Proteomics of Uveal Melanomas Suggests HSP-27 as aPossible Surrogate Marker of Chromosome 3 Loss

Sarah E. Coupland,1,2 Henrik Vorum,2,3,4,5 Nakul Mandal,4 Helen Kalirai,1 Bent Honore,4

Steen Fiil Urbak,3 Sarah L. Lake,1 Justyna Dopierala,1 and Bertil Damato6

PURPOSE. To compare the proteomic profiles of primary uvealmelanomas, with and without loss of chromosome 3.

METHODS. Frozen specimens from three uveal melanomas withdisomy 3 and from four tumors with monosomy 3, accordingto fluorescence in situ hybridization (FISH) analysis, were sub-jected to high-resolution, two-dimensional (2-D) gel electro-phoresis. The protein expression profiles of the two uvealmelanoma cytogenetic groups were compared: Proteins thatdiffered significantly were excised and analyzed by tandemmass spectrometry. Differentially expressed proteins were fur-ther analyzed with Western blot analysis. An independent co-hort of 41 formalin-fixed, paraffin-embedded (FFPE) uveal mel-anomas, whose chromosome 3 status had been determined bymultiplex ligation-dependent probe amplification (MLPA), wasexamined for the appropriate antigens by immunohistochem-istry.

RESULTS. Four protein spots were 1.5-fold (Student’s t-test, P �0.05) differentially expressed in the two uveal melanomatypes: two spots were overexpressed in the disomy 3 groupcompared with the monosomy 3 group, whereas two spotswere underexpressed. Identification of the four spots yieldednine proteins. Western blot analysis confirmed the results forheat shock protein (HSP)-27, vimentin, and pyruvate dehydro-genase � (PDHB), with a statistical significance for the first twoproteins. HSP-27 was significantly downregulated, whereas vi-mentin was upregulated in the monosomy 3 tumors (Student’st-test, P � 0.003 and P � 0.005, respectively). Immunohisto-chemistry confirmed low-to-negative HSP-27 protein expres-

sion in monosomy 3 uveal melanomas (Student’s t-test; P �0.011).

CONCLUSIONS. Low-to-negative HSP-27 protein expression inuveal melanoma correlates strongly with monosomy 3. Furthervalidation is necessary to determine whether immunohisto-chemical assessment of HSP-27 expression correlates with met-astatic mortality. (Invest Ophthalmol Vis Sci. 2010;51:12–20)DOI:10.1167/iovs.09-3913

Uveal melanoma causes metastatic disease in approximately50% of all patients, despite successful treatment of the

primary intraocular tumor.1 Metastatic disease usually involvesthe liver and is invariably fatal, typically within a year of theonset of symptoms.2,3 Uveal melanomas can essentially besubtyped into low- and high-grade metastatic risk groups, ac-cording to their clinical, histologic and genotypic features.2–6

Knowledge of a particular tumor’s characteristics allows for amore accurate prediction of metastasis in the individual patientand aids treatment and management decisions.2,3

Although a large amount of information is available withregard to the molecular biological differences of low- andhigh-grade uveal melanomas, little is known about the corre-sponding protein complement of the genome (proteomics) inthese melanoma subgroups. To date, proteomic analysis hasbeen conducted on uveal melanoma cell lines,7–9 serum,10 andaqueous humor11 of patients with uveal melanoma. Theseinvestigations revealed overexpression and underexpression ofsome proteins; however, no correlation was made with thecytogenetic status of the tumor cells.

The purpose of our study was, therefore, to determinewhether the proteomic profiles of monosomy 3 uveal melano-mas differed significantly from those of disomy 3 tumors and,if so, whether such differences identified possible biomarkersof the high-grade uveal melanomas.

METHODS

The study was approved by the local committee of the NationalResearch Ethics Service and informed consent was obtained from allpatients. The research adhered to the tenets of the Declaration ofHelsinki.

Proteomic Analysis

Seven uveal melanomas were selected for proteomic analysis. Thesefrozen tumor samples had been taken from uveal melanomas treated bylocal resection or enucleation in 2005 and had been selected on thebasis of cytogenetic analysis for chromosome 3 routinely performed byfluorescence in situ hybridization (FISH) at this time.2 The histomor-phologic features of these seven uveal melanomas were noted afterroutine fixation, embedding in paraffin, and staining of sections, usinghematoxylin and eosin (H&E) and periodic acid Schiff (PAS).

Protein Extraction. Frozen uveal melanoma tissue was homog-enized and dissolved in lysis buffer containing 9 M urea, 2% (vol/vol)

From the 1Department of Pathology, School of Cancer Studies,University of Liverpool, Liverpool, United Kingdom; the 3Departmentof Ophthalmology, Aarhus University Hospital, Aarhus, Denmark; the4Department of Medical Biochemistry, Aarhus University, Aarhus, Den-mark; and the 6Ocular Oncology Service, St. Paul’s Eye Clinic, RoyalLiverpool University Hospital, Liverpool, United Kingdom.

2Contributed equally to the work and therefore should be consid-ered equivalent authors.

5Present affiliation: Department of Ophthalmology, Aalborg Hos-pital, Aarhus University Hospital, Aalborg, Denmark.

Supported by research funding from The Danish Eye Health Soci-ety, The Danish Eye Research Foundation, The Danish Medical Re-search Council, the John and Birthe Meyer Foundation, the AarhusUniversity Research Foundation, the Synoptik Foundation, the BeckettFoundation, Foreningen Østifterne, and the Eye Tumour ResearchFund, Royal Liverpool University Hospital.

