Chemokine Receptor CCR7 Expression Predicts Poor Outcome in Uveal Melanoma and Relates to Liver Metastasis Whereas Expression of CXCR4 Is Not of Clinical Relevance

Abstract

Purpose.: To examine the prognostic relevance of expression of the chemokine receptors CCR7 and CXCR4 and its ligand CXCL12 in uveal melanoma in nonmetastatic and metastatic patients with correlation to liver metastasis and overall survival.

Methods.: Primary uveal melanoma specimens from 19 patients with correlating liver metastasis specimens and 30 primary uveal melanoma specimens of patients without metastasis were collected between the years 1988 and 2008. Expression of CCR7, CXCR4, and CXCL12 were studied using immunohistochemistry. Single nucleotide polymorphism (SNP) arrays were used to examine gains or losses of chromosomes 1, 3, 6, and 8 and the regions of CCR7 (17q12-q21.2), CXCR4 (2q21), and CXCL12 (10q11.1) genes.

Results.: Strong cytoplasmic staining for CCR7 correlated with the presence of epithelioid cells (P = 0.037), tumor thickness (P = 0.011), lymphocytic infiltration (P = 0.041), and necrosis (P = 0.045). Nuclear staining for CXCR4 correlated with lymphocytic infiltration (P = 0.017). CXCL12 showed no correlation to histologic parameters. Single nucleotide polymorphism analyses showed no copy number variations in the regions of CCR7, CXCR4, or CXCL12. Strong expression of CCR7 was observed in 76% of the metastatic patients and 0% of nonmetastasis patients. In multivariate analysis, CCR7 staining was inversely correlated to overall survival and disease-free survival, whereas CXCR4 nuclear staining was not.

Conclusions.: Our data suggest that CCR7 plays a role in uveal melanoma metastasis and is associated with poor survival. CCR7 and its involved related pathways are of prognostic value in uveal melanoma and may prove to be a target for therapeutic intervention.

Introduction

Uveal melanoma (UM) is the most common type of primary eye cancer in adults. Approximately 0.7/100,000 of the Western population is affected yearly. ¹ The age-adjusted incidence of UM (5.1 per million) has remained unchanged from 1973 to 2008. Despite new and more conservative treatments, survival has not improved during this time period. ² Well-known clinical prognostic factors are age and location of the tumor. Uveal melanoma patients that are of older age tend to have a worse prognosis. The most important clinical prognostic factor is the tumor size, which is often used for selection of treatment. A large study showed that each increase in millimeter of tumor thickness increased the risk of metastasis by 5%. ³ The most common location for the initial metastasis is the liver. Median survival with initial liver metastasis is 4 to 12 months. ⁴ Uveal melanoma contain fast growing epithelioid cells, has aggressive behavior, and is therefore associated with a poor prognosis. Other histologic features that have been associated with mortality and metastasis are extrascleral extension, extracellular matrix patterns, mitotic count, tumor necrosis, and presence of lymphocytic infiltration. ⁵  

Chemokine receptor CCR7 is mainly expressed in lymphoid cells, its function is mainly mediating cell migration of naïve lymphocytes and mature dendritic cells to secondary lymphoid organs toward chemokines CCL19 and CCL21. CCL19 and CCL21 are the ligands of CCR7 and are both mainly expressed by stroma cells in lymphoid tissues. ⁶ CCR7 has been shown to regulate integrins, which influence the transport of cancer cells and help them migrate through the extracellular matrix ⁷ . Wang et al. found evidence that CCR7 mediates survival and invasion of metastatic squamous cell carcinoma cells through activation of PI3K. ⁸ This indicates that CCR7 is involved in tumor progression and metastasis. It is known that chemokine receptors are expressed widely on different types of cancers such as non-small cell lung cancer, gastric cancer, head/neck cancer, and colon cancer. ^9–12 CCR7 is also expressed on B-cell acute lymphatic leukemia and chronic lymphatic leukemia. ¹³ Strong expression of CCR7 has been associated with poor clinical outcome in patients with cutaneous melanoma. ^14,15 Recently, strong expression of CCR7 was found in UM cells. ¹⁶ The role of chemokine receptor CXCR4 has been described in multiple malignancies that metastasize to the liver, and therefore might provide a pathway for therapeutic intervention. In addition, CXCR4 also localizes on vascular endothelial cells and mediates the angiogenic activity of the chemokine CXCL12. ¹⁷ A recent study shows that CXCR4 is commonly expressed in UM and correlates with epithelioid cell type, a well-established prognostic factor. ¹⁸ High levels of its ligand, CXCL12, in the liver offer an attractive explanation for the selective metastasis of UM to the liver. ¹⁹ It has been hypothesized that the CXCL12/CXCR4 pathway might mediate cancer cells to ‘home' to specific secondary sites, thereby promoting organ-specific hepatic metastasis. ²⁰  

