The Role of Ceroid Lipofuscinosis Neuronal Protein 5 (CLN5) in Endosomal Sorting

Abstract

Mutations in the gene encoding CLN5 are the cause of Finnish variant late infantile Neuronal Ceroid Lipofuscinosis (NCL), and the gene encoding CLN5 is 1 of 10 genes (encoding CLN1 to CLN9 and cathepsin D) whose germ line mutations result in a group of recessive disorders of childhood. Although CLN5 localizes to the lysosomal compartment, its function remains unknown. We have uncovered an interaction between CLN5 and sortilin, the lysosomal sorting receptor. However, CLN5, unlike prosaposin, does not require sortilin to localize to the lysosomal compartment. We demonstrate that in CLN5-depleted HeLa cells, the lysosomal sorting receptors sortilin and cation-independent mannose 6-phosphate receptor (CI-MPR) are degraded in lysosomes due to a defect in recruitment of the retromer (an endosome-to-Golgi compartment trafficking component). In addition, we show that the retromer recruitment machinery is also affected by CLN5 depletion, as we found less loaded Rab7, which is required to recruit retromer. Taken together, our results support a role for CLN5 in controlling the itinerary of the lysosomal sorting receptors by regulating retromer recruitment at the endosome.

INTRODUCTION

Neuronal Ceroid Lipofuscinosis (NCL) diseases represent a group of recessive disorders of childhood characterized by progressive vision loss, seizures, ataxia, deafness, mental retardation, and greatly reduced life span (18, 24). At the cellular level, NCL is characterized by an accumulation of autofluorescent lipopigments with morphological heterogeneity between various forms (12). Several forms of NCL have been identified on the basis of age of onset, progression of disease, and neurophysiological and histopathological findings. These disorders are the result of germ line mutations in at least 10 genes (CLN1 to CLN9 and cathepsin D) (18), but the precise functions of most of these proteins are still unknown, although most encode either soluble or transmembrane proteins localized to either the endoplasmic reticulum (ER) or endosomes-lysosomes.

CLN1 (also known as palmitoyl-protein thioesterase-1 [PPT1]) is a soluble lysosomal palmitoylthioesterase with no known endogenous substrates, but a deficiency in this enzyme causes infant-stage-onset NCL (3). CLN3 is a transmembrane protein that has been shown to have palmitoyldesaturase activity (25) and may play a role in lysosomal acidification, organelle fusion, and apoptosis (28, 29). Mutations in the CLN3 gene cause juvenile NCL, more commonly known as Batten's disease. The exact function of CLN3 is still not fully elucidated; however, it was recently proposed that it affects lysosomal trafficking and sorting in yeast and mammalian cells (9, 23). Moreover, ablation of CLN3 caused an accumulation of the cation-independent mannose 6-phosphate receptor (CI-MPR) lysosomal sorting receptor in the trans-Golgi network (TGN) (23), and that study found a maturation defect of the soluble lysosomal protein cathepsin D, supporting a role for CLN3 in sorting to the lysosomal compartment. Although the function of CLN5 is unknown, germ line mutations in the gene encoding this protein are implicated in Finnish variant late infantile NCL (33). In humans, the CLN5 gene maps to chromosome 13q22, consists of 4 exons spanning 13 kb of genomic DNA, and encodes a protein of 407 amino acids. The predicted amino acid sequence of CLN5 shows no homology to previously reported proteins, and although several studies suggested that CLN5 has at least one transmembrane domain (4, 33, 38), other studies report that it may be a soluble protein (16). Whereas transfection of COS-1 cells with CLN5 cDNA results in the synthesis of a highly glycosylated 60-kDa polypeptide, 47-, 44-, 42-, and 40-kDa polypeptides were produced in cell-free translation assays due to usage of the alternative initiator methionine (17). Finally, previous studies have shown that CLN5 interacts with CLN2 and CLN3 (38) and localizes to the endosomal-lysosomal compartment (17). However, the mechanism cells use to sort CLN5 to lysosomes and the function of this protein remain unknown.

The trafficking of soluble luminal lysosomal cargo such as cathepsin D, CLN1, prosaposin, and β-glucocerebrosidase is sorted by the cation-dependent mannose 6-phosphate receptor (CD-MPR), the cation-independent mannose 6-phosphate receptor (CI-MPR), sortilin, and LIMP-II (7, 10, 20, 30). For anterograde traffic (Golgi compartment to endosome), cargo binds to the receptors in the Golgi compartment and is packaged into clathrin-coated vesicles (5). When the receptor-cargo complex reaches the more acidic environment of the endosomes, the cargo dissociates from the receptor and the majority of the receptor is recycled to the Golgi compartment for another round of sorting while a small percentage is degraded in lysosomes (6). The efficient retrograde traffic (endosome to Golgi compartment) of CI-MPR and sortilin requires the retromer protein complex (1, 35). Mammalian retromer comprises 2 distinct subcomplexes: a dimer of a still-undefined combination of sorting nexin 1 (SNX1), SNX2, SNX5, and SNX6 that can interact with the endosomal membrane via phosphatidylinositol 3-phosphate (PI3P) and a heterotrimer composed of vacuolar protein sorting 26 (Vps26), Vps29, and Vps35 that can bind the cytosolic tails of the lysosomal sorting receptors (2, 6).