Submitted for publication April 25, 2009; revised July 6 and 13,2009; accepted July 13, 2009.

Disclosure: S.E. Coupland, None; H. Vorum, None; N. Mandal,None; H. Kalirai, None; B. Honore, None; S.F. Urbak, None; S.L.Lake, None; J. Dopierala, None; B. Damato, None

Corresponding author: Sarah E. Coupland, Department of Pathology,University of Liverpool, Liverpool, UK; s.e.coupland@liverpool.ac.uk.

Anatomy and Pathology

Investigative Ophthalmology & Visual Science, January 2010, Vol. 51, No. 112 Copyright © Association for Research in Vision and Ophthalmology

Triton X-100, 2% (vol/vol) immobilized pH gradient (IPG) buffer (pH3-10NL), and 2% (wt/vol) dithiothreitol (DTT). The protein sampleswere stored at �80°C. The total protein content in each melanomasample was assayed (Non-interfering Protein Assay; Calbiochem-MerckChemicals, Beeston, UK).12

Two-Dimensional Gel Electrophoresis. One-dimensionalisoelectric focusing (IEF) was performed by using nonlinear pH 3-10NLIPG strips (Amersham Pharmacia Biotech, Uppsala, Sweden). The IPGstrip was rehydrated for 20 hours at room temperature in 200 �L lysisbuffer with approximately 100 �g protein, and 150 �L rehydrationbuffer (8 M urea, 2% [wt/vol] CHAPS, 0.3% [wt/vol] DTT, and 2%[vol/vol] IPG buffer), using a dry strip tray (Immobiline DryStrip Re-swelling Tray; Amersham Pharmacia Biotech). IEF was performed onan electrophoresis unit (Multiphor II; PerkinElmer Windsor, UK) at 500V for 0.01 hour, 500 V for 5 hours, 3500 V for 5 hours, and 3500 V for9.5 hours in a gradient mode at 17°C (MultiTemp III ThermostaticCirculator; GE Healthcare Life Sciences, Buckinghamshire, UK). Beforetwo-dimensional SDS-PAGE, the IPG strip was equilibrated twice: firstfor 10 minutes under gentle agitation in 20 mL of equilibration solutioncontaining 0.6% (wt/vol) Tris-HCl (pH 6.8), 6 M urea, 30% (vol/vol)glycerol, 1% (wt/vol) SDS, and 0.05% (wt/vol) DTT; and second in 4.5%(wt/vol) iodoacetamide and bromophenol blue. For the second dimen-sion, the equilibrated IPG strip was transferred to a polyacrylamide gel.Electrophoresis was then performed vertically at a maximum voltage of50 V, 5 mA for approximately 20 hours on equipment designed andpreviously described by Vorum et al.13

Silver Staining. Gels were visualized by silver staining opti-mized for high-sensitivity protein identification by mass spectrom-etry.14 Briefly, individual gels were fixed in 50% (vol/vol) ethanol,12% (vol/vol) acetic acid, and 0.0185% (vol/vol) formaldehyde over-night. After the gels were washed three times for 20 minutes in 35%(vol/vol) ethanol and pretreated for 1 minute in 0.02% (wt/vol)Na2S2O3;5H2O, they were rinsed in water and stained in 0.2%(wt/vol) AgNO3 and 0.028% (vol/vol) formaldehyde for 20 minutes.Further rinsing with water was performed before development in6% (wt/vol) Na2CO3 and 0.0185% (vol/vol) formaldehyde, 0.0004%(wt/vol) Na2S2O3;5H2O, for approximately 3 minutes. Finally, de-velopment was arrested in a fixative solution of 40% (vol/vol)ethanol, 12% (vol/vol) acetic acid.

Image Analysis. Silver-stained gels were scanned with a densi-tometer (GS-710 Calibrated Imaging Densitometer; Bio-Rad, HemelHempstead, UK). The peptide spots were analyzed by software(PDQuest software; Bio-Rad) that designated a volume to each spotproportional to the amount of protein. All well-separated and clearlyfocused spots that were 1.5-fold (Student’s t-test P � 0.05) differen-tially expressed between the disomy 3 and monosomy 3 uveal mela-

nomas were selected for identification by liquid chromatography–tandem mass spectrometry.12

Protein Identification. The proteins were excised from thegels and subjected to in-gel digestion. The peptide samples wereanalyzed by LC-MS/MS essentially as described.12 In short, digestedpeptides were separated on an inert nano liquid chromatographysystem (LC Packings, San Francisco, CA) connected to a mass spec-trometer (Q-Tof Premier; Waters, Milford, MA), spectra were obtained(MassLynx 4 SP4; Waters), and raw data were processed (ProteinLynxGlobalServer 2.1; Waters). The processed data were used to search thetotal part of the Swiss-Prot database (release 52.4, 265,950 sequences;http://www.expasy.org; provided in the public domain by Swiss Insti-tute of Bioinformatics, Geneva, Switzerland) using the online versionof the Mascot MS/MS Ion Search facility (http://www.matrixscience-.com/ available commercially from Matrix Science, Ltd., London, UK).The search was performed with doubly and triply charged ions withtwo missed cleavages, a peptide tolerance of 50 ppm, one variablemodification, carbamidomethyl-C, and an MS/MS tolerance of 0.05 Da.Contaminating peptides, such as keratins and trypsin, cross-contami-nating peptides from previous samples, and peptides originating fromdifferent sources such as BSA, were disregarded. At least one bold redpeptide was required in the search. The scores of the protein hits wererecorded and are reported herein. Individual peptide ion scores higherthan 36 indicate identity or extensive homology, giving a less than 5%probability that the observed match was a random event.