In this study, we investigated the expression of CCR7 and CXCR4/CXCL12 in both primary and metastatic tumor samples. Furthermore, we examined whether expression levels are correlated with clinical, histopathologic, and chromosomal parameters as well as survival, to gain more insight into prognostic factors and possible opportunities for therapeutic intervention. 

Materials and Method

Patient Samples

Forty-nine UM paraffin samples, fresh UM tissue for DNA isolation, and corresponding metastatic liver paraffin samples were collected at the Department of Pathology of the Erasmus University Medical Center between 1988 and 2008. Classical histopathologic parameters such as cell type, mitotic count (per 8 mm² equal to 50 high-power fields [HPF]), necrosis, lymphocytic infiltration, extra ocular extension, and optical nerve invasion were scored by hematoxylin and eosin (H&E) staining, and closed vascular loops with periodic acid Schiff staining without hematoxylin staining. Cell type was assigned according to the Callender classification system. The research followed the tenets of the Declaration of Helsinki. 

Construction of Tissue Microarray (TMA) Samples

The TMA contained 19 cases of UM patients with histologic confirmed metastasis and 30 patients without metastasis. These were constructed from representative areas of formalin-fixed specimens. The TMA consisted of random located 2-mm core samples of each case. The 4-μm sections were cut from TMA and were stained with H&E to confirm the presence of the expected tissue histology within each tissue core. Additional sections were cut for immunohistochemistry (IHC) analyses and fluorescence in situ hybridization (FISH) analyses. 

Immunohistochemistry CCR7, CXCR4, and CXCL12

In the evaluation set, we assessed expression of CCR7, CXCR4, and CXCL12 by IHC to determine specificity and sensitivity. The samples were scored positive or negative by masked screening. Immunohistochemistry was performed with an automated IHC staining system (Ventana BenchMark ULTRA; Ventana Medical Systems Inc., Tucson, AZ) using alkaline phosphatase method for all antibodies and a red chromogen. Briefly, following deparaffinization and heat-induced antigen retrieval for 64 minutes, with exception of CXCR4 (protease treatment for 4 minutes at 36C°), the tissue sections were incubated with primary antibody CCR7 (clone 150503, 1:32000 dilution), CXCL12 (clone 79018, 1:50 dilution), and CXCR4 (clone 44716, 1:128000 dilution; all from R&D Systems, Minneapolis, MN) for 1 hour at 36C°. A subsequent amplification step was followed by incubation with hematoxylin II counter stain for 8 minutes and then blueing reagent for 8 minutes according to the manufacturer's instructions (Ventana). Liver, tonsil, and breast tissue were used as positive controls for CCR7, CXCR4, and CXCL12. Tumors were scored according to the intensity as negative (–), mildly positive (+), moderately positive (++), or strongly positive (+++) of cytoplasmic staining for CCR7 and CXCL12. Nuclear and cytoplasmic staining was scored in a similar way for CXCR4. The histopathologic characterization of the tissue sections and the IHC staining's were independently evaluated by an ophthalmic pathologist. 

DNA Isolation

Deoxyribonucleic acid was isolated from fresh received primary tumor samples using the QIAamp DNA-mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Deoxyribonucleic acid concentration was measured using the NanoDrop ND-1000 Spectrophotometer (NanoDrop technologies, Wilmington, DE) and Picogreen assay (Molecular Probes, Eugene, OR). Deoxyribonucleic acid was stored at −20C°. 