A recent study found that CI-MPR was not implicated in the lysosomal localization of CLN5 (34). We tested the hypothesis that sortilin is involved in the sorting and trafficking of CLN5 to the lysosomal compartment. We report an interaction between CLN5 and sortilin; however, CLN5 was still properly localized in sortilin-depleted cells. Interestingly, in CLN5-depleted cells, we found that CI-MPR and sortilin were degraded in lysosomes due to the lack of retromer recruitment to endosomes. Taking these data together, we propose that CLN5 is part of an endosomal switch that determines whether the lysosomal sorting receptors are recycled to the Golgi compartment or degraded in lysosomes.

MATERIALS AND METHODS

Antibodies and reagents.

The following mouse monoclonal antibodies were used: anti-CD222 against CI-MPR, anti-myc 9E10, anti-hemagglutinin (anti-HA), and MMS-118P against green fluorescent protein (GFP) (all from Cedarlane Laboratories, Burlington, Ontario, Canada) and anti-CD63, anti-SNX1, and Ab-5 against actin (all from BD Bioscience, Mississauga, Ontario, Canada). The following polyclonal antibodies were used: anti-cathepsin D (Cedarlane Laboratories, Burlington, Ontario, Canada), anti-Lamp2 and anti-TGN46 (Sigma-Aldrich, Oakville, Ontario, Canada), anti-SNX3, anti-AP3, anti-red fluorescent protein (anti-RFP), anti-Vps26, anti-Vps35, and anti-prosaposin (all from Abcam, Cambridge, MA), and anti-CLN5 and anti-CLN1 (both from Santa Cruz Biotechnology, Santa Cruz, CA). The cDNAs for RFP-Rab5 and RFP-Rab7 were acquired from Addgene (Cambridge, MA). The HA-CLN5 construct was purchased from Genecopoeia (Germantown, MD). pGEX-RILP_220-299 was a generous gift from Aimee Edinger (University of California, Irvine, CA). The cDNA for CLN1 and myc-CD63 was purchased from Origene (Rockville, MD). The dominant-active Rab7 (RFP-Rab7Q67L) and the dominant-negative Rab7 (RFP-Rab7T22N) were generated by site-directed mutagenesis. The myc-Rab1a construct was a generous gift from Terry Hebert (McGill University, Montreal, Quebec City, Quebec, Canada).

Cell culture and immunofluorescence.

Cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum and antibiotics. Transfections were performed using 1 μg of DNA per 4-cm² plate with Lipofectamine reagent supplemented with Plus reagent according to the instructions of the manufacturer (Invitrogen, Burlington, Ontario, Canada). Immunofluorescence was performed as previously described (11).

Coimmunoprecipitation, cycloheximide chase, and glutathione S-transferase (GST) pulldown assays.

These methods were previously described (19, 21).

RNA interference.

The small interfering RNAs (siRNAs) against CLN5 and CLN1 were purchased from Invitrogen and transfected using Oligofectamine (Invitrogen, Burlington, Ontario, Canada) according to the manufacturer's instructions. The siRNA against the 3′ untranslated region (3′UTR) of CLN5 was purchased from IDT DNA (Coralville, IA) and used according to the manufacturer's instructions. Cells were depleted of CLN5 and CLN1 by the use of 100 nM siRNA and grown for 48 h before our assays were performed. To deplete sortilin, HeLa cells were transfected using short hairpin sortilin (shSortilin) and Lipofectamine reagent according to the instructions of the manufacturer (Invitrogen, Burlington, Ontario, Canada).

Quantification of immunofluorescence signals at the endosome.

The endosomal compartment was assumed to consist of fluorescent puncta, most of them of sub[diffraction-limit size. Wide-field fluorescence microscopy images were taken using a 63× oil-immersion objective and appropriate dichroics to detect Alexa 594, RFP, or GFP. The exposure time of the camera was chosen in order to avoid saturation and maximize the intensity dynamic range and was kept constant for all images acquired. To accurately analyze the data in a nonbiased way, an algorithm was custom programmed using Matlab (Mathworks), which automatically detects puncta and computes their mean fluorescence intensities in images composed of several cells. Fluorescent puncta were detected using linear band-pass filters that preserved objects of a size window and suppressed noise and large structures. These filters were applied by performing two 2-dimensional convolutions of the image matrix with a Gaussian and a boxcar kernel. First, the image was convolved with a Gaussian kernel of the characteristic length of the noise component. This is considered a low-pass filter, since only fluctuations longer than that given length are kept after the operation. Second, the image matrix was convolved with a boxcar kernel twice as big as the point spread function. This last operation is a low-pass filter for near-diffraction-limit objects. Finally, the subtraction of the boxcar image from the Gaussian images becomes a band-pass filter to choose elements bigger than those representing noise and up to twice the diffraction limit. In order to limit the puncta considered in the calculation to those inside cells, an intensity threshold was established using Otsu's method (27). The cytosolic fluorescence was enough to allow use of this automatic thresholding procedure to assign foreground pixels to cells and background pixels to empty space. This coarse estimation of signal corresponding to the foreground pixels was further refined and used as a mask to consider only puncta inside cells and discard signal arising from culture debris and nonspecific staining. The first refinement was to clean the mask by removing isolated pixels (1s surrounded by eight 0s). Next, a morphological erosion (13) was performed, and holes inside the mask were filled. Finally, the area of each individual object in the mask was measured, and all objects whose size was less than 50 times that of the biggest object (often several cells close together) were removed. Using the value for the already-filtered image multiplied by the cell mask value, an algorithm that finds local maxima of intensity was applied to detect the brightest pixels. Each of these intensity maxima was considered to represent the brightest location of a puncta, and circles 4 pixels in radius were established around the maxima. For each set of images, the mean intensity was calculated for all those circles as a measure of protein expression level. The program opened all images one by one, performed the band-pass filter, established a mask to delineate cells, found endosome circles, computed their mean intensity, and calculated the average intensity for all endosomes in the full set of images.

