The characterization of the invasion phenotype of uveal melanoma tumour cells shows the presence of MUC18 and HMG-1 metastasis markers and leads to the identification of DJ-1 as a potential serum biomarker

Authors


Abstract

Uveal malignant melanoma (UM) is the most frequent primary intraocular tumour in adult humans. Because the survival rate of patients with UM has changed little in the past few decades, a better understanding of the molecular events governing UM development and the identification of markers indicating the potential for metastasis at the time of diagnosis are necessary to design improved and more specific treatments. In this study, we investigated UM tumour development by comparing two recently established UM cultures with different invasion potential by two-dimensional gel electrophoresis. Protein features expressed differentially were identified by mass spectrometric analysis. Potential markers were assayed in both cultures and in long-term established UM cell lines (UW-1, OCM-1, SP6.5 and 92.1) by Western blotting and their role in invasion analysed using Matrigel membranes. Comparative analysis revealed that UM cultures with low- and high-grade invasion potential differ in their cellular metabolism and, more interestingly, in several cancer-associated proteins, including those implicated in cell adhesion and migration, proliferation and various oncogenes. Our data indicate a correlation between MUC18 and HMG-1 expression and the invasiveness of UM cells. We also demonstrate the expression and secretion of DJ-1 oncoprotein by UM cells. We suggest a possible role for MUC18 and HMG-1 proteins in UM cell invasion. The secretion of DJ-1 by UM cells, and the ability to detect this protein in UM patients' sera implicate it as a potential noninvasive biomarker for this malignancy. © 2006 Wiley-Liss, Inc.

Uveal malignant melanoma (UM) is the most frequent intraocular tumour in adult humans.1 Unlike cutaneous melanoma, uveal melanoma disseminates mainly through the blood stream and preferentially establishes metastases in the liver. Metastatic liver disease is the leading cause of death in uveal melanoma and can develop after a long disease-free interval, which suggests the presence of occult micrometastatic disease at the time the primary eye tumour is diagnosed and treated.2 Unfortunately, advances in eye cancer treatment have not paralleled those made in the management of other types of cancer, and the survival rate of patients with uveal melanoma has changed little in the past few decades.3 A better understanding of the molecular events governing uveal melanoma development and the identification of markers indicating the potential for metastasis at the time of diagnosis are necessary to design improved and more specific treatments. Various clinical and molecular prognostic factors have been suggested in uveal melanoma, but none has proved to be sufficiently useful or viable for routine clinical use.4, 5

Currently, there are many challenging technologies and approaches to identify tumour markers for prognostic and therapeutic purposes.6 Transcriptional studies alone are not sufficient, with several investigators reporting a poor correlation between mRNA and protein abundance.7 Furthermore, a single gene can encode for more than one mRNA species through differential splicing, and proteins can undergo as many as 200 posttranslational modifications.8 For this reason, proteomics has become a complementary technique to genomic analysis. Studies of global protein expression by proteomics technology in human tumours have yielded information about tumour heterogeneity and have led to the identification of various polypeptide markers which are potentially useful as diagnostic tools.9 We recently applied this approach to obtain the first proteome analysis of a previously well-characterized uveal melanoma primary cell culture named UM-A.10, 11 This analysis represented the first step towards the utilization of proteomics in the study of uveal melanoma tumours and a novel approach to search for potential markers. The UM-A primary culture, characterized by the presence of hypophosphorylated and unexpectedly inactive antitumoural protein Rb, changed into a phenotypically different cell line after 7 passages.10 On the basis of extracellular matrix (ECM) invasion assays that showed a low and high degree of metastatic potential in the UM primary cell culture (UM-A < 7) and cell line (UM-A > 7), respectively, we present here the differential proteome analysis of both cultures. The identification of differentially regulated features reveals an increase in the cellular metabolism of the UM-A cell line and, interestingly, a different expression profile of cancer-related proteins. Proteins involved in cell adhesion and migration, proliferation, various oncogenes and others, many described for the first time in UM, are listed in this article. A selection of novel proteins is validated by other techniques, showing a possible correlation between MUC18 and HMG-1 expression and UM invasiveness. We also show the overexpression and secretion of DJ-1 oncoprotein in UM cells and the ability to detect it in certain UM patient sera, indicating DJ-1 as a potential biomarker.

