The tyrosine motifs of Lamp 1 and LAP determine their direct and indirect targetting to lysosomes

The lysosomal membrane proteins are the major protein constituents of the lysosomal membrane. On the basis of their sequence and overall structural homology, the family of lysosomal membrane proteins can be divided into five groups named LAP (Lysosomal acid phosphatase), Lamp 1 (Lysosomal associated membrane protein 1), Lamp 2, Lamp 3 (or Limp 1) and Limp II (Lysosomal integrated membrane protein II) [for review see Hunziker and Geuze (Hunziker and Geuze, 1996)]. Although LAP, Lamp 1 and Lamp 2 are type I membrane proteins, Lamp 3 harbours four and Limp II harbours two transmembrane segments.

Following their transit through the Golgi complex, the newly synthesised lysosomal membrane proteins can be delivered to late endosomes/lysosomes by two major pathways. The direct route involves transport from the trans-Golgi network (TGN) via the endosomal system to lysosomes, bypassing the plasma membrane. This applies for most of the lysosomal membrane proteins studied so far, including Lamp 1 (Harter and Mellman, 1992; Höning and Hunziker, 1995) and Limp II (Barriocanal et al., 1986; Sandoval et al., 1994). Trafficking of LAP to lysosomes follows an indirect route. The newly synthesised LAP precursor is first delivered into a recycling loop between endosomes and the cell surface before it is delivered to lysosomes. The half-life of LAP in this recycling loop is around five to six hours (Waheed et al., 1988; Peters et al., 1990), although transport of other lysosomal membrane proteins from the TGN to lysosomes takes less than 90 minutes (Barrioccanal et al., 1986, Green et al., 1987).

The known sorting information for intracellular trafficking of lysosomal membrane proteins is located in their short C-terminal cytoplasmic tails. Although Limp II utilises a leucine-based sorting motif (Sandoval et al., 1994), Lamp 1, Lamp 2, Lamp 3 and LAP carry tyrosine-based sorting motifs. Lamp 1, Lamp 2 and Lamp 3 have an 11 amino-acid cytoplasmic tail with a linker sequence of seven residues separating their tyrosine motifs from the membrane (Höning and Hunziker, 1995; Gough and Fambrough, 1997). By analysing splice variants of Lamp 2, which differ in the amino-acid composition of the tyrosine motif, it has been shown that differences in the steady-state distribution and the internalization rate are dependent on the tyrosine motifs (Gough and Fambrough, 1997; Gough et al., 1999). The cytoplasmic tail of LAP contains 19 amino acids and is composed of a linker sequence of eight residues, followed by a tyrosine motif and a C-terminal extension of seven residues (Peters et al., 1990; Lehmann et al., 1992; Prill et al., 1993).

Intracellular sorting of lysosomal membrane proteins requires their interaction with cytosolic adaptor complexes. LAP binds in vitro to the clathrin-associated adaptor complex AP2, but not to AP1 (Sosa et al., 1993), whereas Lamp 1 binds in vitro to AP1 and to AP2. Neither LAP nor Lamp 1 binds to AP3 in vitro (Höning et al., 1998). With the yeast two-hybrid system, however, a binding of Lamp 1 to the AP3 μ-chain was detectable (Ohno et al., 1998). Furthermore, cells that lack endogenous AP3 or are depleted in the μ3 chain mis-sort Lamp 1 to the cell surface (Le Borgne et al., 1998; Dell’Angelica et al., 1999). Because in cells lacking AP1, sorting of Lamp 1 is not affected (Meyer et al., 2000), it indicates that AP3 but not AP1 is critical for sorting of Lamp 1. It is not known where in the cell these adaptors bind to LAP or Lamp 1, except that AP1 and Lamp 1 colocalise in clathrin-coated structures at the TGN (Höning et al., 1996). AP2 is thought to mediate sorting of lysosomal membrane proteins into clathrin-coated vesicles at the cell surface, as has been documented for the cargo membrane proteins [for review see Marsh and McMahon (Marsh and McMahon, 1999)].

In the present study, we have analysed in BHK-21 cells the recycling of LAP between endosomes and the plasma membrane. By using truncated forms of LAP and chimeras in which the lumenal and transmembrane regions of LAP and Lamp 1 or Limp II and their cytoplasmic tails were mixed, we show that the membrane-proximal 12 residues of the LAP tail are necessary and sufficient to mediate recycling and that the tyrosine motif of LAP (YRHV) is the critical sequence element mediating recycling. When the binding of LAP and Lamp 1 to AP1 was analysed in vitro, we found that the residues in position +2 and +3 of the tyrosine motif determine the differential binding of Lamp 1 and LAP to AP1. The substitution of residues +2 and +3 in the LAP tyrosine motif for those of Lamp 1 and vice versa, which results in vitro in the gain and the loss of AP1 binding, respectively, was paralleled in vivo by a direct and an indirect lysosomal transport of LAP forms carrying the respective tails.