Membrane isolation assay.

This experiment was performed as previously described (15, 36).

Photoaffinity labeling.

At 24 h posttransfection, HeLa cells were resuspended in 60 μl of labeling buffer (10 mM HEPES [pH 7.4], 1 mM MgCl₂, 0.1 mM β-mercaptoethanol) and passed eight times trough a 23-gauge needle at room temperature. GTP ([γ] 4-azidoanilide 2′,3′-biotin-long chain-hydrazone) (Affinity Photoprobes, Lexington, KY) was then added to achieve a final concentration of 0.1 μM. The samples were then incubated at room temperature for 10 min and were then subjected to UV irradiation (254 nm; 6 mW/cm²) with a mounted lamp for 5 min to cross-link the probe to proteins. Labeling buffer was then added to 750 μl, and RFP-Rab7 and RFP-Rab5 were immunoprecipitated using a monoclonal RFP antibody. The labeling reaction was then detected by Western blot analysis using streptavidin-HRP.

RESULTS

CLN5 interacts with but is not trafficked by sortilin.

Although it is well established that CLN5 localizes to the endosomal-lysosomal compartment (17), few studies have addressed its mechanism of trafficking. A recent study showed that CLN5 was properly localized in CI-MPR-deficient fibroblasts (34), so we tested the hypothesis that CLN5 is a cargo of sortilin. Since soluble cargo must bind a lysosomal sorting receptor, we wanted to determine whether or not CLN5 was an interactive partner of sortilin. To test this, we coexpressed HA-CLN5 and sortilin-myc in HeLa cells and immunoprecipitated CLN5 with anti-HA antibody from HeLa cell lysate (Fig. 1A). We observed a specific band corresponding to sortilin-myc in the presence of HA-CLN5 at pH 7 (Fig. 1A) that was absent in the bands corresponding to lysates not expressing HA-CLN5, even though sortilin-myc was expressed in both. Similarly, sortilin-myc was able to coimmunoprecipitate HA-CLN5 when we immunoprecipitated with anti-myc antibody (data not shown), and our immunoprecipitation protocol using anti-HA antibody was clean, as seen by using a Coomassie-stained gel (data not shown). Next, we further investigated this interaction to determine whether or not CLN5 is a cargo of sortilin. Earlier studies have established that lysosomal cargo such as prosaposin coprecipitates with the corresponding receptor(s) (in this case, sortilin) when immunoprecipitation experiments are performed at pH 7 but that the same interactions are lost when performed at a more acidic pH (39). We performed a coimmunoprecipitation of CLN5 and sortilin at pH 5 and found that this interaction was pH independent (Fig. 1A), as CLN5 bound to sortilin at both neutral pH (Fig. 1A) and acidic pH (Fig. 1A), suggesting that CLN5 did not behave as a sortilin cargo protein. However, as predicted, the interaction between prosaposin and sortilin was weakened at pH 5 compared to pH 7 (Fig. 1A), as prosaposin-myc could not efficiently coimmunoprecipitate sortilin-GFP at pH 5 but could at pH 7. The CLN5-sortilin interaction was specific, as HA-CLN5 was not able to coimmunoprecipitate the myc-CD63 lysosomal membrane protein at either pH level (Fig. 1A). To confirm that CLN5 is not a cargo of sortilin, we tested whether or not CLN5 was properly localized to the lysosomal compartment in sortilin-depleted cells. We transfected HeLa cells with short hairpin sortilin (shSortilin) to knockdown sortilin and obtained a 50% decrease in the absolute amount of sortilin following the knockdown (Fig. 1B). An immunofluorescence assay performed on mock-depleted cells (Fig. 1C, D, and E) and sortilin-depleted cells (Fig. 1F, G, and H) showed that CLN5 localization was not affected by the depletion of sortilin (Fig. 1F, arrows). It was previously shown that prosaposin did not localize to lysosomes in sortilin-depleted cells (20), and as expected, the localization of prosaposin was perturbed in sortilin-depleted cells (Fig. 1H), as shown by the lack of punctuate structures compared to mock-depleted cells (Fig. 1E, arrows). On the basis of those data, we concluded that CLN5 is not a cargo of sortilin.

Fig 1.

CLN5 interacts with the lysosomal sorting receptors. (A) HeLa cells were transfected with sortilin-myc, myc-CD63, prosaposin-myc (PSAP-myc), sortilin-GFP, and HA-CLN5 as indicated. Whole-cell lysates were immunoprecipitated (IP) with anti-HA or anti-myc antibodies at the pH shown and subjected to Western blotting (Wb) with anti-myc, anti-HA, or anti-GFP antibody. The amount of sortilin-myc, myc-CD63, or sortilin-GFP preimmunoprecipitation (Pre-IP) is shown and represents 10% of the input. (B) HeLa cells transfected with sortilin-myc were either mock or sortilin depleted with a short-hairpin construct (shSortilin). Whole-cell lysates were run on 12% acrylamide gels and subjected to Western blotting (Wb) with anti-myc and antiactin antibodies. HeLa cells expressing sortilin-myc were mock transfected (C, D, and E) or transfected with shSortilin (F, G, and H) to deplete sortilin. Cells were fixed in 4% paraformaldehyde and immunostained with anti-CLN5 antibody (C and F, green), anti-myc antibody (D and G, green), or antiprosaposin (PSAP) antibody (E and H, red). Arrows indicate the normal localization of CLN5 (C and F), sortilin-myc (D), or PSAP (E). Stars indicate nuclear background staining for antiprosaposin antibody (E and H). Bar, 10 μm.