Material and methods

Cell culture

The UM-A primary cell culture, the resulting cell line and normal melanocytes (NM) were obtained as described previously.10 UM-A < 7, UM-A > 7, NM and UM cell lines (UW-1, OCM-1, SP5.6 and 92.1) were cultured in RPMI medium containing 5% inactivated fetal calf serum (FCS), 2 mM glutamine and standard antibiotics at 37°C in 5% CO2.

Immunoblotting

Whole cell lysates were prepared by direct lysis of subconfluent cells in cold RIPA buffer, as described previously.10 Equal amounts of protein (30 μg/lane) were separated on sodium dodecyl sulphate–polyacrylamide gels (SDS–PAGE) and electroblotted onto nitrocellulose membranes. The membranes were probed successively with primary antibodies and horseradish peroxidase-labelled secondary antibodies (ECL, GE Healthcare, Uppsala, Sweden). Anti-DJ-1 and MUC18 antibodies were purchased from Upstate (Lake Placid, NY) and anti-actin and HMG-1 from Sta Cruz Biotechnology (Santa Cruz, CA). To prepare cell culture secreted proteins for Western blot, the cell culture medium was concentrated in a Centricon YM-10 (Millipore, MA) to a final volume of 100 μl, according to the manufacturer's instructions. All Western blot images shown in the figures are representative of at least 3 independent experiments.

Extracellular matrix invasion assays

The invasion potential of UM cells was assessed using a BD BioCoat™ Matrigel™ Invasion Chamber (Becton Dickinson, MA), following the manufacturer's indications. Briefly, 15 × 104 cells were seeded over an 8-μm pore size PET membrane with ECM or without ECM (control) inserts in 0.1% FCS-RPMI medium for 48–72 hr. FCS-RPMI (5%) was added to the lower chamber as a chemoattractant. Noninvading cells were removed from the upper chamber by gently wiping the upper surface of the membrane with a cotton swab. Membranes were fixed in 100% methanol and stained in 1% toluidine blue. Levels of invasion were assessed by counting the number of cells present in 5 different fields in triplicate on the lower surface of the membrane under the light microscope (40× magnification). Data are expressed as the percent of invasion through the Matrigel matrix and relative to the migration through the control membrane.

Two-dimensional gel electrophoresis

UM-A < 7 and UM-A > 7 protein samples were obtained as described previously.11 Briefly, cells were washed twice in phosphate balance solution and disrupted by gentle sonication in sample buffer (5 M urea, 2 M thiourea, 2 mM tributylphosphine, 65 mM DTT, 65 mM CHAPS, 0.15 M NDSB-256, 1 mM sodium vanadate, 0.1 mM sodium fluoride, 1 mM benzamidine). The supernatant was removed and 600 μg of protein was taken up in a total volume of 375 μl sample buffer. 3–10NL immobilized pH gradient (IPG) strips were hydrated in the sample, and isoelectric focusing was carried out for 70 kV hr at 17°C, according to the method of Sanchez et al.12 Following focusing, the IPG strips were immediately equilibrated for 10 min in 4 M urea, 2 mM thiourea, 12 mM DTT, 50 mM Tris (pH 6.8), 2% (w/v) SDS, 30% (w/v) glycerol and placed on top of the second dimension gels embedded in 0.5% melted agarose. Proteins were separated in the second dimension on SDS-PAGE gradient gels (9–16% T, 2.67% C) under running conditions of 10°C, 20 mA/gel for 1 hr, followed by 40 mA/gel for 4 hr. Following electrophoresis, the gels were fixed and stained with the fluorescent dye OGT MP17 (Oxford Glycosciences, Abingdon, UK), on the basis of Ref.13. 16-Bit monochrome fluorescence images were obtained at 200-μm resolution by scanning gels with an Apollo II linear fluorescent scanner (Oxford Glycosciences).

Differential image analysis

Scanned images were processed with a custom version of MELANIE II (Bio-Rad Laboratories Ltd, Hemel Hempstead, UK). Four pI 3–10 gels were prepared in independent experiments for both UM-A < 7 and UM-A > 7 samples. Internal calibration of the 2DE (two dimensional gel electrophoresis) gel images with regard to pI and molecular weight was carried out, as described previously.14 For differential image analysis, a synthetic gel image was generated by means of accurate spot matching. This synthetic image contained all protein features detected in UM-A < 7 and UM-A > 7 samples. Only the features present in at least 3 of 4 individual gels belonging to either the UM-A < 7 or UM-A > 7 samples were considered for differential analysis. Intensity (optical density) was measured by summing pixels within each spot boundary (spot volume) and recorded as a percentage of the total spot intensity of the gel: %V = spot volume/Σ volumes of all spots resolved in the gel. Variations in protein expression were calculated as the ratio of average volumes (%V) and carefully validated by repeated image analysis by human operators. Differential expression of a protein present in both UM-A < 7 and UM-A > 7 gels was considered significant when the fold change was at least 2 and p was no more than 0.05 after rank-sum test applied on %V values.

