Sialylated Core 1 O-Glycans Influence the Sorting of Pmel17/gp100 and Determine Its Capacity to Form Fibrils*

Pmel17 is a melanocyte/melanoma-specific protein that is essential for the maturation of melanosomes to form mature, fibrillar, and pigmented organelles. Recently, we reported that the less glycosylated form of Pmel17 (termed iPmel17) is sorted via the plasma membrane in a manner distinct from mature Pmel17 (termed mPmel17), which is sorted directly to melanosomes. To clarify the mechanism(s) underlying the distinct processing and sorting of Pmel17, we generated a highly specific antibody (termed αPEP25h) against an epitope within the repeat domain of Pmel17 that is sensitive to changes in O-glycosylation. αPEP25h recognizes only iPmel17 and allows analysis of the processing and sorting of iPmel17 when compared with αPEP13h, an antibody that recognizes both iPmel17 and mPmel17. Our novel findings using αPEP25h demonstrate that iPmel17 differs from mPmel17 not only in its sensitivity to endoglycosidase H, but also in the content of core 1 O-glycans modified with sialic acid. This evidence reveals that iPmel17 is glycosylated differently in the Golgi and that it is sorted through the secretory pathway. Analysis of Pmel17 processing in glycosylation-deficient mutant cells reveals that Pmel17 lacking the correct addition of sialic acid and galactose loses the ability to form fibrils. Furthermore, we show that addition of sialic acid affects the stability and sorting of Pmel17 and reduces pigmentation. Alterations in sialyltransferase activity and substrates differ between normal and transformed melanocytes and may represent a critical change during malignant transformation.

ated with a highly metastatic phenotype (11)(12)(13). Despite the importance of sialic acid addition, no study has reported sialyltransferase activity or its role in the sorting of melanosomespecific proteins in various types of melanocytic cells.
N-Glycans on Pmel17 contains a mixture of complex, hybrid, and mannose-rich carbohydrate chains that distinguish its secreted and cytoplasmic forms (3). In contrast, no information for the O-glycan content of Pmel17 is available. Recently, we reported that a glycoform of Pmel17 (iPmel17) is sorted via the plasma membrane in a manner distinct from mPmel17, which is sorted directly to melanosomes (14). These facts suggest that Pmel17 can be glycosylated differently and that those forms may have distinct functions. The phenomenon that any given glycosylation site on a given protein synthesized by a particular cell type can have a range of variations in the precise glycan structure is known as microheterogeneity (15). To date, there have been no reports that address this issue regarding melanosome-specific proteins.
The putative function(s) of iPmel17, which is the major form of Pmel17 seen in pigmented cells, is a matter of conjecture and controversy at this point (16,17). The rapid processing of Pmel17 and the lack of in depth glycosylation studies have been impediments to elucidating mechanisms involved in the complex processing, trafficking, and functions of the various forms of Pmel17. To develop another tool to resolve some of these issues, we designed a highly specific antibody against a peptide sequence in the core region of Pmel17. That antibody, termed ␣PEP25h, specifically recognizes iPmel17 and has allowed us to characterize the different processing events that generate this important fragment. Reactivity with ␣PEP25h can be blocked by glycosylation at residues adjacent to its epitope, allowing us to study iPmel17 processing and sorting compared with mPmel17 that is recognized by ␣PEP13h. We demonstrate for the first time that iPmel17 differs from mPmel17 by having different sialylated core 1 O-glycans. Furthermore, we show that Pmel17 deficient in sialic acid and galactose loses the ability to form fibrils, and thus the addition of sialic acid at the ␣2,3-position is an important determinant of Pmel17 sorting through the secretory pathway.

Cell Cultures, Human Tissue Samples, and Skin Biopsies-
Pigmented (MNT-1) human malignant melanoma cells and HeLa cells were cultured as described previously (18,19,20). CHO cells as well as Lec2 and Lec8 mutant CHO cells were cultured as described previously (21). Cell extracts from neuroblastoma cell lines (SKNAS and SKNSH) were gifts from Dr. Maria Tsokos, NCI, National Institutes of Health. The human melanocyte cell line M253 was cultured in 154 Medium supplemented with growth factors as described by Cascade Biologics (Portland, OR). Paraffin-embedded samples from human kidney, uterus, and lung were obtained from the tissue bank in Philadelphia. Shave biopsies, 4 mm in diameter, were taken before and 1 day after a 1 minimum erythema dose of UV radiation, as described previously (22).
Antibodies, Peptide Synthesis, and Immunoaffinity Purification-␣PEP13h recognizes the C terminus of human Pmel17 (18). The ␣PEP25h polyclonal antibody was raised against a keyhole limpet hemocyanin-conjugated synthetic peptide (keyhole limpet hemocyanin-CTPEATGMTPAEV-SIVVLSGTT-CO 2 H) by immunization in rabbits as reported previously (23,24). Peptides used for affinity purification were synthesized by the solid phase method with 9-fluorenylmethoxy-carbonyl chemistry using a 431A peptide synthesizer (Applied Biosystems, Foster City, CA). Each peptide was purified by high pressure liquid chromatography on a Vydac C-4 column with 0.05% trifluoroacetic acid/water/acetonitrile. The mass of each peptide was confirmed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (Micromass, Beverly, MA). Sera from immunized rabbits were affinity-purified using each peptide coupled with SulfoLink (Pierce). The specificity of each antibody was confirmed by enzyme-linked immunosorbent assay using synthetic peptides.
Mass Spectrometry-Immunoprecipitated samples were separated on Tris-glycine gels and stained with the colloidal blue staining kit (Invitrogen) according to the manufacturer's instructions. Retrieved protein bands were reduced, carbamidomethylated, and in-gel digested with trypsin using the Montage In-Gel Digest ZP kit (Millipore, Billerica, MA) according to the manufacturer's directions. Peptide separation was performed at 300 nl/min and was coupled to on-line analysis by nanoflow liquid chromatography-electrospray ionization-tandem mass spectrometry on an LTQ ion trap mass spectrometer (ThermoElectron, San Jose, CA) equipped with a nanospray ion source. From the eluted peptides, 2 l were loaded onto a 0.075 ϫ 50 mm PicoFrit BioBasic C18 packed tip column (New Objective, Woburn, MA) using the Paradigm MS4 MDLC (Michrom Bioresources, Auburn, CA). Elution of peptides into the mass spectrometer was performed with a linear gradient from 95% mobile phase A (5% acetonitrile, 0.5% acetic acid, 94.95% water) to 65% mobile phase B (5% water, 0.5% acetic acid, 94.95% acetonitrile) in 45 min and then to 95% mobile phase B in 5 min. Tandem mass spectra were searched against the human NCBI data base using TurboSEQUEST in BioWorks version 3.2 (ThermoElectron, San Jose, CA).
