Acute perturbation of retromer and ESCPE-1 leads to functionally distinct and temporally resolved defects in endosomal cargo sorting

The human retromer is a stable heterotrimer of VPS35, VPS29 and VPS26 whose principal role is to orchestrate the endosomal retrieval of hundreds of internalised cargo and promote their recycling to the cell surface; a prototypical cargo being the glucose transporter GLUT1. Retromer’s role in a distinct endosomal retrieval pathway, the retrograde sorting of the cation-independent mannose 6-phosphate receptor (CI-MPR) to the trans-Golgi network (TGN), remains controversial. Here we have developed and applied knocksideways to acutely inactivate retromer and by visualising the sorting of endogenous GLUT1 and CI-MPR provide insight into the temporal dynamics of endosomal cargo sorting in HeLa and H4 human neuroglioma cells. While retromer knocksideways led to the development of time-resolved defects in cell surface sorting of GLUT1 we failed to observe defects in the sorting of the CI-MPR. In contrast knocksideways of ESCPE-1, a key regulator of retrograde CI-MPR sorting, resulted in a time-resolved defect in CI-MPR sorting. Together these data provide independent evidence consistent with a comparatively limited role for retromer in ESCPE-1 dependent CI-MPR retrograde sorting in HeLa and H4 human neuroglioma cells.


INTRODUCTION
The endosomal pathway functions as a major intracellular hub for the sorting of numerous integral proteins, which include signalling receptors, adhesion molecules, nutrient transporters, ion channels, and their associated proteins and lipids (collectively termed 'cargoes') (Maxfield and McGraw, 2004;Grant and Donaldson, 2009;Cullen and Steinberg, 2018). On entering the pathway, cargoes are sorted between two fates: they are either selected for degradation within the lysosome or retrieved from this fate and promoted for recycling to the plasma membrane and the trans-Golgi network (TGN) (Cullen and Steinberg, 2018). The efficient sorting of cargo is essential for normal cellular homeostasis and defects in sorting are increasingly linked with human physiology and pathophysiology (Schreij et al., 2016;Cullen and Steinberg, 2018).
Together, cargo recognition and organisation of a localised F-actin network leads to the formation of one or more retrieval sub-domains on the cytosolic face of the endosomal membrane that provide platforms for the co-ordinated biogenesis of cargo-enriched transport carriers (Puthenveedu et al., 2010).
At steady state, CI-MPR is enriched at the TGN where it associates with newly synthesised hydrolase precursors (Braulke and Bonifacino, 2009). The resulting CI-MPR:hydrolase complex is transported to the endosomal pathway, where the acidified endosomal lumen induces the release of the hydrolase. While the hydrolase precursors are delivered to the lysosome, where they contribute to the degradative capacity of this organelle, the unoccupied CI-MPR is retrieved and recycled to the TGN for further rounds of hydrolase delivery. Many studies in mammalian cells are consistent with retromer in regulating CI-MPR transport (Arighi et al., 2004;Seaman, 2004;Seaman, 2007;Wassmer et al., 2007;Bulankina et al., 2009;Harbour et al., 2010;Hao et al., 2013;Breusegem and Seaman, 2014;McGough et al., 2014;Cui et al., 2019). However, we and others have recently questioned this model (Kvainickas et al., 2017;Simonetti et al., 2017). Rather, structural, biochemical and functional evidence has associated ESCPE-1 in sequencedependent endosome-to-TGN CI-MPR transport through direct recognition of a bipartite sorting motif (Kvainickas et al., 2017;Simonetti et al., 2017;Simonetti et al., 2019).
Part of this controversy may stem from technical variability and in particular the reliance on the generation of retromer knockdown and knockout cells (Seaman, 2018). These procedures induce the gradual loss of retromer expression over the course of hours and days, a time window that has the potential to initiate the activation of compensatory pathways that suppress phenotypes or result in variable and subtle phenotypes. Here we have applied the 'knocksideways' methodology (Robinson et al., 2010) to acutely remove retromer and trap this complex on an organelle not implicated in retromer function. Using time-resolved analysis of cargo trafficking, we show that while acute retromer inactivation leads to robust defects in the endosomal recycling of the prototypical retromer cargo GLUT1, no detectable perturbation was observed in the distribution of the CI-MPR. In contrast, acute knocksideways mediated inactivate of ESCPE-1 results in perturbed endosome-to-TGN transport of the CI-MPR. Our study therefore defines a method for the acute inactivation of endosomal retrieval and recycling complexes and provides further data to support the need to reflect on the central role of retromer in the retrograde sorting from the endosome to the trans-Golgi network.

