3 distribution, we thus used the constitutively active mutant Rab5Q79L
, which induces the formation of enlarged early endosomes by promoting their homotypic fusion. On these large endosomes, regions containing Hrs, clathrin, and ubiquitinated receptors could be resolved by light microscopy from those containing EEA1, indicating that components of the machinery that sorts down-regulated receptors into ECV/MVB internal vesicles are concentrated in specific early endosomal domains [9
]. When GFP-Rab5Q79L
was coexpressed with monomeric red fluorescent protein (mRFP)-SNX3, both proteins were present on enlarged early endosomes, as expected. Approximately 90% of endosomes containing SNX3 were also labeled with Rab5Q79L
, but some variation in SNX3 association with individual Rab5Q79L
-endosomes was observed (), presumably reflecting different angles of visualization in the confocal planes. SNX3 was clearly seen to colocalize preferentially with Hrs (A), ubiquitinated proteins (B), and clathrin (C) in regions that seemed devoid of EEA1 (D). These observations further demonstrate that ectopically expressed SNX3 accumulates on early endosomal membranes. They also indicate that this accumulation occurs preferentially in multivesicular regions that contain, in addition to SNX3 itself, the protein machinery responsible for sorting into lumenal vesicles. It thus seems that the lumenal invagination process continues in the presence of excess SNX3, leading to the accumulation of multivesicular regions on early endosomes, but that more distal transport events, including ECV/MVB detachment—or maturation—(A–F and B–D), are inhibited, perhaps because excess SNX3 limits the access or binding of downstream machineries.
Since overexpression of SNX3 caused an expansion of multivesicular regions on early endosomes, we investigated the impact of SNX3 down-expression on the formation of lumenal vesicles by electron microscopy. To this end, the lumen of ECV/MVBs was labeled with endocytosed horseradish peroxidase (HRP) pulsed for 15 min and then chased for 30 min at 37 °C, after microtubule depolymerization with 10 μM nocodazole [46
]. As expected [46
], multivesicular structures with the characteristic ECV/MVB morphology were found in controls (A, upper panel), accounting for approximately 70% of the total HRP-positive structures (see quantification in D). By contrast, after SNX3 knockdown to approximately 20% of the control levels (inset in C), approximately 70% of the total HRP-positive profiles did not seem to contain lumenal vesicles, but were otherwise similar to controls (B, upper panel). All SNX3 knockdown experiments were repeated with two small interfering RNA (siRNA) target sequences without significant differences (see A, B, and S8
SNX3 Silencing Inhibits Membrane Formation within MVBs but Does Not Affect EGFR Early-to-Late Endosomal Transport and Degradation
SNX3 Rescues the Formation of Internal Vesicles in Hrs siRNA-Treated Cells
SNX3 Controls the Formation of Intralumenal Vesicles That Incorporate the EGF Receptor
To better visualize the presence or absence of lumenal vesicles, ECV/MVBs were labeled with 5-nm proteinA-gold endocytosed for 30 min at 37 °C with or without nocodazole. In the mock-treated control, the gold particles distributed within ECV/MVBs, which typically contained approximately 20 vesicles per profile, whether microtubules were present (A, lower panel) or not (unpublished data). After SNX3 knockdown, however, 5-nm gold particles labeled vesicles of the same diameter as ECV/MVBs, but with only three to five internal vesicles per profile, whether microtubules were intact (B, lower panel) or not (unpublished data). The appearance and size of internal vesicles were otherwise undistinguishable from the controls. Altogether, our data thus indicate that SNX3 plays a direct and specific role in the formation of intralumenal membrane invaginations within nascent ECV/MVBs, since multivesicular regions are increased by overexpression (F) and decreased by down-expression (B).
