An unconventional dileucine-based motif and a novel cytosolic motif are required for the lysosomal and melanosomal targeting of OA1

Ocular albinism type 1 (OA1), the most common form of ocular albinism, is an X-linked inherited disorder characterized by foveal hypoplasia with severe reduction of visual acuity, nystagmus, strabismus, photophobia, iris translucency, hypopigmentation of the retina, and misrouting of the optic tracts, resulting in loss of stereoscopic vision (King et al., 1995). Patients affected with the disorder also show the presence of melanosomal abnormalities, including giant melanosomes (macromelanosomes), in the pigment cells of the skin and eyes (melanocytes and retinal pigment epithelium, RPE), suggesting an underlying defect in melanosome biogenesis (O'Donnell, Jr et al., 1976; Garner and Jay, 1980; Wong et al., 1983). The OA1 protein is an integral membrane glycoprotein, specifically expressed in melanocytes and RPE, and displaying features of G-protein-coupled receptors (GPCRs) (Schiaffino et al., 1996; Schiaffino et al., 1999). In fact, OA1 has seven transmembrane domains and homologies with several members of the GPCR superfamily, including residues highly conserved in most GPCRs, some of which are the site of albinism-causing mutations (Schiaffino et al., 1999; d'Addio et al., 2000). In addition, OA1 binds heterotrimeric Gi, Go and Gq proteins by co-immunoprecipitation and in vitro binding assays (Schiaffino et al., 1999).

However, at variance with canonical GPCRs, OA1 is not localized to the cell surface, but at steady state is exclusively detectable on the membrane of intracellular organelles. In fact, previous morphological and biochemical studies performed by us and others have established that the endogenous OA1 in melanocytes is excluded from the plasma membrane and is mainly localized to late endosomes, lysosomes and melanosomes (Schiaffino et al., 1996; Schiaffino et al., 1999; Samaraweera et al., 2001; Basrur et al., 2003). Melanosomes represent the specialized subcellular organelles of pigment cells devoted to the synthesis, storage and transport of melanins. Genetic, biochemical and functional evidence suggests that melanosomes are related to lysosomes (Orlow, 1995; Dell'Angelica et al., 2000). Nevertheless, melanosomes also differ from lysosomes by being unique to pigment cells, characterized by specific resident proteins and displaying a distinctive structure. Indeed, melanosomes are classified into four maturation stages based upon melanin content and ultrastructural morphology (stage I-III correspond to immature melanosomes, whereas stage IV corresponds to mature organelles) (Seiji et al., 1963; Marks and Seabra, 2001). In addition, melanosomes co-exist with lysosomes in pigment cells and are inaccessible to fluid-phase markers even at long chase times, thus appearing distinct from endocytic organelles (Raposo et al., 2001).

Little is known about the sorting mechanisms that retain OA1 at intracellular locations and drive it to its bipartite lysosomal/melanosomal compartment. Glycosylation does not seem to be involved (Shen and Orlow, 2001) and, although missense mutations identified in patients with ocular albinism can lead to protein misfolding and endoplasmic reticulum (ER) retention, missense mutations apparently do not provoke other kinds of sorting defects (d'Addio et al., 2000; Shen et al., 2001). In addition, as for other melanosomal proteins including tyrosinase and TRP1 (Winder et al., 1993), OA1 is efficiently delivered to the lysosomes when heterologously expressed in non-melanocytic cells, indicating that it contains trafficking information recognized by the lysosomal sorting machinery even in the absence of melanosomes (Schiaffino et al., 1999; Shen et al., 2001). However, OA1 does not contain any canonical tyrosine or dileucine-based motifs, which often direct the traffic of lysosomal and melanosomal membrane proteins (Setaluri, 2000; Bonifacino and Traub, 2003). To shed light on the targeting determinants of OA1, we generated a series of chimeric and mutant OA1 proteins and analyzed their subcellular distribution when expressed in melanocytic and non-melanocytic cells.

We and others previously reported that, when heterologously expressed at moderate levels in COS7 cells, OA1 is sorted to late endosomes and lysosomes, showing extensive colocalization with cathepsin D, LAMP2 and DAMP, a basic congener of dinitrophenol that accumulates in acidic compartments (Schiaffino et al., 1999; Shen et al., 2001). We obtained similar results in HeLa cells transiently expressing a mycHis-tagged OA1 (OA1-mycHis). Indeed, the recombinant OA1-mycHis displays a vesicular cytosolic distribution and colocalizes largely with the fluid-phase marker Lucifer Yellow, subjected to 1 hour of chase prior to fixation (Fig. 1A). This time interval allows accumulation of the tracer into the late-endocytic pathway, particularly into lysosomes, because, in these chase conditions, its staining overlaps extensively with DAMP and LAMP1 (data not shown). In melanocytes, the endogenous OA1 has been localized to late endosomes and lysosomes (Samaraweera et al., 2001) and to melanosomes (Schiaffino et al., 1996; Schiaffino et al., 1999; Schiaffino et al., 2002; Basrur et al., 2003); however, quantitative studies with multiple markers have never been performed.