Retromer knocksideways design and temporal dynamics.
To design the retromer knocksideways, we first engineered a cassette encoding for the core VPS35 subunit fused through a carboxy-terminal flexible linker to green fluorescent protein (GFP), which was itself linked to the amino-terminus of rapamycin-binding (FRB) domain (resultant construct encode VPS35-GFP-FRB, Figure 1A). In light of evidence linking retromer to aspects of lysosomal function (e.g. Kvainickas et al., 2019), we utilised a modified version of FRB (T2098L) that enables the induction of heterodimerisation by rapalog (AP21967), a compound that has a lower affinity to endogenous mTOR than rapamycin (Clackson et al., 1998). To validate that the VPS35-GFP-FRB chimera was functional we expressed the construct in a previously characterised VPS35 knockout HeLa cell line (Kvainickas et al., 2017). The VPS35-GFP-FRB chimera localised to cytosolic puncta that were identified as endosomes by means of colocalization with the endosome marker EEA1 (Supplementary Figure 1A). GFP-nanotrap immunoprecipitation confirmed that VPS35-GFP-FRB retained the ability to associate with the core retromer components VPS26 and VPS29 (Supplementary Figure 1B). Consistent with retromer assembly, expression of the VPS35-GFP-FRB chimera in the VPS35 knockout HeLa cells reverted the observed lysosomal missorting of GLUT1 and allowed recycling of the transporter back to the cell surface (Supplementary Figure 1C and D). The designed VPS35-GFP-FRB chimera is therefore correctly localised to endosomes and displays the ability to assemble into a retromer complex that retains its function in endosomal cargo retrieval and recycling.
To engineer the acceptor compartment, we fused red fluorescent protein (RFP) to FKBP and linked this to either a mitochondrial targeting sequence (yeast Tom70pforming Mito-RFP-FKBP (Robinson et al., 2010) or a peroxisomal targeting sequence (PEX3 (residues 1-42)forming PEX-RFP-FKBP (Kapitein et al., 2010)). To ensure a balanced coexpression, we cloned the genes encoding for Mito-RFP-FKBP and VPS35-GFP-FRB into a bicistronic vector and generated a corresponding bicistronic vector for PEX-RFP-FKBP and VSP35-GFP-FRB ( Figure 1A). To visualise the temporal dynamics of retromer knocksideways we performed live imaging immediately after the application of rapalog.
For both the mitochondrial and peroxisomal knocksideways systems we observed dynamic accumulation of VPS35-GFP-FRB onto the corresponding acceptor compartment (Supplementary Movies 1A and B), such that approximately 10 minutes after induction of dimerization there was clear co-localization between retromer and the acceptor compartment ( Figure 1B and Supplementary Figure 1E).
Considering that retromer has been implicated in mitochondrial function (Braschi et al., 2010;Tang et al., 2015;Wang et al., 2016), we decided to focus on developing the peroxisomal acceptor compartment system: to date, peroxisomes have not been implicated in retromer biology. To increase the capacity of the acceptor compartment we converted PEX-RFP-FKBP to PEX-Myc-3xFKBP (each FKBP separated by flexible linkers GGSGGGSGGAP) ( Figure 1A). In transiently transfected HeLa cells, the PEX-Myc-3xFKBP chimera displayed colocalization with the known peroxisome marker PMP70 (Supplementary Figure 1F).
In VPS35 knockout HeLa cells transiently transfected to express PEX3-myc-3xFKBP and VPS35-GFP-FRB, the addition of 100 nM of rapalog established that rerouting of VPS35-GFP-FRB from EEA1-positive endosomes to peroxisomes was achieved within 10 minutes and was complete by 30 minutes ( Figure 1C-E) -in the continued presence of rapalog the peroxisome rerouted VPS35-GFP-FRB was retained on this organelle (maximum time studied 24 hours). Together these data establish a method for the acute knocksideways of a functional VPS35-GFP-FRB construct.
Using knocksideways to examine in cellulo retromer assembly.
GFP-nanotrap immunoisolation is an established method for identifying protein-protein interactions, including those of the retromer complex (McGough et al., 2014;McMillan et al., 2016). Here we used knocksideways to analyse protein-protein interactions in cellulo.
Consistent with the assembly of VPS35-GFP-FRB into a functional heterotrimeric complex (Supplementary Figure 1B), analysis of the endogenous localisation of VPS26 revealed that it too was rerouted to peroxisomes with a similar kinetic profile to that observed for VPS35-GFP-FRB (the lack of a suitable antibody precluded the equivalent analysis of VPS29) (Figure 2A and B). In addition, the major retromer accessory complex, the FAM21 containing WASH complex (Derivery et al., 2009;Gomez and Billadeau, 2009;Harbour et al., 2012;Jia et al., 2012) was also rerouted to peroxisomes upon retromer knocksideways ( Figure 2C and D, Supplementary Figure 2A  Given the selectivity of retromer knocksideways we also decided to apply this methodology to examine the in cellulo relationship between retromer and the SNX-BAR proteins that assemble to form the ESCPE-1 complex (Simonetti et al., 2019). In yeast these SNX-BAR proteins associate with the Vps26:Vps35:Vps29 heterotrimer to form the stable pentameric retromer complex (Seaman et al., 1998). In metazoans however, retromer and ESCPE-1 appear to function independently, which is inconsistent with the formation of a long-lived and stable pentameric complex (Kvainickas et al., 2017;Simonetti et al., 2017). Indeed, we failed to observe the rerouting of endogenous SNX1, a component of the ESCPE-1 complex, onto peroxisomes after 24 hours of rapalog treatment ( Figure 2F-H). These data therefore support the in cellulo and in vivo evidence that in metazoans retromer and ESCPE-1 have evolved into functionally distinct complexes (Kvainickas et al., 2017;Simonetti et al., 2017;Simonetti et al., 2019;Strutt et al., 2019). Overall, the designed VPS35 knocksideways provides a method for the acute and selective rerouting of retromer (and its functional coupled accessory proteins) away from endosomes to the neighbouring peroxisome organelle.