The cDNAs for wild-type rat Limp II, LAP carrying the Limp II tail and for Limp II carrying the LAP tail were kindly provided by I. Sandoval (Madrid, Spain). Chimeras between Lamp 1 and LAP were generated using an Afl-II site that was introduced into the six base-pairs coding for the last residue of the transmembrane region and the first cytoplasmic tail residue of LAP and Lamp 1 (Höning and Hunziker, 1995). LAPΔ7, which lacks the C-terminal seven tail residues, was cloned by introducing a premature stop codon replacing the codon for the tail residue alanine 13 (Lehmann et al., 1992). All LAPΔ7 variants in which the tail was substituted for the Lamp 1 tail or in which the tyrosine motif was modified were generated by PCR. The changes were introduced by using the respective coding 3′ primer also harbouring a stop codon followed by a XbaI cloning site and a 5′ primer harbouring the unique Msc-I restriction site of LAP for cloning (the primer sequences are available upon request). The correct sequence of all constructs was verified by sequencing using dye terminator cycle sequencing (Perkin Elmer Biosystems) and cloned into the mammalian expression vector pBEH. BHK-21 cells were stably transfected with the respective cDNAs, together with the vector pSV2pac using Effecten (Qiagen), and single-cell clones were selected with puromycine (Gibco).

LAP constructs harbouring the lumenal domain of LAP were immunoprecipitated with a rabbit serum (LS-4) raised against the soluble form of human LAP. LAP constructs habouring the cytoplasmic tail of LAP were immunoprecipitated using a rabbit serum (1-V) raised against the cytoplasmic tail of human LAP. The serum 1-V was also used for western blotting, whereas for the detection of the lumenal LAP-domain in western blots the serum LS-4 was replaced by the rabbit serum LM-9, which was raised against the membrane-associated form of human LAP.

Cells grown in 35-mm plates to confluency were rinsed twice with phosphate-buffered saline, then preincubated for one hour in methionine and cysteine-free growth medium containing 4% of dialyzed FKS followed by a short pulse of 15 minutes with 100 μCi of [³⁵S]cysteine/methionine (Amersham) in preincubation medium. After a chase of one hour in normal culture medium, the cells were harvested or subsequently exposed to neuraminidase (see below).

After metabolic labelling and a chase for one hour, the cells were treated with neuraminidase (Vibrio cholerae, Boehringer Mannheim). For selective desialylation of the fraction of LAP, Lamp 1 or Limp II located at the plasma membrane, the cells were placed on ice and incubated for one hour with 80 mU neuraminidase diluted in 1 ml PBS⁺⁺, pH 3.5 (PBS containing 0.9 mM CaCl₂ and 0,5 mM MgCl₂) adjusted to pH 5.5. Prior to experiments, the neuraminidase was dialysed overnight against 50 mM Na-acetate, pH 5.5, containing 0.15 M NaCl and 9 mM CaCl₂. To monitor exchange between intracellular- and plasma-membrane-associated LAP, Lamp 1 or Limp II, the cells were exposed to neuraminidase (80 mU/1 ml cell culture medium) at 37°C for varying periods of time. Neuraminidase treatment was terminated by washing the cells twice with ice cold PBS⁺⁺, pH 7.4 supplemented with 0,1 mM 2,3-dehydro-2-deoxy-N-acetylneuraminic acid. After additional washing with PBS⁺⁺, pH 3.0 and PBS⁺⁺, pH 7.4, the cells were harvested and solubilised in 10 mM Tris/HCl, pH 7.4, containing 150 mM NaCl, 2% Triton X-114 and the protease inhibitor cocktail P8340 (Sigma). After sonication, the homogenate was adjusted to 0.03% protamine sulfate, incubated for 10 minutes at 4°C and centrifuged for 30 minutes at 10.000 g. The supernatant was subjected to a Triton X-114 condensation as described in (Braun et al., 1989) to separate membrane-associated proteins from soluble ones. To immunoprecipitate LAP, Lamp 1 or Limp II from the detergent phase, the samples were mixed with 0.8 volumes of 10 mM phosphate, pH 7.4, containing 150 mM NaCl, 1% Triton X-100, 0.5% Na-deoxycholate, 0.5% SDS and 2 mg/ml bovine serum albumin. The samples were treated with Pansorbin prior to immunoprecipitation with the appropriate antiserum and protein A agarose (Sigma) as immunoadsorbent as described in (Waheed et al., 1988). After splitting the immunoprecipitates, four-fifths were washed, solubilised from the immunoabsorbent and separated on 4% isoelectric focusing (IEF) gels as described in (Braun et al., 1989). One-fifth of the immunoprecipitates were used for SDS-PAGE to control the purity of the immunoprecipitated LAP, Lamp 1 or Limp II. The radioactivity incorporated into the samples was quantified as described below.