Single Nucleotide Polymorphism (SNP) Array Analysis

Two hundred nanograms of fresh primary tumor DNA were used as input for whole genome analysis by SNP array for each UM sample, to examine gains or losses of chromosomes 1, 3, 6, and 8 and the regions of CCR7 (17q12-q21.2), CXCR4 (2q21), and CXCL12 (10q11.1) genes (Illumina 610Q BeadChip, Illumina, San Diego, CA). The SNP array data were analyzed with the Nexus 6 software (Biodiscovery, El Segundo, CA). 

FISH Analysis

To validate nonrandom chromosomal anomalies on either the short arm (p) and/or long arm (q) of chromosomes 1, 3, 6, and 8, FISH was performed on fresh tumor material (n = 31) or primary UM paraffin samples (n = 15) as described before. ²¹  

Statistical Analysis

We used several tests to assess whether clinical, histopathologic, and chromosomal parameters were associated with expression levels. The χ² test and the Fisher's exact test were used for categoric variables. The Mann-Whitney U test and the Kruskal-Wallis test were used for two and more than two continuous variables, respectively. For survival analyses both the overall survival and the disease-free survival (DFS) were taken into account. The primary endpoint for overall survival was defined as the time to death due to metastasis, whereby death due to other causes were treated as censored. The DFS was defined as the time to the development of metastatic disease. The influence of single prognostic factors on overall survival and DFS were assessed using the Kaplan-Meier method (for categorical variables) or the Cox proportional hazard analysis (for continuous variables). To identify the independent value of the prognostic factors on overall survival and DFS, we used a multivariate Cox proportional hazard analysis with a forward stepwise method based on likelihood ratios. An effect was considered significant if the P value was less than or equal to 0.05. The statistical analyses were performed with the SPSS software version 20.0 (IBM, Chicago, IL). 

Results

A total of 49 patients were included in the study. Twenty-three of the patients were male and 26 were female. The mean age was 57 years (range 27–84); the mean largest tumor diameter was 12.5 mm (range 5.0–19.0), and the mean tumor thickness was 6.9 mm (range 2.0–20.0). Twenty-three tumors were classified as spindle cell type, five as epithelioid cell type and 21 as mixed. The mean overall survival (follow up) was 171 months for the nonmetastatic group and 69 months for the patients who developed metastatic disease. Four patients died due to another cause. The clinical and histopathologic tumor characteristics for nonmetastatic patients and metastatic patients are shown separately in Tables 1 and 2, respectively. 

Table 1

 

Patient Characteristics of the Investigated Study Group Stratified for the Presence of Metastatic Disease

Patient Characteristics	Patients Without Metastasis, No. of Patients (N)	Patients With Metastasis, No. of Patients (N)
Male	14/30	9/19
Female	16/30	10/19
	Mean (Range)	Mean (Range)
Age at time of diagnosis, y	57.7 (34.0–84.0)	56.2 (27.0–84.0)
Tumor thickness, mm	8.0 (2.0–20.0)	5.1 (2.0–11.0)
Largest tumor diameter, mm	12.2 (5.0–18.0)	12.9 (10.0–19.0)
Overall survival, mo	171.24 (52.93–272.39)	69.02 (6.83–181.82)
Disease–free survival, mo	171.24 (52.93–272.39)	59.78 (6.83–156.88)

Table 2

 

Histopathologic Parameters of the Investigated Study Group Stratified for the Presence of Metastatic Disease

Histopathologic Parameters	Patients Without Metastasis, No. of Patients (N)	Patients With Metastasis, No. of Patients (N)
Localization tumor
Choroid	25/30	14/19
Ciliary body	5/30	5/19
Cell type
Spindle	19/30	4/19
Mixed	10/30	11/19
Epithelioid	1/30	4/19
Epithelioid cells
Absent	18/30	4/19
Present	12/30	15/19
Necrosis
Absent	19/30	11/19
Present	11/30	8/19
Closed vascular loops
Absent	23/30	10/19
Present	7/30	9/19
Lymphocytic infiltration
Absent	25/30	10/19
Present	5/30	9/19
Extra ocular extension
Absent	29/30	17/19
Present	1/30	2/19
Optic nerve invasion
Absent	30/30	19/19
Present	0/30	0/19
Mitotic count (per 8 mm2), mean (range)	4.2 (1.0–22.0)	8.0 (1.0–40.0)