To determine the steady-state localization of endogenous OA1 in melanocytes, we took advantage of the pigmented human melanoma cell line MNT1, which has been previously utilized as a suitable model for melanosome biogenesis (Kushimoto et al., 2001; Raposo et al., 2001). The subcellular distribution of OA1 was analyzed by indirect immunofluorescence and confocal microscopy, using markers of different subcellular compartments: LAMP1, a marker of late endosomes and lysosomes; Pmel17, a marker of immature melanosomes; TRP1, a marker of mature melanosomes; EEA1, a marker of early endosomes; and Lucifer Yellow at 1 hour chase, to label the lysosomes. Previous quantitative ultrastructural studies have shown that, in MNT1 cells, some degree of overlap exists among LAMP1, Pmel17 and TRP1 (Raposo et al., 2001). In particular, LAMP1 is also present to a minor extent on immature and mature melanosomes, whereas Pmel17 labels stage I-III melanosomes and TRP1 labels stages III and IV melanosomes (Raposo et al., 2001). By contrast, melanosomes were found inaccessible to fluid-phase markers (Raposo et al., 2001); therefore, Lucifer Yellow staining is expected to label the late-endocytic pathway only.

The quantitation of colocalization was performed by assessing the percentage of OA1-positive vesicles on confocal micrographs that were also positive for the different markers (Fig. 1B and Fig. S1, supplementary material). This statistical analysis showed that endogenous OA1 mostly co-distributes with LAMP1 (88%) and with Pmel17 (74%). Essentially identical results were obtained in MNT1 cells transiently transfected with the OA1-mycHis expression vector and immunodecorated with anti-His₆ antibodies. Exogenous OA1 colocalizes by 82%±0.1 with LAMP1 (vesicles counted: 305; cells counted: 18) and 74%±0.2 with Pmel17 (vesicles counted: 490; cells counted: 26). Minor colocalizations were found between endogenous OA1 and either TRP1 (21%) or EEA1 (6%), whereas about a third (31%) of the OA1 compartment corresponds to that labeled by Lucifer Yellow (Fig. 1B).

To establish which domains of OA1 are sufficient for intracellular targeting, we generated protein chimeras consisting of the N-terminus and transmembrane domain of rat LAMP1 and the cytosolic regions of OA1 (LAMP1/OA1 chimeras). LAMP1 is a type 1 lysosomal membrane glycoprotein that contains specific sorting signals within its cytosolic C-terminal tail. In fact, previous studies have shown that, when its C-terminus is deleted, LAMP1 accumulates instead at the plasma membrane by means of the default pathway (Williams and Fukuda, 1990) and hence can be conveniently used as a `neutral' reporter (Simmen et al., 1999; Storch et al., 2004). In our case, if any of the cytosolic domains of OA1 contain lysosomal targeting signals, when fused to the C-terminus-deleted LAMP1 they are expected to re-direct targeting of the protein from the cell surface to the lysosomes. The precise regions of OA1 utilized to generate LAMP/OA1 chimeras are indicated in Fig. 2A. Since the limits of transmembrane domains are not predictable with certainty in OA1, few additional amino acids were included over the putative cytosolic loops. For the same reason, the stop transfer signal RKR at the end of the transmembrane domain of rat LAMP1 was retained in the constructs, to guarantee proper insertion of the chimeric proteins within the membrane. Nevertheless, similar results were obtained in the presence or in the absence of the RKR tripeptide either in HeLa or COS-7 cells (data not shown).

We first analyzed the subcellular distribution of LAMP/OA1 chimeras in transiently transfected HeLa cells by indirect immunofluorescence, using an antibody directed specifically against LAMP1 of rat origin and showing no cross-reaction with the protein from other species. Although LAMP/OA1 chimeras did not show any evident ER retention, cells were treated with cycloheximide 1 hour prior to fixation to eliminate the pool of newly synthesized proteins (see Materials and Methods). The results of this analysis are shown in Fig. 2B. Interestingly, we observed that addition of the first or the second cytosolic loops of OA1 to the C-terminus-deleted LAMP1 (LAMP/i1 and LAMP/i2) does not interfere with its plasma membrane localization, visible as a diffuse cell staining that is more intense at margins and membrane spikes and corroborated by colocalization with farnesylated GFP (Fig. 2B and not shown). Conversely, chimeric proteins carrying the third cytosolic loop or the C-terminal tail of OA1 (LAMP/i3 and LAMP/CT) showed a complete or partial vesicular cytoplasmic distribution, respectively (Fig. 2B), and colocalized with Lucifer Yellow at 1 hour chase (not shown), suggesting that these regions contain lysosomal sorting signals.

