“Subcellular Proteomics” of Neuromelanin Granules Isolated from the Human Brain*

“Subcellular proteomics” is currently the most effective approach to characterize subcellular compartments. Based on the powerful combination of subcellular fractionation and protein identification by LC-MS/MS we were able for the first time to 1) isolate intact neuromelanin granules from the human brain and 2) establish the first protein profile of these granules. This compartment containing neuromelanin (NM) is primarily located in the primate’s substantia nigra, one of the main brain regions that severely degenerates in Parkinson disease. We used mechanic tissue disaggregation, discontinuous sucrose gradient centrifugation, cell disruption, and organelle separation to isolate NM granules from human substantia nigra. Using transmission electron microscopy we demonstrated that the morphological characteristics of the isolated NM granules are similar to those described in human brain tissue. Fundamentally we found numerous proteins definitely demonstrating a close relationship of NM-containing granules with lysosomes or lysosome-related organelles originating from the endosome-lysosome lineage. Intriguingly we further revealed the presence of endoplasmic reticulum-derived chaperones, especially the transmembrane protein calnexin, which recently has been located in lysosome-related melanosomes and has been suggested to be a melanogenic chaperone.

Melanins are widely distributed throughout the plant and animal kingdoms. In humans, these heterogeneous, complex polymer pigments occur naturally in the hair, the skin, the inner ear, the iris, and the choroid of the eye (1). Melanin in the brain has an appearance and structure similar to cutaneous melanins and has thus been named neuromelanin (NM) 1 (2). NM is found inter alia in dopaminergic neurons of a small area in the human midbrain important for the control of movement that is known as the substantia nigra pars compacta (SN; from the Latin meaning "black body"). The loss of this dark pigment and the resulting pallor of the SN is one of the most striking features of the common movement disorder Parkinson disease. A relationship between the loss of the dopaminergic SN cells and their NM content (3), a specific affinity to iron (4), and a significant binding of ␣-synuclein to NM in the diseased state (5) suggest a functional role for NM in neurodegeneration in Parkinson disease (6).
Although much is known about the peripheral melanins to which NM is thought to be related, many basic questions remain to be answered about NM in the brain. Thus it is unclear why only some human dopamine neurons produce NM within their cytoplasm (7). Little is also known about the structure of NM, and the understanding of its genesis and function within the cell remains speculative.
Nuclear magnetic resonance spectroscopic studies have shown that NM resembles synthetic cysteinyldopamine melanin more closely than the more simple dopamine melanin; however, human NM appears to be a structurally more complex chemical structure than any of the synthetic models (8). In addition to the melanin backbone, nuclear magnetic resonance spectroscopic studies have demonstrated that cholesterol and other uncharacterized high molecular mass lipid components are closely associated with NM (8 -10). Dolichol was identified as the major lipid component of NM (11). A proteinaceous component making up ϳ5-15% of the isolated molecule is also present that has been suggested to represent an integral component of the polymer (8,10).
A general understanding of neuromelanogenesis could be provided by investigation of the synthetic pathway of peripheral melanins and comparison with what is known about NM. Genetic and enzymatic regulation of melanin production in the periphery has been primarily characterized by the study of fur pigmentation in the mouse. Similar experiments, however, cannot be used to elucidate the pathway of NM synthesis as NM does not occur in rodents. It has long been debated whether the NM synthesis is enzymatically controlled, like all melanins in the periphery, or whether NM arises from a simple autoxidation process (for reviews, see Refs. 12 and 13). In the apparent absence of a role for tyrosinase in neuromelanogenesis, the search for an enzyme associated with NM production has yielded no likely candidates to date (12,13). It is noteworthy that Parkinson disease patients treated with large quantities of the dopamine precursor L-3,4-dihydroxyphenylalanine (L-DOPA, levodopa) do not exhibit increased quantities of NM within their surviving SN neurons as might be expected to be the case if NM represents a product of pure dopamine autoxidation. The time course of NM appearance also supports the hypothesis that NM synthesis is a regulated process. In the human, NM is not present in functional dopaminergic neurons at birth but first appears at around 3-5 years of age.