CLN5 is implicated in the trafficking of the lysosomal sorting receptors CI-MPR and sortilin.

Since CLN5 is not a cargo of CI-MPR (34) or sortilin (our data), we examined the biological significance of the CLN5-sortilin interaction. We tested whether or not CLN5 plays a role in the steady-state localization of CI-MPR and sortilin by using siRNA to deplete CLN5 in HeLa cells. We were able to efficiently deplete both CLN5 (Fig. 2M, top panels) and CLN1 (Fig. 2M, bottom panels) by the use of a pool of 3 siRNAs, as shown by Western blotting. We found that in CLN5-depleted cells (Fig. 2J, K, L, J′, K′, and L′), the intensities of the immunofluorescence signal for endogenous CI-MPR (Fig. 2K, star) and sortilin-myc (Fig. 2K′, star) were significantly reduced compared to mock-treated cell results (Fig. 2H and H′, arrows). However, the depletion of CLN1 (Fig. 2D, E, F, D′, E′, and F′), a known CI-MPR cargo, did not have an effect on the localization or the intensity of CI-MPR (Fig. 2E, arrow) or sortilin-myc (Fig. 2E′, arrow) compared to mock-depleted cell results (Fig. 2B and B′, arrows). To verify that the Golgi compartment was still intact in CLN5-depleted cells, we compared the immunofluorescence staining of the Golgi marker giantin in mock-depleted cells to the staining in CLN5-depleted cells. We found no significant differences in giantin staining in mock-depleted (Fig. 2N, red) compared to the CLN5-depleted (Fig. 2O, red) cells, suggesting that the Golgi compartment was intact in CLN5-depleted cells. However, we found that TGN46 staining was fragmented in CLN5-depleted cells (Fig. 2Q, red) compared to mock-depleted cells (Fig. 2P, red), which was expected, as a similar phenotype was observed in retromer-depleted cells (35). In retromer-depleted cells, the immunofluorescence staining of CI-MPR is significantly reduced compared to mock-depleted cells, and cycloheximide chase experiments have shown degradation of the receptor in 6 h (1, 31, 35). Since CLN5 depletion resulted in a similar immunofluorescence phenotype, we investigated whether the depletion of CLN5 would lead to the degradation of sortilin and CI-MPR with kinetics similar to the kinetics of retromer depletion. To determine this, we performed a cycloheximide chase experiment; after incubation with cycloheximide for 6 h was performed, we found a significant reduction in expression of CI-MPR (Fig. 3A) and sortilin-myc (Fig. 3A) in CLN5-depleted cells compared to mock- or CLN1-depleted cells, which showed no significant degradation (Fig. 3A). The depletion of CLN5 did not affect the total amount of the endosomal-lysosomal membrane protein CD63 (Fig. 3A). Quantification of three separate experiments showed decreases in the levels of CI-MPR and sortilin-myc of 46% and 59%, respectively, in CLN5-depleted cells compared to mock-depleted cells, while we found no significant changes in expression of CD63 (Fig. 3B). Significantly, the degradation kinetics in CLN5-depleted cells is similar to that seen in retromer-depleted cells (1) or in cells deficient in palmitoylation, which is also required for retrograde trafficking of the lysosomal sorting receptors (21). Since the lysosomal sorting receptors were not being efficiently recycled to the Golgi compartment and were being degraded in CLN5-depleted cells, we would expect the missorting of lysosomal cargo proteins such as cathepsin D or prosaposin. We found an increase in the amounts of precursor and intermediate forms of cathepsin D in CLN5-depleted cells compared to mock-depleted cells (Fig. 3C), which is consistent with previously published results from studies performed with cells that had been either Rab7 or retromer depleted (32). Based on these experiments, we concluded that the receptors in CLN5-depleted cells were not recycling to the Golgi compartment but were being degraded in lysosomes, suggesting that CLN5 plays a role in determining the itinerary of the lysosomal sorting receptors at the endosome.

Fig 2.

CLN5 is required for the localization of CI-MPR and sortilin. HeLa cells were mock depleted (A to C, A′ to C′, G to I, and G′ to I′), CLN1 depleted (D to F and D′ to F′), or CLN5 depleted (J to L and J′ to L′) and immunostained with anti-CLN1 (A, C, A′, C′, D, F, D′, and F′, green), anti-CLN5 (G, I, G′, I′, J, L, J′, and L, green), anti-CI-MPR (B, E, H, and K, red), or anti-myc (B′, E′, H′, and K′, red) following sortilin-myc transfection. Arrows indicate the perinuclear staining of CI-MPR (B, E, and H) and sortilin (B′, E′, and H′), while stars indicate lack of perinuclear staining for CI-MPR (K) and sortilin-myc (K′). Bar, 10 μm. (M) HeLa cells were transfected with HA-CLN5 following siRNA treatment (siCLN5 or siCLN1) as indicated. Whole-cell lysates were subjected to Western blotting (Wb) with anti-HA, anti-CLN1, or antiactin antibodies. HeLa cells were either mock depleted (N and O) or CLN5 depleted (P and Q) and immunostained with anti-CLN5 antibody (N to Q, green) and giantin (N and O, red) or TGN46 (P and Q, red). Arrows indicate the normal staining pattern of giantin and TGN46, while stars represent dispersed staining. Bar, 10 μm.