Western Blot Analysis

Samples were loaded and run on 10% to 20% gradient Tris-Glycinepolyacrylamide gels (Novex; Invitrogen Corp.; Paisley, Scotland, UK).After transfer by electroelution to nitrocellulose membranes (Hy-bond-C Extra; Amersham, UK), the blots were blocked with 5% milk in80 mM Na2HPO4, 20 mM NaH2PO4, 100 mM NaCl, and 0.1% Tween 20(PBS-T; pH 7.5) for 1 hour and incubated overnight at 4°C with theappropriate primary antibodies (Table 1). The labeling was visualizedwith horseradish peroxidase (HRP)–conjugated secondary antibodies(P0217, P0449, or P0260; DakoCytomation, Glostrup, Denmark, andab16349-250; Abcam, Cambridge, UK) by enhanced chemilumines-cence (GE Healthcare). In an effort to conserve tissue samples, someblots were reprobed after treatment with Western blot stripping bufferaccording to the manufacturer’s protocol (Restore; Pierce Biotechnol-ogy Inc., Rockford, IL).

Immunohistochemistry

Antigens identified as being significantly differentially expressed betweenmonosomy 3 and disomy 3 tumors in the proteomic studies were further

TABLE 1. Commercial Antibodies Used in Western Blot Analysis

Antibody Name Clone/Peptide Species ManufacturerAntibody

Concentration

Heat shock protein 27 Polyclonal; C terminus Goat Santa Cruz Biotechnology, Santa Cruz, CA 1:100Vimentin Polyclonal; C terminus Rabbit Santa Cruz Biotechnology 1:100Pyruvate dehydrogenase � Monoclonal; partial recombinant Mouse Abnova Corp., Taipei, Taiwan 1:100�-Actin Polyclonal; aa 151–320 Chicken Abcam, Cambridge, UK 1:500Glyceraldehyde-3-phosphate

dehydrogenasePolyclonal; internal Goat Santa Cruz Biotechnology Inc. *

Glyceraldehyde-3-phosphatedehydrogenase

Polyclonal; C terminus Rabbit GenWay Biotech Inc., San Diego, CA *

Fumarate hydratase Polyclonal; recombinant Rabbit Proteintech Group Inc., Chicago, IL *�-Tubulin Polyclonal; N terminus Goat Santa Cruz Biotechnology *Tropomyosin Polyclonal; aa 1-284 Rabbit Santa Cruz Biotechnology *Lysozyme C Polyclonal; Internal Goat Santa Cruz Biotechnology *F-actin capping protein � subunit

(CAPZA 1)Polyclonal; aa 1-286 Chicken Abcam *

aa, amino acid.* No detectable reaction.

IOVS, January 2010, Vol. 51, No. 1 HSP-27 and Uveal Melanoma 13

investigated by immunohistochemistry applied to an independent cohortof 41 uveal melanomas. These 41 tumors were selected on the basis oftheir chromosome 3 status, as determined for prognostication purposesby MLPA4 in a fresh biopsy taken directly after enucleation or tumorexcision between 2006 and 2008.

After conventional staining of these 41 uveal melanomas with H&Eand PAS, immunohistochemistry was performed on 4-�m sections madefrom the corresponding formalin-fixed, paraffin-embedded (FFPE) tissueblocks. Sections were dewaxed, rehydrated, and endogenous peroxidaseblocked with H2O2 in methanol. To detect HSP-27, we also subjected theslides to heat-induced epitope retrieval (pH 7.0) before incubation withthe appropriate antibody. After they were rinsed in running water, slideswere loaded onto an autostainer (DakoCytomation) and washed withphosphate-buffered saline (PBS). The sections were incubated with pri-mary antibody for 40 minutes at room temperature (RT): mouse anti-human HSP-27 antibody, diluted 1:100 (Novocastra, Newcastle, UK), andmouse anti-human vimentin antibody, 1 �g/mL (DakoCytomation). Boundantibody was detected with an HRP system (Advance; DakoCytomation)according to the manufacturer’s instructions. Peroxidase was visualizedwith an AEC substrate kit (Vector Laboratories Ltd, Peterborough, UK),and the sections were counterstained with Mayer’s hematoxylin and

mounted in aqueous medium (Aquatex; Merck Chemicals, Ltd., Notting-ham, UK). Negative control experiments were performed by replacing theprimary antibody with mouse isotype control antibody. Normal humanskin and tonsil were used as positive control tissues for HSP-27 andvimentin antibodies, respectively.

The percentages of melanoma cells staining for HSP-27 and vimen-tin were evaluated by two independent observers who were unawareof the chromosome 3 status. These percentages were scored as fol-lows: 0% tumor cells positive (score 0), 1% to 24% tumor cells positive(score 1), 25% to 49% tumor cells positive (score 2), 50% to 74% tumorcells positive (score 3), and 75% to 100% tumor cells positive (score 4).The intensity of cellular staining was also scored as none (0), weak (1),moderate (2), or strong (3). A combined score was thus obtained bymultiplying the two scores, as previously described in other tumors,15

giving a possible maximum score of 12.