To dissect these signals further, we generated chimeric proteins carrying progressive deletions of the third cytosolic loop (i3) or the C-terminal tail of OA1 and expressed them in HeLa cells. This strategy allowed isolation of the signal-containing regions to the 19 N-terminal amino acids of i3 (i3FY; Fig. 2A) and to the 25 N-terminal amino acids of CT (CT2; Fig. 2A), since the corresponding fusion proteins (LAMP/i3FY and LAMP/CT2) were promptly delivered to the lysosomes (Figs 3, 4). By contrast, the remaining part of i3 (18 C-terminal amino acids of i3) does not appear to contain any signal for intracellular localization, since the protein chimera carrying this portion, LAMP/i3ΔFY, displayed plasma membrane distribution and colocalization with farnesylated GFP (Fig. 3B). Similarly, the remaining part of CT (65 C-terminal amino acids of CT) was found to induce significant cell-surface accumulation of the corresponding chimera, LAMP/CTΔCT2, even though in this case a mixed vesicular and plasma membrane distribution pattern was observed, suggesting that this region of OA1 might contain further signals (Fig. 4B; see below for further description).

Next, we took advantage of the interspecies amino acid conservation of the i3FY and CT2 sequences to generate chimeric proteins carrying progressive deletion of conserved blocks (LAMP/i3FG and LAMP/i3FA; LAMP/CT2M5 and LAMP/CT2M11) (Fig. 3A, Fig. 4A). In addition, we mutated a double leucine (LL) to alanine (AA) in i3FY (LAMP/i3FY LL→AA; Fig. 3A), since dileucine-based motifs have been implicated in the sorting of lysosomal and melanosomal membrane proteins (Bonifacino and Traub, 2003); and we mutated a tryptophan-glutamic acid (WE) to alanine (AA) in CT2 (LAMP/CT2M5 WE→AA; Fig. 4A), since despite poor conservation of this region these two residues are preserved in all species tested. We expressed these chimeras in HeLa cells and observed that LAMP/i3FA distributes predominantly to the cell surface, whereas both LAMP/i3FY LL→AA and LAMP/i3FG display a mixed distribution, characterized by a vesicular pattern colocalizing with Lucifer Yellow, superimposed on a diffuse plasma membrane staining (Fig. 3B and not shown). As mutation of the dileucine does not result in a complete plasma membrane displacement, which is obtained instead with deletion of an additional eight contiguous amino acids in the LAMP/i3FA chimera, it is likely that the relevant signal-containing sequence in i3 is represented by SLLKGRQGIY (the `S-Y' sequence). Furthermore, we found that LAMP/CT2M5 shows a lysosomal pattern comparable with that of LAMP/CT2. By contrast, LAMP/CT2M5 WE→AA and LAMP/CT2M11 were mostly delivered to the plasma membrane (Fig. 4B and not shown), revealing a crucial role for the `WE' doublet.

Four different missense mutations were identified within i3 distally to the dileucine in patients with ocular albinism, including G229V, T232K, E235K and I244V (Fig. 2A) (Schnur et al., 1998; Bassi et al., 2001). The role of these mutations in the pathogenesis of the disease remains to be established, since previous studies have shown that, in the context of the native OA1, they do not determine ER retention of the protein nor other apparent sorting defects (d'Addio et al., 2000; Shen et al., 2001). Since the presence of additional targeting signals in OA1 might have masked the effect of these mutations, we introduced each of them within the LAMP/i3 chimera; moreover, we introduced the G229V mutation within LAMP/i3FY, and the T232K, E235K and I244V mutations within LAMP/i3ΔFY. Upon expression in HeLa cells, we found no difference between the distribution pattern of the original chimeras and that of the mutation-carrying chimeras; thus, either in the absence or in the presence of missense mutations, LAMP/i3 and LAMP/i3FY showed a lysosomal distribution, whereas LAMP/i3ΔFY was delivered to the plasma membrane (data not shown). These findings support the idea that, except for ER retention, missense mutations in ocular albinism do not cause other kinds of trafficking defects and that G229 within the S-Y sequence is not a crucial component of the sorting signal.

To determine whether the dileucine-like motif contained within the S-Y sequence and the WE signal are also relevant for lysosomal/melanosomal targeting in melanocytic cells, we transiently expressed each chimera in MNT1 cells. In this cell type, immunofluorescence patterns were more variable than in HeLa cells, both within the same and among different experiments; therefore, we performed a statistical analysis of multiple transfection experiments by evaluating the relative number of cells showing plasma membrane, vesicular or mixed distribution of the chimeric proteins (Fig. 5A and Fig. S2, supplementary material; see Materials and Methods for details). Similar to the results obtained in HeLa cells, this analysis revealed that most of MNT1 cells transfected with LAMP/i3ΔFY display cell-surface localization of the recombinant protein, whereas most of cells transfected with LAMP/i3FY, LAMP/CT2 and LAMP/CT2M5 show vesicular distribution (Fig. 5A and Fig. S2, supplementary material). LAMP/i3FA, LAMP/CT2M5 WE→AA and LAMP/CT2M11 were predominantly found on the plasma membrane with or without a superimposed vesicular staining. By contrast, LAMP/i3FY LL→AA and LAMP/i3FG were characterized by a high percentage of cells displaying mixed and vesicular labeling, respectively, with only about 10% of cases showing complete cell-surface localization (Fig. 5A).