In-gel digestion and peptide extraction

Protein features assigned to mass spectrometric analysis were excised from the gel by a software-driven robotic cutter. The recovered gel pieces were dried in a speed-vac, and in-gel digestion was carried out by the automated DigestPro workstation (Abimed, Langenfeld, Germany), according to the protocol of Shevchenko et al.15

Mass spectrometric analysis

Mass spectrometric analysis was carried out using a Q-TOF Micro instrument (Micromass, Manchester, UK) coupled to CapLC (Waters, Milford, MA). The tryptic peptides were loaded and desalted on a 300 μm id/5 mm length C18 PepMap column (LC packings, San Francisco, CA). The peptides were resolved on a 75 μm id/150 mm length C18 Atlantis NanoEase column (Waters). The peptide mixture was eluted with 98% acetonitrile containing 0.1% formic acid over 60 min at a flow rate of 250 nl/min. The gradient was as follows: 0–3 min, constant 5% acetonitrile; 3–30 min, acetonitrile increased to 40%; 30–35 min, acetonitrile increased to 90%; 35–40 min, acetonitrile constant at 90%; 40–45 min, acetonitrile decreased to 5%; 45–60 min, acetonitrile constant at 5%. Mass spectrometry data were acquired and analysed by Masslynx software version 4 (Micromass) using automatic switching between MS and MS/MS modes. The survey scan (1 sec) was obtained over the mass range m/z 300–1,600 in the positive-ion mode with a cone voltage of 35 V and a capillary voltage of 3,500 V. When the signal reached a user-defined threshold (10 counts/sec), peptide precursor ions could be selected for MS/MS (8 sec total scan time) over the mass range m/z 50–2,000. Fragmentation was performed using argon as the collision gas and with a collision energy profile (20–40 eV) optimized for various mass ranges of precursor ions. The selected precursor ions were automatically included in the exclusion list. The database search was performed with the MASCOT search tool (Matrix science, London, UK) screening Swiss-Prot (release 45.2 of November 23, 2004) and TrEMBL (release 28.2 of November 23, 2004), and restricting the database search to human taxonomy. Positive identification was accepted only when the data satisfied the following criteria: (i) MS data were obtained for a full-length y-ion series of a peptide comprising at least 8 amino acids and no missed cleavage; (ii) MS data with 1 or 2 missing y-ions were obtained for 2 or more different peptides comprising at least 8 amino acids and no missed cleavage.

Results

Differential analysis of human uveal melanoma primary cell culture (UM-A < 7) and the resulting cell line (UM-A > 7)

We observed previously that cells from the UM-A primary culture became phenotypically different from the cells of origin after passage 7 (UM-A > 7).10 These cells were less firmly attached, exhibiting more rounded, refractile cellular bodies and an accelerated growth rate (Fig. 1a). In this study, we demonstrate that UM-A > 7 cells show a higher invasion potential in Matrigel assays (Fig. 1b), and contrary to the primary cell culture of origin (UM-A < 7), these cells are characterized by a highly phosphorylated form of the antitumoural protein Rb (Fig. 1c).

Figure 1.

Different in vitro behaviour of the human uveal melanoma primary cell culture UM-A < 7 and the resulting cell line UM-A > 7. (a) Phenotype of the UM cultures UM-A < 7 and UM-A > 7. (b) Matrigel invasion assay. (c) Rb phosphorylation status on UM-A < 7 and UM-A > 7.