Acute retromer knocksideways leads to a time-resolved GLUT1 sorting defect.
Retromer and retromer-associated cargo adaptors have been shown to control the endosomal retrieval and recycling of >400 cell surface proteins including the glucose transporter GLUT1 (Steinberg et al., 2013). To define the functional consequences of retromer knocksideways we examined the steady-state distribution of GLUT1 in VPS35 knockout HeLa cells rescued by expression of VPS35-GFP-FRB knocksideways.
Following the addition of rapalog for 24 hours, fixed cell confocal imaging revealed a GLUT1 missorting phenotype, defined by the steady-state loss of GLUT1 at the cell surface and the enrichment of GLUT1 with LAMP1-positive late endosomes / lysosomes ( Figure 3A and B). To time-resolve the appearance of the GLUT1 trafficking phenotype, we fixed cells at various time points following rapalog addition. Quantification established that a statistically significant GLUT1 missorting phenotype began to emerge after 1-3 hours of retromer knocksideways and reached a maximum penetrance after 10 hours ( Figure 3C and D). The difference between the time scales of retromer knocksideways ( Figure 1C-E) compared with the appearance of the GLUT1 missorting phenotype is entirely consistent with the known rate of GLUT1 lysosomal mediated degradation observed upon retromer suppression and reflects the relatively slow rate of endocytosis of this transporter (Steinberg et al., 2013).
The missorting of GLUT1 upon retromer knocksideways was not the result of a global effect on endosomal sorting as the endosomal retrieval and recycling of the retrieverdependent cargo, α5β1-integrin was not affected upon retromer knocksideways (Supplementary Figure 3A and B) -consistent with the lack of effect of retromer knocksideways on the endosomal association of the retriever complex (Supplementary Figure 2E-G). Moreover, the development of the GLUT1 missorting did not stem from the recruiting of 'foreign' proteins to peroxisomes as retromer knocksideways performed in wild-type HeLa cells, which retain expression of endogenous VPS35 that is not subject to knocksideways, did not illicit the development of a GLUT1 missorting phenotype (Supplementary Figure 3C and D). Together, these data support that it is the specific removal and inactivation of retromer that causes the time-resolved development of the observed GLUT1 missorting phenotype.