Three 35-mm dishes of cells expressing LAP or LAP chimeras were chilled on ice and washed five times with ice cold PBS⁺⁺, pH 7.4, followed by the incubation with cold NHS-SS-biotin (5 mg/ml PBS⁺⁺, pH 7.5) to selectively biotinylate LAPs present at the cell surface. Biotinylation was stopped by washing twice with 50 mM glycine in PBS⁺⁺ and twice with PBS⁺⁺ before the cells of the first dish were harvested to quantify the amount of biotinylated cell-surface LAP. The remaining two dishes were incubated for 10 minutes at 37°C with prewarmed growth medium to allow endocytosis. The cells were then placed on ice to stop internalization and incubated twice for 20 minutes each with freshly prepared glutathione buffer (60 mg/ml glutathione, pH 8.0, 83 mM NaCl, 1.1 mM CaCl₂, 1.1 mM MgCl₂) to remove the biotin label from proteins present at the cell surface. After washing five times with PBS⁺⁺, the second dish was harvested. In this dish, the biotinylated LAP corresponds to the LAP fraction that has been internalised during the 10-minute recultivation at 37°C and protected from the biotin stripping. The third dish was again incubated for 10 minutes at 37°C. This allows the biotinylated LAP to recycle back to the cell surface from endosomes. After this incubation, plasma-membrane-associated biotin was removed as described above. In this dish, the biotinylated LAP corresponds to the LAP that had been internalized during first incubation at 37°C and did not recycle to the cell surface during the second incubation at 37°C. The cells were lysed in 10 mM Tris/HCl, pH 7.4, containing 150 mM NaCl and 0.1% TritonX-100. After sonification, the samples were centrifuged for 30 minutes at 100,000 g. The supernatant was split: one tenth was directly analysed by SDS-PAGE, and the remaining sample was subjected to precipitation of the biotinylated proteins by streptavidin agarose. The precipitates were resolved by SDS-PAGE and transferred onto nitrocellulose, followed by the detection of LAP by western blotting. The amount of internalised LAP can be quantified by comparing the signals from dish one and two. No signal in dish two would correspond to zero internalisation, whereas a signal in dish two equivalent to that of dish one would correspond to 100% endocytosis. Recycling is detectable by comparing the signals from dish two and three. If the signals in dish two and three are equal, no recycling has occured, whereas no signal in dish three would indicate 100% recycling of the internalised LAP. The same type of experiment was also performed with cells expressing Lamp 1 and Limp II and their chimeras.

In order to analyse the endosomal localization of LAP reporter proteins that carry mutated cytoplasmic tail sequences of either LAP (RMQAQPPGYRTI in LAPΔ7/TI) or Lamp 1 (RKRSAHAGYQHV in Lamp1/HV), BHK cells expressing the indicated mutants were incubated for 15 minutes with antibodies against the lumenal domain of LAP before fixation. Subsequently, the cells were permeabilised with Triton X-100, followed by incubation with an antibody against EEA1 (Transduction Lab, USA). After washing, and an incubation with 10% goat serum, the primary antibodies were visualised using goat secondary antibodies coupled to either Cy2 or Cy3 (Dianova, Germany). To examine the lysosomal delivery, the cells were incubated with antibodies against LAP (see above) for one hour, fixed and subsequently labelled for endogenous Lamp 1, followed by the incubation with the respective secondary antibodies. All samples were mounted in Fluoromount (DAKO, Danmark) and analysed with a confocal laser-scanning microscope (Zeiss, Germany). All images are confocal sections obtained under identical microscopical settings. For the quantification of the colocalisation of LAP with either EEA1 or Lamp 1, all red-coloured pixels (EEA1 or Lamp 1) and all green pixels (LAP) that were found at identical positions in a confocal section were compared to those green- and red-coloured pixels that could not be located to the same position. This analysis was performed for at least 40 randomly selected single cells.

Radiolabelled and immunoprecipitated proteins were resolved by SDS-PAGE, followed by detection and quantification of the incorporated radioactivity using a Fujix BAS 1000 bioimaging system (Fuji Photo Film Co., Japan) and the Image Gauge software (Version 3.0) supplied by the manufactor. Signals obtained during western blotting were detected using a CCD camera (Cybertech) and quantified using the imaging and documentation software Wincam (Cybertech).

Repeated recycling between endosomes and the plasma membrane before delivery to lysosomes distinguishes the biosynthetic route of LAP from that of other lysosomal membrane proteins such as Lamp 1 and Limp II. Here, we aimed at the identification of the sorting signals that mediate the recycling of LAP. To discriminate between signals in the lumenal or transmembrane region of LAP and those that are located within the cytoplasmic tail, we constructed LAP chimeras in which the tail of LAP was fused to the transmembrane region and the lumenal part of Lamp 1 or Limp II. In addition, we substituted the tails of Lamp 1 and Limp II for that of LAP (Fig. 1). The constructs encoding LAP, Lamp 1, Limp II and their chimeras were stably expressed in BHK-21 cells.

