After treatment with agonist such as isoproterenol, β2ARs trigger a signaling cascade and undergo clathrin mediated endocytosis. β2ARs are then rapidly recycled from the early endosome antigen 1 (EEA1) compartment (, ) to the plasma membrane (), resensitizing the cell5
. Internalized transmembrane proteins are generally thought to leave the endosome through tubules6
. In the case of β2AR, receptor-containing tubular endosomal protrusions can be visualized in living cells ()7
. β2AR recycling is sequence-dependent, requiring a C-terminal PDZ ligand2
. When this ligand is occluded by a HA tag (β2AR-HA), mutant receptors fail to recycle efficiently and are not seen in endosomal tubules ()2
. Therefore, these tubules likely represent the structure responsible for sequence-dependent recycling of β2AR.
Rapid recycling β2ARs selectively enter retromer-associated endosomal tubules
β2AR and CIMPR follow divergent trafficking paths upon exit from the same retromer-associated tubule
Knockdown of retromer by RNAi inhibits β2AR recycling and misroutes internalized β2ARs to lysosomes
Cargo that are capable of recycling efficiently with bulk membrane flux, such as the transferrin receptor (TFR), exit endosomes via multiple dynamic tubules8
. The β2AR was previously shown to enter a specific subset of these tubules7
. To investigate the hypothesis that β2AR-containing tubules are biochemically distinct, we took a candidate-based approach to identify specific tubule components.
Because both recycling (β2AR) and non-recycling (β2AR-HA) receptors were localized to similar endosomes but appeared to differ in lateral distribution near the highly curved neck region of endosomal tubules (), we asked whether curvature sensing/inducing Bin–Amphiphysin–Rvs (BAR) domain-containing proteins might be localized there. We first considered sorting nexins, because several contain a BAR domain and an endosome-associating PX domain9,10
. We looked at four of these BAR domain-containing sorting nexins (SNX1, SNX4, SNX5, and SNX9). We noticed striking, and nearly complete, overlap of SNX1 and SNX5 with β2AR-containing tubules (; data not shown for SNX5). Quantification across multiple cells verified that the vast majority (92.5%; n = 40 tubules) of endosomal tubules containing internalized β2ARs co-localized with a concentrated region of GFP-SNX18
SNX1 and SNX5 have been previously associated with retromer. Retromer is composed of two distinct multi-protein subcomplexes, one containing sorting nexins and the other containing vacuolar protein sorting (VPS) proteins VPS26, VPS29, and VPS3511,12
. Recruitment of the VPS26/29/35 subcomplex to the endosome requires previous recruitment of the SNX-containing subcomplex and Rab713
. Therefore, we next visualized a component of the VPS26/29/35 retromer subcomplex to determine whether the complete retromer complex was forming on these tubules.
VPS29-GFP was also concentrated on β2AR-containing tubules (, Supplementary Movie
); quantification across multiple examples indicated that 96.5% (n=57 tubules) of β2AR-containing endosome tubules were decorated with VPS29-GFP. These VPS29 foci extended off the edge of the endosome limiting membrane and co-localized with tubular β2AR staining. In rapid (2Hz) image series, β2AR could be seen extending distally beyond the SNX1 or VPS29 marked portion of the tubule (, Supplementary Movie
). Essentially all VPS29-GFP puncta co-localized with β2AR on vesicles (95.4%, n=177 spots) (). Additionally, 31.1% of these VPS29-GFP spots were tubular protrusions from an endosome with a resolvable lumen. Consistent with this co-localization, biochemical purification of EEA1 endosomes co-purified β2AR and retromer component VPS35 (; uncut blots for all figures are in Supplementary Figure 2
). Together, this data suggests that β2ARs access the retromer marked tubular endosomal network14
In contrast to β2AR, recycling-defective β2AR-HA did not extend into retromer tubules (). This distinction between β2AR and β2AR-HA localization was quantified by circumferential line scan analysis around the endosome limiting membrane. Whereas β2AR was enriched ~50% at the base of VPS29-GFP coated tubules relative to the rest of the endosome membrane, β2AR-HA was not (). These results indicate that β2ARs specifically enter retromer-associated tubules following endocytosis, and suggest that tubule entry is the primary sorting step required for β2AR recycling to the plasma membrane.