Fluorescence and Electron Microscopy-For dual immunofluorescence staining, cells were grown in 2-well chamber slides (Nalgene, Naperville, IL) and fixed with 4% paraformaldehyde for 15 min at 4°C. Cells were then incubated with primary antibodies overnight at 4°C, followed by incubation with Texas Red anti-rabbit or fluorescein isothiocyanate streptavidin for polyclonal and monoclonal antibodies, respectively (25). Human skin specimens embedded in paraffin were processed as described previously (26). Fluorescence signals were classified according to whether they showed green, red, or yellow fluorescence, the latter being indicative of colocalization of the red and green fluorescence signals. Images were obtained using an LSM 510 confocal microscope (Zeiss, Jena, Germany). Analysis and quantification of the colocalization signal were evaluated under equal magnification, laser intensity, and saturation in the same preparations using Zeiss colocalization software. For each staining condition, 10 optical fields were observed, and representative images from two separate experiments are shown in the figures. For electron microscopy and immunoelectron microscopy, we used a previously published method without modifications (20).
Expression Vector Cloning and Transfection of the MART1 and Pmel17 Genes-The MART1 vector was a gift from Dr. Toshihiko Hoashi (Tokyo, Japan). DNA oligonucleotides for gp100 PCR were synthesized and purified by Operon. The forward primer is 5Ј-GGG ATG GAT CTG GTG CTA AAA AGA TGC CTT CTT C-3Ј, and the reverse primers are 5Ј-CAC CAG CCT TAA GGT GGC TGT ACC ATC CAG-3Ј and 5Ј-GAC CTG CTG CCC ACT GAG GAG GGG GCT ATT-3Ј. These primers were used to construct expression vectors for the full-length gp100 cDNA (pcDNA5/Pmel17fl-V5) and its truncated form (pcDNA5/Pmel17⌬-V5) that contains the N-terminal 468 amino acids. The DNA fragments were amplified from a gp100-containing plasmid (clone IOH4070 open reading frame; Invitrogen) by 35 cycles of PCR as described previously (27). The PCR fragments were subcloned into the pcDNA5/FRT-V5-His6 vector (Invitrogen) confirmed by NheI and BamHI digestion and by DNA sequencing (the DNA Sequencing Facility, NCI, Bethesda). The plasmids were prepared using a Qiagen plasmid extraction kit. Transient transfection of full-length Pmel17 (FL-Pmel17), cleaved N-terminal fragment (cNTF-Pmel17), and a mock FRT control vector all dual tagged with V5 and His into MNT-1 cells, CHO, Lec2, and Lec8 cells was achieved by employing Lipofectamine 2000 (Invitrogen) or by electroporation using the Nucleofector kits and equipment (AMAXA, Gaithersburg, MD) according to the manufacturer's instructions.
Immunoblotting and Sample Preparation-For immunoblotting, cell extracts or isolated fractions were prepared as published previously (20). Briefly, cell extracts were mixed with sample buffer (2ϫ) (Invitrogen) supplemented with 2-mercaptoethanol and heated for 5 min at 100°C. Samples were then separated on 10, 8 -16,% or 4 -20% Tris-glycine gels (Invitrogen) and transferred to polyvinylidene difluoride membranes (Invitrogen). Membranes were blocked with casein for 1 h at 23°C and were then incubated with primary antibodies overnight at 4°C. After washing with T-PBS (1ϫ PBS, 1% Tween 20, pH 7.2), blots were incubated in horseradish peroxidase-linked anti-rabbit or anti-mouse secondary antibodies (Amersham Biosciences) for 1 h at 23°C. Primary antibodies were detected using the ECL plus system (Amersham Biosciences), according to the supplier's instructions.
Sialyltransferase Assays-Assays were performed as described previously (28). Briefly, enzyme activity was measured at 37°C for 4 h in 50 l of reaction mixture comprising 0.1 M cacodylate buffer, pH 6.2, 10 mM MnCl 2 , 0.2% Triton CF-54, 50 M CMP-[ 14 C]Neu5Ac (1.85 kBq) (Amersham Biosciences), with 23 l of the enzyme source and 1 mg⅐ml Ϫ1 of fetuin, asialofetuin, orosomucoid, asialo-orosomucoid, bovine submaxillary mucin (BSM), or asialo-BSM (Sigma) as acceptor substrates. The reactions were stopped by addition of 1 ml of H 2 O. Proteins were precipitated in 1 ml of 5% phosphotungstic acid in 2 N HCl and filtered on GF/A glass microfiber filters (Whatman). The radio-active material present on each filter was counted by liquid scintillation. The rate of this reaction was linear with time at least for 8 h. All reactions were performed in triplicate.
Quantitative Real Time PCRs-Total cytoplasmic RNA was isolated from cells using the RNeasy mini kit (Qiagen, Valencia, CA) according to the supplier's instructions. Reverse transcription was performed using 200 ng of cytoplasmic RNA as reported previously (14). All amplified products were sequence-verified. Quantitative real time PCRs were performed using the Opticon analysis system (MJ Research Inc., Waltham, MA) and a hot-start PCR that contained the double-stranded specific DNA-binding dye SYBR Green I (Sigma). The following primers and their product sizes were used: ␤-actin, 5Ј-CCCTCCATCGTCCACCGCAA-ATGCTTC and 3Ј-GACTGCTGTCACCTTCACCGTTC-CAG, 204 bp; ST3Gal I, 5Ј-TTCCTCACCTCCTTCTTCCTG-AACTAC and 3Ј-TCTTCTCCAGCATAGGGTCCACAT-TCC, 346 bp; and ST6GalNAc II, 5Ј-GGAAATGTCGGTGG-AGTGTTCAGCAAG and 3Ј-AAGCAACTAACCCCCATC-AAGTGCCAGACCCTC, 356 bp. After 5 min at 95°C, 40 cycles were performed as follows: 15 s denaturation at 94°C, 30 s annealing at 60°C, 30 s extension at 72°C, and fluorescence detection at 78°C. A melting curve fluorescence analysis was performed on each sample once the amplification cycles were completed to verify that a single product had been amplified. The Ct is defined as the point when the amplification starts the exponential phase (29). The fold difference was calculated by subtracting the Ct of the test sample from the Ct of actin to give ⌬Ct, and then fold difference ϭ 2 Ϫ⌬Ct .