CCR7 Staining

Variation in intensity of cytoplasmic expression was evaluated for CCR7 and was ranked ranging from complete lack of expression to strong cytoplasmic expression (Fig. 1). No nuclear staining was observed for CCR7. As demonstrated in Table 3, strong CCR7 expression was observed in 76% (13/17) of metastatic patients and 0% (0/26) nonmetastatic patients. When lymphocytic infiltration was observed the lymphocytes showed CCR7 expression. Identical CCR7 expression patterns were found in the metastatic liver specimens. In a multivariate analysis, CCR7 expression was inversely correlated to overall survival (P = 0.000) (Fig. 2A) and correlated with the adverse prognostic histologic parameters such as the presence of epithelioid cells (P = 0.037), tumor thickness (P = 0.011), lymphocytic infiltration (P = 0.041), and necrosis (P = 0.045). The correlations are shown in Table 4. There was no correlation between CCR7 expression and chromosomal abnormalities or losses and gains in CCR7 (17q12-q21.2) related regions using SNP analysis. 

Figure 1

 

Histology and immunohistochemistry of chemokine receptor CXCR4 and CCR7 expression in UM: histologic parameters were scored using H&E staining (A and D). CXCR4 was scored with either strong nuclear staining (B) or cytoplasmic staining (E). CCR7 expression was scored with strong cytoplasmic staining (C, 40×) or negative staining (note positive lymphocytes and macrophages) (F). Original magnification for all panels ×400.

Figure 2. 

 

Overall survival analysis in relation to chemokine expression: overall survival analysis showed a high significant inverse correlation with strong CCR7 expression (A), nuclear CXCR4 expression showed no significance with overall survival (B).

Table 3

 

An Overview of the Numbers of Cases with Chemokine Expression Levels Stratified for Nonmetastatic and Metastatic UM Patients

	–	+	++	+++
Nonmetastatic UM, N = 30
CCR7	12/30	14/30	4/30	0/30
CXCR4 (nuclear)	11/30	8/30	7/30	4/30
CXCR4 (cytoplasm)	5/30	5/30	13/30	7/30
CXCL12	24/30	6/30	0/30	0/30
Metastatic UM, N = 19
CCR7	1/17	3/17	0/17	13/17
CXCR4 (nuclear)	6/17	2/17	1/17	8/17
Missing value, N = 2
CXCR4 (cytoplasm)	7/17	6/17	2/17	2/17
CXCL12	17/17	0/17	0/17	0/17

 The expression levels are classified as: − = negative, + = mildly positive, ++ = moderately positive, +++ = strongly positive.

Table 4

 

Correlations Between CCR7, CXCR4, and CXCL12 Expression and Clinical, Histopathologic, and Chromosomal Data