The recycling of LAP between endosomes and the plasma membrane is accompanied by an elevated cell-surface expression of LAP compared to that of other lysosomal membrane proteins (Braun et al., 1989). This can be assayed by incubating the cells on ice in the presence of neuraminidase. The desialylation of oligosaccharides attached to the lumenal domain of LAP shifts the isoelectric point to a more basic value and can be detected by isoelectic focusing. A typical experiment is depicted in Fig. 2. When BHK cells were metabolically labelled for 15 minutes and chased for 60 minutes, followed by a neuraminidase treatment on ice, 22% of LAPs were recovered as desialylated forms (lane B). In non-treated cells, only 6% of LAPs were detectable as desialylated forms (lane A), indicating that 16% of LAPs are present at the plasma membrane. Treatment of cells with neuraminidase at 37°C also allows the detection of those LAP molecules that are delivered from the TGN or endosomes to the plasma membrane during the incubation period in the presence of neuraminidase. If such a transport occurs, the fraction of desialylated LAP that forms after treatment at 37°C should be higher than after treatment on ice. After neuraminidase treatment for 20 minutes at 37°C, 82% of LAPs were recovered as desialylated forms (lane C). This indicates that 60% of the LAPs had been delivered to the cell surface from internal membranes within 20 minutes at 37°C (compare lanes B and C).

The third assay to monitor trafficking of LAP is based on biotinylation of cell-surface-associated LAP and allows us to monitor directly the internalisation of LAP and its subsequent recycling from endosomes back to the cell surface. Three dishes were biotinylated on ice in parallel. The cells from the first dish were harvested directly after biotinylation and processed for the detection of total and biotinylated LAP (see Materials and Methods for details). The remaining two dishes were incubated for 10 minutes at 37°C to allow internalisation of biotinylated LAP. Subsequently, the dishes were placed on ice and surface-associated biotin was stripped off using reducing conditions. Thus, only biotinylated LAP that had been internalised is protected from biotin stripping. After stripping, the cells from the second plate were harvested to quantify the fraction of internalised biotinylated LAP. The third dish was incubated for a second time for 10 minutes at 37°C, followed by biotin stripping on ice. During the second incubation at 37°C, biotinylated LAP that had been internalised during the first incubation at 37°C can return back to the plasma membrane where it is sensitive to biotin stripping.

A typical experiment obtained for LAP is illustrated in Fig. 3. Lanes I to III in Fig. 3A represent 10% of total LAP in dishes one to three. In Fig. 3B, lane I represents the fraction of LAP susceptible to biotinylation on ice, which corresponded to 12% of the total LAP. After incubation for 10 minutes at 37°C (Fig. 3B, lane II), 78% of the biotinylated LAP was protected from biotin stripping, indicating its efficient internalisation. When the cells were incubated for an additional 10 minutes at 37°C to allow recycling of the biotinylated LAP that had been internalised during the first incubation for 10 minutes at 37°C, 81% of the biotinylated LAP had returned back to the plasma membrane, as indicated by the susceptibility to biotin stripping (compare lanes II and III in Fig. 3B). In control experiments, we confirmed that after biotinylation of the cell surface, the biotin label is completely removed under conditions for biotin stripping (not shown). We also confirmed that recycling of LAP from endosomes back to the plasma membrane was dependent on the incubation at 37°C, as the amount of biotinylated LAP remained constant when the cells were kept on ice between the first and the second biotin stripping (not shown).

When these three assays were applied to cells expressing LAP chimeras, in which the cytoplasmic tail of LAP was replaced with that of Lamp 1 (LAP-Lamp-1) or of Limp II (LAP-Limp-II), it became apparent that the expression of the chimeras at the cell surface, their transport to the cell surface and their recycling back to the cell surface were severely decreased or even abolished (Fig. 2; Fig. 3; Table 1). The neuraminidase assays indicated that the fraction at the cell surface and the fraction of chimeras transported to the cell surface during an incubation for 20 minutes at 37°C were approximately six-fold lower than for LAP. The transport back to the cell surface of biotinylated chimeras that had been internalised was below the limit of detection. These data clearly indicate that transport of LAP to the cell surface is greatly facilitated by the cytoplasmic tail of LAP and that transport from endosomes back to the cell surface requires the cytoplasmic tail of LAP.

In order to test whether the cytoplasmic tail of LAP is sufficient to induce transport of other membrane proteins to the cell surface and their recycling between endosomes and plasma membrane, chimeras were constructed in which the cytoplasmic tails of Lamp 1 (Lamp-1-LAP) and of Limp II (Limp-II-LAP) were substituted by that of LAP. Although expression of Lamp 1 and Limp II at the cell surface and the transfer to the cell surface during a 20-minute incubation period at 37°C were low (below 5%), these values were markedly higher than for Lamp-1-LAP and Limp-LAP chimeras (Table 1). For Lamp-1-LAP, a fraction as high as 52% was recovered at the cell surface. Although the return back to the cell surface of internalised Lamp 1 and Limp II was below the limit of detection, between 45 and 55% of the internalised Lamp-1-LAP and Limp-II-LAP returned back to the cell surface (Table 1). These data clearly indicate that the cytoplasmic tail of LAP is not only necessary for the transport of LAP to the cell surface and its recycling between endosomes and surface but is also sufficient to induce these trafficking routes when fused to other membrane proteins.