Multiplex Ligation-Dependent ProbeAmplification (MLPA)

The genomic abnormalities of the independent cohort of 41 uvealmelanomas were determined with MLPA, as described previously.4,16

FIGURE 1. Two-dimensional poly-acrylamide gel analysis. Histogramsshow the densitometry measure-ments of the significantly up- ordownregulated spots. Blue: disomy3; red: monosomy 3.

TABLE 2. Clinicopathologic Data of Patients from the Proteomic Analysis

Sample ID Sex Age (y)CellType LBD (mm) Height (mm)

Mitosis(no./40 HPF)

ClosedLoops

Chromosome3 Status*

IOM001 F 56 E 15.8 13.7 1 N DIOM002 F 30 E 15.9 10.1 9 N DIOM006 F 81 E 11.4 7.6 2 N DIOM003 M 70 S 19.2 9.2 7 Y MIOM004 F 42 E 19.2 12.7 28 N MIOM005 F 78 E 20.0 9.7 2 Y MIOM007 M 49 E 17.5 7.5 6 N M

E, epithelioid cell type; S, spindle cell type; HPF, high power field; N, no; Y, yes; D, disomy 3; M, monosomy 3.* Chromosome 3 status as determined using FISH.

14 Coupland et al. IOVS, January 2010, Vol. 51, No. 1

MLPA replaced FISH as the diagnostic test used to identify chromo-somal aberrations at the Royal Liverpool Hospital Ocular OncologyCentre in 2007, since, as detailed in our recent report,4 it accuratelydetects monosomy 3 with a greater sensitivity than FISH.

Statistical Analysis

Differences in the protein expression levels for HSP-27 and vimentinwere examined with a Student’s t-test between the uveal melanomasubgroups (i.e., monosomy 3 versus disomy 3). This test was also usedfor any correlations with histomorphologic features (e.g., cell type andpresence or absence of closed loops). Correlations between the pro-tein expression levels for HSP-27 or vimentin and age at primarydiagnosis, largest basal diameter (LBD), tumor height, and mitotic

frequency were determined with the bivariate Spearman’s correlation(all statistical analysis by SPSS, ver.11.0; SPSS Science, Chicago, IL).

RESULTS

Differential Protein Expression According to thePresence or Absence of Monosomy 3

The proteomic profiles of three disomy 3 uveal melanomaswere compared with those of four monosomy 3 tumors. Theclinical, histopathologic, and cytogenetic features of these tu-mors are summarized in Table 2. Two protein spots wereoverexpressed in the disomy 3 melanomas compared with themonosomy 3 tumors, whereas two spots were underexpressed

FIGURE 2. Western blot analysis ofuveal melanoma tissue: three tumorswith disomy 3 and four tumors withmonosomy 3. (A) Blots were incu-bated with commercial antibodiesagainst the appropriate antigens.�-Actin was used as a loading con-trol. (B) Mean densitometry (opticaldensity � area) of the antibody–anti-gen reaction shown on the blots. Stu-dent’s two-tailed t-test: *Only the twostrongest visible reactions in thePDHB disomy 3 were used in thecalculations.

TABLE 3. Mass Spectrometric Analysis of Differentially Expressed Protein Spots

SpotM3/D3Ratio ID MW Score

0306 3.08 Tropomyosin �-4 chain TPM4_HUMAN 28.5 551Vimentin VIME_HUMAN 53.6 343Lysozyme C LYSC_HUMAN 16.5 75

3410 0.53 Tubulin TBAK_HUMAN 50.1 526Pyruvate dehydrogenase El component � subunit

(mitochondrial)ODPB_HUMAN 39.2 126

F-capping protein subunit �-1 CAZA1_HUMAN 32.9 365213 0.47 Heat-shock protein �-1 (HSP 27) HSPB1_HUMAN 22.8 5728006 2.17 Glyceraldehyde-3-phosphate dehydrogenase

(GAPDH)G3P_HUMAN 36.0 164

Fumarate hydratase (mitochondrial) FUMH_HUMAN 54.6 41

IOVS, January 2010, Vol. 51, No. 1 HSP-27 and Uveal Melanoma 15

(Fig. 1). In four spots, tandem mass spectrometry identifiednine proteins, as indicated in Table 3. Western blot analysis ofthe protein extracts showed significant differences for HSP-27and vimentin but not pyruvate dehydrogenase � (PDHB; Fig. 2)between the uveal melanoma groups. Compared to disomy 3uveal melanomas, monosomy 3 tumors showed downregula-tion of HSP-27 and upregulation of vimentin (Fig. 2; Table 4).

Expression of HSP-27 and Vimentin Protein inSections of Uveal Melanoma

HSP-27 and vimentin protein expression were determined byimmunohistochemistry in FFPE sections of an independentcohort of 41 uveal melanomas treated between 2006 and 2008,either by local resection or enucleation. Twenty-one of thesetumors had been classified as disomy 3 and 20 as monosomy 3after MLPA analysis of the corresponding diagnostic frozentumor specimen.

HSP-27 Immunoreactivity. The intrinsic positive controlfor HSP-27 in the enucleated eyes was the pigmented andnonpigmented ciliary body epithelium, which demonstratedcytoplasmic staining, whereas the intrinsic negative controlsincluded macrophages and lymphocytes within the tumors.HSP-27 staining was also cytoplasmic in the uveal melanomacells. The Mayer’s hematoxylin counterstain facilitated assess-ment of the cellular morphology, and, therefore, determination

of immunopositivity in either reactive or tumor cells. HSP-27positivity ranged from 0% to 100% (mean, 64.0%) with tumorcells displaying various degrees of intensity between speci-mens (Fig. 3A–D; Table 5). HSP-27 staining also showed someintratumoral variation—for example, tending to be stronger inmelanoma cells surrounding blood vessels and showing moreheterogeneity in areas of epithelioid cells (Fig. 3).

