Sequence-dependent trafficking of GDE2, a GPI-specific phospholipase promoting neuronal differentiation

GDE2 is a six-transmembrane glycerophosphodiesterase with phospholipase D-like activity that cleaves select glycosylphosphatidylinositol (GPI)-anchored proteins and thereby influences biological signaling cascades. GDE2 promotes neuronal differentiation cell-autonomously through glypican cleavage and is a prognostic marker in neuroblastoma, while GDE2 deficiency causes progressive neurodegeneration in mice and developmental defects in zebrafish. However, the regulation of GDE2 remains unclear. Here we show that in undifferentiated neuronal cells, GDE2 undergoes constitutive internalization and traffics back along both fast and slow recycling routes, while a small percentage is sorted to late endosomes. GDE2 trafficking is dictated by distinctive C-terminal tail sequences that determine secretion, endocytosis and recycling preference, respectively, and thereby regulate GDE2 function both positively and negatively. Our study reveals the sequence determinants of GDE2 trafficking and surface localization, and provides insight into the control of GPI-anchored protein activities with potential implications for nervous system disorders associated with impaired trafficking and beyond.


Introduction
The surface of eukaryotic cells contains a great variety of glycosylphosphatidylinositol (GPI)anchored proteins, many of which are involved in the regulation of vital cellular functions, including receptor signaling, cell adhesion, differentiation and cell-cell communication. GPIanchoring is a highly complex post-translational modification that tethers membrane proteins via their C-terminus to a unique glycosylated phosphatidylinositol (PI) core in the outer leaflet of the plasma membrane (Ferguson et al., 2015;Fujita and Kinoshita, 2010;Paulick and Bertozzi, 2008). Since they lack a transmembrane domain, GPI-anchored proteins cannot transmit signals by themselves but must interact with transmembrane effectors to achieve signaling competence.
Importantly, GPI-anchoring confers a unique property to membrane proteins, namely susceptibility to phospholipase attack. Indeed, GPI-anchored proteins can be released from their anchor and detected as soluble proteins, some of which are considered disease biomarkers.
Yet, identification of the responsible phospholipase(s) has long been elusive. A secreted GPIspecific phospholipase D (GPI-PLD) does not act on intact cells (Low and Huang, 1991), whereas a cell-associated GPI-PLD activity that releases GPI-anchored proteins has remained unidentified to date (Metz et al., 1994).
Recent studies have advanced the field by showing that members of the glycerophosphodiester phosphodiesterase (GDPD) family, notably GDE2 and GDE3, function as GPI-specific phospholipases that cleave select GPI-anchored proteins and thereby alter signaling cascades and cell behavior (Matas-Rico et al., 2016;Park et al., 2013;van Veen et al., 2017). GDE2 (or GDPD5) is the best studied family member and, along with GDE3 and GDE6, is characterized by six transmembrane domains, a catalytic GDPD ectodomain and intracellular N-and Cterminal tails (Fig. 1A,B). GDE2 acts in a phospholipase D (PLD)-like manner towards soluble substrates (i.e. releasing choline from glycerol-3-phosphocholine (Gallazzini et al., 2008)) in common with virtually all GDPD family members (Corda et al., 2014;Ohshima et al., 2015). One notable exception is GDE3 (GDPD2), which functions as a phospholipase C (PLC) showing substrate selectivity different from GDE2 (Corda et al., 2009;van Veen et al., 2017).
GDE2 was originally shown to drive neuronal differentiation and survival in the developing spinal cord through surface cleavage of GPI-anchored RECK, an ADAM metalloprotease inhibitor and Notch ligand regulator, leading to inhibition of Notch signaling and induction of differentiation in adjacent neural progenitors (Park et al., 2013). More recently, we reported that GDE2 promotes neuronal differentiation in a cell-autonomous manner through surface cleavage of Glypican 6 (GPC6), one of the six GPI-anchored heparan sulfate proteoglycans (Matas-Rico et al., 2016).
Enforced GDE2 expression led to altered Rho/Rac signaling, induction of neural differentiation markers, cell spreading, neurite outgrowth and resistance to RhoA-driven neurite retraction (Matas-Rico et al., 2016). Furthermore, GDE2 expression was found to strongly correlate with favorable outcome in neuroblastoma, a childhood malignancy characterized by impaired neuronal differentiation (Matas-Rico et al., 2016). In mice, Gde2 knockout led to progressive neurodegeneration in the spinal cord with pathologies reflecting human neurodegenerative disorders, which was accompanied by reduced glypican shedding (Cave et al., 2017). Finally, depletion of GDE2 in zebrafish embryos led to impaired motility and reduced pancreas differentiation and insulin expression; the latter phenotype could be rescued by human GDE2, indicating conservation of human and zebrafish GDE2 function (van Veen et al., 2018).
Altogether, these results underscore the need for tight control of GDE2 surface expression and activity in vivo. However, it is still unknown how GDE2 surface levels and function are regulated.
Here we report that in undifferentiated neuronal cells, GDE2 undergoes constitutive endocytosis and recycling along distinct Rab GTPase-regulated trafficking routes, while a small part is sorted to late endo-lysosomal compartments. Through progressive C-terminal truncations, we define distinctive sequences that govern GDE2 exocytosis, endocytosis and recycling pathway preference, respectively. We show that these sequences regulate GDE2 surface expression and function, both positively and negatively, as measured by GPC6 shedding and induction of neural differentiation markers. Our study establishes a link between endocytic recycling and GDE2 function in neuronal cells, of potential relevance for nervous system disorders associated with impaired membrane trafficking.