Interestingly, SNX3 knockdown did not affect virus fusion (C) and caused only a small, marginal decrease in nucleocapsid release (D). Previously, we had found that the VSV envelope undergoes fusion primarily with the membrane of ECV/MVB internal vesicles, thus releasing the capsid into their lumen, where it remains hidden [44
]. In late endosomes, back fusion of these vesicles with the endosome-limiting membrane then ensures capsid delivery to the cytoplasm, indicating that VSV fusion and capsid release occur in sequential steps of the pathway. In particular, we found that depolymerization of the microtubules, which reduces early-to-late endosome transport [46
], does not affect viral fusion, but efficiently inhibits VSV delivery to late endosomes and capsid release [44
] (see D). In contrast to controls, capsid release was only marginally affected by microtubule depolymerization in cells treated with SNX3 siRNAs (D)—much like in cells treated with PI 3-kinase inhibitors or Hrs siRNAs [44
], which both decrease intralumenal membranes in endosomes [30
] (see also B). This was not due to some indirect effects of SNX3 siRNAs, since early-to-late endosome transport remained microtubule dependent after SNX3 knockdown (see below and Figure S7
A and S7
B). It thus appears that, when ECV/MVBs lack intralumenal vesicles after SNX3 knockdown, VSV fusion can be triggered at the limiting membrane, thus by-passing the need for transport to late endosomes—again much like after PI 3-kinase inhibition or Hrs knockdown [44
The “empty” endosomes in cells lacking SNX3 closely resemble endosomes observed after Hrs knockdown in mammalian cells [30
] (see B) or mutagenesis in Drosophila
]. In these studies, Hrs depletion also inhibited EGF receptor degradation, supporting the view that sorting into intralumenal invaginations mediates lysosomal targeting [13
]. To our surprise, SNX3 knockdown had little effect on EGF receptor degradation (Figure S7
C), significantly less than Hrs knockdown (see blot in A) in parallel experiments (E, quantification of the blots in F). Consistently, a wave of fluorescent EGF reached late endocytic compartments containing Lamp1 in cells treated with SNX3 siRNAs (H; quantification in G) as in mock-treated cells (Figure S8
A), and this transport required intact microtubules (Figure S7
A and S7
B). Then, EGF no longer colocalized with EEA1 in early endosomes (H; quantification in G) as in mock-treated cells (Figure S8
A). This is in contrast to the inhibition observed after SNX3 overexpression (C–E). SNX3 knockdown thus appears to prevent the formation of intralumenal invaginations within endosomes without interfering with EGF receptor transport and degradation, indicating that the lysosomal targeting of signaling receptors is then uncoupled from sorting into ECV/MVBs.
Both Hrs and SNX3 seem to play a role in the membrane invagination process, but only Hrs, and not SNX3, appears to be involved in EGF receptor targeting to lysosomes, perhaps suggesting that Hrs acts upstream of SNX3 in receptor sorting and multivesicular body biogenesis. Consistent with this notion, Hrs knockdown selectively reduced the expression of SNX3—without affecting any other protein involved in endosome membrane dynamics that we tested (A)—and in particular decreased the membrane-associated pool of SNX3 (Figure S8
B). By contrast, SNX3 knockdown had no effect on Hrs expression (A and see Figure S8
C). We thus wondered whether the known effect of Hrs knockdown could be due, at least in part, to reduced levels of SNX3.
As expected [30
], approximately 60% of the total HRP-labeled endosomes appeared to contain fewer internal membranes (B, left panel; quantification in D) in cells treated with Hrs siRNAs, much like endosomes in cells treated with SNX3 siRNAs (B)—and in contrast to endosomes in mock-treated cells (A). Parallel analysis of cells after proteinA-gold internalization revealed that these endosomes contained five to ten internal vesicles per profile, which were otherwise similar to controls (A), whether microtubules were intact (B, right panel) or not (unpublished data)—again much like after SNX3 knockdown and in contrast to controls (A and B). The multivesicular morphology of endosomes in cells treated with Hrs siRNAs could, however, be restored by overexpression of GFP-SNX3 (C; quantification in D), despite that fact that Hrs expression remained silenced (Figure S8
D). Structures labeled with 5-nm proteinA-gold then contained approximately 25–30 lumenal vesicles per profile, similar to endosomes in mock-treated controls. These data thus strongly suggest that the morphological phenotype of endosomes in Hrs-depleted cells is at least in part due to low SNX3 levels, and also demonstrate that SNX3 is a component of the molecular machinery that drives intralumenal membrane invagination.