We next tried to narrow down the sequence within the LAP tail that mediates recycling. It was shown earlier that a truncated form of LAP lacking the C-terminal seven amino acid residues (LAPΔ7, Fig. 4) harbours the signals for rapid internalisation and also for basolateral sorting (Lehmann et al., 1992; Prill et al., 1993). When we analysed the truncated LAPΔ7 for its ability to recycle, we observed no difference to LAP (Table 2). Also, the expression at the cell surface and the transport to the cell surface within a 20-minute incubation period at 37°C were comparable to that of LAP. This indicates that the first 12 tail residues of LAP are sufficient to mediate recycling of LAP between endosomes and the plasma membrane.

The 12-residues tail sequence of LAPΔ7 can be divided into two parts, one of which is the linker sequence of eight residues separating the C-terminal tyrosine motif (–YRHV) from the membrane. This is similar to the Lamp 1 tail (Fig. 4), which consists of a linker of seven residues and the C-terminal tyrosine motif (–YQTI). We were interested in defining whether the differences between the linker sequences or between the tyrosine motifs cause the differences in trafficking of LAP and Lamp 1. To analyse this, we constructed a chimeric LAP in which the tyrosine motif of LAPΔ7 tail was substituted with that of Lamp 1 (LAPΔ7/YQTI). In a second chimera, the tyrosine motif of Lamp 1 was replaced by that of LAP (Lamp 1/YRHV, Fig. 4). When these chimeras were analysed for their ability to recycle between endosomes and the plasma membrane, recycling of LAPΔ7/YQTI was not detectable, whereas that of Lamp 1/YRHV (58%) was comparable to that of LAP (72%) and LAP Δ7 (70%) (see Table 2). These data show that the tyrosine motif of LAP, but not its linker sequence, is critical for recycling.

The in vivo experiments described above have shown that the tyrosine motif of LAP, but not Lamp 1, mediates recycling between endosomes and the cell surface. Several studies have demonstrated the interaction of these signals with the cytoplasmic adaptor complexes AP2 and/or AP1 in vitro. LAP binds with high affinity to AP2 but not to AP1 (Sosa et al., 1993; Höning et al., 1998), whereas Lamp 1 binds to AP1 and to AP2 (Höning et al., 1996).

Here, we have analysed whether mutating either the tyrosine motif within LAP or Lamp 1 changes the ability of both cytoplasmic tail peptides to bind to the clathrin-associated adaptor complexes AP1 and AP2 by using a biosensor. In agreement with published data, we detected a high-affinity binding of AP1 (38nM) and AP2 (54nM) to the tail of Lamp 1, whereas only AP2 (27 mM) binds to the LAP tail. In agreement with our in vivo results, LAPΔ7 exhibited a high affinity for AP2 (22nM), but no binding to AP1 was detectable (Fig. 5). We next analysed binding of AP1 and AP2 to peptides corresponding to the tails of the LAP-chimeras LAPΔ7/YQTI and Lamp 1/YRHV described above (see Fig. 4). Although both peptides bound AP2 with similar affinities (not shown), binding to AP1 was only detectable for LAPΔ7/YQTI peptide (Table 3A). These results indicate that AP2 does not discriminate between the tyrosine motifs of LAP and Lamp 1. AP1 binding, however, is more selective and depends on the Lamp 1 tyrosine motif –YQTI, but adaptor binding was independent of the Lamp 1 linker sequence.

In additional experiments, we changed the residues +1, +2 or +3 following the tyrosine in the tail of Lamp 1 for the respective residues of the LAP tyrosine motif (Table 3B). When binding of AP1 to these peptides was recorded, it turned out that replacing residues +1 (Gln to Arg) had a minor effect on AP1 binding. However, changing the residue +2 (Thr to His) lowered the affinity about 11-fold. Replacing residue +3 (Ile to Val) had an intermediate effect on AP1 binding. When two of the residues +1, +2 and +3 were replaced simultaneously, replacing residues +2 together with one of the others again had the most dramatic effect, and replacing +2 and +3 together abolished AP1 binding almost completely.

When similar experiments were performed with LAPΔ7 peptides in which residues +1, +2 and +3 were replaced by those of Lamp 1 (Table 3C), we observed that replacing residue +1 (Arg to Glu) did not improve AP1 binding. Replacing residue +2 (His to Thr) improved AP2 binding markedly, whereas replacing residue +3 (Val to Ile) had an intermediate effect. Again when two residues were replaced simultaneously, substitution of residue +2 with one of the two others restored AP1 binding most effectively. These data demonstrate that in addition to the tyrosine the residue +2 in tyrosine motifs of the –YXX→ type is most important for the binding of AP1 in vitro.