GDE2 localization and trafficking: predominance in recycling endosomes
We selected undifferentiated neuronal cells that express very low levels of endogenous GDE2, namely SH-SY5Y and N1E-115 cells (Matas-Rico et al., 2016). In these cells, GDE2 (GFP-, mCherry-or HA-tagged) is detected in discrete microdomains, or clustered nanodomains (rafts), as shown by super-resolution microscopy (Matas-Rico et al., 2016) (Fig. 1C). In addition, GDE2 is abundant in endocytic vesicles, particularly in the perinuclear region ( Fig. 1C) (Matas-Rico et al., 2016). Treatment of the cells with the dynamin inhibitor dynasore resulted in GDE2 accumulation at the plasma membrane with almost complete loss of GDE2-positive vesicles, indicating that GDE2 undergoes dynamin-dependent internalization instead of bulk endocytosis ( Fig. S1A). GDE2-GFP-containing vesicles are highly mobile and show rapid directional movement towards the tips of developing neurites, as shown by fluorescence live-imaging (Movie S1).
Rab GTPases are master regulators of membrane trafficking and show high selectivity for distinct endosomal compartments (Hutagalung and Novick, 2011;Wandinger-Ness and Zerial, 2014). To determine the nature of the GDE2-containing vesicles, we used Rab GTPase markers of early, late and recycling endosomes, respectively. In both neuronal cell lines, GDE2-GFP colocalized with early endosome marker Rab5-mCh ( Fig.1D and Fig. S1B). Rab5 coordinates clathrin-mediated endocytosis and biogenesis of early endosomes and their fusion. The majority of intracellular GDE2 was detected in two distinct populations of recycling endosomes, namely those representing the Rab4-and Rab11-dependent recycling routes ("fast" and "slow", respectively) ( Fig. 1D and S1B). Rab4-dependent fast recycling of membrane cargo involves a half-time of a few minutes, whereas Rab11 regulates slow recycling through perinuclear endosomes with a half-time of about 12 minutes (Maxfield and McGraw, 2004). In addition, GDE2 was found to co-localize with the endogenous transferrin receptor (TrfR), a prototypic cargo in clathrin/dynamin-mediated recycling pathways (Mayle et al., 2012).
Quantification of the results from two independent neuronal cell lines confirmed that GDE2 predominantly localized to Rab11 + endosomes (mean ~60%), somewhat less to Rab5 + and Rab4 + endosomes (mean 30-45%, depending on the cell line), and much less to Rab7 + late endosomes and lysosomes ( Fig. 1D and Fig. S2).

GDE2 interacts with Rab GTPases and undergoes constitutive endocytosis and recycling.
To validate the GDE2 localization data biochemically, we examined the interaction of GDE2 with relevant Rab GTPases in HEK293 cells. When GDE2 immuno-precipitates were blotted for Rab proteins, GDE2 was detected in complex with Rab4, Rab5, Rab7 and Rab11 (Fig. 1E), in support of the co-localization results. We then measured the internalization and recycling of GDE2 using a biotin labeling procedure (Fig. 1F). GDE2-mCh-expressing N1E-115 cells were surface-labeled with NHS-SS-Biotin at 4 o C and endocytosis was initiated by a temperature shift to 37 o C. Cell-surface biotin was stripped and recycling of the internal GDE2 pool was allowed to proceed for 15 or 30 min. Lysates were precipitated using Streptavidin beads and GDE2 was eluted from the beads using anti-mCh antibody. As shown in Fig. 1F, internalized GDE2 was found to recycle back to the plasma membrane within 15-30 min. Of note, the GDE2 intracellular pool was not affected by serum stimulation (Fig. 1F). Similarly, GDE2 co-localization with Rab4, Rab5 and Rab11 was not altered in the presence of serum (Results not shown). We conclude that GDE2 endocytosis and recycling is a constitutive process, insensitive to serum factors.