Metabolic Labeling-Metabolic labeling and immunoprecipitation were performed as described previously (20). Briefly, cells were cultured in 10-cm dishes until 80% confluent. Cells were preincubated in methionine-free medium for 30 min at 37°C and were then labeled for 30 min with 0.5 mCi of [ 35 S]Met/Cys (Redivue Pro Mix; Amersham Biosciences) and chased in methionine-plus medium for the times indicated in the text. After harvesting, cells were washed and incubated in lysis buffer overnight (33). Samples were pre-cleared with normal rabbit serum and protein G-Sepharose 4 fast flow beads (Amersham Biosciences) for 2 h at 4°C. The supernatants were collected and immunoprecipitated with either ␣PEP13h or ␣PEP25h for 2 h at 4°C, and immunocomplexes were then separated with protein G-Sepharose beads (Amersham Biosciences) for 2 h at 4°C. For double immunoprecipitation, we recovered these supernatants and immunoprecipitated them again with either ␣PEP13h or ␣PEP25h for 2 h at 4°C and then recovered the immunocomplexes again with protein-G beads and continued with the protocol. After washing, the final pellets were suspended in SDS sample buffer (Invitrogen) with 2-mercaptoethanol (Sigma), heated for 5 min at 100°C, and separated on 10 or 8 -16% Tris-glycine gels (Invitrogen). Control samples were immunoprecipitated with normal rabbit serum and were processed in parallel.
Lectin Staining-Lectin blot analysis using digoxigenin-labeled lectins (Roche Applied Science) was performed according to the manufacturer's instructions.
Melanin Content Assay-Melanin content was determined as described previously (34). In brief, cell pellets were dissolved in 100 l of 1 N NaOH, and melanin concentrations were quantitated by absorbance at 405 nm using a standard curve generated from synthetic melanin (Sigma). For each condition, the melanin content is expressed as nanograms of melanin divided by total protein concentration in micrograms. Values are then reported as a percentage of those obtained in controls. Each experiment was repeated at least two times.

RESULTS
␣PEP25h Specifically Detects iPmel17-The antibody ␣PEP25h was designed against residues 393-414 ( Fig. 1A) in the core region of Pmel17 (35,36). ␣PEP25h would be expected to recognize only the two major splice forms of Pmel17 and not a recently reported minor third splice variant (37). To assess the sensitivity and specificity of ␣PEP25h, we affinity-purified it from rabbit antisera and compared its reactivity patterns against the well characterized ␣PEP13h (which recognizes the C terminus of Pmel17). An enzyme-linked immunosorbent assay showed that both purified antibodies, ␣PEP25h and ␣PEP13h, have similar binding curves and high specificities for their immunizing peptides (supplemental Fig. 1).
To analyze the specificity of ␣PEP25h for Pmel17, we used MNT-1 melanoma cells, which remain quite differentiated and highly pigmented (18,38), and we compared its immunoreactivity patterns with those of ␣PEP13h. The processing of Pmel17 was assessed by [ 35 S]Met/Cys metabolic labeling of MNT-1 cells for 30 min followed by chases of various times and then immunoprecipitation with ␣PEP13h or ␣PEP25h (Fig.  1B). ␣PEP13h recognized iPmel17 as an ϳ95-kDa band (Fig.  1B, black arrowhead) at the 0-min chase time. After 15 min of chase, minor bands of mPmel17 ( Fig. 1B white arrowhead) and cPmel17 (gray arrowhead) began to appear as ϳ115and ϳ26-kDa bands, respectively, as reported previously (4). Interestingly, iPmel17 was reduced only by 30% after 1.5 h of chase, whereas mPmel17 and cPmel17 decreased by ϳ70% (measured using ScionImage software). In contrast, immunoprecipitation of those same labeled samples with ␣PEP25h detected a major band at ϳ95 kDa that was not further modified, and it was highly stable over time, being reduced by only 45% after 1.5 h of chase. These results indicate that ␣PEP25h identifies only one specific form of Pmel17 (iPmel17) and has a specificity for Pmel17, which is distinct from ␣PEP13h (which recognizes the mPmel17 and cPmel17 forms as well as iPmel17).
To further characterize the specificity of ␣PEP25h, we used immunoblotting with various human melanocytic cell lines (M253 normal melanocytes and MNT-1 melanoma cells) and nonmelanocytic cell lines (SKNAS and SKNSH neuroblastoma cells, HeLa cells, and human primary fibroblasts). ␣PEP13h identified Pmel17 only in melanocytic cells as expected and also detected minor bands of mPmel17 and cPmel17, along with an even smaller fragment (ϳ10 kDa) in MNT-1 cells (Fig. 1C, arrow). In contrast, ␣PEP25h specifically recognized Pmel17 only as a single major band (ϳ95 kDa) and only in melanocytic cells. As detected by ␣PEP25h and by ␣PEP13h, the iPmel17 band was sensitive to digestion with EndoH (which removes high mannose/hybrid N-glycans) and with PNGaseF (which removes all N-glycans) (Fig. 1D).
The band recognized by ␣PEP25h is identical in size to iPmel17 as detected by ␣PEP13h. To determine whether they are in fact the same, we performed sequential immunoprecipitation using the 35 S-labeled samples after a 45-min chase, a time when all Pmel17 bands were observed (Fig. 1E). Fig. 1E, lane 1, shows bands immunoprecipitated with ␣PEP13h alone, but if the sample had been immunodepleted by ␣PEP25h, little of the ϳ95-kDa band remained and could be immunoprecipitated by ␣PEP13h (lane 2). Conversely, immunodepletion with ␣PEP13h first followed by immunoprecipitation with ␣PEP25h resulted in an 85% depletion of the band recognized by ␣PEP25h (Fig. 1E, lane 4). Bands immunoprecipitated by preimmune sera are shown in Fig. 1E, lanes 5 and 6, as controls for nonspecific binding. We conclude that the ϳ95-kDa bands recognized by ␣PEP13h and by ␣PEP25h are in fact the same protein, iPmel17. To further confirm this, we immunoprecipitated Pmel17 with ␣PEP25h and with ␣PEP13h from MNT-1 cell lysates and analyzed the major band detected by each antibody using mass spectrometry. Several Pmel17 tryptic peptides (Table 1), each showing unique tandem mass spectra (supplemental Fig. 2), were identified in those bands with Ͼ95% confidence. The sum of these results confirms that ␣PEP25h specifically recognizes iPmel17.
Because Pmel17 is cleaved into two fragments by a proprotein convertase (5), we examined whether ␣PEP25h detected only iPmel17 and/or the cleaved N-terminal fragment (cNTF) of Pmel17. Thus, we treated MNT-1 cells with 50 M Dec-RVKR-CMK (CMK), a known inhibitor of proprotein convertases, including furin (39), for 24 h and then metabolically pulse-chase-labeled them with [ 35 S]Met/Cys. Pretreatment with CMK eliminated the cPmel17 form detected by ␣PEP13h (Fig. 1F) and actually stabilized mPmel17, as expected. In contrast, ␣PEP25h continued to detect only iPmel17. Digestion with EndoH shifted the mobility of iPmel17 as detected by ␣PEP13h and by ␣PEP25h.
