The cargo-selective retromer complex is a recruiting hub for protein complexes that regulate endosomal tubule dynamics

Abstract

The retromer complex is required for the efficient endosome-to-Golgi retrieval of the CIMPR, sortilin, SORL1, wntless and other physiologically important membrane proteins. Retromer comprises two protein complexes that act together in endosome-to-Golgi retrieval; the cargo-selective complex is a trimer of VPS35, VPS29 and VPS26 that sorts cargo into tubules for retrieval to the Golgi. Tubules are produced by the oligomerization of sorting nexin dimers. Here, we report the identification of five endosomally-localised proteins that modulate tubule formation and are recruited to the membrane via interactions with the cargo-selective retromer complex. One of the retromer-interacting proteins, strumpellin, is mutated in hereditary spastic paraplegia, a progressive length-dependent axonopathy. Here, we show that strumpellin regulates endosomal tubules as part of a protein complex with three other proteins that include WASH1, an actin-nucleating promoting factor. Therefore, in addition to a direct role in endosome-to-Golgi retrieval, the cargo-selective retromer complex also acts as a platform for recruiting physiologically important proteins to endosomal membranes that regulate membrane tubule dynamics.

Fig. 1.

Identification and features of retromer-interacting proteins. (A) HeLa cells stably expressing GFP-tagged VPS29, HRS1 or a mutant of VPS29 (I91S) that cannot bind VPS35 were lysed and treated with anti-GFP. The IPs were analysed by SDS-PAGE. Several proteins associated with VPS29-GFP, but not with HRS1 or the VPS29I91S mutant, are shown. The bands were excised and subjected to MALDI-TOF mass spectrometry. The proteins identified are listed as 1–22. (B) The domain architecture of the retromer-interacting proteins. Globular domains are shown as rectangles and predicted coiled-coil domains are coloured orange; unstructured regions are depicted as straight lines.

Fig. 2.

The catalytic domain of TBC1D5 is required for its retromer-recruitment regulating activity. (A) HeLa cells were transiently transfected with wild-type GFP-TBC1D5 or a mutant (R169A and Q204A, termed TBC1D5 RQ) that lacks catalytic activity. The wild-type GFP-TBC1D5 causes VPS26 to become cytosolic, whereas the RQ mutant colocalises with VPS26 and does not redistribute the cargo-selective complex into the cytoplasm. Scale bar: 20 μm. (B) HeLa cells transiently transfected with GFP-TBC1D5, GFP-TBC1D5 RQ or empty vector were blind scored for ‘normal’ (i.e. membrane-associated) or ‘displaced’ (i.e. cytosolic) VPS26 (n, number of transfected cells scored). In ~75% of cells expressing the wild-type TBC1D5, VPS26 was scored as displaced, whereas <10% of cells expressing GFP-TBC1D5 RQ were scored as displaced.

Fig. 3.

TBC1D5 interacts with VPS29 through the L152-centered hydrophobic patch. (A) TBC1D5 cloned into the Y2H ‘prey’ vector, pGAD424, was tested against VPS35, VPS29 and VPS26 expressed from the Y2H ‘bait’ vector, pGBT9. Growth on –His plates was observed for the pGBT9-VPS29 and pGAD424-TBC1D5 combination. (B) Mutation of the L152 residue on VPS29 blocks the interaction with TBC1D5 but not with VPS35. (C) Truncations of TBC1D5 (tagged with GFP) were tranfected into HeLa cells. Twenty-four hours after transfection, the cells were lysed and the lysates IPed with anti-GFP. Only full-length TBC1D5 retains a robust interaction with retromer.

Fig. 4.

TBC1D5 requires the cargo-selective retromer complex to target to endosomes. (A) HeLa cells stably expressing GFP-TBC1D5 RQ were treated with siRNA to knockdown (KD) TBC1D5, VPS26, VPS35, SNX1 or RAB7. Loss of VPS26, VPS35 or RAB7 results in GFP-TBC1D5 redistributing into the cytoplasm but SNX1 KD does not affect GFP-TBC1D5 localisation. Scale bar: 20 μm. (B) Western blot of lysates from cells treated as in A confirms the efficacy of the siRNA targeting TBC1D5, VPS26, VPS35 and SNX1.