To investigate if retromer is required for β2AR recycling, we depleted retromer components by RNAi. We first assessed trafficking effects using fluorescence microscopy to visualize ligand-dependent redistribution of surface-labeled receptors between the plasma membrane and intracellular membranes. The β2AR agonist isoproterenol induced pronounced redistribution of antibody-labeled β2ARs from the plasma membrane to intracellular puncta, indicative of ligand-induced endocytosis. This process was not detectably inhibited by siRNA-mediated knockdown of VPS35 (). Subsequent recycling of receptors to the plasma membrane after agonist removal was obviously reduced in VPS35-depleted cells ().
We then quantified this effect of retromer depletion using a flow cytometry assay measuring changes in the internalized pool of receptors. β2AR recycling, as measured by antibody efflux following agonist removal, was strongly inhibited by multiple siRNA sequences targeting either VPS26 or VPS35 (). Further, retromer depletion inhibited β2AR recycling at all time points examined (). Consistent with this, VPS35 depletion, which prevents endosome association of VPS29-GFP, resulted in endosomes devoid of β2AR tubules (). Additionally, we verified that retromer is required for β2AR recycling in physiologically relevant A10 cells, a rat atrial-derived vascular smooth muscle cell line (Supplementary Figure 1
Retromer depletion leads β2ARs to be misrouted to the lysosome and degraded (). This effect of retromer depletion on receptor degradation is fully consistent with the observed inhibition of β2AR recycling, and similar to that of disrupting the receptor’s C-terminal PDZ ligand2
To examine the specificity of retromer depletion effects, we tested another protein that traffics through the same endosome as β2AR. For this we used TFR, a constitutively endocytosed nutrient receptor that can recycle with bulk membrane15
. When TFR-GFP or labeled Transferrin was imaged alongside internalized β2AR in living cells, β2AR only appeared in a subset of the TFR containing tubules projecting from the limiting membrane of individual early endosomes (), as shown previously7
. Additionally, only a subset of TFR tubules labeled with VPS29-mCherry (). As TFR can access non-retromer recycling tubules, it should be able to exit the endosome in the absence of retromer tubules. Consistent with this, retromer depletion had no effect on degradation of endogenous TFR () and produced only a small kinetic delay in the recycling of TFR back to the plasma membrane (). This second observation is consistent with the ability of a visible fraction of TFR to enter retromer tubules and further supports the ability of these tubules to mediate rapid plasma membrane recycling.
Retromer depletion preferentially affects β2ARs over TFRs traversing the same endosomes
As retromer complex has not previously been implicated in recycling from endosomes to the plasma membrane, we next sought to compare β2AR localization and trafficking to that of a membrane cargo which exhibits retromer-dependent trafficking to the TGN. For this purpose we focused on the cation-independent mannose phosphate receptor (CIMPR)16,17
. We first evaluated the localization of β2AR relative to CIMPR in fixed cells. β2AR was induced to reach a steady-state of internalization and recycling so that it populated various intermediates in endocytic and recycling pathways. Under this condition, β2AR was localized primarily in EEA1 marked endosomes and at the plasma membrane (, )18,19
. CIMPR co-localized with β2AR in a large fraction of early endosomes but was also prominently localized in perinuclear endomembranes not containing β2AR (). This is consistent with steady state localization of CIMPR in the TGN and late endosomes20
. Supporting this, the localization of perinuclear CIMPR was similar to, or closely adjacent to, that of the trans-Golgi/TGN marker galactosyltransferase-GFP (GalT-GFP) 21,22
. We also note that retromer cargo that traverse the TGN on the way to indirect plasma membrane delivery have shown significant colocalization with markers of the TGN region23
. Importantly, and in marked contrast, the β2AR localization observed in these cells was clearly distinct. β2ARs localized to peripheral early endosomes but not in the perinuclear distribution characteristic of CIMPRs ().