Taken together, ␣PEP25h specifically recognizes an epitope of iPmel17 that is masked in mPmel17, which we hypothesize to result from steric hindrance of the epitope by glycosylation, as discussed below. ␣PEP25h works well in all immunological methods tested, an advantage over limitations in these proce- , after which samples were immunoprecipitated with ␣PEP13h or ␣PEP25h as noted. Immunoreactive bands were analyzed by SDS-PAGE and visualized by autoradiography as detailed under "Materials and Methods." ␣PEP13h precipitated three bands: mPmel17 (white arrowhead), iPmel17 (black arrowhead), and cPmel17 (gray arrowhead), whereas ␣PEP25h identified only iPmel17 (black arrowhead). C, lysates were obtained from melanoma and from non-melanoma cells as noted and were immunoblotted with ␣PEP13h or ␣PEP25h as noted; ␤-actin was used as a loading control. D, lysates of MNT-1 cells were digested with EndoH or PNGaseF for 3 h at 37°C and were immunoblotted with ␣PEP25h or ␣PEP13h as indicated. E, cells metabolically labeled with [ 35 S]Met/Cys were chased for 45 min and were sequentially immunoprecipitated with ␣PEP13h or ␣PEP25h, or with normal preimmune serum, as indicated. F, cells were treated with 50 M CMK for 24 h and were then metabolically labeled with [ 35 S]Met/Cys and chased for 1.5 h. Extracts were digested with EndoH where noted for 3 h at 37°C, after which samples were immunoprecipitated with ␣PEP13h or ␣PEP25h as noted. Immunoprecipitated bands were visualized by autoradiography as above.
␣PEP25h Detects iPmel17 in Early Melanosomes-To evaluate the usefulness of ␣PEP25h in characterizing the intracellular localization and specificity of iPmel17, we performed immunofluorescence and immunoelectron microscopic analysis. Confocal microscopy ( Fig. 2A) revealed that in MNT-1 cells, ␣PEP25h (top) had a granular cytoplasmic staining pattern and a linear distribution near the plasma membrane in 100 and 90% of cells observed, respectively. There was only a minor amount of colocalization of ␣PEP25h with HMB-45 ( Fig. 2A, top left), a stage II melanosome marker, or with clathrin (top right), which was usually near the plasma membrane or in the perinuclear area but not in dendrites. These results indicate that ␣PEP25h detects iPmel17 after it leaves the ER compartment. Immunofluorescence of paraffin-embedded skin specimens unexposed to light ( Fig. 2A, middle) or exposed to UV radiation (bottom) revealed that ␣PEP25h staining was restricted to melanocytes in the basal layer of the epidermis, where it showed a 97% colocalization with MART1, a melanocyte-specific marker, but not with keratin 5, a keratinocyte-specific marker. Interestingly, ␣PEP25h also revealed a fine extracellular granular pattern that was observed in or near suprabasal keratinocytes ( Fig. 2A, insets, arrows), which may represent secreted vesicles (exosomes) or melanosomes containing iPmel17. Indeed, Pmel17 has been identified previously as a component of exosomes released by melanoma cells (40). Tissue sections from kidney, breast, lung, and uterus were processed in parallel with skin sections but showed no staining with ␣PEP25h (data not shown).
To further confirm the intracellular location of ␣PEP25h, we performed dual immunoelectron microscopy (Fig. 2B). ␣PEP25h (20 nm gold) was localized lining the internal matrix of stage I melanosome membranes and occasionally colocalized FIGURE 2. ␣PEP25h detects Pmel17 in stage I melanosomes. A, MNT-1 cells were fixed and dual stained with ␣PEP25h (red) and with HMB-45 or clathrin (green). Note the granular distribution pattern of ␣PEP25h and the lack of colocalization (yellow) with HMB-45 (top row). Skin sections unexposed or exposed to UV light (middle and bottom, respectively) were deparaffinized and stained with ␣PEP25h (red) and for MART1 or keratin 5 (green). Note the colocalization of ␣PEP25h and MART1 (inset, left) and the presence of granular staining outside melanocytes (insets, arrows). B, immunoelectron microscopy of MNT-1 cells double-stained with ␣PEP25h (25 nm gold) and for AP2, HMB-50, or MART1 (10 nm gold). Colocalization of ␣PEP25h with AP2 and localization at the plasma membrane or exocytic vesicles are indicated with arrows. A vesicular exosome containing ␣PEP25h is shown in the inset. Roman numerals refer to melanosome stages. with AP2-containing vesicles (10 nm gold). ␣PEP25h was also identified at the plasma membrane and within small vesicles (ϳ80 nm) near the plasma membrane that resemble exosomes (Fig. 2B, arrow) (16). In addition, ␣PEP25h colocalized with HMB-50 (10 nm gold) in stage I melanosomes and with MART1 (10 nM gold) in stage II melanosomes. The specific localization of ␣PEP25h reactivity in those organelles can be seen in several additional images (supplemental Fig. 3). Thus, we conclude that ␣PEP25h specifically recognizes and identifies iPmel17 localized on the internal side of stage I and II melanosomal membranes and in the plasma membrane, which confirms the sorting of this protein through the secretory pathway to melanosomes (14). Reactivity of ␣PEP25h with intramelanosomal fibrils may reflect the loss of the epitope by cleavage (19) or because of melanin deposition, or it may indicate that iPmel17 is not incorporated into those fibrils.
The iPmel17 Form Is Not Retained in the ER but Is Processed Differently in the Golgi-iPmel17 has been considered either as an ER-retained form, because of its sensitivity to EndoH digestion (5,38), or as the cNTF-Pmel17 form, because of the predicted size of that fragment after cleavage (38). To resolve if iPmel17 represents either of those forms, we examined whether high mannose residues, an indication of ER location, were present on Pmel17 immunopurified using ␣PEP13h from MNT-1 cells. This purified Pmel17 sample was treated with or without a combination of three enzymes that remove hybrid and complex glycans as follows: neuraminidase (which removes sialic acid); ␤-1,4galactosidase, which digests unmodified ␤ (1-4)-linked Gal; and ␤-N-acetylglucosaminidase, which removes ␤-linked GlcNAc (supplemental Fig. 4). Those samples were stained with the lectin Galanthus nivalis agglutinin, which recognizes high mannose type glycans. Carboxypeptidase Y, which contains only high mannose glycans, was used as a positive control. G. nivalis agglutinin reacted only with enzyme-digested Pmel17 (ϳ80 kDa). These results confirm that neither iPmel17 nor mPmel17 contains high mannose residues, showing that iPmel17 is not an ER-retained form and that it contains N-glycan structures normally modified in the Golgi. Furthermore, the difference in size between the band detected (ϳ80 kDa) and its predicted size (Ͻ70 kDa) may be due to the presence of O-glycans added in the Golgi as discussed below.