Vimentin Immunoreactivity. An intrinsic positive con-trol for vimentin in the enucleated eyes was the retina, dem-onstrating positivity within the Muller cells and similar supportcells. Vimentin staining was cytoplasmic and ranged from 10%to 100% (mean; 76.0%) within the tumor cells. As with HSP-27staining, vimentin protein expression showed some intratu-moral variation (Fig. 4).

Correlation of HSP-27 and Vimentin ProteinExpression with Clinicopathologic andCytogenetic Risk Factors for DiseaseProgression and Metastasis

Clinicopathologic information for all patients included in thisstudy is detailed in Table 5. Briefly, these 41 patients with uvealmelanoma had a mean age of 60.9 years (range, 34.0–81.0) atprimary treatment. All patients had been treated by enucleationor local resection (i.e., no uveal melanoma had undergone

TABLE 4. Western Blot Densitometry

Antibody

Disomy 3Mean Volume

(ODxArea)

Monosomy 3Mean Volume

(ODxArea)P

(Student’s t-test)

Factor Up-/Down-regulation

Monosomy 3

Vimentin 5.82 12.20 0.005 X2.1 UPHSP-27 10.51 6.67 0.003 X1.6 DOWNPDHB 3.31* 2.53 0.097 X1.3 DOWN

* Only the two strongest visible reactions in the PDHB disomy 3 blot were used in the calculations.

FIGURE 3. HSP-27 immunoreactivityin FFPE uveal melanomas. (A) The tu-mor cells, which are of the epithelioidtype, demonstrated a strong and ho-mogenous positivity for HSP-27. (B–D)Heterogeneous immunoreactivity forHSP-27 in uveal melanomas, depend-ing on cell type, admixed reactivecells, and distance from blood vessels;(B) mixed-cell uveal melanoma; (C) ep-ithelioid cell type uveal melanomawith admixed melanomacrophages;and (D) spindle cell type uveal mela-noma. Magnification: (A) �40 objec-tive; (B, D) �20 objective; (B inset; C)�60 objective.

16 Coupland et al. IOVS, January 2010, Vol. 51, No. 1

radiotherapy). The tumors had a mean largest basal diameter of16.2 mm (range, 7.8–22.1) and a mean height of 8.8 mm(range, 1.9–15.7). Histologic examination, as described previ-ously,2 showed epithelioid cells in 14 tumors and closed PAS�loops in 21 tumors, with the mitotic rate exceeding 5 per 40high-power fields in 20 tumors.

Statistical analysis of the HSP-27 protein expression score(i.e., percentage of tumor cells stained multiplied by intensityscores) was significantly lower in monosomy 3 uveal mela-noma when compared with disomy 3 tumors (Student’s t-test;P � 0.011; Fig. 5A). This remained statistically significant evenif only the percentage of tumor cells stained was tested (Stu-dent’s t-test; P � 0.013; Fig. 5B). Neither the vimentin proteinexpression score nor the percentage of tumor cells stainedshowed significant differences between monosomy 3 and dis-omy 3 tumors (Student’s t-test; P � 0.64 and P � 0.60, respec-tively).

With respect to the histomorphologic features, statisticalanalysis of the HSP-27 protein expression score in uveal mela-

nomas showed a negative correlation with mitotic frequency(Spearman’s correlation; P � 0.014). The vimentin proteinexpression score was significantly higher in uveal melanomaswith closed loops than in those without (Student’s t-test; P �0.011). In addition, a negative correlation was observed be-tween the vimentin protein expression score and LBD (Spear-man’s correlation; P � 0.043). Neither HSP-27 nor vimentinimmunoreactivity showed a significant association with anyother clinicopathologic variable tested (i.e., age of patient atprimary diagnosis, cell type, or tumor height).

DISCUSSION

The present study demonstrates a statistically significant asso-ciation between chromosome 3 loss in uveal melanomas andthe downregulation of HSP-27 (P � 0.011). This associationwas demonstrated with both proteomics and immunohisto-chemistry. To our knowledge, this correlation has not been

TABLE 5. Clinicopathologic Parameters and HSP-27 and Vimentin Protein Expression Data

SampleID

Age atPrimary

Diagnosis(y)

CellType

LBD(mm)

Height(mm)

MitoticFrequency

(No./40HPF)

ClosedLoops

Chromosome3 Status(MLPA)

HSP-27 Vimentin

%TumorCells

Stained

Intensityof

TumorCells

StainedStaining

Score

%TumorCells

Stained

Intensityof

TumorCells

StainedStaining

Score

1 44 S 17.3 11.0 2 N D 20 Moderate 2 55 Weak 32 72 S 12.2 11.1 14 N D 15 Moderate 2 50 Moderate 63 80 S 15.1 10.7 2 Y D 90 Moderate 8 40 Moderate 84 63 S 16.1 7.5 7 Y D 55 Moderate 6 65 Moderate 65 69 S 12.0 3.7 3 N D 100 Moderate 8 10 Moderate 26 62 E 14.4 6.4 4 Y D 90 Strong 12 90 Strong 127 52 S 21.2 10.8 4 Y D 40 Moderate 4 85 Strong 48 40 S 17.5 7.9 7 N D 60 Moderate 6 90 Weak 49 74 S 7.8 1.9 1 N D 100 Moderate 8 90 Weak 4

