Purpose: Retinoids inhibit proliferation and induce differentiation in melanoma cells. Retinoic acid receptors (RAR) and retinoid X receptors (RXR) mediate the various modulatory effects of retinoids in cells. We have studied the in situ expression of each RAR and RXR protein (α, β, γ) in a large series of melanocytic lesions and correlated the expression with clinicopathologic features and prognosis of the patients.

Experimental Design: Tissue microarray blocks of 226 melanocytic lesions were semiquantitatively evaluated by immunohistochemistry for the cytoplasmic and nuclear expression of RAR and RXR protein (α, β, γ).

Results: A significant decrease of RARβ protein (P < 0.0001), nuclear expression of RARγ (P < 0.0001), and RXRα (P < 0.0001) was found in primary and metastatic melanomas as compared with nevi. Loss of nuclear immunoreactivity for RARγ (P = 0.048) and RXRα (P = 0.001) was observed in the lesions showing vertical growth pattern. In addition, in patients with concomitant loss of cytoplasmic staining for RARα and RXRα, the probability of overall survival (log-rank test, P = 0.002) and disease-specific survival (log-rank test, P = 0.014) was significantly lower.

Conclusions: Aberrant expression of retinoid receptors seems to be a frequent event in melanoma and suggests an impairment of the retinoid pathway in this cancer. Our data indicate the loss of retinoid receptor expression with melanoma progression and suggest a possible prognostic significance of the analysis of retinoid receptors in melanoma.

Melanoma is recognized as one of the most aggressive cancers due to its relatively high propensity for metastasizing. In the United States, its incidence has increased 15-fold in the past 40 years, a rate more rapid than that described for most other malignancies. Melanoma is the fifth leading cancer in men and the sixth leading cancer in women in the United States. In 2007, there will be an estimated 60,000 new cases of invasive melanoma, 48,000 new cases of in situ melanoma, and an estimated 8,000 deaths due to melanoma in the United States (1). Many melanoma cases occur in young individuals, and once metastasis occurs, there is little effective treatment available. Understanding the pathogenesis and progression of melanoma is needed to facilitate the development of new methods of treating the disease.

A normal growth pattern of a melanocyte involves controlled proliferation in the epidermis. On occasion, melanocytes can proliferate forming benign lesions (nevi) in the epidermis and also into the dermis (2). It has been proposed that melanocytes, whether from normal intraepidermal melanocytes or from the cells in melanocytic nevi, may develop malignant transformation into melanoma, which then usually progresses from a radial to a vertical growth phase and finally to regional and distant metastases. As in various malignancies, both genetic predisposition and exposure to environmental agents are risk factors for melanoma development. Melanoma likely arises due to the accumulation of mutations in genes critical for cell proliferation, differentiation, and cell death (3, 4). Currently, there are several validated histologic and clinical factors to provide diagnosis, prognosis, stratification, and management of melanoma patients. However, there is still a need of additional features, such as biomarker expression, to further substratify and help manage patients with melanoma.

Retinoic acid receptors (RAR) and retinoid X receptors (RXR) (each one including three different isotypes—α, β, and γ) are the nuclear receptors belonging to the superfamily of ligand-inducible transcriptional regulatory factors that mediates the actions of retinoid. In the presence of retinoids, these receptors form RAR-RXR heterodimers and modulates the expression of target genes through the interaction with specific DNA sequences, called retinoic acid response element, which are located in the promoter regions. In recent years, various studies have shown that transcriptional activities of the nuclear proteins can be regulated by the modulation of their subcellular localization (5). In some cases, phosphorylation of retinoid receptors causes the protein to be exported to the cytoplasm (6). Consequently, retinoid-dependent transcriptional activity by RAR-RXR is reduced.

As a first effect of the binding to nuclear retinoid receptors, retinoids should induce the transcription of RAR or RXR subtypes (7). A previous study by Boehm et al. (8) showed simultaneous decreased expression of RARβ and RXRα in melanocytic tumors. Retinoic acid has been shown to induce melanin synthesis (9) and inhibits the proliferation of melanoma cells in vitro (10). In clinical trials, retinoic acid has been found to be effective in dysplastic nevi (11) and resulted in the regression of intracutaneous metastases from malignant melanoma in some patients (12). Aberrant expression of retinoid receptors has been reported in various neoplasias and may be a causal factor in retinoic acid resistance and refractoriness in retinoid response. These observations make us postulate that the difference in the response of melanoma patients to retinoic acid therapy clinical trials might be due to the differential expression of retinoid receptors in various patients.

Our working hypothesis was that abnormal expression of retinoid receptors (RAR and RXR) may be involved in melanomagenesis, and that the differential subcellular localization of retinoid receptors may correlate with progression and/or prognosis of melanoma patients. The aims of the present study were then to detect if any changes in the localization of isoforms of RAR/RXR or any loss in their expression during the melanoma progression and to evaluate the expression of retinoid receptors to determine whether differences in that expression could contribute to disease prognosis in malignant melanoma.

Patients and samples. In the present study, a total of 226 cases of melanocytic lesions including primary melanoma (superficial spreading, nodular, acral, and lentigo malignant melanoma; 130 cases), metastases of the above-mentioned cases (47 cases) and benign nevi (49 cases) were obtained from the Department of Pathology of the University of Texas M.D. Anderson Cancer Center. The specimens included small biopsies or excisions. The study was approved by the Institutional Review Board. The median follow-up was 4.8 years (range, 0-17.0).

Tissue microarray construction. For tissue microarray construction, H&E-stained sections were reviewed from each block to confirm the diagnosis and select the areas of interest. Either 0.6-mm (punch biopsies of BN and DN cases) or 1.0-mm (excision specimens) cylindrical cores of tissue were punched out from donor blocks to preserve most of the original tissue. The selected tissue cores were inserted in a standard 4.5 × 2 × 1-cm recipient block using a manual tissue arrayer (Beecher Instruments) with an edge-to-edge distance of 0.1 or 0.15 mm. At least two tissue cores were taken for each case. Three tissue microarray blocks containing a total of 450 cores of tissues were constructed. Each block, in addition to the melanocytic samples, had three cylinders from the same control tissues. Sections measuring 5 μm thick from all three blocks were cut, and one standard H&E-stained slide was examined to verify the presence of diagnostic lesional cells.