Although containing a melanin, neuronal NM granules are organelles different from melanosomes, which are mainly localized in melanocytes. In contrast to NM, skin-and hairbased melanins are contained within discrete regularly sized membrane-bound organelles called melanosomes, which differ structurally depending on which type of melanin they contain. It has been suggested that melanosomes are closely related to lysosomes (14). In contrast, NM is bordered indistinctly, and the NM granules, as they are called, exhibit a wider size range than that of melanosomes (0.5-2.5 m) (15,16). The structurally segregated lipid component of NM that is not found in peripheral melanin pigments has been suggested to originate from the lipid-containing pigment lipofuscin, which also accumulates intracellularly with age (17).
Here we report a method to isolate intact, highly pure NM granules from human SN for subcellular proteomics. Proteomics is the large scale study of gene expression at the protein level that ultimately provides direct measurement of protein expression levels and insight into the activity state of all relevant proteins (18 -20). Subcellular proteomics thus is a powerful approach to gain new insight into functional properties of isolated organelles. We applied one-dimensional (1-D) SDS-PAGE, tryptic in-gel digestion, and nano-LC separation followed by ESI-MS/MS for protein analysis. Following this approach, we identified numerous proteins specific for organelles originating from the endosome-lysosome lineage.

EXPERIMENTAL PROCEDURES
Isolation of NM Granules Using Subcellular Fractionation-We used a sequential top-down approach that simultaneously allows the reduction of the complexity of the sample and the enrichment of the target structures at each isolation stage. Fig. 1 schematically summarizes the approach chosen. Brains were provided from the Austro-German Brain Bank in Wü rzburg. The use of postmortem human brain tissue was approved by the Ethics Committee of the University Clinics of Wü rzburg. The SN was dissected from postmortem brains of subjects with no history of neurological or neurodegenerative diseases within 36 h of death on a cool plate (Ϫ15°C). 1.0 g of frozen SN tissue was thawed in "Separation Buffer" (10 mM HEPES, 10% glucose, pH 6) on ice and carefully passed through a polypropylene mesh into a Petri dish. The whole procedure was performed on a plate cooler set at 4°C. The resulting cell suspension was layered on top of a discontinuous sucrose gradient (1, 1.2, 1.4, and 1.6 M) and separated by centrifugation at 4000 ϫ g at 4°C for 15 min. The pelleted dark cell bodies were recovered and washed with "Isolation Buffer" (10 mM HEPES, 1 mM EDTA, 100 mM KCl, 10% sucrose, pH 7.5) containing a protease inhibitor mixture (0.01% (v/v), Sigma). Subsequently the cell disruption was carried out by 10 passages through a 26-gauge needle to yield a suspension of cellular organelles that was layered on top of an 80% (v/v) Percoll cushion (Fluka, Buchs, Switzerland) and centrifuged at 4000 ϫ g at 4°C for 10 min. The pelleted dark granules were washed once with Isolation Buffer and twice with "Washing Buffer" (10 mM HEPES, 250 mM NaCl, 0.01% (v/v) Triton X-100, pH 7.5) to remove unspecifically associated proteins. The isolated NM granules were stored at Ϫ80°C until analyzed.
Transmission Electron Microscopy-The quality of the granule isolation and the preservation of the ultrastructural features were monitored by transmission electron microscopy. The aspect of cell homogenates and isolated NM granules were compared as a control to monitor the enrichment of the granules. The samples were fixed overnight in 2% (v/v) glutardialdehyde in 0.1 M phosphate-buffered saline, pH 7.4, at 4°C and incubated in 2% OsO 4 , 1,5% (v/v) glutardialdehyde followed by dehydration with increasing concentrations of ethanol. After incubation in 1,2-epoxypropane (Sigma) (2 ϫ 15 min) the samples were embedded in EPON TM epoxy resin (Sigma). Following polymerization at 65°C (48 h) thin sections were prepared that were contrasted with lead citrate and uranyl acetate (21) before being monitored under the transmission electron microscope (LEO 912 AB, LEO Elektronenmikroskopie, Oberkochen, Germany).
1-D SDS-PAGE-The protein samples were separated electrophoretically on 10 -20% Tricine gels (Novex, San Diego, CA) in an XCell II TM Mini-Cell (Invitrogen) using Tricine-SDS running buffer. Following electrophoresis the gel was either stained with colloidal Coomassie Brilliant Blue G-250 (22) or further processed for Western blotting.