10 81 S 15.3 9.5 4 N D 50 Moderate 6 75 Weak 411 42 S 16.7 9.1 1 N D 65 Strong 9 80 Weak 412 34 S 9.6 6.9 7 N D 90 Moderate 8 95 Moderate 813 80 S 15.1 10.7 2 Y D 95 Strong 12 95 Moderate 814 53 S 18.5 13.0 2 N D 75 Moderate 8 50 Moderate 615 67 S 17.0 9.2 2 Y D 100 Strong 12 70 Moderate 616 67 S 14.9 10.5 6 N D 85 Moderate 8 100 Weak 417 80 S 20.6 15.7 15 Y D 90 Moderate 8 95 Weak 418 49 S 18.3 11.7 1 N D 95 Moderate 8 60 Moderate 619 60 S 14.8 3.1 4 Y D 95 Moderate 8 75 Weak 420 46 S 12.3 4.5 4 Y D 100 Moderate 8 100 Strong 1221 62 S 11.1 4.0 2 N D 100 Strong 12 95 Strong 1222 67 E 22.1 10.2 31 N M 75 Moderate 8 40 Weak 223 62 E 19.1 8.4 7 Y M 5 Weak 1 90 Moderate 824 66 E 16.4 6.6 8 Y M 10 Weak 1 85 Weak 425 59 E 20.0 7.5 6 Y M 10 Moderate 2 95 Moderate 826 63 E 17.5 10.3 13 N M 15 Moderate 2 30 Moderate 427 44 E 13.6 6.7 7 Y M 0 Nonee 0 85 Moderate 828 62 E 19.1 6.8 15 Y M 75 Moderate 8 95 Moderate 829 48 S 15.6 8.1 9 Y M 0 Nonee 0 95 Weak 430 68 E 17.3 11.7 10 N M 65 Weak 3 45 Weak 231 71 S 17.2 12.6 7 N M 15 Moderate 2 65 Moderate 632 77 S 15.2 9.7 4 N M 30 Moderate 4 80 Moderate 833 53 S 15.2 9.7 2 Y M 85 Weak 4 80 Moderate 834 61 E 16.3 7.2 11 Y M 55 Moderate 6 95 Weak 435 53 E 20.0 8.3 12 Y M 25 Weak 2 85 Moderate 836 63 S 17.3 10.3 2 N M 85 Moderate 8 90 Weak 437 73 E 14.8 7.7 3 Y M 95 Strong 12 90 Moderate 838 53 S 16.6 13.6 4 N M 85 Moderate 8 80 Weak 439 64 E 19.6 9.8 17 Y M 85 Moderate 8 90 Weak 440 62 S 17.4 7.3 2 N M 90 Moderate 8 50 Weak 341 50 E 14.5 7.5 14 Y M 100 Strong 12 95 Moderate 8

E, epithelioid cell type; S, spindle cell type; HPF, high power field; N, no; Y, yes; D, disomy 3; M, monosomy 3.

IOVS, January 2010, Vol. 51, No. 1 HSP-27 and Uveal Melanoma 17

reported previously. Our study raises the possibility of usingimmunohistochemical staining for HSP27 as a prognostic indi-cator when cytogenetic studies are not possible.

HSPs are a class of functionally related proteins with expres-sion that increases when cells are exposed to elevated temper-atures or stress. Intracellular HSPs, including HSP-27 (locatedon chromosome 7), are overexpressed in a variety of cancercells, are essential for the survival of these cell types, and havebeen demonstrated as being prognostically significant in somecarcinomas. For example, HSP-27 overexpression is associatedwith a poor prognosis in gastric, prostate, and node-negativebreast carcinomas.17–19 In contrast, high levels of HSP-27 ex-

pression indicate a good prognosis in non–small-cell lung car-cinomas and ovarian carcinomas.20,21

HSPs have been reported previously in a limited number ofuveal melanomas and cell lines.7,9,22 Pardo et al.7 describedseveral HSPs with the exception of HSP-27, in the proteome ofa single patient with uveal melanoma. Zuidervaart et al.9 re-ported relative HSP-27 overexpression in a cell line derivedfrom a uveal melanoma metastasis compared with the cell linederived from the corresponding primary uveal melanoma. Highlevels of HSP-27 expression were also reported in 20 enucle-ated eyes with uveal melanoma.22 As in our study, this inves-tigation found variability in HSP-27 staining across the uvealmelanomas, but did not find any correlation between HSP-27expression and known histopathologic prognostic parameters.We did, however, find a negative correlation between HSP-27and the mitotic rate of the uveal melanomas. This finding willbe examined in more detail in a larger study.

Our finding of a decreased expression of HSP-27 in mono-somy 3 uveal melanomas may be indirectly supported by re-cent studies on cutaneous melanoma. Of interest, in the humancutaneous melanoma cell line A375,23 HSP-27 overexpression

FIGURE 5. Boxplots showing the relationship between chromosome 3status and (A) the HSP-27 protein expression score (B) the percentageof tumor cells that stained positively for HSP-27. Probabilities show thestatistical significance of the differences between monosomy 3 (M3; n� 20) and disomy 3 (D3; n � 21) uveal melanomas, as determined byStudent’s t-test.

FIGURE 4. Vimentin immunoreactivity in FFPE uveal melanomas. (A)The melanoma cells, which are of the spindle type, demonstrated astrong immunoreactivity for vimentin. (B, C) Heterogenous immuno-reactivity for vimentin was present in these uveal melanomas (ofepithelioid and mixed cell type, respectively), depending on cell type,the presence/absence of closed loops (inset, PAS without counter-stain), and the distance from blood vessels. Magnification: (A, B inset,C) �20 objective; (B) �40 objective.