On the basis of a recently performed proteome study of the UM-A primary cell culture, which revealed several candidates as potential cancer markers,11 we present here the proteome differences between UM-A < 7 and UM-A > 7 samples that may give relevant clues to further understand disease processes in these tumour cells. Comparing 2D gels of UM-A < 7 and UM-A > 7 cell lysates (Fig. 2a), we found that most of the protein features identified in the UM primary cell culture proteome11 were also present in the derived cell line (UM-A > 7). These include spots containing proteins related to the oncogenesis process, such as oncogene DJ-1, oncoprotein 18/Stathmin and tumour-rejection Ag, among others (Fig. 2b). However, we were specifically interested in a differential analysis between both sets of gels, focusing on the identification of disappearing and appearing spots, and at least a 2-fold up- or downregulation of spot intensities (with p < 0.05). Approximately 85% of protein features were present in at least 3 of the 4 gels per group (UM-A < 7 or UM-A > 7), a requirement for being considered for the analysis. These strict criteria for the inclusion of proteins into the analysis were used to avoid misidentifications due to gel-to-gel variations. Applying these criteria, 290 differentially regulated protein features from ∼1,100 spots were detected in the pI 3–10 analytical range. Ninety percent of the differentially expressed proteins were successfully identified by liquid chromatography and mass spectrometry (see Supplementary Table). From those, 41% were present only in the cell line (UM-A > 7), 31% were present only in the primary culture (UM-A < 7), 12% were upregulated in UM-A > 7 and 10% upregulated in UM-A < 7.

Figure 2.

Differential proteome analysis of human uveal melanoma primary cell culture (UM-A < 7) and the resulting cell line (UM-A > 7). (a) Representative images of the pI 3–10 proteome coverage 2DE map of UM-A < 7 and UM-A > 7. (b) Synthetic 2DE image representing all protein features present in the analysis. The location of some relevant cancer-related proteins on the UM 2D-proteome map is shown. Proteins present only in UM-A < 7 are shown in bold characters and proteins present only in UM-A > 7 are shown in italic; all other proteins are present in both cultures.

The differential analysis showed that UM-A < 7 contained increased levels of actin-related and actin-binding proteins compared to that UM-A > 7 cells contained. However, cytokeratines 8 and 18, together with vimentin, a combination characteristic of uveal melanoma tumours, were present in both cell cultures (see Supplementary Table). Interestingly, several features corresponding to proteins implicated in cell adhesion present in the primary culture disappeared in the cell line, e.g., vinculin and ezrin. The UM-A > 7 culture was characterized by the presence of proteins related to the disassembly of actin filaments (e.g., macrophage capping protein, WD-repeat protein) and also by various forms of tubulins. Furthermore, cellular metabolism seems to be increased in UM-A > 7 cells, as reflected in the appearance of new spots related to glycolysis enzymes (e.g., alpha enolase, fructose-biphosphate aldolase A-C, gamma-enolase, phosphoglycerate kinase, glyceraldehide 3 phosphate dehydrogenase) and energy metabolism (e.g. ATP synthases). Other spots containing nuclear proteins decreased their expression or disappeared in UM-A > 7 culture, such as CENP-F, Oncogene FUS and nuclear protein Hcc, whereas new spots were detected in UM-A > 7, such as BRCA-1, HMG-1, nuclear protein SkiP, protein associated with myc and new isoforms of nucleophosmin (Table I, Fig. 2b). In addition, various annexins were present in the differential analysis, since the corresponding spots were either absent or less intense in the cell line compared to that in the primary cell culture (e.g., annexin A1, A2, A5, A6). Similarly, LIM and SH3 domain protein 1, Rho GDP dissociating inhibitor, tumour protein D54 and S100A6 and A6 were identified only in spots from early passages of the culture (Table I, Fig. 2b and Supplementary Table). Interestingly, other new proteins like HS1-binding protein, related to the promotion of cell survival, CTCL tumour antigen se33-1, pelota homolog, new isoforms of GTP-binding nuclear protein RAN, HSP60β and the cell adhesion protein MUC18 were identified as exclusive features present in the UM cell line (Table I, Fig. 2b and Supplementary Table). The location of some relevant cancer-related proteins on the UM 2D-proteome map is shown in Figure 2b.

Table I. Relevant Cancer-Related Proteins Identified in the UM-A < 7 Versus UM-A > 7 Differential Analysis
Protein nameProtein IDMWpIFoldBiological function
  1.  Note: Protein ID, Swiss-Prot accession number. MW and pI, molecular weight and isoelectric point for identified protein features. Fold represents fold change (UM-A < 7 versus UM-A > 7); a negative value indicates that the feature is present in a higher extend in the UM-A < 7; em, spot only present in UM-A < 7 and lm, spot only present in UM-A > 7. Biological function based on Swiss-Prot and PudMed databases.