Fig. 5.

In vivo analysis of the interactions between the cargo-selective complex and the various retromer-interacting proteins identified in Fig. 1. (A) Cells stably expressing GFP-TBC1D5 RQ, VPS29-GFP, VPS29I91S-GFP, VPS29L152E-GFP or GFP-VPS35 were lysed and the lysates were treated with anti-GFP. After washes, the IPs were analysed by SDS-PAGE. GFP-TBC1D5 RQ co-IPed the retromer cargo-selective (VPS35-VPS29-VPS26) complex. Endogenous TBC1D5 (arrowhead) co-IPed with VPS29-GFP and GFP-VPS35 but was not observed in the VPS29I91S or the VPS29L152E lanes. Bands corresponding to the other retromer-interacting proteins (FAM21, FKBP15, KIAA1033 and strumpellin, highlighted with dots) were observed in the VPS29, VPS29L152E and VPS35 lanes. (B) Samples identical to those in A were analysed by western blotting, confirming that TBC1D5 does not co-IP with the VPS29L152E mutant. (C) Samples identical to those in A were analysed by LC-MSMS. The different proteins have been colour-coded as follows: red, retromer; green, GFP; blue, TBC1D5; pink, FKBP15; grey, CAPZa and CAPZb; and yellow, strumpellin, KIAA1033, FAM21 and WASH. Proteins considered to be contaminants were left unshaded, e.g. TRIM21 is a contaminant in these native IPs and is present owing to its IgG binding domain (Keeble et al., 2008). No peptides corresponding to TBC1D5 were detected in the VPS29L152E sample and no peptides from SNX proteins were detected in any samples.

Fig. 6.

Yeast two-hybrid (Y2H) and biochemical analysis of interactions between the retromer cargo-selective complex and retromer interacting proteins. (A) A cartoon diagram of the Y2H interactions of the retromer cargo-selective complex and the various retromer-interacting proteins reported herein. (B) Cells expressing VPS29-GFP were lysed and subjected to centrifugation on a 5-30% sucrose velocity gradient. Fractions were collected and analysed by SDS-PAGE (lower panel) and western blotting (upper panel). VPS35 and VPS26 co-fractionated. WASH1 and strumpellin also co-fractionated in the later (larger complex) fractions. FKBP15 was detected in multiple fractions. (C) VPS29-GFP cells were treated with siRNA to KD various proteins and then lysed. Anti-GFP was used to IP the VPS29-GFP and associated proteins. The IPs were analysed by western blotting. The association of strumpellin with the retromer cargo-selective complex is most strongly affected by KD of KIAA1033, RAB7 or SNX1.

Fig. 7.

WASH1 and FKBP15 require the cargo-selective retromer complex for their membrane association. (A) HeLa cells treated with siRNA were fixed and labelled with antibodies against WASH1 and SNX1. WASH1 localisation is lost after KD of KIAA1033, VPS26 or VPS35 but SNX1 or FKBP15 KD did not affect WASH1 localisation. Scale bar: 20 μm. (B) HeLa cells were transiently transfected with GFP-tagged constructs of strumpellin, KIAA1033 and FAM21. In control cells, the GFP-tagged constructs colocalise with VPS26 but, after VPS26 KD, no membrane association is observed. Scale bar: 20 μm. (C) Cells were treated as in A but this time were labelled with antibodies against FKBP15 and SNX1. FKBP15 localisation was most strongly affected by KD of VPS35, VPS26 and WASH1. Scale bar: 20 μm. (D) Cells treated with siRNA to silence VPS26 or VPS35 were labelled with antibodies against EEA1 and WASH1 or FKBP15. In control cells, EEA1 partially colocalises with both WASH1 and FKBP15 (indicated with arrowheads), but in VPS26 KD or VPS35 KD cells, the membrane association of WASH1 and FKBP15 is lost. EEA1 remains associated with punctate endosomal structures. Scale bar: 20 μm.