To investigate the specificity of these signals for the OA1 compartment, we colocalized LAMP/OA1 chimeras with endogenous OA1 in transiently transfected MNT1 cells. The quantitation of colocalization was performed by assessing on confocal micrographs the percentage of rat LAMP1-positive vesicles that were also positive for endogenous OA1. Interestingly, as shown in Fig. 5B, over 80% of intracellular vesicles enriched in LAMP/CT2 also contain endogenous OA1, indicating that CT2 exhibits a significant ability to direct the C-terminus deleted LAMP1 toward the OA1 compartment. This capacity is maintained by LAMP/CT2M5, but is abolished by mutation or deletion of the WE doublet, since LAMP/CT2M5 WE→AA and LAMP/CT2M11 are mostly delivered to the cell surface and their residual intracellular vesicles co-segregate with endogenous OA1 by less than 30% (Fig. 5B and Fig. S2, supplementary material). Modest levels of colocalization with endogenous OA1 compared with LAMP/CT2 were also observed with LAMP/i3FY and its mutants (Fig. 5B and Fig. S2, supplementary material).

To establish whether the peptide motifs identified above are relevant for the sorting of the native OA1, we generated missense or deletion mutants of full-length mycHis-tagged OA1 and explored their localization in transiently transfected HeLa and MNT1 cells. In particular, we produced a mutant in which the entire S-Y sequence in i3 was deleted (OA1/ΔSY; Fig. 2A, round brackets in i3), a mutant carrying a LL→AA amino acid substitution in i3 (OA1/LL→AA; Fig. 2A, arrowheads in i3), and a mutant carrying a WE→AA amino acid substitution in CT (OA1/WE→AA; Fig. 2A, arrowheads in CT). In addition, we created two double mutants, carrying either the ΔSY deletion or the LL→AA substitution in combination with the WE→AA mutation (OA1/ΔSY/WE→AA and OA1/LL→AA/WE→AA, respectively). To obtain a quantitative comparison among the distributions of different mutants, transiently transfected HeLa and MNT1 cells were counted and classified into three categories, based on the vesicular, plasma membrane or mixed distribution of the recombinant proteins, as performed previously with LAMP/OA1 chimeras (see Materials and Methods for details). The results of this analysis are shown in Figs 6 and 7.

Both in transfected HeLa and MNT1 cells, we observed that neither the deletion of the S-Y sequence nor the mutation of the dileucine or of the WE motif alone result in significant missorting of the corresponding OA1 variants to the plasma membrane (Figs 6, 7). By contrast, the concomitant abrogation of the S-Y sequence and of the WE motif resulted in a clear plasma membrane displacement of the double mutant, which showed labeling of the cell surface in 76% and 45% of transfected HeLa and MNT1 cells, respectively (Figs 6, 7). Plasma membrane localization was confirmed by colocalization with farnesylated GFP or Na⁺/K⁺ ATPase and was usually accompanied by a persistent vesicular staining (Fig. 6A, Fig. 7A), so that most cells displayed a mixed distribution pattern. Similar, although less striking, results were obtained with the OA1/LL→AA/WE→AA double mutant, confirming that other amino acids of the S-Y sequence in addition to the dileucine are relevant for sorting.

The LAMP/CTΔCT2 chimera, carrying the last 65 amino acids of CT fused to LAMP1, displays a mixed vesicular/plasma membrane distribution when expressed in HeLa cells (Fig. 4B), suggesting that this region of OA1 might contain additional signals for intracellular localization. Thus, we tested its contribution to the sorting of the full-length protein by generating a deletion mutant of OA1 lacking the 65 C-terminal amino acids (OA1CT2-mycHis; Fig. 2A, bar in CT). When expressed in HeLa cells, OA1CT2-mycHis showed a subcellular distribution indistinguishable from the full-length OA1 and colocalized extensively with Lucifer Yellow at 1 hour chase (Fig. 8A). In transiently transfected MNT1 cells, OA1CT2 displayed a vesicular distribution and could be compared with the endogenous OA1 staining, since it is not recognized by the W7 anti-OA1 antibody (tested by western blotting and immunofluorescence analyses, not shown). OA1CT2 colocalizes by 76% with endogenous OA1 and to a significant extent with LAMP1, Pmel17 and Lucifer Yellow (Fig. 9A and Fig. S3, supplementary material), displaying a behavior very similar to that of the endogenous and exogenous full-length OA1 (Fig. 1B).