Proteins present or over-expressed in UM-A<7
 Annexin A1P0408332,1536.03emRegulation of membrane trafficking and cellular adhesion.
31,5015.75em
 Annexin A2P0735536,0616.27−2.05Plays a role in the regulation of cellular growth and in signal transduction pathways.
35,8646.47em
25,5925.41em
28,5126.97−2.08
 Annexin A5P0875833,4065.27−2.16Potential role in cellular signal transduction, inflammation, growth and differentiation.
 Annexin A6P0813379,5215.76−2.09Implicated in membrane-related events along exocytotic and endocytotic pathways.
 Calcyclin (S-100A6)P0670310,5425.12emInvolved in calcium signalling. Loss of expression in advanced stage cancers.
10,4775.19em
 Cathepsin B precursorP0785825,7845.34emParticipates in intracellular degradation and turnover of proteins. Related to tumour invasionand metastasis.
33,4065.27−2.16
 CENP-F kinetochore proteinP4945479,5215.76−2.09Chromosome segregation during mitosis. Interacts with retinoblastoma protein (Rb), CENP-E and BUBR1.
 EzrinP1531181,1616.07−3Cell adhesion, mobility and cell survival might participate in tumour progression.
 LIM and SH3 domain 1Q1484736,0616.27−2.05Focal adhesion protein necessary for cell migration and survival.
35,8646.47em
 Multidrug resistance  associated protein MGr1-AgP0886534,7084.66emBelongs to the S2P family ofribosomal proteins.
 Nuclear protein Hcc-1P8297932,1536.03emTranslational control of cell growth, metabolism and carcinogenesis.
 Oncogene FUSP3563762,8745.98emMay play a role in maintenance of genomic integrity.
 Rho GDP-dissociation inhibitor 1P5256523,2075.03−3.02Belongs to the Rho GDI family.
 Rho GDP-dissociation inhibitor 2P5256626,3675.15−2.05Belongs to the Rho GDI family.
 S100 calcium-binding protein A16Q96FQ69,0966.37emBelongs to the S-100 family. Up-regulated in tumours.
 Tumour protein D54O4339928,6695.24emMay play roles in calcium-mediated signal transduction and cell proliferation.
30,1135.11em
30,1135.07em
 UV excision repair protein RAD23 homolog BP5472755,4314.86−2.36Involved in DNA excision repair.
 VinculinP18206149,3905.92emInvolved in cell adhesion. May allow cancer cells to move awayfrom tumours.
135,2735.94em
131,4136.09−2.09
134,7156.21em
Proteins present or over-expressed in UM-A>7
 Breast cancer type 1 susceptibility proteinP3839847,5236.15lmPlays a central role in DNA repair by facilitating cellular response to DNA repair.
 CTCL tumour antigen se33-1Q9H2G132,8087.04lmTumour-associated antigen found in several cutaneous T-cell lymphomas (CTCL).
 DNA excision repair protein ERCC-6Q0346817,7466.32lmIs involved in the preferential repair of active genes.
 GTP-binding nuclear protein RANP6282625,4186.91lmRequired for the import of proteins into the nucleus and also for RNA export. Involved in chromatin condensationand control of cell cycle.
 Heat shock protein HSP 90-βP0823835,9205.31lmMolecular chaperone. Related to various cancers.
108,6465.18em
 Heterogeneous nuclear  ribonucleoproteins A2/B1P2262631,0217.112.04Pre-mRNA processing.
 High mobility group  protein 1 HMG-1P0942928,2586.33lmHas a role in the transcription of many genes involved at different steps in the metastatic cascade and has beenlinked with cancer in human and animal models.
 HS1-binding proteinO0016532,5464.86lmMay function in promoting cell survival.
 Macrophage capping proteinP4012140,4996.03lmHelps to modify actin structures in response to external signals, which permit rapid changes in shape during development.
42,0395.482.31
46,5216.89lm
39,0155.60lm
14,2626.42lm
 Melanoma-associated antigen MUC18P4312136,5976.31lmMay allow melanoma cells to interact with cellular elements of the vascular system thereby enhancing hematogeneoustumour spread.
 Nuclear protein SkiPQ1357372,6105.74lmInteracts with the ski oncogene.
 NucleophosminP0674835,8255.11lmMay function in the assembly or transport of ribosomes.
34,8824.73lm
 Pelota homologQ9GZS644,3415.98lmPossibly participates in cell cycle regulation.
 Protein associated with MycO7559234,2376.31lmInteracts directly with the transcriptional activating domain of the oncogene–Myc.
 Uveal autoantigenQ9BZF945,6605.41lmMyosin heavy chain with Ankyrin repeat.