As described above, it is possible to induce AP1 binding of LAPΔ7 by substituting the residues +2 and +3 of its tyrosine signal with the respective residues of the Lamp 1 tyrosine signal (LAPΔ7/TI, Table 2). On the other hand, the Lamp 1 tail lost its AP1-binding capacity when the residues +2 and +3 of its tyrosine signal were substituted with the respective residues of LAP (Lamp 1/HV, Table 2). Although it is not absolutely clear with which adaptor complexes Lamp 1 and LAP interact during their intracellular itinerary, a more general interpretation of these results is that differences in the affinity of sorting signal binding to adaptors may result in sorting of the respective proteins into diverse intracellular pathways.

In order to test the idea that the gain or the loss of adaptor binding correlates with the inability or ability to recycle between endosomes and the cell surface, chimeric LAP carrying a LAPΔ7/TI or a Lamp 1/HV cytoplasmic tail (see Fig. 4) was expressed in BHK cells. Recycling between endosomes and the plasma membrane was assayed as described above. As shown in Fig. 6, only 17% of the internalised LAPΔ7/TI chimera was recycled back to the plasma membrane within 10 minutes. However, when the Lamp 1/HV chimera was analysed, 64% of the internalised fraction was found to recycle from endosomes back to the plasma membrane within 10 minutes. These results demonstrate that in the case of the tyrosine motifs of LAP and Lamp 1, recycling from endosomes to the plasma membrane in vivo correlates with differences in adaptor binding in vitro.

These results were further supported by experiments in which the endosomal localization and the lysosomal delivery of the LAPΔ7/TI and Lamp 1/HV chimera were analysed by immunofluorescence. In the first set of experiments, cells expressing the two chimeras were incubated with antibodies against the lumenal LAP reporter domain for 15 minutes prior to fixation. Following permeabilization, the cells were stained for the endocytosed anti-LAP antibodies and the early endosomal marker EEA1 (Mu et al., 1995; Christoforidis et al., 1999). Colocalisation of LAPΔ7/TI (green colour) with EEA1 (red colour) results in staining of the respective structures in a yellow colour. As shown in Fig. 7A, only a small amount of LAPΔ7/TI colocalised with EEA1 compared to Lamp 1/HV (Fig. 7B), which was much more abundant in EEA1-positive endosomes. The quantification of the degree of colocalisation (see Materials and Methods for details) revealed that only 20% of LAPΔ7/TI but more than 60% of Lamp 1/HV could be colocalised with EEA1 in early endosomes after 15 minutes of anti-LAP antibody endocytosis, suggesting that substituting the last two residues within the Lamp 1 tyrosine motif with those of LAP leads to the retention of the mutant within endosomes. If this is really the case, lysosomal delivery of this chimera should be less efficient compared to LAPΔ7/TI. Indeed, when we analysed the localization of the anti-LAP antibodies after one hour of endocytosis, more than 80% of LAPΔ7/TI could be localised to lysosomes, as revealed by the colocalisation with endogenous Lamp 1 (Fig. 7C), although less than 35% of Lamp 1/HV was detectable in lysosomes (Fig. 7D). In conclusion, Lamp 1/HV is not only retained in endosomes but also exhibits an inefficient lysosomal delivery compared to LAPΔ7/TI.

In most cells analysed so far, newly synthesized lysosomal membrane proteins like Lamp 1 or Limp II are delivered to lysosomes via a strictly intracellular pathway avoiding transient appearance at the cell surface. Under certain conditions, such as overexpression or mutation of cytoplasmic tail residues (e.g. the glycine residue preceding the tyrosine motif in Lamp 1), a variable fraction appears at the cell surface during transport from the endoplasmic reticulum to lysosomes. The molecules appearing at the cell surface are internalised and delivered to lysosomes [(Höning and Hunziker, 1995); this study for overexpressed Lamp 1 and Limp II]. This is in contrast to newly synthesized LAP, which enters a recycling loop between endosomes and the plasma membrane before it is delivered to lysosomes. Here we show that the cytoplasmic tail of LAP is necessary and sufficient to mediate the recycling of LAP and Lamp 1 or Limp II chimeras carrying the LAP cytoplasmic tail. The sequence necessary for the recycling could be narrowed down to the tyrosine motif YRHV in the LAP tail, although the related tyrosine motif YQTI of Lamp 1 was unable to support recycling.

Differential sorting of LAP and Lamp 1 mediated by their tyrosine motifs may occur at more than one site within the cells. The observation that overexpressed Lamp 1 and LAP are both internalised and then separated, Lamp 1 being transported to lysosomes and LAP being recycled to the cell surface, clearly identifies the endosomes as one site where the tyrosine motifs of LAP and Lamp 1 mediate differential sorting. The sequence requirements for this differential sorting were therefore the focus of our interest and analysed in more detail.