Fig 3.

CI-MPR and sortilin are degraded in CLN5-depleted cells. (A) HeLa cells transfected with sortilin-myc were mock, CLN5, or CLN1 depleted and incubated with cycloheximide (50 μg/ml) for the times indicated. Total cell lysates were subjected to Western blotting (Wb) with anti-CI-MPR, anti-myc, anti-CD63, or antiactin antibodies. (B) Quantification of 3 separate cycloheximide chase experiments performed with cells that were mock, CLN1, or CLN5 depleted and subjected to Western blotting for endogenous anti-CI-MPR, anti-myc, or anti-CD63 antibodies. (C) Total cell lysates from HeLa cells that were mock or CLN5 depleted were subjected to Western blotting with anti-cathepsin D or antiactin antibodies.

CLN5 is required for the recruitment of Vps26 to endosomes.

We next wanted to determine the mechanism that leads to the degradation of the lysosomal sorting receptors in CLN5-depleted cells. Since retromer depletion leads to a phenotype similar to that seen with CLN5 depletion in terms of both stability and localization of CI-MPR and sortilin, we investigated the effect of CLN5 depletion on the recruitment of retromer to endosomal membranes. We found a significant decrease in the intensity of Vps26 immunofluorescence staining (Fig. 4E, star) in CLN5-depleted cells (Fig. 4D, E, and F) compared to mock-depleted cells (Fig. 4A, B, and C) or CLN1-depleted cells (Fig. 4J, K, and L). To rigorously quantify this observation, we developed an ad hoc algorithm to detect endosomes and quantify their fluorescence intensity. We obtained an unbiased image sample by randomly acquiring images from the total cell population. A representative immunofluorescence image is shown (Fig. 4M, left panel) along with the mask that the software identifies as representing Vps26-positive structures (Fig. 4M, right panel, red circles). This quantification revealed a 50% decrease in Vps26 immunofluorescence staining in the CLN5-depleted cell population (Fig. 5A, black bar) compared to mock-depleted cells (Fig. 5A, white bar) or CLN1-depleted cells (Fig. 5A, gray bar). This observed decrease in endosomal Vps26 intensity could indicate either degradation of the subunit (Vps26) or a block in recruitment of retromer to endosomal membranes. Thus, in order to differentiate between the two possibilities, we performed Western blotting to detect the absolute amount of the retromer subunit Vps26 and found no changes in expression in comparisons of mock-, CLN1-, and CLN5-depleted cells (Fig. 5B). To determine whether or not the depletion of CLN5 affected the intensity of immunofluorescence staining in endosomes-lysosomes in general, we compared the results of immunofluorescence staining of CD63. The localization of this protein to the lysosomal compartment is retromer independent. We found no significant difference in the immunofluorescence staining intensity of CD63 in mock-depleted cells (Fig. 5C, white bars) compared to CLN5-depleted cells (Fig. 5C, black bars). It has previously been proposed that the Vps26, -29, and -35 trimer is recruited independently of the SNX dimer (22). To gain further insight into the role of CLN5 in the recruitment of retromer to endosomal membranes, we investigated the effect of CLN5 depletion on the recruitment of SNX1 to endosomal membranes. Interestingly, we found no significant changes in the amount of SNX1 recruited to endosomal membranes in CLN5-depleted cells (Fig. 5D, black bar) compared to mock-depleted cells (Fig. 5D, white bar) by immunofluorescence. To confirm our immunofluorescence data, we performed a membrane isolation experiment to determine the amounts of Vps26, Vps35, AP3, SNX1, and SNX3 in the cytosol and on membranes in mock-, CLN5-, and Rab7-depleted cells. We found less membrane-bound and more cytosolic Vps26 and Vps35 in CLN5- and Rab7-depleted cells compared to mock-depleted cells (Fig. 6A). However, we found no change in the distribution of SNX3, which is known to recruit retromer to endosomes to traffic the Wnt-binding protein Wntless (15) and blocks the efficient retrograde trafficking of the CD8-CI-MPR chimera (14), or in the distribution of SNX1, while AP3, an adaptor protein complex recruited to endosomes, seemed to be more cytosolic in Rab7-depleted cells whereas CLN5 depletion had no effect (Fig. 6A). Lamp2 staining was used to identify the membrane fraction (Fig. 6A). The quantification of 3 separate experiments shows the distribution of Vps26 and Vps35 in mock-, CLN5-, and Rab7-depleted cells (Fig. 6B). In mock-treated cells, 60% and 70% of Vps26 and Vps35, respectively, were found in the membrane fractions (black bars), whereas in CLN5-depleted cells, the membrane fractions of Vps26 and Vps35 were reduced to 35% and 50%, respectively, and in Rab7-depleted cells the proportions of membrane association were 35% and 45% for Vps26 and Vps35, respectively. The changes in membrane association of retromer subunits obtained in the CLN5- and Rab7-depleted cells are comparable to those found in previously published reports (15, 36) and support a role for CLN5 in the recruitment of retromer to endosomal membranes.

Fig 4.