Validation of potential markers for uveal melanoma: Role of HMG-1 and MUC18 in in vitro cell invasion assays

We analysed further the potential biological relevance of some interesting proteins identified in the differential proteome analysis. Because of our interest in studying the dissemination of cancerous cells in UM patients, we focused on the expression of 2 metastasis-associated proteins, the gene expression-regulating protein HMG-1 and the cell adhesion protein MUC18 (Fig. 3a), by validating their presence in whole cell extracts by Western blot (Fig. 3b). To eliminate the possibility of cell line-specific expression, we included standard cell lines established by other groups used for the study of uveal melanoma, namely UW-1, SP6.5, OCM-1 and 92.116 and a primary cell culture of NM established previously by our group.10 HMG-1 protein was observed at higher levels in the UM-A cell line than in the primary cell culture, and overall, was overexpressed in all the UM cell lines assayed (Fig. 3b). In the case of MUC18, immunodetection showed its presence at low levels in UM-A < 7, UW-1 and 92.1; it was overexpressed in UM-A > 7, OCM-1 and SP6.5 and was not present in NM (Fig. 3b). Matrigel invasion assays showed a correlation between the invasion potential and MUC18 expression in the UM cultures (Fig. 3c). This experiment showed a large difference in the invasion potential between the UM primary culture (20%) and the cell line (70%), paralleling the expression levels of MUC18 in these cells (Figs. 3b and 3c). The same correlation was found with UW-1, OCM-1 and SP6.5, where expression levels of MUC18 was correlated with the invasion potential; however, the UM cell line 92.1 did not show this pattern even though the invasion potential was similar to OCM-1 (Figs. 3b and 3c). Although the levels of HMG-1 were high in all the UM cultures when compared to the normal cells, a clear correlation with the invasion potential assayed in Matrigel on these cells was not found (Figs. 3b and 3c).

Figure 3.

Validation of potential invasion markers for uveal melanoma. (a) Zoom images from 2DE gels showing the differentially regulated proteins HMG-1 and MUC18 in UM-A cultures. (b) Representative Western blot images showing the levels of expression of HMG-1 and MUC18 in NM, UM-A < 7, UM-A > 7 and other UM cell lines (UW-1, OCM-1, SP6.5, 92.1). Equal protein loading was confirmed by measuring the amount of actin in the different cell extracts. (c) Graph showing the percentage of UM cells able to migrate through Matrigel matrix and relative to the migration through a control membrane without ECM.

Validating the presence of DJ-1 oncogene in uveal melanoma: DJ-1 as a potential biomarker in uveal melanoma patients' sera

DJ-1 oncogene was not differentially regulated between UM-A < 7 and UM-A > 7 (Fig. 4a). However, we found that DJ-1 was overexpressed in all the UM cell lines, including the UM primary cell culture, when compared to NM (Fig. 4b). It has been described that this protein can be secreted from malignant cells;17 therefore, we studied whether this was the case in our UM model. We analysed the presence of DJ-1 in the culture medium of UM-A and other UM cell lines by Western blot. All these cultures except the NM secreted DJ-1 into the cell culture media (Fig. 4b). This result indicated that DJ-1 may be secreted by tumour cells into the blood stream; and indeed, preliminary trials revealed its presence in certain UM patients' sera (data not shown).

Figure 4.

Validation of the presence of DJ-1 oncogene as a potential biomarker in uveal melanoma. (a) Zoom images from 2DE gels showing spots containing DJ-1 protein. (b) Representative Western blot images showing the levels of expression and the secretion into the cell culture media of DJ-1 for NM, UM-A < 7, UM-A > 7 and other UM cell lines (UW-1, OCM-1, SP6.5, 92.1). Equal protein loading was confirmed by measuring the amount of actin in the different cell extracts.

Discussion

Differential proteome analysis of human uveal melanoma primary cell culture (UM-A < 7) and the resulting cell line (UM-A > 7)

To unravel alterations related to tumour progression, proteomics has recently attracted great attention, because it allows the identification of qualitative and quantitative changes in protein composition, including posttranslational modifications, when comparing healthy and cancerous cells. Recently, Hayashi et al. proved the value of this approach, comparing regressive and progressive cancer cell lines, for the identification of molecular abnormalities in tumour progression.18 Molecular determinants for UM have been studied at the genomic level from 2 UM cell lines with different invasion potentials.19 However, to our knowledge, the comparison of UM cell lines by proteomics has not been performed to date.