C-terminal tail truncations of GDE2 disclose unique regulatory sequences.
Many integral membrane proteins contain linear sequence motifs that determine endocytosis, recycling or degradation (Bonifacino and Traub, 2003;Cullen and Steinberg, 2018). However, sequence inspection did not reveal canonical sorting motifs in the cytoplasmic regions of GDE2, such as tyrosine-or leucine-based motifs. To explore the sequence determinants of GDE2 trafficking we focused on the C-terminal tail (CT; aa 518-605) ( Fig. 2A). Of note, the distal CT region of GDE2 shows marked sequence divergence among vertebrates. As shown in Fig. 2A, the aa 570-605 region is poorly conserved between mammalian, chicken and zebrafish GDE2, suggesting that the last 35 CT residues do not play a key regulatory role.
We made GDE2 CT truncations at aa 570, 560, 550, 540 and 530 (HA-tagged), respectively fast recycling endosomes at the expense of Rab11 co-localization ( Fig. 2B-F). This suggests that the aa (561-570) region of GDE2 is required for endocytosis and, in particular, determines recycling pathway preference, shifting endocytosed GDE2 from slow to fast recycling. Strikingly, when truncated at aa 550, GDE2(C550) accumulated at the plasma membrane, with little or no expression detected in early and recycling endosomes ( Fig. 2C-G). It thus appears that the GDE2 aa (551-560) region is required for GDE2 endocytosis and consequent recycling. Misfolded protein aggregates are usually routed to the autophagosome-lysosome degradation pathway. Indeed, GDE2(C530) was found to accumulate in lysosomes (Fig. S2). We therefore conclude that the CT (aa 541-550) region is required for proper GDE2 expression, folding and transport through the early secretory pathway. The respective GDE2-Rab co-localizations are quantified in Fig. 2F. Together, these results disclose unique regulatory sequences in the GDE2 CT.
After Dox treatment (24 hrs) of SH-SY5Y cells, the GDE2 truncation mutants showed subcellular localizations similar to those in N1E-115 cells (Fig. 3B). We next measured surface levels of GDE2 and its mutants relative to their intracellular accumulation by quantitative analysis of confocal images (Fig. 3C); moreover, we quantified GDE2 surface levels by FACS analysis (Fig.   3D). From both figures, it is seen that GDE2-FL, GDE2(C570) and GDE2(C550) show similar surface levels, while GDE2(C550) shows a much higher surface to cytoplasm ratio, consistent with its failure to undergo endocytosis. Finally, GDE2(C560) showed reduced surface expression, consistent with its preferred accumulation recycling endosomes, particularly the Rab4 + fast recycling ones (Fig. 2F)

GDE2 signaling activity: GPC6 shedding and induction of NEUROD1 and SNAP25
Catalytic activity of GDE2 and its truncation mutants was measured by loss of GPC6 from the plasma membrane, using GDE2(H233A) as a negative control (( Fig. 3D and Fig. S3C). GPC6 is the only endogenously expressed glypican in SH-SY5Y cells (Matas-Rico et al., 2016). Activity of GDE2 and its truncation mutants towards GPC6 correlated with their respective surface levels, except for GDE2(C560) (Fig. 3D). Strikingly, GDE2(C560) was virtually inactive despite significant surface expression, but correlating with its preference for the Rab4 + fast recycling pathway (Fig. 2F). This suggests that the aa 551-560 region negatively regulates GDE2 function. We next investigated GDE2-induced transcriptional responses in SH-SY5Y cells.
GDE2 upregulates the expression of various neuronal differentiation markers, including NEUROD1 and SNAP25 (Matas-Rico et al., 2016). NEUROD1 is a transcription factor that drives both neurogenesis and pancreas development (Cho and Tsai, 2004), while SNAP25 encodes a t-SNARE synaptic vesicle fusion protein (Kasai et al., 2012). How GPI-anchor surface cleavage leads to induction of neural differentiation genes is a key open question. We examined the expression of NEUROD1 and SNAP25 and found that their induction closely mirrored the degree of GPC6 shedding (Fig. 3E). GDE2(570) was fully active, similar to GDE2-FL, as was the GDE2(550) mutant that accumulates on the surface but lacks endocytosis and recycling competence. Strikingly again, GDE2(560) showed hardly any transcriptional responses, correlating with its lost ability to cleave GPC6, but at odds with its surface expression (Fig. 3E). Fig. 4 summarizes our assignment of consecutive CT sequences to GDE2 secretion, endocytosis and recycling preference, as well as to negative regulation of GDE2 activity.