	CCR7				CXCR4 (Nuclear)			CXCR4 (Cytoplasm)				CXCL12
	–	+	++/ +++		–/+	++/ +++		–	+	++/ +++		–	+
	N = 13	N = 17	N = 17	P	N = 17	N = 30	P	N = 18	N = 11	N = 18	P	N = 23	N = 8	P
Sex
Male	6	7	9	0.788†	7	15	0.560†	9	8	5	0.059†	13	1	0.045†
Female	7	10	8		10	15		9	3	13		10	7
Epithelioid cells
Absent	9	9	4	0.037†	10	12	0.214†	7	5	10	0.602†	14	5	1.000*
Present	4	8	13		7	18		11	6	8		9	3
Necrosis
Absent	4	12	12	0.045†	13	15	0.076†	10	5	13	0.328†	12	7	0.108*
Present	9	5	5		4	15		8	6	5		11	1
Closed vascular loops
Absent	11	11	10	0.302†	11	21	0.708†	13	7	12	0.879†	18	6	1.000*
Present	2	6	7		6	9		5	4	6		5	2
Optic nerve invasion
Absent	13	17	17	-	17	30	-	18	11	18	-	23	8	-
Present	0	0	0		0	0		0	0	0		0	0
Mean age at time of diagnosis, y	54.2	56.9	59.8	0.206‡	58.0	56.7	0.674‡	58.1	54.0	58.2	0.680‡	56.4	58.9	0.603§
Mean tumor thickness, mm	9.0	6.9	5.5	0.011‡	7.3	6.8	0.463‡	6.5	7.6	7.1	0.624‡	8.1	7.6	0.285§
Mean largest tumor diameter, mm	13.2	11.8	12.5	0.602‡	12.7	12.3	0.867‡	11.4	13.4	12.9	0.182‡	12.6	11.4	0.601§
Mean mitotic count (per 8 mm2)	6.1	5.2	6.1	0.807‡	4.5	6.5	0.621‡	6.7	4.3	5.8	0.244‡	4.6	3.0	0.443§
	CCR7				CXCR4 (Nuclear)			CXCR4 (Cytoplasm)				CXCL12
	–/+	++/+++			–/+	++/+++		–/+	++/+++			–	+
	N = 30	N = 17	P		N = 17	N = 30	P	N = 29	N = 18	P		N = 23	N = 8	P
Localization tumor
Choroid	25	12	0.460*		12	25	0.460*	22	15	0.719*		19	7	1.000*
Ciliary body	5	5			5	5		7	3			4	1
Lymphocytic infiltration
Absent	25	9	0.041*		16	18	0.017*	18	16	0.091*		19	7	1.000*
Present	5	8			1	12		11	2			4	1
Extra ocular extension
Absent	30	15	0.126*		17	28	0.528*	27	18	0.517*		23	7	0.258*
Present	0	2			0	2		2	0			0	1
Chromosome 1p loss
Yes	7	5	0.438*		6	6	0.784†	7	5	0.647†		5	2	1.000*
No	15	5			9	11		10	10			11	4
Chromosome 3 loss
Yes	10	5	1.000*		9	6	0.062†	6	9	0.126†		6	4	0.341*
No	13	5			5	13		12	6			11	2
Table 4 . Continued
	CCR7				CXCR4 (Nuclear)			CXCR4 (Cytoplasm)				CXCL12
	–/+	++/+++			–/+	++/+++		–/+	++/+++			–	+
	N = 30	N = 17	P		N = 17	N = 30	P	N = 29	N = 18	P		N = 23	N = 8	P
Chromosome 6p gain
Yes	10	1	0.055*		7	4	0.208†	5	6	0.611†		7	3	0.624*
No	11	9			8	12		11	9			10	2
Chromosome 6q loss
Yes	5	2	1.000*		6	1	0.037*	2	5	0.220*		5	0	0.290*
No	16	8			9	15		14	10			12	5
Chromosome 6q gain
Yes	3	0	0.533*		2	1	0.600*	1	2	0.600*		2	1	1.000*
No	18	10			13	15		15	13			15	4
Chromosome 8p loss
Yes	2	2	0.567*		1	3	0.620*	1	3	0.288*		1	0	1.000*
No	21	8			13	16		18	11			17	5
Chromosome 8p gain
Yes	8	1	0.217*		4	5	1.000*	6	3	0.698*		5	2	0.621*
No	15	9			10	14		13	11			13	3
Chromosome 8q gain
Yes	11	7	0.283*		8	10	0.797†	10	8	0.797†		7	2	1.000*
No	6	9			9	6		11	3

 * The Fisher-Exact test was used to calculate correlations between two categorical variables if the assumptions for the χ²-test were not met.  † The χ²-test was used to calculate correlations between categorical variables.  ‡ The Kruskal-Wallis test was used to calculate correlations between more than two continuous variables.  § The Mann-Whitney test was used to calculate correlation between two continuous variables.

CXCR4/CXCL12 Staining

Expression of CXCR4 showed both cytoplasmic and nuclear expression (Fig. 1). The nuclear expression of CXCR 4 was correlated with lymphocytic infiltration (P = 0.017) (Table 4). Nuclear expression of CXCR4 was not significantly related to overall survival (P = 0.088) as shown in Figure 2B. When lymphocytic infiltration was observed the lymphocytes showed CXCR4 expression. There was an identical CXCR4 expression pattern in the metastatic liver specimens. There was no significant correlation between CCR7 expression and CXCR4 expression either cytoplasmic (6/47, P = 0.628) or nuclear (11/47, P = 0.838). CXCL12 was only mildly positive in the cytoplasm in a minority of the tumor samples (Table 3) and showed mild expression in the biliary epithelium of the liver samples. There were no correlations with chromosomal abnormalities or losses and gains in CXCR4 (2q21)- or CXCL12 (10q11.1)-related regions in SNP analysis. 