We have successfully established a UM primary cell culture (UM-A) that proved to be a valuable tool for the study of this neoplasia in vitro.10, 11 Seike et al. studied different cancer cell lines and found that most tissue-cultured cells retain characteristics that reflect their in vivo origin and differentiation phenotype after long-term culture.20 This is also consistent with previous reports that tissue-cultured cells retain the gene expression signature of their original tissues.21, 22 Taking into account the low incidence of UM (7–10 per million per year) and that only about 5% of patients with newly diagnosed disease undergo enucleation, the acquirement of significant numbers of fresh tissue samples for the establishment of new primary cultures has been traditionally problematic in the study of UM.23 Mainly for this reason, the molecular signature of UM is still very poorly known, and may explain why the survival rate of patients with UM has not paralleled that of other cancer types, with no improvement made in the last decades.24

Here we continue a previous study whose aim was to better characterize UM-A, by studying the molecular progression from the primary cell culture to cell line after 7 passages in vitro. The occurrence of differences between primary cell cultures and cell lines is generally known, but very few reports have studied these changes in more depth.25, 26 Interestingly, invasion assays with both UM-A cultures showed a significant increase in the percentage of cells able to degrade the ECM in the UM cell line (UM-A > 7) when compared to the primary culture (UM-A < 7). Taking into consideration that more UM tumour samples and cultures need to be analysed, and the difficulty in linking the enormous amount of information yielded by proteomics, we believe that this analysis may give relevant clues about disease stages, including dissemination and metastasis.

Differentially expressed proteins between the UM-A primary cell culture and the resulting cell line correlate well with differences observed when culturing these cells in vitro. The increased amount of mitochondrial, protein processing and general metabolism proteins observed in the cell line explains their higher proliferation rate and consequently more active cell metabolism compared to that of the cells of origin. The fact that the Rb antitumoural protein is inactivated by phosphorylation in the UM-A > 7 also reflects the increment on cell division of these cells. In agreement with the differences found in the invasion potential of UM-A cultures, we found that some proteins related to cell adhesion were shown to be present only in the UM primary cell culture, while various actin-disassembling features were present exclusively in the cell line. In addition, several of the identified protein features correlated well with proteins described previously at different stages of cancer. The presence of spots containing Annexin A1, A5 and A6 was significantly decreased in the UM-A cell line, as described previously in various oncogenesis processes, including skin melanoma.27, 28, 29 The differentially expressed proteins Rho GDPI 1 and 2, mostly present in the UM primary cell culture, also correlated well with previous reports involving the downregulation of those proteins in late stages of cancer.30, 31

On the other hand, we found proteins previously linked to tumour development or related to cell migration and invasion that were upregulated or exclusively present in the primary cell culture. Interestingly, new proteins were differentially present in the UM cell line like HMG-1, nuclear protein SkiP, and hnRNPA2/B1. This phenomenon may be related to new chromosome aberrations and to the cell selection and dedifferentiation processes inherent in the establishment of cell lines in vitro. However, both UM-A primary cells and the resulting cell line showed the presence of cytokeratins 8 and 18 together with vimentin, constituting the interconverted phenotype characteristic of uveal melanoma tumours.32

HMG-1 and MUC18 as potential invasion markers in uveal melanoma

The presence of a new isoform of the non-histone chromosome-binding protein HMG-1 in the UM cell line is interesting. We described previously the presence of HMG-1 in the UM-A primary cell culture, which coimmunoprecipitated with the hypophosphorylated form of the antitumoural protein Rb;10 however, the role of HMG-1 in UM is still unknown. We validated the presence of HMG-1 and found high levels of this protein in UM culture extracts. Most normal, differentiated mammalian cells express extremely low levels of HMG-1. Only cells experimentally transformed and many neoplasms are characterized by high levels of this protein.33, 34 In addition, HMG-1 is involved in the transcriptional regulation of a number of genes reported to play key roles in different biological processes of neoplasm progression and metastasis.35 According to our results, HMG-1 may play a role in UM, even though we could not show a clear relationship between HMG-1 levels in UM cultures and its invasion potential in Matrigel membranes. This result agrees with those reports that do not clarify whether HMG-1 has a metastasis-associated or metastasis-inducing role.34