Tissue Homogenate-To provide a positive control for the Western blot analysis, 0.5 g of SN tissue was disrupted using a Potter-Elvehjem homogenizer in Lysis Buffer containing protease inhibitor mixture (0.01%, v/v), and proteins were extracted with 16 mM 3-[(3cholamidopropyl)dimethylamino-1-propanesulfonate (Calbiochem).
Western Blot Analysis-The separated proteins were transferred electrophoretically onto nitrocellulose membranes (Invitrogen) using the XCell II blot module. Nonspecific binding was blocked with 5% (w/v) nonfat dried milk, 0.5% (v/v) Tween 20 in Tris-buffered saline, pH 7.3, for 1 h at 20°C. Immunoblots were probed with primary antibodies at the appropriate dilutions at 4°C overnight or at room temperature for 1 h. Membranes were washed in Tris-buffered saline containing 0.1% (v/v) Tween 20 (3 ϫ 10 min) followed by incubation with the secondary antibody at 20°C for 1 h. Additional washing was performed with Tris-buffered saline containing 0.1% (v/v) Tween 20 (3 ϫ 10 min), and the immunocomplexes were visualized by enhanced chemiluminescence (ECL TM system, Boehringer Ingelheim). Stripping of immunoblots for repeated probing was performed by incubating the membranes at 50°C for 15 min in 100 mM 2-mercaptoethanol, 2% SDS, 62.5 mM Tris-HCl, pH 6.7.
Detection of Mannosylated Proteins by GNA Lectin-The biotinylated GNA lectin was applied to visualize mannosylated proteins (23,24) of isolated NM granules and total SN tissue homogenate. The proteins were transferred electrophoretically onto polyvinylidene difluoride membranes (Invitrogen) using the XCell II blot module. The membranes were blocked with Tris-buffered saline containing 0.1% (v/v) Tween 20 at 20°C for 1 h and incubated with biotinylated GNA lectin overnight at 4°C followed by incubation with horseradish peroxidase-linked streptavidin at 20°C for 1 h. The bands of mannosylated proteins were visualized by enhanced chemiluminescence (ECL system).
In-gel Digestion with Trypsin-An entire lane of a gel previously stained with colloidal Coomassie Brilliant Blue G-250 was sliced into 4-mm cubes, and each of these was placed into a separate quartz reaction tube (Sigma) (25). The gel cubes were washed three times with 10 mM NH 4 HCO 3 , pH 7.8, and 10 mM NH 4 HCO 3 , pH 7.8, acetonitrile (1:1, v/v) each for 10 min. The gel cubes were subsequently reswollen by addition of 2 l of modified trypsin (Promega, Madison, WI; 0.05 g/l in 10 mM NH 4 HCO 3 , pH 7.8). The digestion was performed overnight at 37°C.
Sample Preparation for LC Separation-10 l of 0.1% (v/v) trifluoroacetic acid/acetonitrile (1:1, v/v) were added to each gel slice followed by sonication for 10 min. This step was repeated twice, and the supernatants containing the extracted peptides were combined in separate quartz tubes.
Mass Spectrometric Analysis-Nano-HPLC-ESI-MS/MS analysis of the tryptically generated peptides was carried out as described previously (26). Spectra were recorded on a Finnigan LCQ TM Classic (Thermo Electron, San Jose, CA) ion trap mass spectrometer equipped with a nanoelectrospray ion source (Pico View TM 100, New Objective Inc., Woburn, MA). The peptides were preconcentrated by loading onto a -precolumn (0.3-mm inner diameter ϫ 5 mm, Pep-Map TM , LC Packings Dionex, Idstein, Germany) before being separated by reverse-phase nano-LC (75-m inner diameter ϫ 250 mm, PepMap, LC Packings Dionex) using a precolumn split.