Given that β2AR and CIMPR achieve different steady state localization patterns, we next used live imaging to ask if these distinct membrane cargoes localize to the same endosomes and, if so, if they localize to the same retromer tubules. Confocal microscopy revealed that both FITC-conjugated anti-CIMPR antibody (CIMPR AB), used to detect endogenous receptor, and a GFP-tagged CIMPR construct (cCIMPR-GFP) clearly co-localized with β2AR in the same endosomes and in the same tubules (). This overlap was extensive, as >90% of the β2AR-positive tubules resolved also contained GFP-tagged CIMPR.
We next asked how the retromer tubule could support rapid endosome-to- plasma membrane traffic, evidently independent of traversing the TGN. Knowing that the small GTPase Rab4A had previously been implicated in this process we tested if it associated with β2AR containing endosomes. GFP-Rab4A localized to these compartments and was visibly enriched on the β2AR-containing retromer tubules ()7
. Further, as expected, depletion of Rab4A by siRNA inhibited recycling of β2AR (data not shown)1
. In contrast, depletion of several components that function in endosome-to-TGN trafficking of CIMPR (Rab6A’, Rab7b (a distinct gene from Rab7), and GCC185) failed to significantly affect β2AR recycling24–28
(). As a control for depletion, we established in parallel experiments that these siRNAs caused a pronounced increase in cell surface expression of CIMPR, consistent with net disruption of normal endosome-TGN cycling of CIMPRs established previously ()16
Despite the co-occurrence of both membrane cargoes in a large fraction of endosomal tubules, β2AR and CIMPR were not identically localized in individual tubules imaged at high temporal resolution. Specifically, CIMPR was typically enriched near the distal end of tubules or in protrusions that extended beyond the β2AR labeled areas (). Apparent separation of cargo was visualized for both endogenous CIMPR and the GFP CIMPR transgenic protein (). These results support the previously proposed idea that an elaborated tubular endosomal network, capable of mediating the trafficking of cargoes to multiple destinations, is fed by the retromer tubule14
. Further work will be required to investigate this, as there is also evidence that a significant fraction of CIMPRs can traffic to the plasma membrane following endosome exit20
We next investigated the mechanistic basis for the role of retromer in β2AR recycling. For CIMPR, it is proposed that a direct interaction between the cytoplasmic tail and retromer complex is required for proper trafficking29
. β2AR trafficking is dependent on a PDZ motif present in the cytoplasmic tail that is both necessary and sufficient to mediate its plasma membrane recycling, but the core retromer complex is devoid of any recognizable PDZ domain. SNX27 contains a PDZ domain that binds the β2AR tail, and has recently been shown to be essential for PDZ-directed recycling of the β2AR30
. Verifying this, knockdown of SNX27 robustly inhibited recycling of β2ARs (). Despite this pronounced decrease in β2AR recycling, CIMPR distribution in the same cells appeared unaffected. Furthermore, SNX27 depletion did not alter CIMPR surface expression () or CIMPR turnover in the presence of cyclohexamide (data not shown).
SNX27 serves as an adapter for β2AR, sorting it into the retromer tubule
Live imaging of SNX27 depleted cells revealed that VPS29-GFP puncta still associate with endosomes but that β2ARs no longer appear in these tubular structures (). Depletion of SNX27 also decreased the aggregation of β2ARs at the retromer tubule. In control siRNA treated cells there was a relative receptor density of 1.51±.061 (n=16 endosomes) at the base of the VPS29 tubule, while in SNX27 depleted cells this dropped significantly to 1.15±.055 (n=19 endosomes, p=.0002), phenocopying disruption of the PDZ motif (β2AR-HA). Together, this data suggests that SNX27 acts as a cargo adaptor to retromer, but is not a core component.