To further assess those structures, we used enzymes that remove complex and hybrid structures, with or without neuraminidase treatment (Fig. 3A). As detected by ␣PEP25h, treatment with neuraminidase plus ␤1,4-galactosidase (Fig. 3A, lane  4) and those two enzymes plus ␤-N-acetylglucosaminidase (lane 5) generated novel bands (asterisks). In contrast, detection with ␣PEP13h revealed that treatment with ␤-N-acetylglucosaminidase alone generated a strong band below iPmel17 at ϳ75 kDa (Fig. 3A, lane 3), which represents a net loss of ϳ25 kDa. These results suggest that ␣PEP25h recognizes a form of Pmel17 that is terminated with at least galactose in a Gal-Glc-NAc structure, whereas ␣PEP13h detected N-glycans terminated with GlcNAc (not extended with Gal or sialic acid). These data support the existence of similarly sized iPmel17 forms, not seen in mPmel17, which have different glycosylation structures, as predicted earlier. In contrast, cPmel17 (Fig. 3A, bottom panel) was reduced in size by digestion either with neuraminidase plus ␤1,4-galactosidase (lane 4) or by those enzymes in combination with ␤-N-acetylglucosaminidase (lane 5), which confirmed the presence of highly sialylated complex type N-glycans.
To evaluate whether iPmel17 corresponds to the cNTF form of Pmel17, we made two cDNA plasmids containing the fulllength (FL) or the cNTF sequences of Pmel17, both being modified with a V5-His tag at their C-terminal domain (Fig. 3B). Those constructs were transiently transfected into MNT-1 cells. Immunoblotting analysis (Fig. 3C) revealed that cNTF-Pmel17 appeared as an ϳ75-kDa band (lanes 1 and 2), which was smaller than expected for the fully processed form (ϳ85 kDa based on size of the predicted protein and its possible posttranslational modifications). In contrast, FL-Pmel17 appeared as strong ϳ95-kDa bands and weak ϳ110-kDa bands (Fig. 3C,  lane 4). The ϳ28-kDa band of cPmel17 was only observed after prolonged exposure (Fig. 3C, bottom, gray arrowhead, lanes 3  and 4). ␤-Actin was used as a loading control. Next, we deter-mined the intracellular localization of cNTF-Pmel17, as detected with a V5 antibody, in transiently transfected MNT-1 cells using immunofluorescence (Fig. 3D). The transfected cNTF-Pmel17 (Fig. 3D, green) showed a granular distribution that was distinct from that of ␣PEP25h (red). In fact, the staining of cNTF-Pmel17 was observed in larger structures than those positive for ␣PEP25h or in granules surrounded by ␣PEP25h (Fig. 3D,  inset). The transfection efficiency was ϳ24%, and 92% of the transfected cells showed this pattern. Representative images from two separate experiments are shown. Taken together, these results indicate that iPmel17 is not retained in the ER, but sorts to intracellular organelles in the cytoplasm, and does not correspond to cNTF Pmel17.

The Extent of Core 1 O-Glycans Distinguishes Pmel17 Forms and
Influences the Reactivity of ␣PEP25h-The results above suggest that iPmel17 is O-glycosylated, which is another type of post-translational modification that may alter the properties and functions of glycoproteins. It has been suggested recently that O-glycosylation occurs at the repeat domain (RPT) of Pmel17 (19). This creates glycoepitopes detectable by antibodies that have a double specificity toward both the peptide backbone and the carbohydrate structure. Therefore, we investigated the presence of O-glycans on iPmel17, as detected by ␣PEP25h. Samples were digested with neuraminidase or with endo-␣-N-acetylgalactosaminidase (O-glycanase), which removes unmodified core 1 type O-glycan structures (Fig. 4A). ␣PEP25h detected only the single iPmel17 band after digestion with neuraminidase or O-glycanase, but multiple bands were detected after digestion with both enzymes (Fig. 4A, lane 4). Those multiple bands resembled that pattern detected by HMB-45 in undigested extracts (Fig. 4A, lane 1), although reactivity with HMB-45 was abrogated by neuraminidase digestion (lanes 2 and 4). This result indicates the presence of sialylated core 1 O-glycans on iPmel17. Interestingly, ␣PEP13h detected a band shift of mPmel17 after digestion with both enzymes (Fig.  4A, lane 4). To confirm that, Pmel17 immunopurified using ␣PEP13h was stained with peanut agglutinin that recognizes unmodified core 1 (Gal␤1-3GalNAc-R) in O-glycans. Peanut agglutinin detected bands in samples digested with neuraminidase with or without O-glycanase (Fig. 4A, lanes 2 and 4). Note the disappearance of mPmel17 in Fig. 4A, lane 4. These results indicate that both mPmel17 and iPmel17 contain sialylated core 1 O-glycans and constitute further evidence that iPmel17 is processed through the Golgi but in a fashion distinct from mPmel17. Because ␣PEP25h detects only iPmel17 and not mPmel17, we hypothesized that is because of O-glycan modification. To confirm that O-glycosylation, but not N-glycosylation, masks the ␣PEP25h epitope on Pmel17, we removed all N-glycan structures with PNGaseF followed by neuraminidase and O-glycanase digestion (Fig. 4B). Immunoblotting analysis revealed that iPmel17, as detected by ␣PEP25h, was further reduced to the size of the predicted peptide (ϳ75 kDa) after digestion with all three enzymes (Fig. 4B, arrow, lane 4). In contrast, staining with HMB-45 revealed a similar pattern for iPmel17 (Fig. 4B, upper band), indicating that this particular form still contains some sialylated O-glycans, although those bands were far less intense than those seen at the bottom of the gel. In contrast, all bands below iPmel17 were PNGaseF-resistant, showing their lack of N-glycan structures. Thus, we hypothesize that N-glycosylation in the ER affects only a small proportion of the Pmel17 and partially impairs its subsequent O-glycosylation in the Golgi, hence making the epitope available for detection by ␣PEP25h. To examine the presence of complex N-glycans, Pmel17 immunopurified with ␣PEP13h or with ␣PEP25h was digested with neuraminidase and/or PNGaseF and then was stained with the lectin Dathura stramonium agglutinin, which binds galactose in Gal␤1-4GlcNAc-R structures in unsialylated complex type N-glycans and partially extended core 2 O-glycans (Fig. 4C). D. stramonium agglutinin detected bands in undigested Pmel17 immunopurified with ␣PEP13h or ␣PEP25h (Fig. 4C, lanes 1 and 5) and in neuraminidase-digested samples (lanes 2 and 6), but not in PNGaseF-treated samples. This indicates that iPmel17, and to a lesser extent mPmel17, contains some hybrid or complex type N-glycans that are not modified with sialic acid in MNT-1 cells. Asialofetuin was used as a positive control. These findings confirm not only the presence of sialylated core 1 O-glycans in iPmel17 but demonstrate that those structures influence the reactivity of ␣PEP25h, presumably by steric hindrance of the epitope.
