Young MeLiM developed melanoma metastases in lymph nodes, liver, heart and lung. To isolate melanoma cells from lung metastases, primary cultures of tumors were performed under conditions optimised for pig melanocyte proliferation 7. After 48 hr in culture, adherent cells were predominantly melanocytes. SAGE libraries were constructed from 2.5 millions of these metastatic melanoma cells (MMC) and control PigMel normal melanocytes (NM). A total of 11,700 and 11,300 tags were sequenced from the MMC and NM libraries, corresponding to 6,131 and 5,466 different tags (transcripts), respectively. Our data have been deposited at NCBI's Gene Expression Omnibus 8 and are accessible through the GEO Series number GSE5982.

To identify genes potentially involved in malignant progression, we compared the two libraries. Monte Carlo simulations yielded 70 tags statistically significant at p value < 0.05. Fifty-five (79%) matched expressed sequence tags (EST), the remaining 15 (21%) tags presented no matches. A majority of tags matched genes expressed at high levels. Among the EST, 39 (56% of tags) matched to known cDNAs, the remaining 16 (23%) could not be identified. The identified genes are involved in RNA processing and protein synthesis (20% of the 70 tags), transcription (7%), signalling (4%) and the rest corresponded to scattered functional classes (24%). The list of tags increased and decreased in MMC compared to NM, arbitrarily ordered by functional classes of the genes they represent, are shown in Tables 1 and 2, respectively. Several of the genes in Table 1 have been shown to be differentially expressed in various tumors, compared to their normal counterparts. RPS12, Secernin, CDC10 were found to be upregulated in human colorectal tumors, gastric cancers, diffuse large B-cell lymphomas, respectively 91011. Similarly, genes listed in Table 2 like COXIII, were found to be downregulated in human glioblastoma 12. The mRNA of GNB2L1 corresponded to a tag abundant 31 and 13 counts in MMC and NM, respectively. We chose to study GNB2L1 expression in melanocytes and melanomas in more detail because GNB2L1 encodes RACK1, receptor for activated C kinase, whose mRNA was found to be up-regulated in human carcinomas 13.

To define the distribution pattern of RACK1 mRNA, we performed in situ hybridisation onto pig sections of normal skin, and on samples of cutaneous melanoma and metastatic melanoma samples from lung, liver and lymph nodes. To avoid background in heavily melanogenic tumor areas, a bleaching treatment was added. Film autoradiography obtained with the antisense and sense probes showed a faint signal of RACK1 mRNA in healthy tissues, except in the lymph nodes where the signal was strong (Figure 1A), as reported in human lymph nodes 13. By contrast, an intense signal was observed in tumoral regions of cutaneous melanoma, lung and liver metastasis samples; non-tumoral regions displayed a much lower signal (Figure 1A). The sense probe autoradiographic signal was almost negligible (Figure 1B). Darkfield illumination on emulsion autoradiography highlighted the silver grains on the tumoral region of lung melanoma (Figure 1C). These results confirmed the overexpression of RACK1 mRNA in melanoma, as predicted by our SAGE data.

On tissue sections, identification of melanoma cells was achieved with an antibody against the microphthalmia transcription factor MITF, which produces a specific nuclear signal in melanocytes. Immunohistochemistry showed that normal melanocytes in the basal layer of control skin epidermis, were labelled for MITF (Figure 2A). Tumoral cells in cutaneous melanoma as well as in metastases were also labelled with the MITF antibody (Figure 2B, C). Unspecific labelling was not detected with this antibody (Figure 2D–F). Thus, MITF is a sensitive marker of the melanocytic lineage, useful to study melanoma progression in the pig.

To explore whether overexpressed RACK1 mRNA was translated, we analysed the cellular distribution of RACK1 protein by confocal microscopy, with double immunostaining for MITF. In control Meishan and healthy MeLiM skins, RACK1 protein was expressed in the epidermis, and found in the cytoplasm of keratinocyte (Figure 3A–D). In MITF-positive (MITF+) melanocytes, RACK1 expression was not detected (Figure 3A, C). Consistent with this, when testing dopachrome tautomerase (DCT), a melanogenic enzyme restricted to melanosomes, double labelling of RACK1 and DCT did not overlap in normal skin (Figure 3D).

In cutaneous melanoma, nests of MITF+ cells expressed RACK1 protein (Figure 4A). The RACK1 signal was scattered in the cytoplasm, mostly in the perinuclear area (Figure 4A). In metastases of melanoma to lymph nodes, lung and heart, RACK1 protein was abundant on MITF+ cells (Figure 4B–D). The RACK1 signal was cytoplasmic in MMC. However, an additional labelling on nuclear punctae was observed in 15% of MMC (Figure 4B–D, yellow arrowheads). Thus, in MeLiM, RACK1 overexpression in tumoral tissues was observed in the cytoplasm of melanoma cells at different stages of progression, from cutaneous melanoma to melanoma metastases, with an additional nuclear localisation in MMC.