Discussion
Proper trafficking and surface localization of GDE2 is vital for its GPI-hydrolyzing activity and biological outcome. Our results disclose unique C-terminal sequences that determine GDE2 trafficking, surface expression and function, both positively and negatively. In summary (Fig. 4): (i) CT region (aa 541-550) is essential for proper expression, exocytosis and membrane insertion but lacks endocytosis information; (ii) region (aa 551-560) confers endocytosis and ensuing recycling competence with a clear preference for the Rab4 + fast recycling route but, (iii), it renders GDE2 dysfunctional and thus qualifies as a negative regulatory sequence. Finally, the region (aa 561-570) is required for proper surface expression, endocytic recycling and biological activity, thus overriding negative regulation by upstream CT sequences. The finding that 10-aa sequence 551-EKLIFSEISD-560 confers loss of function to GDE2 is striking and warrants further investigation (Fig. 4B). One explanation could be that GDE2(560), in contrast to GDE2(570), prefers the Rab4 + fast recycling pathway and is mislocalized to membrane nanodomains that are short of GPC6, resulting in loss of GDE2 function. In this light, it will be crucial to elucidate how, and precisely where, GDE2 recognizes and attacks its substrate(s), to which we have currently no clue.
The present findings should pave the way to a mechanistic understanding of GDE2 endocytic trafficking. In particular, it will be essential to identify adaptor proteins and effectors that interact with the GDE2 CT and drive the endocytic recycling machinery, such as the retromer and related multi-protein complexes (Burd and Cullen, 2014;Gautreau et al., 2014;McNally and Cullen, 2018), and map the critical CT residues involved.
In a broader context, the question arises how GDE2 function is regulated in addition its dynamic trafficking behavior. Under oxidative conditions, newly synthesized GDE2 is post-translationally modified and fails to enter the secretory pathway resulting in loss of function (Yan et al., 2015).
Under physiological conditions, GDE2 catalytic activity could be controlled by as-yet-unknown binding partners and/or post-translational modifications, while its functional outcome will critically depend on the availability of specific substrates. Candidate GPI-anchored substrates that are known to determine neuronal fate include the neurodegenerative prion protein (PrPc), contactin family proteins and neurotrophic receptors, all of which exist as soluble forms . Determining GDE2's substrate specificity has therefore high priority.
Regarding pathophysiological implications, GDE2 deficiency predicts poor prognosis in neuroblastoma (Matas-Rico et al., 2016), while loss of GDE2 causes progressive neurodegeneration in mice with pathologies analogous to human disease (Cave et al., 2017).
This suggests that loss of GDE2 function might underlie aspects of human neurodegenerative disease and contribute to the pathogenesis of neuroblastoma. Yet, to our knowledge, diseaseassociated GDE2 dysfunction has not been documented to date, neither in neurodegenerative disease nor in nervous system malignancies. Given the present findings, however, GDE2 dysfunction could result from impaired endocytic trafficking rather than from loss-of-function mutations. Importantly, impaired trafficking is a hallmark of human neurodegenerative disease, including amyotrophic lateral sclerosis (ALS), Parkinson's and Alzheimer's disease (De Vos and Hafezparast, 2017;Kiral et al., 2018;McMillan et al., 2017;Schreij et al., 2016;Xu et al., 2018).
It is therefore tempting to speculate that disease-associated defects in the endocytic sorting machinery, even if subtle, may lead to GDE2 mis-localization and thereby contribute to neurodegeneration and other disorders.
In conclusion, while GPI-anchor cleavage by cell-intrinsic ecto-phospholipases has long been understudied, our study is the first to define sequence determinants of GDE2 trafficking, localization and activity. It thus opens new avenues to elucidate how GDE2 function is normally regulated and explore its potential dysregulation in disease.