To test whether deletion of the distal portion of the CT would hamper the intracellular localization of OA1/ΔSY, OA1/LL→AA or OA1/WE→AA single mutants, we introduced within OA1CT2 the same mutations generated within the full-length OA1 and expressed the resulting mutants in HeLa and MNT1 cells. As previously observed with full-length OA1 mutants, despite the absence of the 65 C-terminal residues, only OA1CT2/ΔSY/WE→AA and to a minor extent OA1CT2/LL→AA/WE→AA double mutants displayed cell-surface localization, whereas none of the single mutants did (Fig. 8B, Fig. 9B). In addition, both in HeLa and MNT1 cells, full-length OA1 and OA1CT2 double mutants showed plasma membrane displacement in a comparable percentage of cells (compare Fig. 6B to Fig. 8B, and Fig. 7B to Fig. 9B).

Finally, to test whether abrogation of the dileucine-like or WE motifs might interfere with the specific subcellular localization of OA1 in melanocytic cells, we colocalized OA1CT2 mutants with endogenous OA1 in MNT1 cells. Similarly to OA1CT2 (Fig. 9A), OA1CT2/LL→AA and OA1CT2/ΔSY displayed a high degree of colocalization with endogenous OA1 (OA1CT2/LL→AA: 81%±0.1; OA1CT2/ΔSY 83%±0.1) (vesicles counted: 1100; cells counted: 40). By contrast, OA1CT2/WE→AA revealed a significantly lower ability to redistribute to the endogenous OA1 compartment, since its colocalization with OA1 dropped to 43%±0.2 (vesicles counted: 1000; cells counted: 40) (see Fig. S4, supplementary material), suggesting that the WE motif plays a crucial role in the specialized sorting of the protein.

In this study, we sought to identify the determinants responsible for the intracellular distribution of OA1 in melanocytic and non-melanocytic cells. To solve previous discrepancies on the localization of the endogenous protein in melanocytes, we performed a quantitative confocal immunofluorescence analysis in the human melanoma line MNT1. By using two distinct antibodies - anti-OA1 to label the endogenous protein and anti-His₆ to detect the exogenous protein - OA1 was found to colocalize extensively with markers of late endosomes and lysosomes (LAMP1, Lucifer Yellow at 1 hour chase) and melanosomes (Pmel17, TRP1). Thus, OA1 appears to be equally distributed along both the lysosomal and the melanosomal sorting pathways. However, the high level of co-segregation with LAMP1 and Pmel17 suggests that a major fraction of the protein localizes to an intermediate compartment, sharing features of late-endosomes and immature melanosomes. This subcellular distribution probably depends on direct sorting from the trans-Golgi network, since OA1 is only minimally found in EEA1-positive early endosomes (Fig. 1) and does not localize to transferrin-labeled recycling endosomes, nor is OA1 detected at the cell surface, even in MNT1 cells transfected with dynamin K44A (our unpublished results), supporting the notion that the protein does not normally transit through the plasma membrane.

To search for the sorting signals responsible for the lysosomal/melanosomal targeting of OA1, we utilized both LAMP1/OA1 chimeras and OA1 mutants and analyzed their steady-state distribution in melanocytic and non-melanocytic cells. By this strategy, we identified two independent signals that are necessary and sufficient for lysosomal/melanosomal localization in HeLa and MNT1 cells. The first signal is contained within the third cytosolic loop and is characterized by a dileucine, although additional adjacent residues included in the highly conserved sequence SLLKGRQGIY (the S-Y sequence) appear to be important (Fig. 2A). Even if dileucine-based motifs are well-known lysosomal/melanosomal sorting signals, the motif carried by OA1 appears unconventional, since it is located in a cytosolic loop instead of a C-terminal tail, and is not preceded by proline and acidic residues (Setaluri, 2000; Bonifacino and Traub, 2003). In addition, at variance with typical dileucine-based signals, mutation of a single leucine of the doublet within a LAMP/OA1 chimera (LAMP/i3FY LL→LA) does not interfere with its intracellular localization (our unpublished results). Finally, the dileucine itself seems to have only a partial role, since entire deletion of the S-Y sequence determines a more dramatic missorting effect both in the context of LAMP/OA1 chimeras and of the native OA1, either in HeLa or MNT1 cells.