Sialylation Influences the Stability and Sorting of Pmel17-Our data indicate that sialic acid is added mostly to core 1 O-glycans of Pmel17 in MNT-1 cells. Therefore, we further investigated the role of sialic acid in the processing of Pmel17. First, we inhibited the addition of sialic acid by blocking the enzyme activity of ST3Gal I, the most common sialyltransferase active on core 1 O-glycans, with the specific inhibitor 2,3-dehydro-2-deoxy-N-acetylneuraminic acid (DANA) overnight and then metabolically labeled MNT-1 cells with [ 35 S]Met/Cys for 30 min (Fig. 5A). As detected by ␣PEP25h and by ␣PEP13h, treatment with DANA reduced the band intensity and stability of all Pmel7 bands after a 90-min chase compared with the control (Fig. 5A, lane C). To determine whether Pmel17 degradation occurs in a pre-or post-Golgi compartment, we treated cells overnight with DANA and then metabolically labeled them with [ 35 S]Met/Cys for 30 min. Following that, we added brefeldin A (BFA, an agent that blocks anterograde traffic from the Golgi) or MG132 (a proteasome inhibitor) (Fig. 5B). The addition of either of those compounds increased Pmel17 stability after a 90-min chase, as detected by ␣PEP25h and by FIGURE 5. Sialylation influences the stability and sorting of Pmel17. A and B, MNT-1 cells were treated overnight with DANA and were 35 S-labeled for 30 min only (A) or were also treated with 10 g/ml BFA or with 50 mM MG132 (B); labeled extracts were then immunoprecipitated with ␣PEP25h or ␣PEP13h, and visualized after electrophoresis on 8 -16% Tris-glycine gels. C and D, cells were treated with DANA (C) or DANA plus BFA or MG132 (D) and were then fixed with 4% paraformaldehyde and dual stained with ␣PEP25h or ␣PEP13h (red) and organelle markers (vti1b, BIP, or clathrin) as indicated (green).
␣PEP13h, which suggests that degradation of Pmel17 occurs in a post-Golgi compartment.
Because there is evidence that sialylated O-glycans can carry plasma membrane apical sorting information (41,42), we analyzed the intracellular distribution of Pmel17 in melanoma cells treated with DANA using immunofluorescence (Fig. 5C). Surprisingly, ␣PEP25h and ␣PEP13h (Fig. 5C, red) showed reduced signals and distributions restricted to the perinuclear area where it colocalized moderately with the Golgi marker Vti1b (green). In contrast, cells treated with DANA plus BFA or MG132 (Fig. 5D) showed a recovery of the iPmel17 signal detected by ␣PEP25h (red). Note the colocalization of Pmel17 with BIP (an ER marker, Fig. 5D, green) after treatment with BFA and its localization in dendrites after treatment with MG132. Taken together, these results indicate that the addition of sialic acid on core 1 type O-glycans is important for the stability of Pmel17 and for its correct sorting through the secretory pathway.
Pmel17 Glycoforms Deficient in Sialic Acid and Galactose Lose the Ability to Form Fibrils-Ever since it was first demonstrated that fibrils within stage II melanosomes react positively with HMB-45, which recognizes an epitope modified with sialic acid (4,43), it has been assumed that sialic acid plays a role in fibril formation. However, ␣PEP25h detects band patterns similar to HMB-45 after treatment with neuraminidase, which suggests that sialic acid alone may not be critical for fibril formation. To address this, we transfected Pmel17 into CHO cells and into Lec2 and Lec8 mutants of CHO cells with disrupted glycan functions, and we evaluated the processing and fibril formation capability of Pmel17. Lec2 mutant CHO cells have a deletion mutation in the CMP-sialic acid transporter which results in N-and O-glycans with a Ͼ90% decrease in sialic acid content (44). Lec8 mutant CHO cells have a deletion mutation in the UDP-galactose transporter which results in a truncated protein with a greatly reduced ability to translocate UDP-Gal inside the Golgi (45). Thus, Lec8 cells generate nongalactosylated and nonsialylated N-and O-glycans (46). Immunoblotting analysis of those transfected cells revealed differences in the sizes of iPmel7 produced in CHO, Lec2, and Lec8 cells (ϳ3-4 kDa each) as detected by ␣PEP25h and by ␣PEP13h (Fig. 6A). This result further confirms that iPmel17 is modified with sialic acid. Analysis of the glycan structures of Pmel17 in these cells was assessed using digestion with EndoH or PNGaseF (supplemental Fig. 5). mPmel17 and iPmel17 were insensitive or sen- sitive to EndoH in all three cell lines, respectively, as expected. In contrast, all Pmel17 bands were sensitive to PNGaseF, also as expected. These results indicate that the mPmel17 and iPmel17 glycoforms expressed in CHO cells are processed similarly as in human melanocytic cells.
To analyze modifications to N-and O-glycan chains, we took advantage of the fact that Lec2 cells are unable to add GalNAc to complex-type N-glycans (47) and that Lec8 cells only form O-GalNAc (48) (Fig. 6B). As detected by ␣PEP25h, digestion with neuraminidase and O-glycanase produced an ϳ80-kDa band (Fig. 6B, asterisk) in CHO cells (lane 3). A similar band was observed in Lec2 cells only after digestion with O-glycanase alone (Fig. 6B, lane 2), whereas no similar changes were observed in Lec8 cells. These results further confirm that iPmel17 is modified with sialic acid and that ϳ10 kDa of core 1 O-glycans are added. Analysis of Pmel17 with ␣PEP13h revealed that the same combination of neuraminidase and O-glycanase reduced the size of mPmel17 (Fig. 6B, "Ͻ"), as had been seen in MNT-1 cells (compare with Fig. 4A). Interestingly, ␣PEP13h detected a pair of novel bands after digestion with neuraminidase, ␤1,4-galactosidase, and ␤-N-acetylglucosaminidase (Fig. 6B, lane 10, arrows). Thus, these data indicate that two differently glycosylated forms of Pmel17 are produced, and those forms have similar sizes to iPmel17 as detected in MNT-1 cells. These results are consistent with our previous observations that iPmel17 is a glycoform of Pmel17.
The primary biological role of Pmel17 is the formation of fibrils within melanosomes, which is essential for the maturation and pigmentation of those organelles. Our group has reported that MART1 interacts with Pmel17 and facilitates fibril formation in transfected cell lines (32). Therefore, Pmel17 and MART1 were transiently transfected together into CHO, Lec2, and Lec8 cells and were analyzed by electron microscopy (Fig. 6C). In mock-transfected cells, large multivesicular bodies containing electron-dense material were frequently observed in all three cell lines. Transfected CHO and Lec2 cells showed fragments of organized fibrils (Fig. 6C, insets, arrowheads) within multivesicular bodies, but those structures were observed less frequently in Lec2 cells. However, the formation of thick fibrils was observed only in Lec2 cells (supplemental Fig. 6), suggesting that the lack of sialic acid promotes the accumulation of Pmel17 without its further organization. Surprisingly, we were unable to identify any fibrils or similar structures in Lec8 cells. The sum of these results suggests that the addition of both sialic acid and galactose to Pmel17 is critical to its capacity to form fibrils and that those changes may be related to the capacity to form sialylated core 1 O-glycans.