Statistical Analysis

Univariate analysis showed that the overall survival was significantly shorter in UM patients with tumors with the presence of epithelioid cells (P = 0.010), large tumor thickness (P = 0.026), presence of vascular loops (P = 0.021), high mitotic count (8 mm² equal to 50 HPF) (P = 0.003), presence of lymphocytic infiltration (P = 0.011), chromosome 8p loss (P = 0.004), chromosome 8q gain (P = 0.005), and CCR7 expression (P = 0.000). Using multivariate analysis only high mitotic count (hazard ratio [HR] 1.2, P = 0.007) and strong positive CCR7 expression (HR 479.1, P = 0.003) showed to be highly significant and independent adverse prognostic factors for overall survival. 

For DFS the univariate analysis showed that tumors with the presence of epithelioid cells (P = 0.006), tumor thickness (P = 0.018), presence of vascular loops (P = 0.026), high mitotic count (P = 0.009), and presence of lymphocytic infiltration (P = 0.012), chromosome 8p loss (P = 0.008), chromosome 8q gain (P = 0.005), and CCR7 expression (P = 0.000) were significantly and inversely correlated with DFS. In multivariate analysis, only high mitotic count (HR 1.1, P = 0.051), chromosome 8p loss (HR 8.4, P = 0.054), and a strong positive CCR7 expression (HR 435.3, P = 0.004) were independent predictors for DFS. 

Discussion

In this study, we found a highly significant inverse correlation between both overall survival and DFS, and CCR7 expression, which shows that CCR7 is associated with poor survival in UM. In our IHC experiments, we demonstrated that chemokine receptors CCR7 and CXCR4 are expressed on UM samples. There were cases with negative CCR7 staining, but these cases showed positive internal controls in lymphocyte infiltration as shown in Figure 1. Negative staining for CCR7 did not significantly correlate with negative staining for CXCR4. Neither was there a significant correlation between CCR7 expression and CXCR4 expression in either cytoplasmic or nuclear pattern. The chemokine CXCL12 is rarely expressed in UM at low levels. It is possible that immunohistochemistry might not be sufficiently sensitive to evaluate the differential expression of CXCL12 in formalin-fixated, paraffin-embedded tissues. Although CCR7 is linked mainly to lymphogenic metastasis, ¹¹ our study strongly indicates that lymphatic dissemination is not the exclusive route for CCR7 associated metastasis. No lymphatic vessels are present in the choroid of the eye. Others have shown that CCR7 regulates migration and adhesion processes of metastatic squamous cell carcinoma cells via αvβ3 integrin, ¹² these findings may be applicable to UM cells. In cutaneous melanoma αvβ3 has been correlated to metastatic behavior, ²² and αvβ3 expression has been shown in primary uveal melanoma and cell lines derived from the same tumors. ^23–25 It can be hypothesized that CCR7 regulates migration and adhesion of UM cells through vascular endothelial cells, and thereby enable metastasis to secondary organs such as the liver. It is known that ligands of CCR7 (CCL19/CCL21) are expressed in the liver, ^26,27 thus, it may be possible that the CCR7 pathway is activated in metastatic events in patients with UM. Expression of CCL19 and CCL21 was not investigated in this study since no significant correlation between these chemokines and metastasis was described by Dobner et al. ¹⁶ Moreover, Rubie et al. showed no significant differences in CCL19 protein and CCL19 gene expression between hepatocellular carcinoma, colon cancer tumor tissue, and normal liver tissue. ²⁸ The authors hypothesize that IHC might not be sufficiently sensitive to evaluate the differential expression of CCL19/CCL21 in formalin-fixated, paraffin-embedded tissues. A mixed or epithelioid cell type was significantly correlated with CCR7 expression (P = 0.006), which is a known histologic parameter associated with poor survival. ²⁹ We also explored CXCR4 and its ligand CXCL12. CXCR4 is expressed in breast, prostate, pancreatic, renal, gastric carcinoma, skin melanoma, glioma, and leukemias. ³⁰ We found CXCR4 expression in UM both cytoplasmic and/or nuclear localized, and similar expression patterns in associated liver biopsy specimens (n = 19). In contrast to studies in breast carcinoma ³¹ and melanoma cell lines, ¹⁹ we did not find loss of CXCR4 expression or shifts from cytoplasmic to nuclear expression in liver metastases of CXCR4 positive primary tumors. Other studies also observed no significant difference between expression levels of CCR7 and/or CXCR4 in paired primary and metastatic breast cancer ³² or in gastric cancer. ³³ In our analyses, strong nuclear CXCR4 expression was not correlated with epithelioid cell type in contrast to previous work by Scala et al. ¹⁸ This may be related to sample size. CXCR4 expression showed no correlation with overall survival or DFS. These last findings do not contradict earlier immunohistochemical studies of CXCR4 expression in primary UM samples. ^16,34,35 However, we provide the first study to compare selected groups of histologically confirmed metastasis and a large group of long-term progression-free survival cases. Our data suggest the more important role for metastatic disease to be with CCR7 expression as compared with CXCR4 expression in UM. On the other hand, immunohistochemistry may not be a sensitive enough tool for the detection of differential chemokine and chemokine receptor expression in tumors for the identification of prognostically significant differences. Other molecular techniques, such as mRNA nor even epigenetic regulators such as microRNA expression may provide more insight into the importance of chemokines in metastatic spread of UM. 