The second signal identified is located within the CT and is characterized by a tryptophan-glutamic acid doublet (WE motif; Fig. 2A). This motif appears novel and differs from all previously reported lysosomal/melanosomal sorting signals. Given that adjacent residues are poorly conserved and their deletion in LAMP/OA1 chimeras does not result in evident sorting effects (LAMP/CT2M11 behaves as LAMP/CT2M5 WE→AA), it is likely that the WE dipeptide represents the relevant component of the motif. Interestingly, sequence analysis of known melanosomal membrane proteins revealed that tryptophan-glutamic acid doublets conserved in humans and mice are also present within the cytosolic domains of the pink-eyed dilution protein. Both dileucine-like and WE motifs do not seem to be dependent on membrane distance or topology, as previously reported for tyrosine-based or dileucine-based signals (Rohrer et al., 1996; Bonifacino and Traub, 2003). Indeed, the dileucine-like signal works similarly when inserted in a cytosolic tail in LAMP/OA1 chimeras or in a loop in the native OA1. Deletion of the RKR stop transfer signal in LAMP/OA1 chimeras did not change the ability of dileucine-like and WE motifs to determine lysosomal sorting in HeLa cells; similarly, addition or deletion of four consecutive residues at the N-terminus of the WE motif did not perturb its ability to determine intracellular localization of the LAMP/CT2 chimera in HeLa or MNT1 cells (our unpublished results).

Both motifs are sufficient and necessary on their own to mediate lysosomal sorting in HeLa cells, thus functioning in a similar way to lysosomal sorting signals in non-melanocytic cell types. However, the WE motif, but not the dileucine-like signal, showed a significant capacity to redirect the corresponding LAMP/OA1 chimera towards OA1-positive compartments in MNT1 cells. Moreover, abrogation of the WE motif in the OA1CT2 construct resulted in a significant reduction of its colocalization with endogenous OA1, whereas mutation of the dileucine or deletion of the entire S-Y sequence produced essentially no effect. Therefore, although we could not demonstrate by confocal immunofluorescence analysis a lysosomal versus melanosomal specificity of any of these signals, our results suggest that they might mediate the targeting of OA1 to different subcompartments in melanocytic cells and that the novel WE motif has a key role in the intracellular transport of OA1.

The dileucine-like and WE motifs must be mutated simultaneously to observe a substantial plasma membrane appearance of the native OA1, suggesting that they are able to facilitate the intracellular targeting of the protein independently and also compensate for each other. However, at variance with the findings obtained using LAMP/OA1 chimeras, a `pure' plasma membrane distribution was rarely observed with OA1 double mutants, which mostly show cell-surface localization associated with a persistent vesicular staining. Thus, although the two motifs identified are clearly implicated in its intracellular targeting, OA1 might contain additional sorting determinants that could justify the incomplete cell-surface misrouting of double mutants. Furthermore, the definitively minor plasma membrane displacement exhibited by double mutants in MNT1 compared with HeLa cells suggests that melanocytic cells possess specific sorting machineries, possibly based on specialized adaptor proteins.

Our screening was restricted to the cytosolic domains of OA1 because they would be accessible to adaptor binding during transit along the secretory pathway, as previously described for canonical tyrosine-based or dileucine-based motifs, typically located in the cytosolic domains of single-spanning lysosomal and melanosomal membrane proteins (Setaluri, 2000; Bonifacino and Traub, 2003). However, additional signals involved in the subcellular targeting of OA1 might be located in other regions, including transmembrane domains and lumenal loops. It is also possible that some other signals lie within the cytosolic loops and were not detected, because they are either spread into multiple loops or dependent on the precise topology of the protein, which we could not reproduce in LAMP/OA1 chimeras. Nevertheless, crucial signals responsible for the steady-state distribution of OA1 do not appear confined within the distal portion of the CT, since our results show that this region is not necessary for the efficient intracellular retention and specific subcellular targeting of the protein in HeLa or MNT1 cells. The C-terminus of OA1 is not highly conserved among species and is enriched in serine and threonine residues. Although its deletion in the OA1CT2 construct does not result in major sorting defects, the mixed vesicular/plasma membrane distribution displayed by the LAMP/CTΔCT2 chimera (Fig. 4B) and the slight reduction of Pmel17 colocalization showed by OA1CT2 (Fig. 9A) suggest that it might contain some supplementary trafficking information, possibly dependent on the activity of the receptor and regulated by phosphorylation or other post-translational modifications, as typically observed in canonical GPCRs (Marchese et al., 2003).

In summary, we show here that, similar to other recently characterized polytopic lysosomal proteins (Cherqui et al., 2001; Kyttala et al., 2004), and in contrast to single-spanning lysosomal/melanosomal proteins (which are typically targeted by unique motifs in their cytosolic tails), OA1 contains at least two sorting signals that are responsible for its efficient intracellular retention and delivery to lysosomes and melanosomes. These signals are unconventional, lie in topologically separate protein domains, and function independently as lysosomal sorting signals in non-melanocytic cells; however, they might have different and more-specialized roles in melanocytic cells. Our findings provide additional evidence that OA1 is a resident lysosomal/melanosomal GPCR, since it possesses specific signals for these subcellular compartments, supporting the idea that a putative ligand should come from the organelle lumen. The identification of such OA1 activator/s might hopefully be facilitated by the present study, which provides the means to deliver OA1 efficiently to the plasma membrane and therefore to take advantage of the widely used second-messenger activation assays requiring cell-surface localization of the receptors.