Retromer-independent CI-MPR retrograde trafficking.
Next, we investigated the role of retromer in the retrograde trafficking of CI-MPR from endosomes to the TGN (Arighi et al., 2004;Seaman, 2004;Kvainickas et al., 2017;Simonetti et al., 2017;Cui et al., 2019). In VPS35 knockout HeLa cells rescued through expression of VPS35-GFP-FRB, the CI-MPR is chiefly localized to the perinuclear TGN as defined through co-localization with TGN markers Golgin97 and TGN46 ( Figure 4A and B, Supplementary Figure 4A and B). After the addition of rapalog and initiation of retromer knocksideways we failed to observe a detectable alteration in the steady-state distribution of the CI-MPR ( Figure 4A and B, Supplementary Figure 4A and B) over a time frame where the endosomal missorting of internalized GLUT1 was readily observed ( Figure 3C and D). Given that the endosome-to-TGN transport of the CI-MPR is considered to occur over a period of approximately 20 -30 minutes (Seaman, 2004), the acute perturbation of retromer function does not appear to lead to a detectable defect in the endosomal sorting of the CI-MPR. These data are consistent with a limited role for retromer in endosome-to-TGN retrograde transport of this receptor (Kvainickas et al., 2017;Simonetti et al., 2017).

Knocksideways of ESCPE-1 leads to a time-resolved defect in CI-MPR sorting.
The ESCPE-1 complex regulates sequence-dependent endosome-to-TGN transport of the CI-MPR (Simonetti et al., 2019). ESCPE-1 comprises a heterodimer of SNX1 or SNX2 (these proteins are functionally redundant) associated with either SNX5 or SNX6, which are also functionally redundant (Wassmer et al., 2009). Of these proteins, it is the PX domains of SNX5 and SNX6 that directly bind to the ΦxΩxΦ(x)nΦ sorting motif in CI-MPR to mediate endosome-to-TGN transport (Simonetti et al., 2019). To provide a positive control for the lack of detectable effect of retromer knocksideways on CI-MPR trafficking we therefore constructed a bicistronic vector encoding for PEX3-myc-3xFKBP and GFP-FRB-SNX5 ( Figure 1A). When expressed in HeLa cells SNX5-GFP-FRB localised to endosomes as defined by co-localization with EEA1 (Supplementary Figure 5A).
Interestingly, after rapalog addition we observed a slight recruitment of EEA1 to the peroxisomal hook indicating a movement of the endosomal compartment to the peroxisomal compartment ( Figure 5A and B). However, this 'dragging' of endosomes was not complete as there was still a loss of co-localization between GFP-FRB-SNX5 and EEA1 ( Figure 5A and C). In GFP-FRB-SNX5 knocksideways endogenous SNX1 was recruited to peroxisomes after rapalog treatment with no loss of co-localization between SNX5 and SNX1 indicating a recruitment of the functional ESCPE-1 complex ( Figure 5D-F).
Next we used the GFP-FRB-SNX5 knocksideways system to time-resolve CI-MPR endosome-to-TGN trafficking. Expression of the GFP-FRB-SNX5 chimera in a previously isolated and characterised SNX5/SNX6 double knockout HeLa cell line (Simonetti et al., 2017) reverted the observed missorting of the CI-MPR and allowed the receptor to return to its normal steady-state localisation (Supplementary Figure 6A and B). Consistent with the role of SNX5 in the ESCPE-1-mediated endosome-to-TGN transport of the CI-MPR (Kvainickas et al., 2017;Simonetti et al., 2017;Simonetti et al., 2019), SNX5 knocksideways in SNX5/SNX6 double knockout HeLa cells led to the time-resolved appearance of a CI-MPR missorting phenotype as defined by a reduced enrichment of the CI-MPR at the Golgin97 or TGN46-labelled TGN with a maximum penetrance at 6 hours ( Figure 6A and B, Supplementary Figure 6C and D). There was no observed defect in α5β1-integrin recycling during GFP-FRB-SNX5 knocksideways indicating the selective nature of this procedure ( Figure 6C and D). Together these data establish that acute perturbation of the ESCPE-1 complex leads to a missorting of CI-MPR, thereby reinforcing the functional significance of the lack of observed effect of retromer knocksideways on the trafficking of this receptor.