Antibodies and immunohistochemical analysis. The avidin-biotin-peroxidase complex method was used to detect immunoreactivity for retinoid receptors (RARs and RXRs) as described previously (13). Briefly, for all TMAs, histologic section was deparaffinized and rehydrated in graded alcohols and distilled water. After blocking the endogenous peroxidase (3% hydrogen peroxide), the heat-induced antigen retrieval was done in a water bath using citrate buffer at pH 6.0. The sections were incubated overnight at 4°C with affinity-purified rabbit polyclonal antibodies against RARα, RARβ, and RARγ and RXRα, RXRβ, and RXRγ of human origin (Santa Cruz Biotechnologies). The sections were then washed in TBS (pH, 7.4) and incubated with goat anti-rabbit biotinylated immunoglobulins (DAKO Corp.). After incubation with the secondary antibody, the sections were reacted with avidin-biotin-peroxidase complex (DAKO) and developed with 3,3′-diaminobenzidine. Afterward, the sections were rinsed in distilled water, lightly counterstained with Mayer's hematoxylin, and mounted for evaluation. The specificity of immunohistochemical procedures was checked by using negative and positive control sections. For negative control of immunoreactions, the primary antibodies were replaced with preimmune rabbit serum. As positive controls, sections of human skin were incubated with the same antibody.

The distribution of immunoreactivity was analyzed by consensus between two of the investigators (A.H.D. and V.G.P.) to score both the nuclear and cytoplasmic immunolabeling. Labeling intensity was also categorized into negative, weak, moderate, or strong. The investigators independently reviewed and scored slides by estimating the percentage of cells exhibiting characteristic labeling. The entire tissue cylinder from each case was considered in the evaluation. Immunoreactive cells were counted randomly over superficial and deep dermal foci in primary melanomas and nevi to detect possible maturation changes within the lesions. When present, intraepidermal melanocytes were also considered. The individual values of the two cores per lesion were averaged to address possible labeling variation within the lesions.

Statistical analysis. Tissue samples were obtained from four disease groups—primary melanoma without metastasis, primary melanoma with metastasis, metastasis, and nevi. For each sample, cytoplasmic intensity of marker expression was scored from 0 (negative) to 3, and percentages of cytoplasmic and nuclear expression were defined as 0 to 3 for 0-5%, 6-25%, 26-75%, and 76-100%, respectively. Disease groups were compared on cytoplasmic intensity and cytoplasmic and nuclear percentage expression of each marker using the Fisher's exact test. We compared marker expression for the following subgroups: (a) primary melanoma without metastasis versus primary melanoma with metastasis; (b) primary melanoma versus nevi; (c) primary melanoma versus metastasis; and (d) primary melanoma or metastasis versus nevi.

For survival analyses, we considered only melanoma and metastasis samples (i.e., excluding the nevi samples). To determine whether expression of retinoid receptors was associated with overall survival and disease-specific survival, we used the Kaplan-Meier method with log-rank test and Cox proportional-hazards regression techniques. Overall survival time was computed from the date of melanoma diagnosis to the date of death for patients who died or to the date of last follow-up for those still alive. Patients still alive at last follow-up were considered censored. Disease-specific survival was computed from the date of melanoma diagnosis to the date of disease progression or to the date of last follow-up. Patients without disease progression at last follow-up were considered censored. Independent variables considered for modeling included retinoid expression variables alone and in combination. Other independent variables considered were Breslow thickness, Breslow thickness log transformed, number of mitoses, number of mitoses log transformed, ulceration, and Clark level. Clark level was dichotomized such that levels 2 and 3 were combined and levels 4 and 5 were combined. All test results were considered significant at P < 0.05. Individual significance levels were not adjusted for multiple comparisons. All analyses were done using SAS 9.1 (SAS Institute).

In the present study, we have examined the expression of various isoforms (i.e., α, β, and γ of RARs and RXRs) in the melanocytic lesions by immunohistochemistry. Nuclear and cytoplasmic labeling was observed for some of these receptors (Figs. 1 and 2).

Fig. 1.

Expression of RARα and RARβ in melanocytic lesions as determined by immunohistochemistry. Original magnification, ×400. A, nevus cells express a high level of RARα in the nucleus. There was a gradual decrease in primary melanoma (B) without metastasis and (C) with metastasis, whereas (D) metastasis showed a robust cytoplasmic staining for RARα. E, RARβ in nevus was found to be in cytosol with no detectable nuclear staining. F, primary melanoma without metastasis had strong cytosolic expression as compared with (G) a primary melanoma with subsequent metastasis and (H) metastasis, with complete loss of the expression.

Fig. 1.

Expression of RARα and RARβ in melanocytic lesions as determined by immunohistochemistry. Original magnification, ×400. A, nevus cells express a high level of RARα in the nucleus. There was a gradual decrease in primary melanoma (B) without metastasis and (C) with metastasis, whereas (D) metastasis showed a robust cytoplasmic staining for RARα. E, RARβ in nevus was found to be in cytosol with no detectable nuclear staining. F, primary melanoma without metastasis had strong cytosolic expression as compared with (G) a primary melanoma with subsequent metastasis and (H) metastasis, with complete loss of the expression.

Close modal
Fig. 2.

Expression of RARγ and RXRα in melanocytic lesions as determined by immunohistochemistry. Original magnification, ×400. A, nevus cells express a high level of RARγ in the nucleus; no expression of RARγ albeit faint cytoplasmic staining was found in primary melanoma (B) without metastasis (C) with metastasis, and (D) metastasis. E, strong nuclear expression of RXRα was found in benign nevus. F, primary melanoma without metastasis had strong cytosolic labeling as compared with (G) the melanoma with metastasis, and (H) metastasis, with no nuclear labeling.

Fig. 2.

Expression of RARγ and RXRα in melanocytic lesions as determined by immunohistochemistry. Original magnification, ×400. A, nevus cells express a high level of RARγ in the nucleus; no expression of RARγ albeit faint cytoplasmic staining was found in primary melanoma (B) without metastasis (C) with metastasis, and (D) metastasis. E, strong nuclear expression of RXRα was found in benign nevus. F, primary melanoma without metastasis had strong cytosolic labeling as compared with (G) the melanoma with metastasis, and (H) metastasis, with no nuclear labeling.

Close modal

Melanoma without metastasis versus melanoma with metastasis. Percentage of cells expressing RARα and RXRα in cytoplasm differed significantly between primary melanomas with metastasis and those without metastasis (Table 1). Compared with melanoma samples without metastasis, samples of melanoma with metastasis tended to be lower in cytoplasmic percent of RARα (P = 0.0002) and cytoplasmic intensity of RARα (P = 0.007), but higher in cytoplasmic percent of RXRα (P = 0.017). None of the other markers differed significantly between these two melanoma groups.