Transfer of Sialic Acid to Core 1 O-Glycans Is Reduced in MNT-1 Melanoma Cells-Our results showing that O-glycan chains on Pmel17 are usually modified with sialic acid, whereas N-glycan chains are not, suggest a more efficient transfer of sialic acid to these O-glycan structures. The addition of sialic acid to nascent O-glycan chains stops the further extension of those structures. Changes in sialyltransferase expression and activity have already been associated with the regulation of glycoepitopes (49). Interestingly, high metastatic potential has been associated with increased ␤1,6-branching in N-glycans of melanoma cells (50,51). Thus, we hypothesize that melanoma cells may exhibit altered sialyltransferase activities compared with normal melanocytes, which would affect the glycosylation pattern of most proteins, including Pmel17. To examine that possibility, we checked the expression of ST3Gal I and ST6GalNAc II, which commonly act on core 1 O-glycan structures (Gal␤1-3GalNAc␣-R). Quantitative reverse transcription-PCR showed that mRNAs encoding both sialytransferases were expressed by melanocytic cells (supplemental Fig. 7). We then analyzed levels of sialyltransferase activity in MNT-1 cells compared with normal human melanocytes using different glycoproteins as acceptor substrates. As shown in Fig. 7A, both types of cells had sialyltransferase activity toward asialofetuin, which contains unsialylated N-glycans and unsialylated core 1 O-glycans, although MNT-1 cells were less active than normal melanocytes. In contrast, very high levels of sialyltransferase activities were detected using fully sialylated fetuin as an acceptor, showing that these cells can substitute sialic acid on already sialylated carbohydrates, an ability known as oligo-or polysialylation (52). Thus, normal melanocytes are 5 times more active than MNT-1 cells and 25 times more active than the breast cancer cell line T47-D, which expresses ST3Gal I and ST6GalNAc II (53), used as controls (data not shown). To discriminate between sialic acid transferred to N-glycans or to O-glycans, fetuin and asialofetuin were treated with PNGaseF, which allows the precipitation of proteins without N-glycans and the measurement of radioactivity transferred only to O-gly-cans (Fig. 7B). Note the reduced levels of sialic acid after treatment with PNGaseF, which corresponds to sialic acid transferred to N-glycans in asialofetuin and in fetuin. Nevertheless, normal melanocytes still exhibited four times more sialyltransferase activities toward fetuin than did MNT1 cells. This result also suggests that normal melanocytes express some polysialyltransferase (ST8sia) active on sialylated O-glycans that is down-regulated in the melanoma cells. We then assessed the transfer of sialic acid to N-or O-glycans using the following acceptors: asialo-orosomucoid, orosomucoid (which contains only N-glycans), BSM, and asialo-BSM (which carries core 3 O-glycans) (supplemental Fig. 8). Normal melanocytes and MNT1 cells exhibited low levels of activity toward N-glycans with or without sialylation, as assessed using orosomucoid and asialo-orosomucoid as acceptors, and toward core 3 O-glycans with or without sialylation, as assessed using BSM or asialo-BSM as acceptors. Therefore, we conclude that normal melanocytes and MNT-1 cells preferentially transfer sialic acid to core 1 O-glycan structures.
To further confirm these novel findings, we used immunoblotting to analyze the products of fetuin and asialofetuin with or without PNGaseF digestion in lysates of T47-D cells (as a positive control), NHM, MNT-1 cells, and water (as a negative control) (Fig. 7B). Ponceau staining showed various patterns that reflect differences between cell lysates, except in Fig. 7B, lanes 4, 8, 12, and 16, which contained only fetuin. Lectin staining with Maackia amurensis agglutinin, which recognizes ␣2,3linked sialic acid to galactose, revealed key differences between normal melanocytes and MNT-1 cells. Surprisingly, ␣2,3linked sialic acid was detected only in N-glycans of normal melanocytes (Fig. 7B, lanes 6 versus 10 and 14 after PNGaseF digestion). In contrast, ␣2,3-sialylation was present both in N-glycans and in O-glycans of MNT-1 cells (Fig. 7B, lanes 7  versus 11 and 15). Note the removal of ␣2,3-linked sialic acid after PNGaseF digestion (Fig. 7B, lane 12) from the N-glycans in fetuin (lane 4) indicates the specificity of the method. In addition, detection of 14 C-labeled sialic acid on the membrane using a PhosphorImager confirmed that most sialic acid transferred during the assay was transferred to O-glycans because the intensity of the signal did not change after PNGaseF digestion (Fig. 7B, right versus left half). Note that asialofetuin can be an acceptor for sialyltransferases in these cell lines, but the transfer is very low compared with fetuin.
The high levels of sialyltransferase activity, their predilection to transfer ␣2,3-sialic acid to N-glycans, and the reduced stability of Pmel17 after inhibition of ␣2,3-linked sialic acid transfer to core 1 O-glycans suggest that treatment with DANA would Based on data in this study, the divergence into iPmel17 and mPmel17 may occur in the medial-or trans-Golgi compartments because of different sialylation and glycosylation patterns. Those distinct forms of Pmel17 then sort directly or indirectly to melanosomes as proposed previously (14). Once in stage I melanosomes, mPmel17 is immediately cleaved and forms fibrils critical to the maturation of stage II melanosomes. After reaching the plasma membrane, iPmel17 may either be sorted to melanosomes or may be secreted or degraded. affect melanogenesis, especially in normal melanocytes. Thus, normal melanocytes were untreated or were treated with DANA for 6 days, and the melanin contents in the cell pellets were then measured (Fig. 7C). Interestingly, a highly significant 20% decrease (p Ͻ 0.001) in melanin content was observed following treatment of normal melanocytes with 20 M DANA. This result probably reflects an overall reduction in melanosome formation because of the decreased stability of Pmel17. However, we cannot exclude that melanin forming enzymes, e.g. tyrosinase, may also be affected.
Taken together, these results demonstrate that MNT-1 melanoma cells have reduced levels of sialyltransferase expression and activity compared with normal melanocytes. Most importantly, we determined that normal melanocytes and MNT1 cells preferentially transfer sialic acid to O-glycan chains.