Recruitment of the Vps26 subunit of retromer to endosomes requires CLN5. HeLa cells were mock depleted (A to C and G to I), CLN5 depleted (D to F), or CLN1 depleted (J to L) and immunostained with anti-CLN5 (A, C, D, and F, green), anti-CLN1 (G, I, J, and L, green), and anti-Vps26 (B, C, E, F, H, I, K, and L, red) antibodies. Arrows represent normal staining for CLN5, CLN1, and Vps26, while stars highlight lack of CLN5 (D), CLN1 (J), and Vps26 (E) staining. Bar, 10 μm. (M) Representative image of Vps26-positive structures identified (red circles) by our ad hoc algorithm to determine the intensity of Vps26 staining in HeLa cells.

Fig 5.

CLN5 is required to recruit retromer to endosomal membranes. (A) Quantification of the relative fluorescence intensities of Vps26 staining in mock-depleted cells, CLN5-depleted cells, and CLN1-depleted cells. Data represent the relative fluorescence intensities of Vps26 staining for 1,400, 1,300, and 3,000 endosomes per condition, respectively, with error bars representing ± standard errors of the means (SEM). (B) HeLa lysates from mock-, CLN-, and CLN5-depleted cells were run on a 12% polyacrylamide gel and subjected to Western blotting (Wb) with anti-Vps26 and antiactin antibodies. (C) Quantification of the relative fluorescence intensities of CD63 staining in mock-depleted (white bar) and CLN5-depleted (black bar) cells. Data represent quantification of the relative intensities of CD63 staining for 4,800 and 4,000 endosomes, respectively, with the error bars representing ±SEM. (D) Quantification of the relative fluorescence intensities of SNX1 staining in mock-depleted and CLN5-depleted cells. Data represent quantification of the relative intensities of SNX1 staining for 4,000 and 2,500 endosomes, respectively, with the error bars representing ±SEM.

Fig 6.

The cytosolic distribution and membrane distribution of Vps26 and Vps35 are altered in CLN5-depleetd cells. (A) Cytosolic (C) and membrane (M) fractions from mock-, CLN5-, or Rab7-depleted cells were stained with anti-Vps35, anti-Vps26, anti-SNX3, anti-SNX1, anti-AP3, and anti-Lamp2 antibodies. (B) Quantification of the cytosolic fraction and membrane fraction of Vps26 and Vps35 from 3 separate experiments.

In order to demonstrate that the effect seen on Vps26 recruitment is specific to CLN5 depletion and not an off-target effect, we adopted a rescue strategy in CLN5-depleted cells. For this, we knocked down endogenous CLN5 by the use of a pool of siRNA directed against the 3′UTR of the designated gene and then reintroduced wild-type CLN5 to test for the rescued phenotype. Using this siRNA strategy, we obtained a 70% reduction in the amount of CLN5 (Fig. 7A), which we could rescue with the expression of HA-CLN5 (Fig. 7A). The knockdown of CLN5 against the 3′UTR resulted in the decreased recruitment of Vps26 (Fig. 7B, black bar) compared to the results seen with mock-treated cells (Fig. 6B, white bar). Reintroducing wild-type HA-CLN5 into these CLN5-depleted cells rescued the recruitment of Vps26 (Fig. 7B, gray bar). Taken together, these results point to a specific role for CLN5 in influencing the trafficking of lysosomal sorting receptors by controlling recruitment of retromer to endosomal membranes.

Fig 7.

Transient expression of wild-type HA-CLN5 rescues the recruitment of Vps26 in CLN5-depleted cells. (A) HeLa cells were either mock or CLN5 depleted with siRNA against the 3′UTR of CLN5. Cells were also transfected or not with HA-CLN5, and total cell lysates were subjected to Western blotting (Wb) for endogenous CLN5 by the use of anti-CLN5, transfected HA-CLN5 with anti-HA antibody, or antiactin. (B) Quantification of the relative fluorescence intensities of Vps26 staining in mock-depleted cells, CLN5-depleted cells, and CLN5-depleted cells transfected with HA-CLN5. Data represent the relative intensities of Vps26 staining for 3,400, 2,200, and 4,700 endosomes per condition, respectively, with error bars representing ±SEM.

CLN5 regulates the localization of Rab7.

Next, we attempted to identify the molecular mechanisms by which CLN5 controls the recruitment of retromer to endosomes. Recently, it was shown that activation of the small G proteins Rab5 and Rab7 is required for recruitment of the Vps26 subunit of retromer (32, 36). Since we obtained similar results by depleting cells of CLN5, we investigated whether CLN5 is implicated or not in Rab5 and Rab7 localization. We depleted cells of CLN5 and looked at the localization of RFP-Rab7 and RFP-Rab5. Our results showed that, in CLN5-depleted HeLa cells (Fig. 8D, E, and F), the intensity of RFP-Rab7 is lower than that seen in mock-depleted cells (Fig. 8B and H) or in CLN1-depleted cells (Fig. 8K). Quantification using our ad hoc algorithm showed a 40% decrease in the intensity of RFP-Rab7 at endosomes in CLN5-depleted cells (Fig. 8M, black bar) compared to mock-depleted cells (Fig. 8M, white bar) or CLN1-depleted cells (Fig. 8M, gray bar). Western blotting on whole-cell lysate determined that the absolute amounts of RFP-Rab7 in mock-, CLN1-, and CLN5-depleted cells were similar (Fig. 8N), suggesting that the decrease in endosomal intensity was due to lack of recruitment and not to degradation. We then compared the fluorescence intensities of RFP-Rab5 in mock- and CLN5-depleted cells and found a slight (18%) decrease in the intensity of RFP-Rab5 in CLN5-depleted cells (Fig. 8O, black bar) compared to mock-depleted cells (Fig. 8O, white bar). Consistent with our Rab7 data, we found no significant differences in the absolute amounts of RFP-Rab5 in either mock- or CLN5-depleted cells (Fig. 8P) as shown by Western blotting. We next tested if CLN5 was in a protein complex with Rab5 and/or Rab7 by cotransfecting HeLa cells with HA-CLN5 and myc-Rab1a, RFP-Rab5, or RFP-Rab7. Following an immunoprecipitation with anti-HA or anti-RFP antibodies, we found an interaction between HA-CLN5 and RFP-Rab5 and between HA-CLN5 and RFP-Rab7 (Fig. 9A) but not between HA-CLN5 and myc-Rab1a (Fig. 9A). This suggested that CLN5 may act as a scaffold for the site of recruitment of Rab7 and subsequently retromer onto endosomal membranes.