Table 1.

Cytoplasmic marker distribution by diagnostic group

VariableDiagnosisnValue, n (%)
P
0123
Cytoplasmic percent RARα Melanoma without mets 68 7 (10.3) 1 (1.5) 27 (39.7) 33 (48.5) 0.0002 
 Melanoma with mets 43 7 (16.3) 10 (23.3) 17 (39.5) 9 (20.9)  
 Melanoma 111 14 (12.6) 11 (9.9) 44 (39.6) 42 (37.8) 0.0044 
 Metastasis 45 11 (24.5) 5 (11.1) 24 (53.3) 5 (11.1)  
Cytoplasmic percent RARβ Melanoma 104 10 (9.6) 6 (5.8) 24 (23.1) 64 (61.5) <0.0001 
 Metastasis 42 6 (14.3) 7 (16.7) 21 (50.0) 8 (19.0)  
 Melanoma 104 10 (9.6) 6 (5.8) 24 (23.1) 64 (61.5) <0.0001 
 Nevi 41 1 (2.4) 9 (21.9) 26 (63.4) 5 (12.2)  
 Melanoma/metastasis 146 16 (11.0) 13 (8.9) 45 (30.8) 72 (49.3) <0.0001 
 Nevi 41 1 (2.4) 9 (21.9) 26 (63.4) 5 (12.2)  
Cytoplasmic percent RXRα Melanoma without mets 65 0 (0.0) 1 (1.5) 24 (36.9) 40 (61.5) 0.0166 
 Melanoma with mets 43 0 (0.0) 1 (2.3) 6 (13.9) 36 (83.7)  
 Melanoma 108 0 (0.0) 2 (1.9) 30 (27.8) 76 (70.4) 0.004 
 Metastasis 31 1 (3.2) 0 (0.0) 17 (54.8) 13 (41.9)  
 Melanoma 108 0 (0.0) 2 (1.9) 30 (27.8) 76 (70.4) 0.0078 
 Nevi 41 1 (3.0) 1 (3.0) 17 (51.5) 14 (42.4)  
Cytoplasmic percent RXRβ Melanoma 110 29 (26.4) 16 (14.5) 34 (30.9) 31 (28.2) 0.0086 
 Metastasis 31 8 (25.8) 8 (25.8) 14 (45.2) 1 (3.2)  
 Melanoma 110 29 (26.3) 16 (14.5) 34 (30.9) 31 (28.2) 0.0038 
 Nevi 33 5 (15.1) 6 (18.2) 20 (60.6) 2 (6.1)  
 Melanoma/metastasis 141 37 (26.2) 24 (17.0) 48 (34.0) 32 (22.7) 0.0172 
 Nevi 33 5 (15.1) 6 (18.2) 20 (60.6) 2 (6.1)  
Cytoplasmic intensity RARα Melanoma without mets 68 49 (72.1) 19 (27.9) 0 (0.0) 0 (0.0) 0.0071 
 Melanoma with mets 43 40 (93.0) 3 (7.0) 0 (0.0) 0 (0.0)  
Cytoplasmic intensity RARβ Melanoma 104 45 (43.3) 45 (43.3) 14 (13.5) 0 (0.0) <0.0001 
 Metastasis 42 35 (83.3) 6 (14.3) 1 (2.4) 0 (0.0)  
 Melanoma 104 45 (43.3) 45 (43.3) 14 (13.5) 0 (0.0) 0.0002 
 Nevi 41 32 (78.0) 9 (21.9) 0 (0.0) 0 (0.0)  
 Melanoma/metastasis 146 80 (54.8) 51 (34.9) 15 (10.3) 0 (0.0) 0.0084 
 Nevi 41 32 (78.0) 9 (21.9) 0 (0.0) 0 (0.0)  
Cytoplasmic intensity RXRα Melanoma 108 30 (27.8) 60 (55.6) 18 (16.7) 0 (0.0) 0.0181 
 Nevi 33 17 (51.5) 15 (45.4) 1 (3.0) 0 (0.0)  
VariableDiagnosisnValue, n (%)
P
0123
Cytoplasmic percent RARα Melanoma without mets 68 7 (10.3) 1 (1.5) 27 (39.7) 33 (48.5) 0.0002 
 Melanoma with mets 43 7 (16.3) 10 (23.3) 17 (39.5) 9 (20.9)  
 Melanoma 111 14 (12.6) 11 (9.9) 44 (39.6) 42 (37.8) 0.0044 
 Metastasis 45 11 (24.5) 5 (11.1) 24 (53.3) 5 (11.1)  
Cytoplasmic percent RARβ Melanoma 104 10 (9.6) 6 (5.8) 24 (23.1) 64 (61.5) <0.0001 
 Metastasis 42 6 (14.3) 7 (16.7) 21 (50.0) 8 (19.0)  
 Melanoma 104 10 (9.6) 6 (5.8) 24 (23.1) 64 (61.5) <0.0001 
 Nevi 41 1 (2.4) 9 (21.9) 26 (63.4) 5 (12.2)  
 Melanoma/metastasis 146 16 (11.0) 13 (8.9) 45 (30.8) 72 (49.3) <0.0001 
 Nevi 41 1 (2.4) 9 (21.9) 26 (63.4) 5 (12.2)  
Cytoplasmic percent RXRα Melanoma without mets 65 0 (0.0) 1 (1.5) 24 (36.9) 40 (61.5) 0.0166 
 Melanoma with mets 43 0 (0.0) 1 (2.3) 6 (13.9) 36 (83.7)  
 Melanoma 108 0 (0.0) 2 (1.9) 30 (27.8) 76 (70.4) 0.004 
 Metastasis 31 1 (3.2) 0 (0.0) 17 (54.8) 13 (41.9)  
 Melanoma 108 0 (0.0) 2 (1.9) 30 (27.8) 76 (70.4) 0.0078 
 Nevi 41 1 (3.0) 1 (3.0) 17 (51.5) 14 (42.4)  
Cytoplasmic percent RXRβ Melanoma 110 29 (26.4) 16 (14.5) 34 (30.9) 31 (28.2) 0.0086 
 Metastasis 31 8 (25.8) 8 (25.8) 14 (45.2) 1 (3.2)  
 Melanoma 110 29 (26.3) 16 (14.5) 34 (30.9) 31 (28.2) 0.0038 
 Nevi 33 5 (15.1) 6 (18.2) 20 (60.6) 2 (6.1)  
 Melanoma/metastasis 141 37 (26.2) 24 (17.0) 48 (34.0) 32 (22.7) 0.0172 
 Nevi 33 5 (15.1) 6 (18.2) 20 (60.6) 2 (6.1)  
Cytoplasmic intensity RARα Melanoma without mets 68 49 (72.1) 19 (27.9) 0 (0.0) 0 (0.0) 0.0071 
 Melanoma with mets 43 40 (93.0) 3 (7.0) 0 (0.0) 0 (0.0)  
Cytoplasmic intensity RARβ Melanoma 104 45 (43.3) 45 (43.3) 14 (13.5) 0 (0.0) <0.0001 
 Metastasis 42 35 (83.3) 6 (14.3) 1 (2.4) 0 (0.0)  
 Melanoma 104 45 (43.3) 45 (43.3) 14 (13.5) 0 (0.0) 0.0002 
 Nevi 41 32 (78.0) 9 (21.9) 0 (0.0) 0 (0.0)  
 Melanoma/metastasis 146 80 (54.8) 51 (34.9) 15 (10.3) 0 (0.0) 0.0084 
 Nevi 41 32 (78.0) 9 (21.9) 0 (0.0) 0 (0.0)  
Cytoplasmic intensity RXRα Melanoma 108 30 (27.8) 60 (55.6) 18 (16.7) 0 (0.0) 0.0181 
 Nevi 33 17 (51.5) 15 (45.4) 1 (3.0) 0 (0.0)  