MUC18 (also named MelCAM/CD146/S-Endo 1-associated antigen), a cell adhesion molecule (CAM) belonging to the immunoglobulin superfamily, is constitutively expressed in the whole human endothelium.36, 37, 38 MUC18 is upregulated during skin melanoma development in a stepwise fashion and coincides with the separation of nevus cells from keratinocytes.39 Despite originating from common precursor cells, UM and cutaneous melanomas exhibit several notable differences in tumour biology and behaviour, including their molecular signatures.40 However, similar to skin melanoma, we found spots containing MUC18 in the UM-A cultures and also the expression of this protein in these and other UM cell lines. A positive association between the loss of the adhesion protein ICAM-1 expression and increased risk of metastasis has been suggested in UM;41 but to our knowledge, no relationship has been described to date between UM and MUC18. The expression of MUC18 is limited not only to melanoma neoplasias, but it has also been implicated with an increase of metastasis potential in human prostate and bladder cancer cells.42, 43 In agreement with these data, we also show a correlation between MUC18 expression in UM-A cultures and their invasion potential in vitro. This correlation was also found in all the UM cell lines except 92.1, which might suggest that more than one molecular event govern UM invasion. More research is necessary to determine if the role of MUC18 in UM agrees with previous data suggesting that MUC18 may play a major role in metastasis by mediating not only melanoma cell–cell interactions, but also melanoma–endothelial cell adhesion.44 This process can be reverted in animal models by the use of fully human antibodies against MUC18.45 Thus, the possibility of using this strategy in UM may be considered in the future. To follow up this further and to correlate it with clinical parameters, we are in the process of immunodetecting MUC18 in UM tumour sections (not shown).

DJ-1 oncogene as a potential serum marker in uveal melanoma

Finally, we show the expression of DJ-1 protein in both UM-A cultures and in different UM cell lines as well as its secretion to the cell culture media. Preliminary results show that DJ-1 is also detectable in the serum of certain UM patients (data not shown). DJ-1 was originally cloned as a putative oncogene capable of transforming NIH-3T3 cells in cooperation with H-ras.46 It has also been implicated in fertilization, the regulation of androgen receptor signalling and oxidative stress.47, 48, 49 Further, mutations of the DJ-1 gene are associated with autosomal early-onset Parkinson's disease.50 Several lines of evidence suggest that DJ-1 plays a role in human tumourigenesis, including breast cancer, nonsmall cell lung carcinoma and prostate cancer.17, 51, 52 Very recently, DJ-1 was identified as a negative regulator of the tumour suppressor PTEN, promoting cell survival in primary breast and lung cancer patients.53 The role of a soluble form of DJ-1 is still unknown, but it has been described as a circulating tumour antigen in serum from 37% of newly diagnosed patients with breast cancer.17 Our preliminary data detecting DJ-1 in UM patients' sera is now being analysed further by including a greater number of patients and correlating the presence of DJ-1 with the clinical data. The presence of biomarkers in serum samples, indicating the potential for metastasis at the time of diagnosis, would be very useful in UM, particularly since occult micrometastases are thought to be present when the primary tumour manifests.2 There is a need to distinguish 2 well-defined tumour groups previously described in UM, presenting high and low risk of metastasis.5, 54 Crucial questions now are how large a tumour mass has to be in order to be detectable by serum screening, and how does the sensitivity of this approach compare with imaging methods.

In summary, we present here the first comparative analysis of 2 UM cell cultures with different invasion potentials. This approach is a valuable tool for the study of the molecular events governing UM development and for the identification of potential markers. As a result of this study, we suggest the involvement of the cell adhesion protein MUC18, and to a lesser extent HMG-1 protein, in the invasion potential of UM cells. We also describe for the first time the overexpression of the oncogene DJ-1 in UM and initial results showing its implication as a potential serum biomarker for this malignancy.

Acknowledgements

The authors thank Prof. Capeans-Tomé (Unidad de Oncoloxía Ocular, Servicio de Oftalmoloxía, Complexo Hospitalario Universitario de Santiago de Compostela, Spain) for her support in obtaining the UM samples. They further thank Dr. Blanco for kindly selecting the UM patients and providing the blood samples, and Dr. de la Fuente for obtaining the serum. The UM cell lines (UW-1, SP6.5, OCM-1 and 92.1) were kindly donated by Prof. Burnier (The Henry C. Witelson Ocular Pathology Laboratory, McGill University, Montreal, Canada). N.Z. is a Senior Research Fellow at Linacre College, Oxford.

Ancillary