Mass Spectrometry Data Analysis-The analysis of the raw MS/MS data occurred automatically based on the Sequest TM algorithm (27,28). The data were searched against the human NCBInr data base (www.ncbi.nlm.nih.gov) using the following parameters: average masses, partial oxidation of methionine (ϩ16 Da), a mass tolerance of Ϯ 1.5 Da, trypsin was used as a specific protease, and a maximum of two missed cleavage sites was tolerated. Furthermore the analysis was restricted to ions in the mass range of 500 -5000 Da and a total ion current greater than 3 ϫ 10 5 . In general, a cross correlation value (X corr ) of greater than 2.0 and a ⌬ correlation score (⌬Cn) greater than 0.1 was accepted for confident identification; inspection of the spectra was performed to confirm the Sequest results.

FIG. 1. Outline of isolation of neuromelanin from human SN.
The isolation procedure includes two consecutive steps. In the first step the tissue is disaggregated leading to the enrichment of pigmented neuronal cell bodies after centrifugation through a discontinuous sucrose gradient. In the second step, these pigmented cell bodies are disrupted and subjected to subcellular fractionation to yield NM granules.

Subcellular Fractionation to Isolate NM-containing Gran-
ules-We developed a mild procedure for the isolation of intact and pure NM granules from human SN to enable subcellular protein analysis. Fig. 1 shows a schematic summary of the approach used. As a first step, the tissue was disaggregated by mechanical sieving into a cell suspension that was subsequently fractionated by centrifugation through a discontinuous sucrose gradient. This step allowed the enrichment of dark cell bodies as a pellet at the bottom sucrose layer. In the second step, these dark cell bodies were disrupted and subjected to an additional centrifugation step for finally isolating the NM granules by subcellular fractionation.  GM130)), mitochondria (a, induced myeloid leukemia cell differentiation protein (Mcl-1)), early endosomal compartments (b, early endosomal antigen 1 (EEA1)), plasma membrane (b, integrin ␣ 2 (VLA-2␣)), and nucleus (c, d; nucleoporin p62 (np62)). The protein extract of isolated NM granules compared with SN homogenate shows the presence of cathepsin B (d), a lysosomal proteinase, and LAMP-1 (e), a marker for late endosomes, lysosomes, and lysosome-related organelles. Dynamin (f) and clathrin (g), which are involved in vesicular traffic and are suggested to be associated to endosomal compartments, are detected. The melanogenic chaperone calnexin (h) is present in NM granules, although the marker for the endoplasmic reticulum (i, 78-kDa glucose-regulated protein (BiP/GRP78)) is absent after stripping. B, blotted proteins were probed with GNA lectin, which specifically binds to mannosylated proteins found in lysosomes or lysosome-related organelles.

Quality Control of Isolated Specimens by Transmission
Electron Microscopy-The purity and quality of the granule isolation were monitored by transmission electron microscopy to evaluate the level of enrichment achieved by this approach as well as the structural and morphological appearance of the isolated granules (Fig. 2). Compared with the homogenates of pigmented cell bodies prior to the NM granule isolation, the isolated NM granules were virtually free from contaminating organelles; this was attributed to the exceptional density of NM granules. Up to now the essential density to penetrate an 80% (v/v) Percoll cushion has only been reported for highly melanized stage IV melanosomes isolated from Xenopus laevis melanophores (29,30).
Purified granules (Fig. 3) displayed all the morphological structures described previously for primate brain tissue by electron microscopy studies (16,31) showing 1) the highly electron-dense patches attributable to the iron-rich NM, 2) the medium electron-dense protein matrix, and 3) vacuolar lipid bulbs. These characteristics were well preserved during isolation underscoring the potential of this strategy for isolation of NM granules.
Quality Control by Western Immunoblotting-The level of enrichment was additionally assessed by Western immunoblotting applying antibodies against "marker proteins" specific for cell organelles (Fig. 4A) such as the Golgi network (GM130), mitochondria (Mcl-1), early endosomal compartments (EEA1), the plasma membrane (VLA-2␣), the nucleus (nucleoporin p62), lysosomes (cathepsin B and LAMP-1), and the endoplasmic reticulum (BiP/grp78). The proteins extracted from NM granules were compared with control tissue homogenate and show the presence of lysosomal markers, whereas the other "organelle marker proteins" are absent (Fig. 4A, a-e and i).