Primary melanoma versus nevi. Melanoma samples (with or without metastasis) differed significantly from nevi samples for percentage of cells showing cytoplasmic immunolabeling of RARβ (P < 0.0001), RXRα (P = 0.008), and RXRβ (P = 0.004, Table 1). Compared with the nevi samples, the melanoma samples had a higher percentage of cases that showed the highest as well as the lowest proportion of cells having cytoplasmic staining of RARβ and RXRβ; furthermore, cases with RXRα cytoplasmic localization were higher in the melanoma samples. Melanoma samples also exhibited significantly higher cytoplasmic intensity of RARβ (P = 0.0002) and RXRα (P = 0.018), but cytoplasmic intensity of RXRβ did not differ significantly between the two groups (Table 1). Nevi samples showed a significantly higher percent of cells showing nuclear localization of both RARγ (P < 0.0001) and RXRα (P < 0.0001; Table 2).

Table 2.

Nuclear marker distribution by diagnostic group

VariableDiagnosisnValue, n (%)
P
0123
Nuclear percent RARγ Melanoma 107 90 (84.1) 14 (13.1) 3 (2.8) 0 (0.0) <0.0001 
 Nevi 32 14 (43.8) 14 (43.7) 4 (12.5) 0 (0.0)  
 Melanoma/metastasis 134 110 (82.1) 19 (14.2) 4 (3.0) 1 (0.8) <0.0001 
 Nevi 32 14 (43.7) 14 (43.8) 4 (12.5) 0 (0.0)  
Nuclear percent RXRα Melanoma 108 82 (75.9) 19 (17.6) 7 (6.5) 0 (0.0) 0.0208 
 Metastasis 31 17 (54.8) 7 (22.6) 7 (22.6) 0 (0.0)  
 Melanoma 108 82 (75.9) 19 (17.6) 7 (6.5) 0 (0.0) <0.0001 
 Nevi 33 11 (33.3) 8 (24.2) 14 (42.4) 0 (0.0)  
 Melanoma/metastasis 139 99 (71.2) 26 (18.7) 14 (10.1) 0 (0.0) <0.0001 
 Nevi 33 11 (33.3) 8 (24.2) 14 (42.4) 0 (0.0)  
VariableDiagnosisnValue, n (%)
P
0123
Nuclear percent RARγ Melanoma 107 90 (84.1) 14 (13.1) 3 (2.8) 0 (0.0) <0.0001 
 Nevi 32 14 (43.8) 14 (43.7) 4 (12.5) 0 (0.0)  
 Melanoma/metastasis 134 110 (82.1) 19 (14.2) 4 (3.0) 1 (0.8) <0.0001 
 Nevi 32 14 (43.7) 14 (43.8) 4 (12.5) 0 (0.0)  
Nuclear percent RXRα Melanoma 108 82 (75.9) 19 (17.6) 7 (6.5) 0 (0.0) 0.0208 
 Metastasis 31 17 (54.8) 7 (22.6) 7 (22.6) 0 (0.0)  
 Melanoma 108 82 (75.9) 19 (17.6) 7 (6.5) 0 (0.0) <0.0001 
 Nevi 33 11 (33.3) 8 (24.2) 14 (42.4) 0 (0.0)  
 Melanoma/metastasis 139 99 (71.2) 26 (18.7) 14 (10.1) 0 (0.0) <0.0001 
 Nevi 33 11 (33.3) 8 (24.2) 14 (42.4) 0 (0.0)  

Primary melanoma versus metastasis. Primary melanoma samples (with or without metastasis) and metastasis samples differed significantly in the percentage of cells expressing RARα, RARβ, RXRα, and RXRβ in the cytosol (Table 1). Compared with the metastasis samples, the melanoma samples tended to have a higher percent cytoplasmic expression of RARα (P = 0.004) and RXRβ (P = 0.009). In addition, the primary melanoma samples tended to have a higher cytoplasmic percentage of RARβ (P < 0.0001) and cytoplasmic intensity of RARβ (P < 0.0001). In the primary melanoma samples, cytoplasmic percentage of RXRα (P = 0.004) was significantly higher, whereas nuclear percentage of RXRα was significantly lower (P = 0.021, Table 2) than in the metastasis samples.

Melanoma/metastasis versus nevi. Melanoma/metastasis samples differed significantly from nevi samples for percentage of cells showing cytoplasmic RARβ (P < 0.0001) and RXRβ (P = 0.017, Table 1). Compared with the nevi samples, the melanoma/metastasis samples had a higher percentage of samples that showed the highest as well as the lowest expression of each marker. Cytoplasmic intensity of RARβ was also significantly higher in the melanoma/metastasis versus nevi samples (P = 0.008, Table 1), but no significant group difference was detected for cytoplasmic intensity of RXRβ. Nevi samples showed a significantly higher percentage of cells with nuclear immunoreactivity of both RARγ (P < 0.0001) and RXRα (P < 0.0001; Table 2).