Full-length rat LAMP1 and C-terminus-deleted rat LAMP1 were obtained by reverse transcriptase (RT)-PCR, using primers LAMP/F2-LAMP/R2 and LAMP/F2-LAMP/R5, respectively, to amplify a rat insulinoma line (RIN5F) cDNA prepared with the Superscript Preamplification System kit (Gibco-BRL). The amplified LAMP1 cDNA was then digested with HindIII and BamHI, and cloned into the same sites of the mammalian expression vector pCR3 (Invitrogen). LAMP/OA1 chimeras were generated by digesting the C-terminus-deleted rat LAMP1 construct with BamHI and EcoRI, and by cloning into these sites the cytosolic regions of OA1, previously amplified using specific primers carrying BamHI and EcoRI flanking sites. For primers and PCR conditions used to generate LAMP1/OA1 chimeras see Fig. S5A, supplementary material. OA1-mycHis (full-length wild type) and OA1CT2-mycHis (with deletion of the C-terminal amino acids 340-404) constructs were generated by amplifying the human OA1 cDNA in the presence of 5% DMSO with primers OA1/FXba-OA1/RHind and OA1/FXba-OA1CT2/Hind, respectively. Intramolecular deletions or mutations of OA1 (OA1/ΔSY, with deletion of amino acids 222-231; OA1/LL→AA, mutated at amino acids 223-224; and OA1/WE→AA, mutated at amino acids 329-330) were generated using a two-step PCR method (Deminie and Emerman, 1993) using specific primers. Briefly, first-step amplifications were performed with two pairs of primers, forward and reverse primers A, and forward and reverse primers B, in two separate reactions. The products of first-step amplifications A and B were subsequently purified and utilized together as template in second-step amplifications, which were performed with primers OA1/FXba-OA1/RHind and OA1/FXba-OA1CT2/Hind, as described above. Wild-type and mutant OA1 amplification products were then digested with XbaI and HindIII, and cloned into the same sites of the mammalian expression vector pcDNA 3.1/myc-His(-) C (Invitrogen). For primers and PCR conditions used to generate OA1 mutants, see Fig. S5B, supplementary material. Proofreading Pfu DNA polymerase (Promega) was used to avoid synthesis errors in all amplifications. DNA sequences of all constructs were confirmed by sequencing (Bio Molecular Research, DNA sequencing service at the University of Padua, Italy; http://bmr.cribi.unipd.it/). Plasmid DNA was purified using the Qiagen Midi kit (Qiagen), and DNA concentration was double checked by ethidium bromide and spectro-photometric analyses.

HeLa cells (American Type Culture Collection) were maintained in Dulbecco's modified Eagle's medium (Gibco, Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Euroclone) and antibiotics (100 units/ml penicillin, 100 μg/ml streptomycin). MNT1 cells (a gift of P. Natali, Regina Elena Institute, Rome, Italy) were maintained in DMEM containing 10% AIM-V medium (Gibco, Invitrogen), 20% FBS, 0.1 mM non-essential amino acids (Gibco, Invitrogen), 1 mM sodium pyruvate (Gibco, Invitrogen) and antibiotics. HeLa cells were transiently transfected on glass coverslips placed in 24-well plates using FuGENE 6, according to the manufacturer's instructions (Roche). To avoid undesirable overexpression, each well was transfected with 60 ng of `expressor' plasmid DNA and the balance was made up with empty vector up to 600 ng total. This protocol was adapted from previous studies on human tyrosinase (Calvo et al., 1999) and allowed detectable expression of wild-type LAMP1 or OA1 without significant overexpression (revealed by appearance of the proteins at the cell surface). MNT1 cells were transiently transfected on glass coverslips placed in 24-well plates using LipoFectamine 2000 reagent, according to the manufacturer's instructions (Invitrogen). To avoid undesirable overexpression, each well was transfected with 270 ng of `expressor' plasmid DNA and the balance was made up with empty vector up to 800 ng total. For transfections with OA1CT2 constructs, DNA quantities were kept higher as a result of the lower expression level of the recombinant proteins (120/600 and 800/800 ng of `expressor' plasmid/empty vector in HeLa and MNT1 cells, respectively). Transfection efficiencies with the different constructs were comparable, as assessed in each experiment by the detection of similar numbers of transfected cells expressing the recombinant proteins.