Protein Identification by Mass Spectrometry and Western Immunoblotting-As depicted in Fig. 5, the proteins extracted    Fig. 6. We found a variety of transmembrane proteins specific for organelles originating from the endosome-lysosome lineage, such as LIMP II (Protein 2) and LAMP-1 (Protein 3) but also subunits of the vacuolar ATPase (Proteins 4 -6) responsible for the acidification of intracellular compartments ( Fig. 6 and Table I). LIMP II was the only lysosomal marker membrane protein unequivocally identified by mass spectrometry. Although our analytical approach is compatible with membrane proteins (25,32), LAMP-1 (Protein 3) and LAMP-3 (LIMP I) (Protein 1) could not be identified by a sufficient amount of peptides by mass spectrometry. As both proteins are glycosylated, LAMP-1 and -3 might be present in a much lower amount in NM granules than LIMP II. The presence of LAMP-1 was substantiated by Western blot analysis (Fig. 4A, e). We further identified several proteases known to be localized in the lysosome (e.g. cathepsin B (Protein 7) (Fig. 4A, d), cathepsin D (Protein 8), and tripeptidyl-peptidase I (Protein 10)) and enzymes mediating the catabolism of sphingolipids or glycoproteins (e.g. acid ceramidase (Protein 15), palmitoyl protein thioesterase 1 (Protein 17), and chondroitinase (Protein 22)).
To further examine the lysosomal traits of NM granules, we used GNA, a mannose-binding lectin, to identify proteins targeted to lysosomal compartments. Because of their selective binding to glycan structures lectins are a powerful tool in the analysis of glycoproteins (39,40). As shown in Fig. 4B, numerous mannosylated proteins were found in NM granules and SN tissue homogenate. However, only a subset of mannosylated proteins was found in NM granules compared with the total tissue homogenate.
Interesting findings, however, are the obvious presence of four ER-derived proteins (Proteins 69 -72, most notably calnexin ( Figs. 6 and 4A, h), whereas by Western immunoblotting we could not show the presence of the ER marker protein 78-kDa glucose-regulated protein (BiP/grp78) (Fig. 4A, i). DISCUSSION This study demonstrated the utility of a new approach to analyze purified NM granules that sheds light on the biogenesis and biology of these unique organelles. We developed a new method for isolation of intact and pure NM granules from human SN without using detergents, organic solvents, and proteases as applied in the isolation of NM from crude tissue homogenates (9). Then we applied 1-D SDS-PAGE followed by in-gel digestion and nano-HPLC coupled to ESI-MS/MS to identify novel constituents of NM granules. This approach overcomes the difficulties in separation of basic and hydro-phobic proteins as with two-dimensional electrophoresis (32,41). In addition, we used Western immunoblot analysis to investigate the potential link to lysosomes. Table I summarizes the proteins identified by ESI-MS/MS. The function and the subcellular localization of some proteins identified by our approach are not yet entirely known.
By using this approach we identified numerous proteins attributable to the endosome-lysosome lineage as well as four proteins from the ER. A link of NM granules to lysosomes has been assumed previously; however, the morphological appearance and the phylogenetic limitation of the pigment NM are their most striking differences to conventional lysosomes.
In eukaryotic cells, lysosomes serve as the terminal sites for delivery of material targeted for removal (42)(43)(44). For this purpose, lysosomes are equipped with a variety of more than 50 soluble acid-dependent hydrolases engaged in final degradation of both exogenous and endogenous macromolecules. Lysosomes are highly dynamic and heterogeneous organelles and thus cannot be identified solely by morphology (45)(46)(47)(48). Therefore, lysosomes are described by the presence of highly glycosylated integral membrane proteins known as LAMPs, LIMPs, and lysosomal membrane glycoproteins. In contrast to late endosomal compartments, they lack mannose 6-phosphate receptors. Lysosomes maintain an acidic pH (4.5-5.0) required for the hydrolytic activities of the luminal enzymes. All or most of these features are shared with specialized cell type-specific "lysosome-related organelles" including melanosomes, platelet-dense granules, lytic granules, and major histocompatibility complex class II compartments among others, which mostly perform physiological functions different from biomolecular degradation (49). Various studies on genetic multiorganellar disorders substantiated the biogenetic relationship between lysosome-related organelles and lysosomes.