Expression of retinoid receptors and patient characteristics. Statistically significant associations were found between the location of the melanocytic lesion and cytoplasmic immunoreactivity for RARα (P = 0.005) and RARβ (P = 0.001). Melanocytic lesions from acral and mucosal regions had higher proportion of patients showing loss of RARα (41.67%) compared with that of lesions of the head and neck/arms (11.11%) or trunk/legs (7.5%). Similarly, acral and mucosal lesions had a higher proportion of patients showing loss of RARβ (18.2%) compared with that of lesions of the head and neck (8.3%) or trunk or legs (4.8%). Perineural invasion was noted for only two samples, both of which showed loss or low expression of RARα in cytosol (P = 0.03). Loss of RARβ was observed in 2 of 11 lesions (18.2%) with vascular invasion compared with 3 of 81 lesions (3.7%) with no vascular invasion noted (P = 0.034). Lesions with predominant spindle cell morphology were more likely to show loss of cytoplasmic RARβ (P = 0.018). Lesions having vertical growth also showed loss of nuclear staining for RARγ (P = 0.048). A higher percentage of RXRα cytoplasmic immunoreactivity was observed in the lesions with higher Clark levels (P = 0.005) and showing vertical growth pattern (P = 0.001). In addition, regressed melanocytic lesions were more likely to have a higher proportion of cells showing RXRα in the nucleus (P = 0.045). Loss of or low cytoplasmic staining for RXRγ was more prevalent among the five lesions with satellitosis (60%) than among the 59 lesions without satellitosis (36%; P = 0.014).

Retinoid receptors associated with survival time. As expected, analysis of samples of primary or metastatic lesions together revealed that patients with Clark levels IV-V lesions had 2.61 times greater hazard of disease-specific death [95% confidence interval (95% CI), 1.17-5.79; P = 0.019; Table 3] compared with that of patients with Clark levels II-III lesions. Examination of various retinoid receptors suggested that patients with lesions of low cytoplasmic RARγ immunoreactivity (scores 0 or 1) had 3.19 (95% CI, 1.32-7.75) times greater hazard of death (P = 0.010, Table 4) and 2.93 (95% CI, 1.21-7.11) times greater hazard of disease-specific death (P = 0.017, Table 3) compared with that of patients with a high cytoplasmic percentage of RARγ (scores 2 or 3). RXRα did not show an association with disease-specific survival, but was significantly associated with overall survival. Patients with lesions that showed a low percentage of cytoplasmic RXRα labeling (scores 0-2) had a significantly increased hazard of death (HR, 1.99; 95% CI, 1.05-3.78; P = 0.035; Table 4) compared with that of patients with a high percentage (score 3) of cytoplasmic RXRα labeling. Conversely, patients with a higher percentage of cells showing nuclear expression for RXRα had nearly thrice (HR, 2.93; 95% CI, 1.16-7.39; Table 3) the hazard of death as did patients with 0-5% of nuclear RXRα (P = 0.023). For disease-specific survival only, patients who had lesions with negative cytoplasmic intensity of RARβ had 2.22 (95% CI 1.18-4.17) times the hazard of death as did patients with lesions that showed positive RARβ cytoplasmic intensity (P = 0.014).

Table 3.

Cox proportional-hazards regression models for disease-specific survival

VariableLevelnHR (95% CI)P
Breslow thickness (mm)  99 1.06 (0.99-1.14) 0.086 
Mitoses  77 1.08 (0.97-1.20) 0.159 
Clark level IV-V 46 2.61 (1.17-5.79) 0.019 
 II-III 49 1.00  
Cytoplasmic intensity RARβ+ 66 2.22 (1.18-4.17) 0.014 
 1-2 52 1.00  
Cytoplasmic percent RARγ+ 0-1 10 2.93 (1.21-7.11) 0.017 
 2-3 98 1.00  
Cytoplasmic percent RARα/RXRα* −/− 13 2.96 (1.25-7.02) 0.014 
 −/+ 17 0.51 (0.12-2.23) 0.371 
 +/− 25 0.87 (0.38-1.96) 0.730 
 +/+ 49 1.00  
VariableLevelnHR (95% CI)P
Breslow thickness (mm)  99 1.06 (0.99-1.14) 0.086 
Mitoses  77 1.08 (0.97-1.20) 0.159 
Clark level IV-V 46 2.61 (1.17-5.79) 0.019 
 II-III 49 1.00  
Cytoplasmic intensity RARβ+ 66 2.22 (1.18-4.17) 0.014 
 1-2 52 1.00  
Cytoplasmic percent RARγ+ 0-1 10 2.93 (1.21-7.11) 0.017 
 2-3 98 1.00  
Cytoplasmic percent RARα/RXRα* −/− 13 2.96 (1.25-7.02) 0.014 
 −/+ 17 0.51 (0.12-2.23) 0.371 
 +/− 25 0.87 (0.38-1.96) 0.730 
 +/+ 49 1.00  
*

RARα: 0-1, negative; 2-3, positive. RXRα: 0-2, negative; 3, positive.

Table 4.

Cox proportional-hazards regression models for overall survival

Independent variableLevelnHR (95% CI)P
Breslow thickness (mm)  99 0.99 (0.88-1.11) 0.845 
Mitoses  77 1.07 (0.95-1.20) 0.274 
Clark level IV-V 46 1.56 (0.72-3.38) 0.258 
 II-III 49 1.00  
Cytoplasmic Percent RARγ 0-1 10 3.19 (1.32-7.75) 0.010 
 2-3 98 1.00  
Cytoplasmic percent RXRα 0-2 45 1.99 (1.05-3.78) 0.035 
 67 1.00  
Nuclear percent RXRα 13 2.93 (1.16-7.39) 0.023 
 22 1.24 (0.57-2.70) 0.591 
 77 1.00  
Nuclear percent RARα/RXRα* +/+ 16 2.74 (1.16-6.46) 0.021 
 −/+ 17 0.99 (0.38-2.54) 0.979 
 +/− 12 1.12 (0.41-3.06) 0.829 
 −/− 59 1.00  
Cytoplasmic percent RARα/RXRα −/− 13 3.43 (1.45-8.07) 0.005 
 −/+ 17 0.32 (0.07-1.39) 0.128 
 +/− 25 1.05 (0.47-2.37) 0.898 
 +/+ 49 1.00  
Independent variableLevelnHR (95% CI)P
Breslow thickness (mm)  99 0.99 (0.88-1.11) 0.845 
Mitoses  77 1.07 (0.95-1.20) 0.274 
Clark level IV-V 46 1.56 (0.72-3.38) 0.258 
 II-III 49 1.00  
Cytoplasmic Percent RARγ 0-1 10 3.19 (1.32-7.75) 0.010 
 2-3 98 1.00  
Cytoplasmic percent RXRα 0-2 45 1.99 (1.05-3.78) 0.035 
 67 1.00  
Nuclear percent RXRα 13 2.93 (1.16-7.39) 0.023 
 22 1.24 (0.57-2.70) 0.591 
 77 1.00  
Nuclear percent RARα/RXRα* +/+ 16 2.74 (1.16-6.46) 0.021 
 −/+ 17 0.99 (0.38-2.54) 0.979 
 +/− 12 1.12 (0.41-3.06) 0.829 
 −/− 59 1.00  
Cytoplasmic percent RARα/RXRα −/− 13 3.43 (1.45-8.07) 0.005 
 −/+ 17 0.32 (0.07-1.39) 0.128 
 +/− 25 1.05 (0.47-2.37) 0.898 
 +/+ 49 1.00  
*