Fig. 7.

WASH1 and FKBP15 require the cargo-selective retromer complex for their membrane association. (A) HeLa cells treated with siRNA were fixed and labelled with antibodies against WASH1 and SNX1. WASH1 localisation is lost after KD of KIAA1033, VPS26 or VPS35 but SNX1 or FKBP15 KD did not affect WASH1 localisation. Scale bar: 20 μm. (B) HeLa cells were transiently transfected with GFP-tagged constructs of strumpellin, KIAA1033 and FAM21. In control cells, the GFP-tagged constructs colocalise with VPS26 but, after VPS26 KD, no membrane association is observed. Scale bar: 20 μm. (C) Cells were treated as in A but this time were labelled with antibodies against FKBP15 and SNX1. FKBP15 localisation was most strongly affected by KD of VPS35, VPS26 and WASH1. Scale bar: 20 μm. (D) Cells treated with siRNA to silence VPS26 or VPS35 were labelled with antibodies against EEA1 and WASH1 or FKBP15. In control cells, EEA1 partially colocalises with both WASH1 and FKBP15 (indicated with arrowheads), but in VPS26 KD or VPS35 KD cells, the membrane association of WASH1 and FKBP15 is lost. EEA1 remains associated with punctate endosomal structures. Scale bar: 20 μm.

Fig. 8.

Increased retromer-positive tubules observed after loss of strumpellin or KIAA1033. (A) HeLa cells treated with siRNA to KD strumpellin were labelled with antibodies against VPS26 and SNX1. Fine tubules observed after strumpellin KD are readily observed but less frequently seen in control cells. Arrows indicate tubules labelled with SNX1; arrowheads indicate VPS26-labelled tubules. Scale bar: 20 μm. (B) HeLa cells treated with various siRNA were labelled with antibodies against VPS26 or SNX1 (as in A) and then were scored for SNX1- or VPS26-positive tubules in a blind experiment. More than 100 cells for each KD were observed. The strumpellin or KIAA1033 KDs increase the frequency of SNX1- and VPS26-positive tubules but the tubules still require EHD1 function. (C) Tubules were measured in 10 cells that had at least three SNX1-tubules. Although WASH1 KD did not increase the number of tubules observed, it did result in the observed tubules being much longer.

Fig. 9.

The VPS35-VPS29-VPS26 retromer complex and SNX1 interact separately with the strumpellin complex. (A) Native IPs were performed on lysates from HeLa cells. Both VPS26 and SNX1 co-IP strumpellin but they do not co-IP each other. (B) Native IPs of VPS26 and SNX1 were performed on control HeLa cells or cells in which strumpellin and KIAA1033 or TBC1D5 expression had been silenced with siRNA. Loss of TBC1D5 does not increase the interaction between the cargo-selective complex and SNX1. Therefore, it seems probable that the cargo-selective complex is not linked to the SNX1-complex through strumpellin and that TBC1D5 does not compete with SNX proteins for binding to VPS29.

Fig. 10.

Analysis of the role of the retromer-interacting proteins in endosome-to-Golgi retrieval. (A) CD8-CIMPR-expressing cells were treated with siRNA to double KD VPS26 and VPS35 or strumpellin and KIAA1033. The cells were incubated with anti-CD8 for 15 minutes and then chased for 30 minutes. In control cells, endocytosed anti-CD8 colocalises with TGN46. Double KD of VPS26 and VPS35 causes the anti-CD8 to accumulate in peripheral structures but double KD of strumpellin and KIAA1033 does not appear to block retrieval. Scale bar: 20 μm. (B) Quantitative analysis of endosome-to-Golgi retrieval. In control cells, 75% of cell-associated anti-CD8 is coincident with the GFP-GMX33, giving a retrieval ratio of 0.75. RNAi KD of VPS26, SNX1, RAB7 or SNX3 inhibits retrieval, lowering the ratio; however, KD of TBC1D5, strumpellin, KIAA1033 or FKBP15 also did not affect endosome-to-Golgi retrieval at 37°C. (C) As in B, but the assay was performed at 32°C.