To investigate further how SNX27 was serving as a cargo adaptor to the retromer tubule, we used an affinity tagging/purification-mass spectrometry approach to determine which other proteins physically associate with it. In all 5 independent SNX27 purifications, components of the WASH actin nucleation complex were predominant hits (Supplementary Table
). However, WASH complex was not detected in control purifications or in those of other bait molecules run at the same time. WASH has been shown to physically link to the retromer complex31
. Therefore this association with SNX27 supports the adaptor hypothesis by establishing protein connectivity.
Association of SNX27 with the retromer tubule was further supported by extensive colocalization between SNX27 and VPS29-GFP in fixed cells (). Moreover, SNX27 co-immunoprecipated endogenous VPS35 but not EEA1, an abundant protein that localizes to the same endosomes. Interestingly, VPS35 co-immunoprecipitated with SNX27 only under crosslinking conditions, in contrast to WASH complex components that co-purified with SNX27 in the absence of crosslinking (). These observations suggest that SNX27 does not interact with the core retromer complex directly, but that its association with the retromer tubule is mediated by other interaction(s) including through the WASH complex. Together, these results indicate that SNX27 acts, through novel connectivity involving the WASH complex, as a specific adapter to promote PDZ-directed plasma membrane sorting through the retromer tubule ().
To probe the generality of the SNX27/retromer -mediated recycling pathway we next examined two other GPCRs that are known to recycle rapidly to the plasma membrane after endocytosis. Efficient recycling of the beta-1 adrenergic receptor (β1AR) requires a class 1 PDZ motif present in its distal cytoplasmic tail, but this motif differs in primary structure and binding properties from the PDZ motif present in the β2AR tail32,33
. The D1 dopamine receptor (D1R) recycles efficiently after endocytosis but does not contain a C-terminal PDZ motif, and truncation of its distal cytoplasmic tail does not disrupt receptor recycling34
. We observed a significant inhibition of β1AR recycling following knockdown of either VPS35 (33.7%±0.1 inhibition, n=9, p= 0.0013) or SNX27 (19.5%±0.1 inhibition, n=9, p= 0.0293). Recycling of FLAG-D1Rs, in contrast, was not significantly inhibited by either manipulation. Thus, retromer-dependent recycling is not unique to the β2AR and can be specified by distinct PDZ motifs. This suggests that many other membrane cargoes which contain C-terminal PDZ ligands, including a large group of G-protein coupled receptors (GPCRs), may traffic via this SNX27/retromer pathway35
Retromer has been shown to function in trafficking of various cargo from endosomes to the TGN, and to function in transcytosis in polarized cells12,36
. Recently several retromer cargo have been identified that traffic to the plasma membrane, though it is unclear whether they do so directly from the endosome or via the TGN37–39
. In this paper we identify an essential role of retromer in mediating rapid recycling of a prototypical GPCR apparently directly to the plasma membrane. We also establish that β2ARs can enter the same retromer tubules as a canonical TGN-directed cargo, and identify a distinct adaptor protein that links signaling receptors specifically to the retromer-dependent plasma membrane recycling pathway. Our results support a revised view of retromer tubules as a multi-functional exit port supporting diverse membrane itineraries, including direct endosome-to-plasma membrane trafficking.
Our results have fundamental implications in the field of cell signaling because they indicate that retromer plays a critical role in mediating the sorting of a prototypical GPCR between the functionally opposing pathways of resensitization and down-regulation, the distinct physiological consequences of which are well established4,40
. Consistent with this, depletion of VPS35 significantly decreased isoproterenol mediated signaling from the β2AR, as measured by cellular cAMP accumulation (). Interestingly, retromer depletion also affected receptor-independent activation of adenylyl cyclase by forskolin, suggesting that retromer tubules have other role(s) in cell signaling.