Nuclear percent RARα/RXRα: RARα: 0, negative; 1-2, positive. RXRα: 0, negative; 1-2, positive.

Cytoplasmic percent RARα/RXRα: RARα: 0-1, negative; 2-3, positive. RXRα: 0-2, negative; 3, positive.

In the Cox proportional hazards models (Tables 3 and 4), RARα and RXRα in combination was significantly associated with overall survival and disease-specific survival. Patients with a loss of cytoplasmic expression for both of these markers had 3.43 times (95% CI, 1.45-8.07; P = 0.005; Table 4) the hazard of death overall and 2.96 times (95% CI, 1.25-7.02; P = 0.014; Table 3) the hazard of disease-specific death as did those with positive expression of both markers. The median overall survival for patients negative for both markers in the cytoplasm (median = 4.5 years, 95% CI, 0.9-5.0) was significantly lower than that for all other groups (P = 0.002; Table 5; Fig. 3A). Overall survival differences between patients positive for one or both of the markers in the cytoplasm did not differ significantly in pairwise comparisons. This same pattern with regard to cytoplasmic expression of RARα and RXRα in combination was seen for disease-specific survival (P = 0.014; Fig. 3B).

Table 5.

Median survival by cytoplasmic expression of RARα and RXRα

SurvivalMarker combinationTotalDeathsMedian, y (95% CI)Log-rank test P value
Overall Cytoplasmic RARα−/RXRα− 13 4.5 (0.9, 5.0) 0.002 
 Cytoplasmic RARα+/RXRα− 25 8.3 (5.2, not attained)  
 Cytoplasmic RARα−/RXRα+ 17 Not attained (9.3, not attained)  
 Cytoplasmic RARα+/RXRα+ 49 17 13.1 (7.1, not attained)  
Disease specific Cytoplasmic RARα−/RXRα− 13 1.8 (0.7, 5.0) 0.014 
 Cytoplasmic RARα+/RXRα− 25 8.3 (5.2, not attained)  
 Cytoplasmic RARα−/RXRα+ 17 Not attained (not attained, not attained)  
 Cytoplasmic RARα+/RXRα+ 49 16 8.1 (5.2, not attained)  
SurvivalMarker combinationTotalDeathsMedian, y (95% CI)Log-rank test P value
Overall Cytoplasmic RARα−/RXRα− 13 4.5 (0.9, 5.0) 0.002 
 Cytoplasmic RARα+/RXRα− 25 8.3 (5.2, not attained)  
 Cytoplasmic RARα−/RXRα+ 17 Not attained (9.3, not attained)  
 Cytoplasmic RARα+/RXRα+ 49 17 13.1 (7.1, not attained)  
Disease specific Cytoplasmic RARα−/RXRα− 13 1.8 (0.7, 5.0) 0.014 
 Cytoplasmic RARα+/RXRα− 25 8.3 (5.2, not attained)  
 Cytoplasmic RARα−/RXRα+ 17 Not attained (not attained, not attained)  
 Cytoplasmic RARα+/RXRα+ 49 16 8.1 (5.2, not attained)  
Fig. 3.

Kaplan-Meier estimation of survival probability of melanoma patients based on RAR and RXR status. The median survival time of patients negative for both cytoplasmic RARα and RXRα was (A) 4.5 y for overall survival and (B) 1.8 y for disease-specific survival.

Fig. 3.

Kaplan-Meier estimation of survival probability of melanoma patients based on RAR and RXR status. The median survival time of patients negative for both cytoplasmic RARα and RXRα was (A) 4.5 y for overall survival and (B) 1.8 y for disease-specific survival.

Close modal

In addition, nuclear expression of this marker combination (RARα and RXRα) was significantly associated with overall but not disease-specific survival; however, it is the positive expression of both markers that showed an increased hazard of death (HR, 2.74; 95% CI, 1.16-6.46; P = 0.021; Table 4) compared with that of patients negative for expression of both markers. The median overall survival for patients positive for both markers in the nucleus was 5.2 years (95% CI, 3.6-8.1) compared with 13.1 years (95% CI, 8.0-not attained) for patients negative for both markers (P = 0.009).

The study evaluates the expression of various subtypes of retinoid receptors in melanocytic lesions. It shows that there is abnormal expression of these receptors in various stages of melanomas and their metastasis. Furthermore, we were able to show a multifaceted relationship between aberrant expressions of RAR and RXR subtypes in melanoma patients, suggesting involvement of cross-talk among these receptors in the tumorigenesis of this disease.

We show in this study that there is a significant loss of cytoplasmic expression of RARα protein in melanomas with metastasis in comparison to those melanomas that had not metastasized at last follow-up. We were not able to detect any nuclear labeling for RARβ protein; there was only weak cytoplasmic labeling in the nevi or melanoma cells. Our finding corroborates earlier observations by Boehm et al. (8) showing loss of expression of RARβ protein in melanoma cells. Loss of expression of RARβ has been observed in various premalignancies and malignant carcinomas (head and neck, lung, esophagus, mammary glands, pancreas, cervix; ref. 14). In melanoma cells, induction of RARβ blocked the cellular proliferation and induced apoptosis (15), supporting the notion that it is a target receptor in many cancers and might also be acting as a tumor suppressor in melanoma. RARγ, a predominant retinoid receptor subtypes in skin, was found to be down-regulated in melanoma lesions as compared with the nevi. In mouse, loss of RARγ predisposed keratinocytes to v-Ha–Ras-induced squamous cell carcinoma (16). It is noteworthy that melanocytes expressing an activated Ha-ras in the TPras transgenic mice are susceptible to induction of melanoma by 7,12-dimethylbenz(a)anthracene (17). It has been shown that RARγ has a critical role in a genetic switch between normal melanocytes and melanoma (15), and the RARγ agonist causes additional differentiation toward a melanocytic phenotype and consistently inhibited growth (18). In the light of above studies, our observations showing loss of RARβ and RARγ in benign nevi suggest a plausible predisposition of benign lesions toward the development of melanoma.