It has been suggested that EGF receptor is trafficked through a subpopulation of multivesicular endosomes in a process that involves annexin A1 [21
]. Annexin A1 is structurally, functionally, and biochemically related to annexin A2 [47
], and both proteins colocalize on early endosomes [48
]. Annexin A2 was also proposed to play a role in multivesicular endosome biogenesis, but not in the invagination process [22
]. Consistently, endogenous annexin A1 colocalized with annexin A2-GFP and mRFP-SNX3, and endogenous annexin A2 with mRFP-SNX3 (Figure S9
A). These observations show that SNX3 is present on endosomes that contain both annexin A1 and annexin A2.
Next, we investigated whether SNX3 plays a role in the formation of intralumenal vesicles that mediate EGF receptor sorting into multivesicular endosomes. To this end, we made use of the ability of the active Rab5 mutant Rab5Q79L to form enlarged early endosomes that provide high spatial resolution by light microscopy [9
], as in . When mock-treated cells were challenged with EGF for 15 min at 37 °C, the EGF receptor was endocytosed into these enlarged endosomes, where greater than 50% of the endocytosed receptor accumulated in the lumen (A, quantification in B), as expected [50
]. Similarly, EGF colocalized with the receptor in the lumen of these enlarged endosomes (unpublished data). Knockdown of Hrs with either one of two siRNAs significantly reduced EGF receptor sorting into the lumen of enlarged endosomes (A, quantification in B), consistent with our electron microscopy analysis (B–D) and in agreement with previous findings [30
]. Similarly, SNX3 depletion with either one of two siRNAs inhibited EGF receptor incorporation in the lumen of large endosomes to the same extent as Hrs knockdown (A, quantification in B). Finally, SNX3 re-expression in the Hrs knockdown background restored EGF receptor accumulation in the lumen of enlarged endosomes to the same extent as observed in mock-treated controls (A, quantification in B). These observations unambiguously demonstrate that SNX3 controls the formation of lumenal membranes that carry the EGF receptor, further confirming the role of SNX3 in the lumenal invagination process.
Although reduced levels of SNX3 seemed to account for the invagination defect in Hrs knockdown cells, SNX3 siRNAs did not affect EGF receptor transport to late endosomes (G–H) and degradation (E and F). Similarly, a wave of endocytosed EGF receptor was exported from EEA1-positive early endosomes and reached Lamp1-positive late endosomes (A, quantification in B) under our conditions of Hrs knockdown (≈80%, A) much like in mock-treated cells (Figure S8
A) or in cells treated with SNX3 siRNAs (H). Since, under the same Hrs knockdown conditions, the formation of internal vesicles (B and A) and EGF receptor degradation were inhibited (E) and SNX3 levels reduced (A), these observations strongly suggest that Hrs is an essential component of the lysosome targeting machinery, which can function independently of receptor sorting into and incorporation within multivesicular endosomes.
MEK1 Cleavage by Anthrax Toxin Lethal Factor
To further discriminate between Hrs and SNX3 functions, we made use of anthrax toxin, which is translocated across the membrane of ECV/MVB intralumenal vesicles. Like VSV nucleocapsids, the toxin hijacks these vesicles to reach late endosomes, where back fusion with the limiting membrane releases the lethal factor into the cytosol, leading to the cleavage of mitogen-activated protein kinase kinases (MAPKKs), and in particular MEK1 [51
]. After addition of anthrax toxin, MEK1 cleavage was slightly retarded in cells treated with Hrs or SNX3 siRNAs (C). Presumably, in the absence of intralumenal membranes (B, B–D, A, and B), toxin translocation could then occur across the limiting membrane of these “empty” ECV/MVBs—in good agreement with our observations on VSV capsid release after Hrs [44
] or SNX3 (D) knockdown. Interestingly, re-expression of GFP-SNX3 in the Hrs knockdown background prevented toxin translocation (C). Presumably, the toxin, once released, remained trapped in the lumen of intralumenal vesicles. Similarly, EGF receptor degradation remained inhibited after re-expression of GFP-SNX3 in the Hrs knockdown background (Figure S9
B). Indeed, excess SNX3 not only restored intralumenal vesicles (C, A, and B), but also inhibited ECV/MVB detachment (or maturation), and thus transport beyond early endosomes towards late endosomes ( and ).