H4 human neuroglioma knocksideways is consistent with a limited role for retromer in retrograde CI-MPR trafficking.
Our study of endosomal cargo sorting associated with depletion or knocksideways of sorting machinery has so far been limited to a single non-neuronal cell type. To extend these observations, we therefore turned to H4 neuroglioma cells and generated both retromer knockout (targeting VPS35) and ESCPE-1 knockout cells (dual targeting of SNX5 and SNX6). Interestingly, confocal imaging of the retromer knockout cells revealed an enhanced intensity in the staining of endogenous CI-MPR that was not observed in the ESCPE-1 knockout cells ( Figure 7A and B). Despite the increase in the CI-MPR signal intensity, retromer knockout cells did not display a significant change in the quantified Pearsons correlation coefficient between CI-MPR and Golgin97. This is consistent with a limited role for retromer in CI-MPR retrograde trafficking in H4 neuroglioma cells ( Figure   7A  Together these data are more suggestive of the observed up-regulation in expression of CI-MPR in retromer knockout cells arising as an indirect affect of a compensatory pathway(s) rather than a direct endosome-to-TGN sorting phenotype.

DISCUSSION
Here we have developed and applied knocksideways to acutely inactivate retromer and the ESCPE-1 complex. Previously developed to inactivate the AP1 and AP2 clathrin adaptors (Robinson et al., 2010;Hirst et al., 2012), this approach provides a method to acutely perturb sorting complex function in a time frame that better aligns with the dynamic nature of endosomal membrane trafficking. By visualising the sorting of endogenous GLUT1 and CI-MPR, our data provide insight into the temporal dynamics of endosomal cargo sorting and support the established role of retromer in cell surface recycling (Chen et al., 2010;Temkin et al., 2011;Steinberg et al., 2013). In contrast, we have provided independent evidence consistent with a comparatively limited role for retromer in ESCPE-1 dependent CI-MPR retrograde sorting in HeLa and H4 human neuroglioma cells (Kvainickas et al., 2017;Simonetti et al., 2017;Simonetti et al., 2019). In contrast to retromer's pentameric assembly in yeast (Seaman et al., 1998) The development of retromer knocksideways has also added to our understanding of the endosomal association of the actin polymerising WASH complex. Direct binding of FAM21 to VPS35 is a major mechanism for the retromer-dependent association of the WASH complex to endosomes (Gomez and Billadeau, 2009;Harbour et al., 2012;Jia et al., 2012). That said, increasing evidence suggests that a sub-population of the WASH complex is associated to endosomes independent of retromer (McNally et al., 2017;Kvainickas et al., 2017;Simonetti et al., 2017;MacDonald et al., 2018). Consistent with these data, acute knocksideways of retromer induces a redistribution of a major proportion of endogenous WASH but a significant sub-population retains an endosomal association.
In summary, by applying knocksideways we have visualized the acute inactivation of retromer and ESCPE-1 and, through quantification of the temporal development of resulting cargo sorting defects, provided clarification of the role of these complexes in the sorting of CI-MPR and GLUT1. Moving forward, we aim to combine knocksideways and our established unbiased proteomic quantification of the cell surface proteome to timeresolve the functional role of endosomal sorting complexes in global cargo retrieval and recycling.

Cell culture and DNA transfection
HeLa and H4 neuroglioma cells were cultured in DMEM (Sigma) supplemented with 10% (vol/vol) FCS (Sigma) and grown using standard conditions. Lipofectamine LTX was used in DNA transfections. Per six well dish 2 µg of DNA was mixed with 4 µL of the LTX supplement into 100 µL of Opti-Mem. In another incubation 100 µL of Opti-Mem was mixed with 8 µL of Lipofectamine LTX. After a 5-minute incubation, the two 100 µL Opti-Mem mixes were combined and incubated for a further 20 minutes. The 200 µL mix was then added dropwise onto 60-80% confluent HeLa cells and transfected cells were left for 48 hours for DNA expression. VPS35 knockout HeLa cells and SNX5/SNX6 double knockout was generated as previously described and cultured as stated above for wildtype HeLa (Simonetti et al., 2017).