Ocular and cutaneous melanomas show different preferential sites for metastatic spread. Skin melanomas spread to lymph nodes, distant skin sites, lung, liver, central nervous system, and bone. Uveal melanoma spreads hematogenous, with a high tendency to metastasize to the liver in 90% to 95% of the patients. The absence of lymphatics in the eye is one of the most important factors for the difference in metastatic spread. The predominance of liver metastasis cannot be solely explained by circulation because the lungs provide the first capillary bed that these cells would encounter. Therefore, it will most probably be a reflection of both preferential homing of cells to the liver combined with preferential growth and survival of disseminated UM cells in the hepatic microenvironment. The metastatic disease occurs almost exclusively in patients whose tumor show chromosome 3 loss. Tumors with concurrent loss of chromosome 1p and 3 are at higher risk of metastasizing than the tumors with other aberrations. ³⁶ Partial deletions or translocations have rarely been described on these chromosomes making it difficult to map putative tumor suppressor genes. However, recently a mutation in the BAP1 gene, located on chromosome 3, has been identified in UMs and this gene seems to play an important role in the tumor progression, ³⁷ but most likely not in metastatic potential. ³⁸ Apart from the CXCR4/CXCL12 and CCR7/CCL19 CCL21 pathways that we have investigated in this report, other pathways have been implicated in the preferential homing of tumor cells to the liver, such as hepatocyte growth factor (HGF) and it's corresponding receptor c-Met, and insulin-like growth factor 1 (IGF-1). ³⁹ Although these pathways are not uniquely restricted to melanoma in combination, they may offer an explanation for the preferential metastatic spread of UM to the liver, and more importantly may offer potential for therapeutic interventions. 

In summary, our study shows that CCR7 expression in UM plays a role in UM metastasis, is independently associated with poor patient survival and can be used as a prognostic marker. CCR7 expression is not exclusively related to lymphogenic metastasis. Further research to validate our findings could be directed to the evaluation CCR7 and CXCR4 mRNA expression by RT-PCR. Further research on CCR7 and associated pathways in UM is necessary for therapeutic targeting. 

Acknowledgments

The authors thank Sharmiela Ramlal for her excellent technical support in immunohistochemistry and Frank van de Panne for his assistance with preparation of the Figures. 

This study was carried out by the Rotterdam Ocular Melanoma Studygroup, which is supported by a grant of the Combined Ophthalmic Research Rotterdam and Stichting Nederlands Oogheelkundig Onderzoek. 

Disclosure: T. van den Bosch, None; A.E. Koopmans, None; J. Vaarwater, None; M. van den Berg, None; A. de Klein, None; R.M. Verdijk, None