Fig 8.

CLN5 is required for the localization of Rab7. (A to L) HeLa cells were mock depleted (A to C and G to I), CLN5 depleted (D to F), or CLN1 depleted (J to L) and transfected with RFP-Rab7 (A to L). Following transfection, cells were fixed in 4% paraformaldehyde and then immunostained with anti-CLN5 (A, C, D, and F, green) or anti-CLN1 (G, I, J, and L, green) antibodies. Arrows indicate the normal distribution of CLN5 (A), CLN1 (G), and Rab7 (B, H, and K). Stars indicate the lack of CLN5 expression (D) and the lack of recruitment of RFP-Rab7 (E) in CLN5-depleted cells. (M) Quantification of the fluorescence intensity of RFP-Rab7 in mock-depleted, CLN5-depleted, and CLN1-depleted cells. Data represent 5,700, 9,300, and 1,600 endosomes per condition, respectively, with error bars representing ±SEM. (N) Expression of RFP-Rab7 in mock-, CLN1-, or CLN5-depleted cells was examined by Western blotting (Wb) with anti-RFP antibody. Antiactin staining served as a loading control. (O) Quantification of the fluorescence intensity of RFP-Rab5 in mock-depleted and CLN5-depleted cells. Data represent the quantification of 2,800 and 4,000 endosomes per condition, respectively. (P) Expression of RFP-Rab5 in mock- and CLN5-depleted cells was examined by Western blotting (Wb) with anti-RFP antibody. Antiactin staining served as a loading control.

Fig 9.

CLN5 is required for the activation of Rab7. (A) HeLa cells were transfected with HA-CLN5 and myc-Rab1a, RFP-Rab5, or RFP-Rab7 as indicated. Total cell lysates were immunoprecipitated (IP) with anti-RFP or anti-HA antibodies and subjected to Western blotting (Wb) with anti-HA, anti-myc, or anti-RFP antibodies. The preimmunoprecipitation (pre-IP) is shown and represents 10% of the input. (B) HeLa cells were cotransfected with HA-CLN5 and wild-type RFP-Rab7, dominant-active RFP-Rab7Q67L, or dominant-negative RFP-Rab7T22N. Total cell lysate was immunoprecipitated (IP) with anti-HA antibody and subjected to Western blotting (Wb) with either anti-RFP or anti-HA antibodies. The amount of RFP-Rab7, RFP-Rab7Q67L, or RFP-Rab7T22N preimmunoprecipitation (Pre-IP) is shown and represents 10% of the input. (C) The amounts of GTP-loaded Rab7 and Rab5 were determined using a nonhydrolysable biotin-conjugated probe (GTP; [γ] 4-azidoanilide 2′,3′-biotin-long chain-hydrazone) in mock- or CLN5-depleted cells. The amount of loaded GTP probe was determined using streptavidin (stav), while the amounts of RFP-Rab7 or RFP-Rab5 were determined using anti-RFP antibody (Wb: RFP). (D) The relative intensities of GTP-loaded Rab7 in mock- and CLN5-depleted cells from 3 independent experiments, with the error bars representing ±SEM. (E) The relative intensities of GTP-loaded Rab5 in mock- and CLN5-depleted cells from 3 independent experiments, with error bars representing ±SEM. (F) GST-RILP_220–299 bound to glutathione-Sepharose beads was incubated with HeLa lysates expressing RFP-Rab7 that were mock, CLN1, or CLN5 depleted. The amount of bound RFP-Rab7 was detected by Western blotting (Wb) with anti-RFP antibody. The Coomassie-stained gel shows the amount of bound GST-RILP_220–299.

CLN5 is required to activate Rab7.

We next tried to determine whether CLN5 was an effector of activated Rab7 or if CLN5 was required to activate Rab7. We tested whether dominant-active Rab7 (RFP-Rab7Q67L) could interact more strongly with CLN5 than wild-type Rab7, which would suggest that it would be an effector like Rab-interacting lysosomal protein (RILP) (8), rubicon (37), or retromer (32). Following an immunoprecipitation with anti-HA antibody, we found that both wild-type Rab7 and dominant-active Rab7 (Rab7Q67L) interacted with CLN5 (Fig. 9B). Moreover, we also found an interaction between dominant-negative Rab7 (RFP-Rab7T22N) and CLN5 (Fig. 9B), suggesting that CLN5 may be part of the Rab7 activation machinery and not an effector. Next, we compared the abilities of cells to load Rab5 and Rab7 with GTP in the presence or absence of CLN5 by measuring the amount of a cross-linkable GTP analogue incorporated into these Rab proteins. We found that, compared to mock-depleted cells, CLN5-depleted cells had significantly less GTP-loaded Rab7 (Fig. 9C) whereas the amount of GTP-loaded Rab5 was not significantly different (Fig. 9C). Quantification of the GTP-loading experiments showed that Rab7 loading was reduced by 76% in CLN5-depleted cells (Fig. 9D, black bar) compared to mock-depleted cells (Fig. 9D, white bar), while the amount of GTP-loaded Rab5 was not changed in mock-depleted cells (Fig. 9E, white bar) compared to CLN5-depleted cells (Fig. 9E, black bar). In support of this data, Rab7 binding to Rab-interacting lysosomal protein (RILP), a known Rab7 effector (8), in a GST pulldown assay was less efficient in cells depleted of CLN5 compared to mock- or CLN1-depleted cells (Fig. 9F). Taken together, these results show that CLN5 is required to recruit and activate Rab7 to subsequently recruit retromer to endosomal membranes.