18 Coupland et al. IOVS, January 2010, Vol. 51, No. 1

has been shown to inhibit cell proliferation and reduce cellinvasiveness. Similarly, in a knockdown model of secretedprotein acidic and rich in cysteines (SPARC) expression incutaneous melanoma,24 there is an increased HSP-27 expres-sion, resulting in inhibited tumor progression. It could bepostulated that underexpression of HSP-27 in melanoma cellsresults in increased tumor cell motility and invasiveness.

Vimentin, whose gene is located on chromosome 10, is amember of the intermediate filament family of proteins, which,along with microtubules and actin microfilaments, make up thecytoskeleton of cells. Vimentin is usually expressed in malig-nancies of mesenchymal cell origin (e.g., melanomas) andrarely, in those tumors arising from epithelium (e.g., carcino-mas). Overexpression of vimentin intermediate filaments isassociated with increased tumor invasiveness and migration ina variety of tumor types in vitro,25–28 because of its role in theprocess of epithelial-to-mesenchymal transition. There is acomplex and dynamic relationship between EMT markers,such as vimentin, smooth-muscle actin and cadherins, as wellas other proteins involved in extracellular matrix remodelingand invasion (e.g., SPARC and laminin).29

Fuchs et al.30 examined 52 uveal melanomas for vimentinexpression, and correlated the findings with histomorphologicfeatures and clinical course.30 Similar to our present study,vimentin expression was found in all uveal melanomas studied,irrespective of cell type, with immunopositivity being demon-strated in more than 50% of the neoplastic cells. While Fuchset al. could not demonstrate any relationship between vimen-tin expression levels and tumor size, cell type, pigmentation, ormetastatic progression, we found a significantly higher vimen-tin expression in those uveal melanomas containing closedloops.

One of the main strengths of our study is that in the uvealmelanomas examined immunohistochemically, chromosome 3status was determined using MLPA, which was more sensitiveand informative than FISH, providing data on 13 loci on chro-mosome 3.4 The relatively low sensitivity of FISH in detectingpartial deletions in chromosome 3 in uveal melanoma and insubdividing these tumors into high- and low-risk metastaticsubtypes, has been reported by our group2,4 and others.31

Despite the statistical correlation between loss of chromosome3 and HSP-27 underexpression shown in the present study, ourdata give only an approximate indication of the sensitivity andspecificity with which HSP-27 expression predicts monosomy3, because of the small number of tumors examined and therelatively short follow-up period. We are currently studying alarger cohort of patients, with longer follow-up. We are par-ticularly interested in metastatic death occurring in patientswith a disomy 3 uveal melanoma showing underexpression ofHSP-27 protein.

In conclusion, our data suggest that low-to-negative HSP-27protein expression levels in uveal melanoma are strongly asso-ciated with monosomy 3. Our preliminary results must bevalidated by larger studies to determine whether measurementof HSP-27 expression by immunohistochemistry can reliablypredict survival when cytogenetic studies are not possible.Additional investigations are also needed to understand thefunctional significance of HSP-27 in uveal melanoma and itsrelationship to other proteins implicated in the process ofepithelial-to-mesenchymal transition.

Acknowledgments

The authors thank Inge Kjærgaard and Mona Britt Hansen for experttechnical assistance.

References

1. Damato B. Treatment of primary intraocular melanoma. Expert RevAnticancer Ther. 2006;6:493–506.

2. Damato B, Duke C, Coupland SE, et al. Cytogenetics of uvealmelanoma: a 7-year clinical experience. Ophthalmology. 2007;114:1925–1931.

3. Damato B, Eleuteri A, Fisher AC, Coupland SE, Taktak AF. Artificialneural networks estimating survival probability after treatment ofchoroidal melanoma. Ophthalmology. 2008;115:1598–1607.

4. Damato BE, Dopierala J, Klaasen A, van Dijk M, Sibbring J, Coup-land S. Multiplex ligation-dependent probe amplification of uvealmelanoma: correlation with metastatic death. Invest OphthalmolVis Sci. 2009;50:3048–3055.

5. Onken MD, Ehlers JP, Worley LA, Makita J, Yokota Y, Harbour JW.Functional gene expression analysis uncovers phenotypic switchin aggressive uveal melanomas. Cancer Res. 2006;66:4602–4609.

6. Onken MD, Worley LA, Davila RM, Char DH, Harbour JW. Prog-nostic testing in uveal melanoma by transcriptomic profiling offine needle biopsy specimens. J Mol Diagn. 2006;8:567–573.

7. Pardo M, Garcia A, Thomas B, et al. Proteome analysis of a humanuveal melanoma primary cell culture by 2-DE and MS. Proteomics.2005;5:4980–4993.

8. Pardo M, Garcia A, Thomas B, et al. The characterization of theinvasion phenotype of uveal melanoma tumour cells shows thepresence of MUC18 and HMG-1 metastasis markers and leads tothe identification of DJ-1 as a potential serum biomarker. Int JCancer. 2006;119:1014–1022.

9. Zuidervaart W, Hensbergen PJ, Wong MC, et al. Proteomic analysisof uveal melanoma reveals novel potential markers involved intumor progression. Invest Ophthalmol Vis Sci. 2006;47:786–793.