DISCUSSION
The design and specificity of ␣PEP25h allow for a better understanding of the complex processing, sorting, and trafficking of Pmel17. In depth analysis of post-translational modifications in Pmel17 has been long overdue. ␣PEP25h is the first antibody that recognizes a known peptide sequence in the luminal domain of Pmel17, which constitutes an advantage compared with monoclonal antibodies such as HMB-45 or HMB-50 that recognize undefined epitopes (54,55). Because the peptide sequence recognized by ␣PEP25h is at the RPT domain, we confirmed that extensive O-glycosylation, as we suppose occurs in mPmel17, hides the epitope and thus abrogates reactivity by ␣PEP25h (as proposed in Fig. 1A). This constitutes an advantage to study the dynamics of iPmel17, which was previously and incorrectly proposed to represent an ER form of Pmel17 based only on its sensitivity to EndoH. Interestingly, Pmel17 is not the only melanosomal protein with this type of sensitivity to EndoH digestion, because similar patterns have been described for correctly processed, although inactive, tyrosinase (56) and for DOPAchrome tautomerase (57). Fig. 8 presents a schematic of Pmel17 processing through cytoplasmic compartments.
The various results obtained with ␣PEP25h in this study confirm that iPmel17 is not retained in the ER but is glycosylated in a manner distinct from mPmel17. Interestingly, immunoelectron microscopy with ␣PEP25h detects iPmel17 on the inner side of early stage melanosomal membranes. These results indicate that Pmel17 is intact, not cleaved, at the time it arrives to stage I melanosomes, an observation consistent with the highly amyloidogenic properties of cNTF-Pmel17 (58).
In this study, we found evidence that N-glycans can be differently N-glycosylated and sialylated in Pmel17, and we have clearly demonstrated that all its glycoforms are substituted with sialylated core 1 O-glycans, although the exact structure of those O-glycans remains to be determined. Interestingly, high metastatic potential of melanoma cells has been associated with increased ␤1-6 branching of N-glycans, because of high activity of N-acetylglucosaminyltransferase V (50,51). This constitutes an altered processing of N-glycans, and one could thus address the question about the impact of such an alteration on Pmel17 membrane expression in melanoma cells.
In light of our results, it is also critical to understand the role of high mannose/hybrid-type glycans in iPmel17 and mela-noma progression. In murine melanoma cells, proteins modified with mannose-type glycans interact with cell surface lectins that are necessary to initiate the spread of murine melanoma (59,60) and its metastasis to the liver (61). Furthermore, high mannose-type glycans are also involved in the functionality and cell surface expression of the mature human transferrin receptor (62,63). Therefore, the mannose-rich/hybrid type N-glycans on iPmel17 may play an active role in its sorting to the cell surface and a similar mechanism involving a mannose-binding lectin may also be involved in its active internalization. The overall reduction in sialyltransferase activity and the predilection to sialylated O-glycan chains in melanoma cells provide for the first time a mechanistic basis for these processes in melanocytic cells.
These findings also reveal a new factor to consider in the processing of melanoma-specific proteins. Differences in the pattern of expression of sialyltransferases, such as ST3 Gal I, between normal and malignant cells had already been associated with a cancer-specific regulation of glycoepitopes (49). In addition, the activities of sialyltransferases, although high in these cells, may be regulated both in normal and in malignant cells by factors that also regulate pigmentation, such as ␣-melanocyte-stimulating hormone. 4 Sialic acids are often present at the nonreducing ends of glycans, conferring strong negative charges on the protein. Melanins are polyanions with a relatively high content of negatively charged carboxyl groups and ortho-semiquinones at physiological pH (64). In this context, how does sialic acid addition then influence the structural role of Pmel17 and its ability to bind melanin to melanosome fibrils? We hypothesize that sialic acid, when added to N-glycans on Pmel17, makes it hydrophilic and leads to the exposure of hydrophobic protein domains, as occurs after the cleavage of Pmel17; this would then favor the formation of polymers through hydrophobic interactions. In contrast, sialic acid added to O-glycans may have a dual role as follows: 1) carrying plasma membrane sorting signals, and 2) protecting Pmel17 against early cleavage and degradation, a function that has been reported previously in other cell lines (41,42). Several studies support these roles as follows. The extent and type of O-glycans modified with sialic acid in the RPT domain of Pmel17 play key roles to determine the proper glycoform of Pmel17 destined to form fibrils (19) or to be targeted to the plasma membrane (41,42). Interestingly, the cNTF fragment reconstituted from E. coli has been shown to be amyloidogenic (58). In this context, it seems that glycosylation may not be required but in fact glycosylation and especially sialylation regulate this process actively (65). Our findings in this study indicate that in addition to sialic acid, galactose also plays a role in regulating fibril formation, but it may contribute in various ways. Pmel17 sorted to the plasma membrane may have a counter-receptor that interacts both with the protein and with its glycan moieties similar to what has been shown for the interaction of P-selectin with its ligand (66) or the mannose 6-phosphate receptor system that targets proteases to lysosomes (67). On the other hand, the presence of sialic acid and galactose could stabilize Pmel17 to achieve a defined conformation that may favor its safe transport to avoid polymerization. Such a stabilizing effect by sialic acid on protein structure has been observed previously (68).
A fair question to ask is why do melanocytes require different types of Pmel17? We hypothesize that this reflects the multifunctional nature of Pmel17. Once mPmel17 is processed and trafficked to stage I melanosomes, it is immediately processed and its C-terminal and cNTF fragments are cleaved to allow fibril formation (19). Following that, melanin is deposited on the fibers, and it is essentially trapped and covered by melanin. Despite that, Pmel17 is also present on the plasma membrane and is one of the most common melanoma antigens detected. Thus, Pmel17 secreted or at the plasma membrane may play other roles, which will require further study to determine. With that in mind, glycosylation seems to play a critical role in determining the trafficking of Pmel17 to various subcellular compartments and thus regulates its functions. Some glycosylated forms of Pmel17 (i.e. mPmel17) will sort directly to melanosomes to initiate melanosome biogenesis, whereas other glycosylated forms (i.e. iPmel17) are sorted through the secretory pathway for secretion or recycling.
Taken together, ␣PEP25h has proven to be a useful antibody that allows specific detection of iPmel17 and allows characterization of its processing and trafficking. Our findings demonstrate that iPmel17 is a glycosylated form that is distinct from mPmel17 and that it is not an ER-retained form. Such differences and the distinct sorting patterns that result in the different glycoforms raise the possibility that Pmel17 may have other important functions in addition to its well known role in generating the structural fibrillar matrix of melanosomes. Our novel findings confirm that the addition of sialic acid to O-glycans on Pmel17 is involved in the stability and sorting of this protein through the secretory pathway, and that the modification with sialic acid and galactose to Pmel17 glycans is critical to its ability to form fibrils, a process that directly regulates pigmentation in mammals. Furthermore, alterations in sialyltransferase activity and substrates in melanoma cells provide a mechanistic explanation for the microheterogeneity observed in the glycan structures of Pmel17.