In the present study, some melanoma cells showed intense RXRα cytoplasmic staining accompanied by decreased immunoreactivity in their nuclei. Similar observation has also been reported by Takiyama et al. (19) in human thyroid carcinoma. The movement of transcription factors between the nucleus and cytoplasm is important in regulating their activity (5). Therefore, entrapment of RXRα in the cytoplasm might be preventing its heterodimerization with other nuclear receptors and, hence, effecting the gene expression. Furthermore, a recent report showed that mice lacking RXRα in epidermal keratinocytes resulted in the progression of benign nevi to invasive human-melanoma–like tumors (20).

Because differential expression of RARs and RXRs was observed in the various stages of melanoma, we wanted to see whether these receptors could be used as potential prognostic markers in melanoma. We found that cytoplasmic localization of RXRα was correlated with better prognosis in melanoma patients. This could be explained by the virtue of RXR nongenotropic pathway, which involves cytoplasmic translocalization of RXRα from the nucleus resulting in the inhibition of cell growth and induction of apoptosis in various cancer cells (21, 22). Therefore, it is possible that those melanoma patients who had more cytoplasmic RXRα did also have cell growth inhibition. In continuation, aberrant expression of more than one receptor subtype in this cohort of patient was associated with poor prognosis of the disease. Because most of these receptors transcend the effect of ligands as heterodimers, it seems plausible that combined loss of receptor expression may render retinoids unable to turn on normal cellular programs. However, patients with RARα/RXRα+ phenotype had better prognosis than any other phenotypes. RXRs are known to bind with other members of the nuclear receptor superfamily [i.e., vitamin D, peroxisome proliferator-activated receptor (PPAR) etc.]. The expression of vitamin D receptors (VDR) and growth inhibition induced by 1,25-dihydroxyvitamin D3 have been noted in certain human malignant melanoma cell lines (23). PPARγ ligands, ciglitazone and PGJ2, inhibited melanoma cell proliferation in a dose-dependent manner (24). In continuation, a recent report by Grabacka et al. (25) shows that fenofibrate, a ligand of PPARα, could decrease metastatic potential of melanoma cells in vitro. Thus, it is plausible that in the subset of patients lacking expression of RARs, RXRα could be heterodimerizing with either VDR or PPAR and inducing growth inhibition via a totally different pathway. In fact, 9-cis-retinoic acid and 1α,25-dihydroxyvitamin D3 in combination showed growth inhibition of prostate cancer cells and of xenografts in nude mice (26).

Our results indicate that melanoma progression may be characterized by a simultaneous loss/decrease in the expression of various RAR and RXR receptors, which may then result in deficient RAR/RXR heterodimers. As a result, such anomalies would alter a variety of pathways under the control of retinoic acid (i.e.,) growth, differentiation, and genes that are regulated by these various receptors. Recently, a study from our group showed that melanocytes accumulate galectin-3 with tumor progression (27). Earlier, our group has also shown that retinoic acid inhibits the expression of galectin-3 and causes differentiation in mouse embryonal carcinoma F9 cells (28). Thus, loss of retinoid receptors leading to an impairment of retinoid pathway during the progression of melanoma could plausibly also be contributing toward the overexpression of galectin-3. Furthermore, widely coregulated down-regulation of expression of most of the retinoid receptor subclasses suggests a fundamental dysregulation of the retinoid pathway in melanoma. This aberration might be due to several reasons: first, melanomas have been shown to have overexpression of tyrosine kinase receptor that signal through the Ras-mitogen-activated protein kinase (Ras-MAPK) pathway (29). As the turnover of retinoid receptors is regulated by their phosphorylation status, it is plausible that in this study, decreased expression of these receptors might be due to their aberrant phosphorylation status by MAPKs. A recent report by Piu et al. (30) shows that RARβ2 is phosphorylated by extracellular signal-regulated kinase 2; second, in various cancers including melanoma loss of RARβ, expression has been correlated to the hypermethylated status of RARβ promoter (31) and/or to a deficient acetylation of histones (32), which in both cases result in repressed chromatin. Apparently, the presence of an epigenetically silent RARβ2 promoter correlates with the lack of RARα and causes resistance to the growth-inhibitory effect of retinoic acid. Conversely, restoring retinoic acid signal at epigenetically silent RARβ2 through RARα leads to RARβ2 reactivation (33). Indeed, in the current study, we have observed the loss of both RARα and RARβ proteins in melanocytic lesions.

In conclusion, we have comprehensively shown in a large cohort of patients for the first time the pattern of expression of various retinoid receptors at the protein level in nevi and primary and metastatic melanomas, encompassing different stages in development and progression of the disease. This study also underscores the possible prognostic significance of analysis of retinoid receptors in melanomagenesis because simultaneous loss of more than one retinoid receptor seems to be a significant independent unfavorable prognostic factor. It may also give a rationale of novel therapies involving retinoids/rexinoids.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank the Melanoma Tissue Core (partly supported by the Skin Cancer Specialized Programs of Research Excellence at M.D. Anderson Cancer Center P50 CA093459-02) for the clinical and histologic information on the melanocytic lesions included in the study.