In all experiments, at 24 hours following transfection, cells were incubated for 1 hour with cycloheximide (50 μg/ml as final concentration) and fixed with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS), pH 7.4. Fixed cells were blocked and permeabilized for 1 hour with 0.2% saponin (Sigma) and 10% porcine serum (Sigma) in PBS. Primary and secondary antibodies were incubated for 2 hours and 1 hour, respectively, with 0.1% saponin and 1% porcine serum in PBS. We used the following primary antibodies: W7 anti-OA1, rabbit anti-human OA1 polyclonal antiserum (Schiaffino et al., 1996); H4A3, mouse anti-human LAMP1 monoclonal antibody (mAb; developed by J. T. August and J. E. K. Hildreth, and obtained from the Developmental Studies Hybridoma Bank, University of Iowa, Dept of Biological Sciences, Iowa City, IA); LY1C6, mouse anti-rat LAMP1 mAb (Stressgen Biotechnologies); HMB45, mouse anti-Pmel17 mAb (DAKO); TA99, mouse anti-human TRP1 mAb (a gift of A. N. Houghton, Memorial Sloane-Kettering Cancer Center, New York, NY); anti-EEA1, mouse anti-EEA1 mAb (BD Biosciences); anti-EEA1, rabbit anti-EEA1 polyclonal antiserum (Affinity Bioreagents); and anti-His₆, mouse anti-His₆ mAb (Roche). For lysosomal labeling, cells were pulsed for 1 hour at 37°C with medium supplemented with 4 mg/ml of Lucifer Yellow CH (LY; Sigma) and, following extensive washing (at least five times) with warm medium, were chased for an additional 1 hour at 37°C with normal medium, to accumulate the tracer into the late-endocytic pathway (Miller et al., 1983). For plasma membrane labeling, cells were either immunodecorated prior to permeabilization with mouse anti-Na⁺/K⁺ ATPase mAb (clone C464.6; Upstate), or were transfected with a vector encoding farnesylated GFP (pEGFP-F, Clontech) that, thanks to the farnesylation signal from c-Ha-Ras, accumulates specifically at the inner face of the plasma membrane. Indeed, farnesylated GFP was found to colocalize extensively with endogenous Na⁺/K⁺ ATPase both in HeLa and MNT1 cells (data not shown). Secondary antibodies were immunoadsorbed Cy2-conjugated donkey anti-rabbit IgG and Cy3-conjugated donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories) or Alexa Fluor 488/594-conjugated donkey anti-rabbit IgG and Alexa Fluor 488/594-conjugated donkey anti-mouse IgG. Labeled coverslips were mounted using Mowiol 4-88 reagent (Calbiochem) and visualized using either an Axiophot microscope equipped with a charge-coupled device camera (Carl Zeiss) for subcellular distribution analyses or a TCS-SP2 AOBS confocal microscope (Leica) for colocalization studies.

Adobe Photoshop and Microsoft Excel softwares were used for image processing and statistical analyses, respectively. Quantitative immunofluorescence analysis of the subcellular distribution displayed by LAMP/OA1 chimeras or OA1 mutants was performed by counting in each transfection experiment the number of cells showing vesicular, plasma membrane or mixed staining patterns. Classification of cells into the three categories was according to the following criteria. Vesicular distribution: bright cytoplasmic vesicular staining more prominent in the perinuclear area on a black background, cell margins are not detectable. Plasma membrane distribution: diffuse and often finely granular staining throughout the cell, intense labeling at margins and membrane spikes, cell margins are clearly visible; occasionally few vesicles may be present, probably as a result of the normal protein turnover from the cell surface. Mixed vesicular-plasma membrane distribution: vesicular pattern comprising lots of bright vesicles, superimposed on a diffuse staining throughout the cell, with labeling of cell margins and membrane spikes. Cell-surface distribution was also confirmed by colocalization with plasma membrane markers (see above). A minor number of cells (fewer than 5%) that could not be classified into these categories (because of being out-of-focus, too confluent or superimposed, etc.) were omitted from the count. Quantitative immunofluorescence analysis of the colocalization of endogenous endogenous OA1 and various markers, recombinant OA1 and various markers, and LAMP/OA1 chimeras and various markers, was performed on digital confocal micrographs by marking within individual cells several (20-50) distinct single OA1- or LAMP1-positive vesicles and subsequently assessing their colocalization with the markers. Vesicles were chosen randomly throughout the cells, with the exception of areas overcrowded with organelles (e.g. perinuclear areas, tips of dendrites in melanocytes). The latter were avoided because of the risk of false positives despite the confocal analysis (see Figs S1 and S2, supplementary material). Quantitative immunofluorescence analyses were performed blind (without knowing the name of constructs analyzed) by the first author and counterchecked for verification by a second independent investigator.

We thank G. Schiavo for constant advice and support during the course of this work, and M. Alessio, G. Raposo, R. Sitia, C. Tacchetti and D. Zacchetti for helpful suggestions and critical reading of the manuscript. Part of this work was carried out in Alembic, an advanced microscopy laboratory established by the San Raffaele Institute and the Vita-Salute San Raffaele University, Milan, Italy. This work was supported in part by the generous donation of the Vision of Children Foundation-San Diego (to M.V.S.), the National Institutes of Health/National Eye Institute (grant N. 5R01EY014540 to M.V.S.), and Telethon-Italy (grant N. F3 to M.V.S.).