Generation of H4 clonal cells
H4 cells were seeded the day prior to transfection, then transiently transfected with CRISPR plasmids encoding the Cas9 enzyme, a puromycin resistance marker and specific gRNA guides against VPS35, SNX5 or SNX6 (Kvainickas et al., 2017;Simonetti et al., 2017) using FuGENE® 6 (Promega). The day after transfection, cells are incubated with 1 μg/mL puromycin for 24 hours to select for knockout cells. Following puromycin selection, H4 cells were seeded into a 96-well plate at a density of 1 cell per well in 200 μL Iscove's Modified Dulbecco's Medium (Thermo Fisher). Clones were grown to confluency then expanded and screened for successful knockout deletion by Western blotting.

GFP trap and Western blot analysis
Cells were lysed in GFP trap buffer (50 mM Tris-HCl, 0.5% NP-40 and Roche protease inhibitor cocktail) and the lysate was added to pre-equilibrated GFP trap beads (ChromoTek). Beads were washed 3 times in the GFP trap buffer and then lysates were diluted in 2x sample buffer and boiled for 10 minutes. Proteins were resolved on a NuPAGE 4-12% gels (Invitrogen) and transferred onto polyvinylidene fluoride membrane (EMD Millipore). Once transferred membranes were blocked in TBS 5% milk and the primary antibody (see antibody section) was diluted in TBS 0.1% (v/v) Tween-20 (TBS-T) 5% (w/v) milk and incubated with the membrane for 1 hour at room temperature or overnight at 4 o C. Membranes were washed 3 times in TBS-T. Secondary antibodies (see antibody section) were diluted into TBS-T with 5% milk and 0.01% SDS and incubated with the washed membrane for 1 hour at room temperature. TBS-T was used to wash the membrane (3 times) prior to quantitative imaging using an Odyssey scanning system (LI-COR Biosciences). Analysis was performed on Image Studio lite (LI-COR Biosciences).

Knocksideways.
pIRESNEO3 bicistronic vectors encoding the knocksideways peroxisomal/mitochondrial acceptor compartment and either VPS35-GFP-FRB or GFP-FRB-SNX5 were transfected into cells. Either 24 hours of 0.1% (v/v) ethanol vehicle was added to the 0 timepoint or rapalog (100 nM) was added to the 24-hour timepoint. The following day rapalog was added to the different timepoints and then cells were fixed and stained.

Immunofluorescent staining
Cells were washed once in PBS before fixation for 8 minutes in 4% PFA (16% PFA stock diluted in PBS). Three washes in PBS were performed and then a 5-minute incubation with PBS 100 mM glycine was used to quench the PFA. After three more PBS washes cells were left in PBS overnight. Cells were incubated with PBS + 3% BSA and 0.1% Triton-X100 for 10 minutes and then with PBS + 3% BSA for a further 10 minutes. Primary antibodies (see antibody section) were diluted in PBS + 3% BSA and incubated for 1 hour.

Cells were washed 3 times with PBS with the secondary antibody (see antibody section)
being diluted into PBS + 3% BSA and incubated for 1 hour. Cells were washed 3 times with PBS and washed once with distilled water before mounting the coverslips in Fluoromount-G.

Microscopy and image analysis
For image acquisition a Leica SP5-AOBS confocal laser scanning confocal microscope was used attached to a Leica DM I6000 inverted epifluorescence microscope. The 63x HCX PL APO oil lens and standard acquisition software and detectors were used. Once acquired Pearson's correlation colocalization and signal intensity analyses were quantified using Volocity 6.3 software (PerkinElmer). Image and line scan analysis was completed using ImageJ FIJI software. GraphPad Prism 7 was used for presentation and statistical analysis of data.
Live cell imaging was performed using a Leica SP8 AOBS confocal laser scanning microscope attached to a Leica DM I6000 inverted epifluorescence microscope. The adaptive focus control was used to prevent drift of the Z plane over time. The two hybrid GaAsP detectors were used to enable low laser settings. Images were acquired using the 63x HC PL APO CS2 lens and a speed of one image per 10 seconds. Imaging was performed at 37 o C and 2x rapalog DMEM complete media was added to the pre-selected cell.