1
Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer statistics, 2007.
CA Cancer J Clin
2007
;
57
:
43
–66.
2
Prieto VG, McNutt NS, Lugo J, Reed JA. Differential expression of the intermediate filament peripherin in cutaneous neural lesions and neurotized melanocytic nevi.
Am J Surg Pathol
1997
;
21
:
1450
–4.
3
Clark WH, Jr., Elder DE, Guerry DT, Epstein MN, Greene MH, Van Horn M. A study of tumor progression: the precursor lesions of superficial spreading and nodular melanoma.
Hum Pathol
1984
;
15
:
1147
–65.
4
Miller AJ, Mihm MC, Jr. Melanoma.
N Engl J Med
2006
;
355
:
51
–65.
5
Nigg EA. Nucleocytoplasmic transport: signals, mechanisms and regulation.
Nature
1997
;
386
:
779
–87.
6
Yasmin R, Williams RM, Xu M, Noy N. Nuclear import of the retinoid X receptor, the vitamin D receptor, and their mutual heterodimer.
J Biol Chem
2005
;
280
:
40152
–60.
7
Chambon P. A decade of molecular biology of retinoic acid receptors.
FASEB J
1996
;
10
:
940
–54.
8
Boehm N, Samama B, Cribier B, Rochette-Egly C. Retinoic-acid receptor β expression in melanocytes.
Eur J Dermatol
2004
;
14
:
19
–23.
9
Lotan R, Lotan D. Stimulation of melanogenesis in a human melanoma cell line by retinoids.
Cancer Res
1980
;
40
:
3345
–50.
10
Fligiel SE, Inman DR, Talwar HS, Fisher GJ, Voorhees JJ, Varani J. Modulation of growth in normal and malignant melanocytic cells by all-trans retinoic acid.
J Cutan Pathol
1992
;
19
:
27
–33.
11
Halpern AC, Schuchter LM, Elder DE, et al. Effects of topical tretinoin on dysplastic nevi.
J Clin Oncol
1994
;
12
:
1028
–35.
12
Levine N, Meyskens FL. Topical vitamin-A-acid therapy for cutaneous metastatic melanoma.
Lancet
1980
;
2
:
224
–6.
13
Chakravarti N, Saadati HG, El-Naggar AK, Lotan R, Diba RE, Esmaeli B. Retinoid receptor subtypes in sebaceous cell carcinoma of the eyelid.
Ophthal Plast Reconstr Surg
2005
;
21
:
292
–7.
14
Xu XC, Wong WY, Goldberg L, et al. Progressive decreases in nuclear retinoid receptors during skin squamous carcinogenesis.
Cancer Res
2001
;
61
:
4306
–10.
15
Spanjaard RA, Ikeda M, Lee PJ, Charpentier B, Chin WW, Eberlein TJ. Specific activation of retinoic acid receptors (RARs) and retinoid X receptors reveals a unique role for RARγ in induction of differentiation and apoptosis of S91 melanoma cells.
J Biol Chem
1997
;
272
:
18990
–9.
16
Chen CF, Goyette P, Lohnes D. RARγ acts as a tumor suppressor in mouse keratinocytes.
Oncogene
2004
;
23
:
5350
–9.
17
Broome Powell M, Gause PR, Hyman P, et al. Induction of melanoma in TPras transgenic mice.
Carcinogenesis
1999
;
20
:
1747
–53.
18
Emionite L, Galmozzi F, Raffo P, Vergani L, Toma S. Retinoids and malignant melanoma: a pathway of proliferation inhibition on SK MEL28 cell line.
Anticancer Res
2003
;
23
:
13
–9.
19
Takiyama Y, Miyokawa N, Sugawara A, et al. Decreased expression of retinoid X receptor isoforms in human thyroid carcinomas.
J Clin Endocrinol Metab
2004
;
89
:
5851
–61.
20
Indra AK, Castaneda E, Antal MC, et al. Malignant transformation of DMBA/TPA-induced papillomas and nevi in the skin of mice selectively lacking retinoid-X-receptor α in epidermal keratinocytes.
J Invest Dermatol
2007
;
127
:
1250
–60.
21
Cao X, Liu W, Lin F, et al. Retinoid X receptor regulates Nur77/TR3-dependent apoptosis [corrected] by modulating its nuclear export and mitochondrial targeting.
Mol Cell Biol
2004
;
24
:
9705
–25.
22
Lin XF, Zhao BX, Chen HZ, et al. RXRα acts as a carrier for TR3 nuclear export in a 9-cis-retinoic acid-dependent manner in gastric cancer cells.
J Cell Sci
2004
;
117
:
5609
–21.
23
Evans SR, Houghton AM, Schumaker L, et al. Vitamin D receptor and growth inhibition by 1,25-dihydroxyvitamin D3 in human malignant melanoma cell lines.
J Surg Res
1996
;
61
:
127
–33.
24
Placha W, Gil D, Dembinska-Kiec A, Laidler P. The effect of PPARγ ligands on the proliferation and apoptosis of human melanoma cells.
Melanoma Res
2003
;
13
:
447
–56.
25
Grabacka M, Plonka PM, Urbanska K, Reiss K. Peroxisome proliferator-activated receptor α activation decreases metastatic potential of melanoma cells in vitro via down-regulation of Akt.
Clin Cancer Res
2006
;
12
:
3028
–36.
26
Ikeda N, Uemura H, Ishiguro H, et al. Combination treatment with 1α,25-dihydroxyvitamin D3 and 9-cis-retinoic acid directly inhibits human telomerase reverse transcriptase transcription in prostate cancer cells.
Mol Cancer Ther
2003
;
2
:
739
–46.
27
Prieto VG, Mourad-Zeidan AA, Melnikova V, et al. Galectin-3 expression is associated with tumor progression and pattern of sun exposure in melanoma.
Clin Cancer Res
2006
;
12
:
6709
–15.
28
Lu Y, Amos B, Cruise E, Lotan D, Lotan R. A parallel association between differentiation and induction of galectin-1, and inhibition of galectin-3 by retinoic acid in mouse embryonal carcinoma F9 cells.
Biol Chem
1998
;
379
:
1323
–31.
29
Shen SS, Zhang PS, Eton O, Prieto VG. Analysis of protein tyrosine kinase expression in melanocytic lesions by tissue array.
J Cutan Pathol
2003
;
30
:
539
–47.
30
Piu F, Gauthier NK, Wang F. β-Arrestin 2 modulates the activity of nuclear receptor RAR β2 through activation of ERK2 kinase.
Oncogene
2006
;
25
:
218
–29.
31
Hoon DS, Spugnardi M, Kuo C, Huang SK, Morton DL, Taback B. Profiling epigenetic inactivation of tumor suppressor genes in tumors and plasma from cutaneous melanoma patients.
Oncogene
2004
;
23
:
4014
–22.
32
Suh YA, Lee HY, Virmani A, et al. Loss of retinoic acid receptor β gene expression is linked to aberrant histone H3 acetylation in lung cancer cell lines.
Cancer Res
2002
;
62
:
3945
–9.
33
Ren M, Pozzi S, Bistulfi G, Somenzi G, Rossetti S, Sacchi N. Impaired retinoic acid (RA) signal leads to RARβ2 epigenetic silencing and RA resistance.
Mol Cell Biol
2005
;
25
:
10591
–603.