The observation that the linker of Lamp 1 (seven residues) can replace the linker of LAP (eight residues) and vice versa without affecting the differential sorting does not imply that the tyrosine motifs operate independently of these linker sequences. For example, substituting the proline residue in position six of the LAP-linker sequence with alanine severely impairs internalisation and basolateral sorting of LAP and results in its accumulation at the cell surface (Lehmann et al., 1992; Prill et al., 1993). Residues such as proline 6 in the LAP tail presumably help to expose the tyrosine motif to the cytoplasmic adaptors rather than being a determinant that is recognised by the adaptors. Residues with a similar function have also been identified in the C-terminal extension comprising residues 13 to 19 of the LAP tail (Lehmann et al., 1992; Prill et al., 1993).

Although the data in this study identified endosomes as a site where the tyrosine motif of LAP and Lamp 1 mediate differential sorting of both membrane cargo proteins, the cytoplasmic adaptors involved in this sorting remain to be identified. However, it was tempting to speculate that the observed differences in sorting between LAP and Lamp 1 are attributable to differences in the binding to adaptor complexes in vitro. If this were so, we would want to characterise the important sequence determinants. It was already known that peptides corresponding to the cytoplasmic tails of LAP and Lamp 1 bind in vitro to AP2 (Sosa et al., 1993; Höning et al., 1996; Höning et al., 1998), whereas only the Lamp 1 peptide binds to AP1 (Höning et al., 1996). By examining peptides corresponding to mutant forms of the LAP and Lamp 1 tail, we could show that binding of AP1 was dependent on the Lamp-1–YQTI tyrosine motif. Single and double amino-acid substitutions identified the threonine in position +2 as the most important residue for AP1 binding. In addition, the type of hydrophobic residue in position +2 of the tyrosine motif had a moderate effect on AP1 binding. In fact, replacing the histidine in positon +3 of the LAP–YRHV tyrosine motif by threonine restored AP1 binding to the level observed for the Lamp 1 tail. In this context, it was interesting to note that AP2 binding was only slightly affected by the amino-acid changes in the position +2 and +3 of the LAP and Lamp 1 tyrosine motifs, showing that AP2 is not as selective as AP1. In summary, our in vitro analyses have revealed that changes in the residues +2 and +3 of a tyrosine motif can determine the loss or gain of adaptor binding. In this context, we would like to point out that the present analysis can not be used to predict which adaptor complex(es) is/are recognised by LAP or Lamp 1 in the living cell or where other adaptor complexes such as AP3 or AP4 or other adaptor-like proteins may discriminate between similar sorting determinants, as shown here for AP1, thereby mediating a selective cargo recognition. However, we think that the in vitro and the in vivo data presented here show that variations not only in the +3 position but also in the +2 position in one type of tyrosine-sorting motif (–YXX→) affect adaptor binding in vitro and in vivo, and this correlates with differences in the efficiency of sorting into a specific transport route.

Indeed, the relevance of the in vitro data for the ability of LAP to recycle between endosomes and the plasma membrane in vivo was confirmed by the observation that a LAPΔ7 tail, in which the residues +2 and +3 correspond to those of the Lamp 1 tyrosine motif (LAPΔ7/TI), was no longer able to mediate recycling, whereas a Lamp 1 tail, in which the residues +2 and +3 correspond to those of the LAP tyrosine motif (Lamp 1/HV), mediates recycling of the LAP reporter between endosomes and the cell surface. Furthermore our immunofluorescence studies on the localisation of internalised LAPΔ7/TI and Lamp 1/HV show that the latter mutant, in contrast to wild-type Lamp 1 (not shown) and LAPΔ7/TI, is not efficiently delivered through the endocytic pathway to lysosomes. The observation that Lamp 1/HV exhibits a high degree of colocalisation with EEA1 suggests that the mutant is retained in the endosomal system, although wild-type Lamp 1 and the respective LAP mutant LAPΔ7/TI are more efficiently targeted in the direction of lysosomes.

In vivo, the differential sorting of LAP and Lamp 1 following their internalisation from the cell surface, such as after recycling of LAP to the cell surface or the rapid delivery of Lamp 1 to lysosomes, has identified endosomes as a crucial site for sorting of the two proteins. In this context, it is interesting to note that the recycling of a truncated transferrin receptor with only four tail residues favoured the idea that recycling from endosomes to the cell surface occurs by default (Johnson et al., 1993; Mayor et al., 1993). This view has been challenged recently by the observation that mutations of a tyrosine within the β2 integrin tail affects transport from the endosome to the cell surface, although internalisation is not affected (Fabbri et al., 1999). Thus, one can conclude that Lamp 1 is efficiently recognised by a sorting factor for further transport to lysosomes, whereas LAP, owing to a lower affinity, is retained in the endosome, which indirectly increases the availability of LAP for recycling to the cell surface.

In this study, we have identified the residues within the respective tyrosine motifs of LAP and Lamp 1 that mediate this differential endosomal sorting; however, the interacting cytosolic adaptor or sorting factor to which the two proteins may bind to remains to be characterised. Several cytosolic factors that were found on endosomes including the adaptor complexes AP1 to AP4 may be candidates for mediating endosomal sorting events. In addition other factors such as β-COP (Aniento et al., 1996), specific rab proteins or adaptor-associated proteins such as PACS-1 or TIP47 may also be considered in this context.