Recently various models of lysosomal biogenesis have been established (46 -48), but all of them include the existence of very specific sorting mechanisms. Membrane proteins such as LAMP-1 and LIMP II exit the trans-Golgi network on a secretory route to the endosomal compartments that is mediated by recognition of their tyrosine-based (LAMP-1) and dileucine-based (LIMP II) sorting motifs (50 -53) by adaptor complex AP-3 (51). This adaptor complex AP-3 is also crucial for the biogenesis of lysosome-related melanosomes by targeting tyrosinase gene family proteins according to their dileucine-based sorting motifs (54). However, in our study on NM granules we did not find the melanogenic enzymes tyrosinase and tyrosinase-related proteins 1 and 2, which are located in melanosomes of cutaneous melanocytes and retinal pigment epithelial cells. Although low levels of tyrosinase mRNA have been reported in human SN (55), tyrosinase protein does not appear to be expressed in the human brain (56). 2 The current evidence thus suggests that tyrosinase does not appear to have a role in the synthesis of human brain NM. Regardless it cannot be excluded that such an enzyme is covalently bound to the framework of the highly abundant pigment, which renders the protein insoluble and inaccessible to detergents (58).
In the context of membrane protein sorting, we want to point out that surprisingly we did not detect the "classical" lysosomal marker protein lysosomal acid phosphatase. At first, lysosomal acid phosphatase is transiently integrated into the limiting membrane as a type I membrane protein and is further proteolytically processed to gradually release the luminal active site of the enzyme into the lysosomal matrix (59).
A distinct pathway guides mannose 6-phosphate-tagged acid-dependent hydrolases to endosomes and lysosomes in clathrin-coated vesicles, which bud from the trans-Golgi network as the mannose 6-phosphate tag is recognized by mannose 6-phosphate receptors. Dynamin has been suggested recently to be associated with late endosomes, tubulovesicles, and clathrin-coated vesicles as well as being involved in the recycling of mannose 6-phosphate receptors after cargo delivery (60). Interestingly by proteomic mapping we identified an unexpectedly small repertoire of mannosylated lysosomal enzymes (Table I), such as glycosyl hydrolases (EC 3.2.1.) (61). Again a similar situation occurred showing a small subset of proteins when tested for mannosylation by GNA recognition (Fig. 4B).
A few proteins specifically facilitating the degradation of glycosphingolipids (62) were identified, such as acid cerami-dase, etc. (Table I). Glycosphingolipids are predominantly found in neuronal tissue; it will be the task of further investigations to elucidate whether or not sphingolipids are found in the lipidic bulbs of the NM granules. Mutated forms of tripeptidyl-peptidase I (Protein 10) and dipeptidyl-peptidase II (Protein 14) have been found to be involved in some types of neuronal ceroid lipofuscinosis, characterized by an accumulation of autofluorescent inclusion bodies (63). These forms of neurodegenerative disorders are regarded as lysosomal storage diseases evoked by defective proteolysis.
Our findings of ER proteins associated with NM granules substantiate their multitopological subcellular localization. We assume a specific localization of the ER proteins in NM granules because major ER constituents, e.g. cytochrome P-450 isoforms or NADPH cytochrome c reductase (64), have not been detected, and we could not identify BiP/grp78 by MS and Western immunoblotting (Fig. 4B). This is in line with recent findings that suggest a direct involvement of the ER in the establishment of early lysosomal structures including phagosomes in neutrophiles (65) and stage I melanosomes in melanocytes (14,66). During the maturation of phagosomes to phagolysosomes the quantity of ER proteins gradually decreases; calnexin especially is lost rapidly (67). In classical lysosomes calnexin is absent but is rather associated with lysosome-related organelles such as melanosomes (68 -70). There is increasing evidence that calnexin functions as a melanogenic chaperone (54,71) currently suggested to mediate the proper folding and sorting of the melanogenic enzymes toward their target compartment (70,72). In summary, our results provided important insight into the cellular machinery used to generate NM granules, revealing an admirable conservation of cellular processes. At this stage of analysis, it has become clear that the majority of the components of NM granules, similar to those of melanosomes, are shared with lysosomes (Fig. 7). The putative precursor compartment may have endosomal traits; however, an additional direct involvement of the ER is suggested to contribute to the constitution of NM granules. Future studies are warranted to clarify the nature of the lipidic bulbs attached to NM granules.