Plasmids and transfections
Human GDE2 cDNA was subcloned in pcDNA3.1 as described (Matas-Rico et al., 2016). Truncated versions of GDE2 were generated by amplification of full-length GDE2-HA or GDE2-mCherry using reverse primers for the last residues of each truncation. This was followed by a digestion with BamHI and EcoRV, after which the amplified inserts were cloned into digested and gel-purified pcDNA3.1, and selected by Ampicilin. GDE2 point mutants were generated by site-directed mutagenesis using two complementary oligonucleotides with the desired mutated bases at the center of their sequences. A temperature gradient from 55 to 60 degrees was used during the PCR amplifications. The PCR products were digested with DpnI and transformed into DH5-α competent bacteria and screened for the expected mutated bases.

Confocal and super-resolution microscopy
Cells cultured on 24 mm, #1,5 coverslips were washed and fixed with 4% PFA, permeabilized with 0.1% Triton X-100 and blocked with 5% BSA for 1 hr. Incubation with primary antibodies was done for 1 hr, followed by incubation with Alexa-conjugated antibodies for 45 min at room temperature. For confocal microscopy, cells were washed with PBS, mounted with Immnuno-MountTM (Thermo Scientific) and visualized on a LEICA TCS-SP5 confocal microscopy (63x objective). Superresolution imaging was done using an SR-GSD Leica microscope equipped with an oxygen scavenging system, as previously described (Matas-Rico et al., 2016). In short, 15000 frames were taken in TIRF or EPI mode at 10 ms exposure time. Movies were analyzed and corrected using the ImageJ plugin Thunderstorm (http://imagej.nih.gov/ij/), followed by correction with an ImageJ macro using the plugin Image Stabilizer.

Live-imaging
Live-cell imaging was done on a Leica TCS SP5 confocal microscope equipped with 63x oil immersion lens (numerical aperture 1.4; Leica, Mannheim, Germany). Coverslips were mounted on a metal ring system and exposed to buffer solution (140 mM, NaCl, 5 mM KCl, 2 mM MgCl2, 1 mM CaCl2, 23mM NaHCO3, 10 mM HEPES, 10 mM glucose). N1E-115 cells were selected randomly and images were collected at appropriate time intervals (5-15 sec). GDE2-mCh was visualized by exciting cells at 561 nm, while emission was detected at 610 ± 10 nm

GDE2 plasma membrane localization
We performed image analysis for plasma membrane localization of HA-tagged GDE2 constructs by using public domain software IMAGEJ. Shortly, confocal images stained for GDE2-HA were segmented and analysed using Fiji software and a macro that automated the process. First, images were thresholded by the MaxEntropy algorithm to delimit single cells and filtered by Gaussian Blur (radius = 2) and smoothed for segmentation with a median radius of two. Using the Region of Interest manager on Fiji, the background was delimited by using the Li algorithm for thresholding. The cytoplasmic regions were selected by subtracting the plasma membrane thickness (fixed to 0.5 μm, but adjustable from a 0.2-5.0 range) and eroded with a pixel width of one to avoid having empty membranes in segmented cells. Next, the plasma membrane region was obtained by subtracting the background to the cytoplasmic region. Finally, the ratio membrane/cytoplasm was calculated from the median of these regions.

Western blotting
For Western blotting, cells were washed with cold PBS, lysed in NP-40 buffer supplemented with protease inhibitors and spun down. Protein concentration was measured using a BCA protein assay kit (Pierce) and LDS sample buffer (NuPAGE, Invitrogen) was added to the lysate or directly to the medium. Equal amounts were loaded on SDS-PAGE pre-cast gradient gels (4-12% Nu-Page Bis-Tris, Invitrogen), followed by transfer to nitrocellulose membrane. Non-specific protein binding was blocked by 5% skimmed milk in TBST; primary antibodies were incubated overnight at 4°C in TBST with 2.5% skimmed milk. Secondary antibodies conjugated to horseradish peroxidase (DAKO, Glostrup, Denmark) were incubated for 1 hr at room temperature; proteins were detected using ECL Western blot reagent.