DISCUSSION

Several conclusions can be drawn from the data presented in this work. First, CLN5 interacts with sortilin. However, this interaction is not required to traffic CLN5 to the lysosomal compartment; rather, it enables the lysosomal sorting receptors (CI-MPR and sortilin) to recycle to the Golgi compartment from endosomes, preventing their degradation. Second, CLN5 is implicated in the recruitment of retromer to endosomes by regulating the localization and activation of Rab7, which has previously been shown to be implicated in retromer recruitment (32, 36), to enable retrograde trafficking of the lysosomal sorting receptors. While it has been known that CLN5 is localized to the endosomal-lysosomal compartment, its function and mechanism of trafficking have not been elucidated. Since a previous report found that CI-MPR was not implicated in the trafficking of CLN5 (34), we tested if sortilin was a trafficking receptor for CLN5. We found that CLN5 binds sortilin; however, CLN5 can interact with sortilin at a more acidic pH, a condition that usually inhibits cargo-receptor interactions such as prosaposin binding to sortilin. This suggested that the CLN5-sortilin interaction may not be required for the trafficking of CLN5 to the lysosomal compartment; in support of this idea, depletion of sortilin by shRNA had no effect on the cellular localization of CLN5, although it did prevent the proper localization of a known cargo, prosaposin, as was previously shown (20). Since neither CI-MPR nor sortilin seems to be implicated in the trafficking of CLN5, it is possible that LIMP-II is required, as it was recently shown that this protein can act as the sorting receptor for β-glucocerebrosidase (30). Moreover, since CLN5 is a potential transmembrane protein, it is therefore possible that CLN5, like other lysosome integral membrane proteins such as CD63 and Lamp2, can interact directly with cytosolic trafficking components to form its own trafficking vesicles. More work is required to elucidate this sorting and trafficking mechanism.

To elucidate the biological significance of the CLN5 interaction with lysosomal sorting receptors, we used siRNA to deplete CLN5 in HeLa cells. In CLN5-depleted cells, we found that the retrograde trafficking of both sortilin and CI-MPR to the Golgi compartment was impeded, which led to their degradation. Interestingly, CI-MPR and sortilin are degraded with similar kinetics in both CLN5- and retromer-depleted cells. On the basis of that finding, we tested the effect of CLN5 depletion on the recruitment of retromer to endosomes. Compared to mock- and CLN1-depleted cells, the recruitment of the Vps portion of retromer to endosomes is significantly reduced in CLN5-depleted cells, as shown by results revealing a decrease in Vps26 localization to endosomal membranes by immunofluorescence and Vps26 and Vps35 by membrane isolation.

The recruitment of retromer to endosomal membranes is a tightly regulated process that requires the small G proteins Rab5 and Rab7 (32, 36). Our results show a significant reduction in the amount of Rab7 and a slight reduction in the amount of Rab5 found on endosomal membranes in CLN5-depleted cells. Interestingly, we also found that the activation of Rab7 was significantly impaired but found no changes in the activation of Rab5. A recent paper demonstrated that the Mon1-Ccz1 complex is a Rab guanine nucleotide exchange factor (GEF) for the yeast homologue of Rab7 (26). It is possible that CLN5 could recruit this GEF to localize and/or activate Rab7. The molecular details of this event need to be further elucidated. If CLN5 is a transmembrane domain protein, it is possible that CLN5 could interact directly with Mon1 and/or Rab7. Alternatively, the interaction could be indirect and mediated via CLN3, a known interactive partner of CLN5. Further work is required to determine this.

Taken together, our results support a role for CLN5 in the retrograde trafficking of the lysosomal sorting receptors in mammalian cells. We propose that, upon arrival in the more acidic environment of the endosome (Fig. 10, step 1), cargo dissociates from the lysosomal sorting receptors and is replaced by CLN5 (Fig. 10, step 2). This provides a signal and/or scaffold to recruit and activate Rab7 (Fig. 10, step 2) followed by the recruitment of retromer (Fig. 8, step 3). This then enables the receptor to traffic to the Golgi compartment for another round of sorting and trafficking. In conclusion, results presented in this study are consistent with a new model suggesting that CLN5 acts as an endosomal switch, allowing lysosomal sorting receptors to recycle to the Golgi compartment for another round of vesicular trafficking and cargo sorting.

Fig 10.

Model showing the role of CLN5 in sorting from the endosomes. When the cargo-loaded lysosomal sorting receptor (green line) arrives at the endosome (Step 1), the change in pH causes a dissociation of the cargo from the receptor that subsequently leads to interaction with CLN5 (Step 2). This enables the recruitment and activation of Rab7 and the recruitment of retromer (Step 3) for recycling to the Golgi compartment, where it can interact with more cargo.