Although PACS-1 and TIP47 mediate endosomal-sorting events, an involvement for sorting of lysosomal proteins within the endosomal system lacks any experimental support. PACS-1 binds to acid cluster motifs present in the cytoplasmic tails of proteins, such as the mannose-6-phosphate receptors, furin or PC6B. More importantly, PACS1 is thought to mediate its function in concert with adaptor complexes such as AP1 and AP3. But, in contrast to sorting on route to lysosomes, the data available today suggest that PACS-1 mediates retrieval of proteins from endosomes to the TGN (Wan et al., 1998; Crump et al., 2001). This is similar to the function of TIP47, a protein that has been shown to bind to the diaromatic phenylalanine/tryptophane motif of the small mannose-6-phosphate receptor and a membrane-proximal signal in the 300kDa mannose-6-phosphate receptor mediating the return of the receptors from late endosomes to the TGN. In addition, rab9 directly participates in the binding of TIP47 to MPR300, possibly regulating this interaction (Diaz and Pfeffer, 1998; Orsel et al., 2000; Carroll et al., 2001). However, as neither LAP nor Lamp 1 harbour a motif known to be recognised by TIP47, its involvement in lysosomal targeting is hypothetical.

Numerous studies have shown that the heterotetrameric adaptor complexes are important for the targeting of lysosomal membrane protein to lysosomes. Although we reported the interaction of Lamp 1 with the AP1 complex in vitro and by electron microscopy in MDCK epithelial cells (Höning et al., 1996), the involvement of AP1 in the endosomal sorting of Lamp 1 for further delivery to lysosomes is questionable. Recent studies could show that in endosomes, AP1 is important for the transport of MPRs (Meyer et al., 2000) and of Shiga toxin (Mallard et al., 1998) to the TGN. Thus, available evidence indicates that AP1 is involved in transport from endosomes to the TGN rather than from endosomes to the cell surface or to lysosomes. Whether AP2, the adaptor complex that particpates in cargo selection during the formation of clathrin-coated vesicles from the plasma membrane, has a functional role on endosomes is not clear today. However such a function should not be fully excluded, as the dominant role of AP2 in endocytosis causes experimental difficulties for testing its role in endosmal sorting. AP3, however, plays a role in the sorting of lysosomal membrane proteins, both in vitro and in vivo. It has been shown in vitro to bind via its μ-subunit to the tyrosine motif of Lamp 1 (Ohno et al., 1998), and, in vivo, the microinjection of a μ3 antisense construct results in mis-sorting of Lamp 1 to the cell surface (Le Borgne et al., 1998). Additionally, in mouse and human cells that lack a functional AP3 complex owing to mutations in either the δ-subunit or β3, Lamp 1 indirectly reaches the lysosome via the cell surface, strongly favouring a function of AP3 in sorting of Lamp 1 (Kantheti et al., 1998; Dell’Angelica et al., 1999). Although AP3 was localised at the TGN and also on endosomes, it is not clear yet where in the cell the interaction between AP3 and Lamp 1 takes place. In addition to AP1 and AP3, Bonifacino and coworkers recently showed in vitro that AP4 is able to bind to tyrosine-based sorting motifs, especially to one that is similar to a motif found in Lamp 2 (–YEQF) (Aguilar et al., 2001). As this motif is binding with higher affinity to μ2 and μ3A compared to μ4, the authors further analysed the intracellular sorting of a reporter protein harbouring an artificial μ4 selective tyrosine motif (–DLYYDPM). Lysosomal sorting of the reporter was independent of AP3, suggesting that AP4 may have a functional role in lysosomal targeting, although direct experimental evidence for AP4 binding to a lysosomal membrane protein is still missing.

Another group of adaptor-interacting proteins, the Golgi-localized, gamma-ear-conataining, ARF binding proteins, or GGAs (GGA 1, 2 and 3), are important for the TGN sorting of mannose-6-phosphate receptors, although lysosomal membrane protein localisation was not dependent on GGA function (Hirst et al., 2000; Puertollano et al., 2001; Zhu et al., 2001). In conclusion there is no evidence so far for a direct involvement of GGAs in the sorting of lysosomal membrane proteins, especially not at the level of endosomes.

The further analysis of LAP and Lamp 1 trafficking, namely the loss or induction of recycling in cell lines that lack one of the candidate adaptors associated with endosomes, such as AP1, AP3 and AP4, may help to identify the adaptor that is responsible for the differential endosomal sorting of LAP and Lamp 1. Such cell lines will also allow us to define whether the differential sorting of LAP and Lamp 1 requires an adaptor for back transport to the cell surface or/and for forward transport to late endosomes/lysosomes.

We thank I. Sandoval (Spain) for providing us with the Limp II cDNA and antibodies against Limp II. We gratefully acknowledge the help of A. Fingerhut for her help during the adaptor purification. This work is supported by the D.F.G. (S.F.B. 523, A5).