Biotin labeling
For quantitation of GDE2 internalization and recycling, we used a biotin labeling assay. GDE2-mChexpressing N1E-115 cells were serum starved for 1hr., transferred to ice, washed in ice-cold PBS, and surface labeled at 4 o C with 0.2 mg/ml NHS-SS-biotin (Pierce). For GDE2 internalization, cells were exposed to serum-free medium at 37°C for the indicated time periods. Cells were transferred to ice and washed with PBS, remaining surface biotin was reduced with sodium 2-mercaptoethane sulfonate (MesNa), and the reaction was quenched with iodoacetamide (IAA) prior to cell lysis. For recycling assays, cells were labeled with biotin as above, and incubated in serum-free medium at 37°C for 30 min to allow internalization of GDE2. Cells were returned to ice, washed with PBS, and biotin was reduced using MesNa. Recycling of the internal GDE2 pool was induced by a temperature shift to 37°C for 0-30 min. Cells were returned to ice, washed with PBS and surface biotin was reduced by MesNa. MesNa was quenched by IAA and the cells were lysed. Biotin-labeled GDE2 was detected using Streptavidin beads and anti-mCh antibody.

Immunoprecipitation
For co-immunoprecipitation of GDE2 and Rabs, HEK293Tcells were plated on plastic dishes of 10 cm diameter and transient co-transfected with GDE2-mCh or -GFP and Rab4-GFP, Rab5-mCh, Rab7-GFP or Rab11-mCh. After 24 hrs cells were lysed using RIPA buffer. Protein concentration was determined using Protein BCA protein assay kit (Pierce). Immunoprecipitation was carried out incubating 500 μg -1 mg cytoplasmic extracts with GFP-or mCherry Trap® beads(ChomoTek) at 4°C for 1hr. Beads were washed three times and eluted by boiling in SDS sample buffer for 10 min. at 95°C. Supernatants were applied onto an SDS gel and subjected to immunoblot analysis Inducible GDE2 expression SH-SY5Y cells with inducible expression of GDE2 constructs were generated using the Retro-X™ Tet-On® Advanced Inducible Expression System (ClonTech), as described (Matas-Rico et al., 2016). After retroviral transduction, the cells were placed under selection with G418 (800 mg/ml) supplemented with puromycin (1 μg/ml) for 10 days. GDE2 induction (in the presence of 1 μg/ml doxycycline) was verified by Western blot and confocal microscopy. Transient transfection was performed with Fugene 6 reagent (Invitrogen) according to the manufacturer's instructions.

Flow Cytometry
For GPC6 and GDE2-HA surface expression analysis, cells were grown in complete medium with 10% FCS with or without doxycycline. Cells were trypsinized into single-cell suspensions and then 8x10 5 cells were incubated with 5 μl of anti-GPC6 antibody LS-C36518 (LifeSpan Bioscience) and in 4 μl of APC anti-HA antibody (Biolegend). Bound GPC6 antibody were detected by incubating with a 1:200 dilution of goat anti-mouse Alexa-488 secondary antibody in 2% BSA for 45 min on ice. Fluorescence measurements were performed using BD LSRFORTESSA and using Flow Jo software.

Induction of NEUROD1 and SNAP25
Total RNA was extracted using the GeneJET purification kit (Fermentas). cDNA was synthesized by reverse transcription from 5 μg RNA using First Strand cDNA Syntesis Kit (Thermo Scientific). As a negative control, the cDNA was replaced by milliQ. Cyclophilin was used as reference gene. Each sample was analyzed in triplo. The normalized expression (NE) data were calculated by the equation NE= 2(Ct target -Ct reference).

Supplemental Information
Supplemental Information includes one movie and three figures.   E. GDE2 (fused to GFP or mCh) associates with the indicated Rab GTPases in HEK293T cells.

Author Contributions
GDE2 was immunoprecipitated (IP) and subjected to immunoblotting (IB) using either anti-GFP or anti-mCh antibody.    B. Scheme of membrane trafficking, localization and signalling output of GDE2 and the indicated truncation mutants. GDE2 is constitutively internalized while the majority of endocytosed GDE2 traffics along Rab4 + and Rab11 + recycling routes in a sequence-dependent manner, with Rab4+ recycling being the preferred route of GDE2(ΔC560). A smaller part of GDE2 is routed to Rab7 + late endosomes for lysosomal degradation. Signalling efficacy is inferred from GPC6 shedding cleavage and induction of NEUROD1 and SNAP25. Note dysfunction of GDE2(ΔC560).
Disease-associated trafficking defects may similarly lead to GDE2 dysfunction. See Discussion.

Supplemental Information
Movie S1. Live imaging of GDE2-containing vesicle movements in SH-SY5Y cells.