10. Pardo M, Dwek RA, Zitzmann N. Proteomics in uveal melanomaresearch: opportunities and challenges in biomarker discovery.Expert Rev Proteomics. 2007;4:273–286.

11. Missotten GS, Beijnen JH, Keunen JE, Bonfrer JM. Proteomics inuveal melanoma. Melanoma Res. 2003;13:627–629.

12. Østergaard M, Hansen GA, Vorum H, Honore B. Proteomic profil-ing of fibroblasts reveals a modulating effect of extracellular calu-menin on the organization of the actin cytoskeleton. Proteomics.2006;6:3509–3519.

13. Vorum H, Madsen P, Svendsen I, Cells JE, Honore B. Expression ofrecombinant psoriasis-associated fatty acid binding protein inEscherichia coli: gel electrophoretic characterization, analysis ofbinding properties and comparison with human serum albumin.Electrophoresis. 1998;19:1793–1802.

14. Mortz E, Krogh TN, Vorum H, Gorg A. Improved silver stainingprotocols for high sensitivity protein identification using matrix-assisted laser desorption/ionization-time of flight analysis. Pro-teomics. 2001;1:1359–1363.

15. Ghadjar P, Coupland SE, Na IK, et al. Chemokine receptor CCR6expression level and liver metastases in colorectal cancer. J ClinOncol. 2006;24:1910–1916.

16. van Dongen JJ, Langerak AW, Bruggemann M, et al. Design andstandardization of PCR primers and protocols for detection ofclonal immunoglobulin and T-cell receptor gene recombinationsin suspect lymphoproliferations: report of the BIOMED-2 con-certed action BMH4-CT98–3936. Leukemia. 2003;17:2257–2317.

17. Cornford PA, Dodson AR, Parsons KF, et al. Heat shock proteinexpression independently predicts clinical outcome in prostatecancer. Cancer Res. 2000;60:7099–7105.

18. Kapranos N, Kominea A, Konstantinopoulos PA, et al. Expressionof the 27-kDa heat shock protein (HSP27) in gastric carcinomasand adjacent normal, metaplastic, and dysplastic gastric mucosa,and its prognostic significance. J Cancer Res Clin Oncol. 2002;128:426–432.

19. Thanner F, Sutterlin MW, Kapp M, et al. Heat shock protein 27 isassociated with decreased survival in node-negative breast cancerpatients. Anticancer Res. 2005;25:1649–1653.

20. Geisler JP, Tammela JE, Manahan KJ, et al. HSP27 in patients withovarian carcinoma: still an independent prognostic indicator at 60months follow-up. Eur J Gynaecol Oncol. 2004;25:165–168.

21. Malusecka E, Krzyzowska-Gruca S, Gawrychowski J, Fiszer-Kierz-kowska A, Kolosza Z, Krawczyk Z. Stress proteins HSP27 and

IOVS, January 2010, Vol. 51, No. 1 HSP-27 and Uveal Melanoma 19

HSP70i predict survival in non-small cell lung carcinoma. Antican-cer Res. 2008;28:501–506.

22. Missotten GS, Journee-de Korver JG, de Wolff-Rouendaal D,Keunen JE, Schlingemann RO, Jager MJ. Heat shock protein ex-pression in the eye and in uveal melanoma. Invest Ophthalmol VisSci. 2003;44:3059–3065.

23. Aldrian S, Trautinger F, Frohlich I, Berger W, Micksche M, Kindas-Mugge I. Overexpression of Hsp27 affects the metastatic pheno-type of human melanoma cells in vitro. Cell Stress Chaperones.2002;7:177–185.

24. Sosa MS, Girotti MR, Salvatierra E, et al. Proteomic analysis identi-fied N-cadherin, clusterin, and HSP27 as mediators of SPARC (se-creted protein, acidic and rich in cysteines) activity in melanomacells. Proteomics. 2007;7:4123–4134.

25. McInroy L, Maatta A. Down-regulation of vimentin expressioninhibits carcinoma cell migration and adhesion. Biochem BiophysRes Commun. 2007;360:109–114.

26. Gilles C, Polette M, Mestdagt M, et al. Transactivation of vimentinby beta-catenin in human breast cancer cells. Cancer Res. 2003;63:2658–2664.

27. Hendrix MJ, Seftor EA, Seftor RE, et al. Biologic determinants ofuveal melanoma metastatic phenotype: role of intermediate fila-ments as predictive markers. Lab Invest. 1998;78:153–163.

28. Seftor EA, Meltzer PS, Kirschmann DA, et al. Molecular determi-nants of human uveal melanoma invasion and metastasis. Clin ExpMetastasis. 2002;19:233–246.

29. Rodriguez-Pinilla SM, Sarrio D, Honrado E, et al. Vimentin andlaminin expression is associated with basal-like phenotype in bothsporadic and BRCA1-associated breast carcinomas. J Clin Pathol.2007;60:1006–1012.

30. Fuchs U, Kivela T, Summanen P, Immonen I, Tarkkanen A. Animmunohistochemical and prognostic analysis of cytokeratin ex-pression in malignant uveal melanoma. Am J Pathol. 1992;141:169–181.

31. Onken MD, Worley LA, Person E, Char DH, Bowcock AM, HarbourJW. Loss of heterozygosity of chromosome 3 detected with singlenucleotide polymorphisms is superior to monosomy 3 for predict-ing metastasis in uveal melanoma. Clin Cancer Res. 2007;13:2923–2927.

20 Coupland et al. IOVS, January 2010, Vol. 51, No. 1