Nuclear translocation of RACK1 has already been recorded 14 and was shown to mediate the induction of BDNF 15. Since melanoma metastases express BDNF more frequently than primary melanomas 16, we checked whether BDNF would be expressed in MeLiM metastases. No BDNF expression was found in melanoma cells of MeLiM metastases (data not shown).

To further investigate the expression of RACK1 in human melanoma, we ascertained its presence in a series of samples. Formalin fixed tissue from 4 normal skin biopsies, 14 nevi, 5 cutaneous and 18 metastases of melanoma were analysed using MITF as melanocytic marker 17. The results are summarized in Table 3. In normal skin, RACK1 was present in negligeable amounts in melanocytes whereas adjacent keratinocytes displayed a strong cytoplasmic signal (Figure 5A). In nevi, RACK1 signal was heterogenous within a given nevus and between nevi showing two patterns: in 7 out of 14 nevi, RACK1 signal was either not detectable (arrowheads, Figure 5B–D) or faint (arrow, Figure 5D). When detected in epidermal or dermal melanocytes, the signal was essentially membranous. In the 7 other nevi, a cytoplasmic or perinuclear signal was observed in nests of epidermal (arrow, Figure 5C) as well as in dermal melanocytes. By contrast, in cutaneous melanoma, whether superficial spreading melanoma or recurrent dermal melanoma, MITF+ fusiform dermal cells as well as dermal nest of melanocytic cells displayed a strong RACK1 cytoplasmic signal which was homogeneous over the whole lesion (Figure 5E, F respectively). In summary, RACK1 signal was low in nevi and very much increased in cutaneous melanoma.

In lymph nodes as well as in liver metastases, cells presented a strong granular, regular cytoplasmic pattern of RACK1 distribution underlining cell shape (Figure 6A–D). RACK1 overexpression was consistently observed in all MITF+ cells from each of the 23 malignant melanoma samples examined. No nuclear RACK1 labelling was found in these melanoma samples. In summary, overexpression of RACK1 was detected in human melanoma with no apparent changes between cutaneous lesions and metastatic melanoma.

Altogether, screening of human lesions indicates a differential expression of RACK1 in nevi and melanoma. In benign nevi, RACK1 signal was low and heterogeneous on melanocytic cells. By contrast, RACK1 signal presented two distinct features in melanoma: a dramatic increase of intensity and an homogeneous cytoplasmic distribution over the lesion.

Finally, we checked whether activated protein kinase C (PKC) was a partner of RACK1 in melanoma cells. Double labelling for MITF and phospho-PKCαβ showed a distinct increase in PKC signal in metastases compared to nevi (Figure 7A, B). When double labelling for RACK1 and phospho-PKCαβ was performed, both signals were observed in the cytoplasm of MMC (Figure 7C, D). Both proteins were expressed at higher levels in MMC compared to nevi and there was a partial cytoplasmic colocalisation. An additional nuclear signal for phospho-PKCαβ was detected in epidermal (not shown) and dermal (Figure 7C) melanocytes and in metastatic melanoma cells (arrow in Figure 7D). In cutaneous melanoma, phospho-PKCαβ signal was heterogeneous, lower or at the same level than in MMC (data not shown). These data suggest that PKCα and/or β are involved in the functional role of RACK1 in metastatic melanoma.

Taken together, these results demonstrate that RACK1 mRNA and protein are up-regulated in human melanomas as found in pig melanomas, and demonstrate a correlation between the melanoma tumoral status and high levels of RACK1.

Affected MeLiM males were mated with healthy Duroc or MeLiM sows at the National Institute for Agricultural Research (INRA, Jouy-en-Josas, France) 14. Animal care and use in this study were approved by the INRA ethics committee, in accordance with European Union standards. Biopsies from 3 month-old or younger MeLiM, bred either in France (n= 13) or in the Czech Republic (n = 3), were used. They included cutaneous melanoma consisting of superficial spreading melanoma (n = 2) and nodular melanoma (n = 7) and metastases in lymph nodes (n = 13), lung (n = 10), liver (n = 1), heart (n = 2) and spleen (n = 3), as well as healthy skin (n = 10). Samples of dorsal cervical skin of a healthy pigmented Meishan pig were used as controls. Collected tissues were fixed in 4% buffered paraformaldehyde (PFA) and embedded in paraffin.

Tumor biopsies of lung from a young MeLiM were used to isolate melanoma cells. Conditions for primary cultures of pig melanocytic cells were as described 7. TPA was added on the second day of culture. After 48 hours cells were rinsed, lysed in Dynabead mRNA direct kit binding buffer (Dynal, Invitrogen, Cergy Pontoise, France) and frozen in liquid nitrogen. Control melanocytes were the non transformed PigMel cells at passage 37 7.

Libraries were generated using the SAGE adaptation for downsized extracts (SADE) method using Sau3A as the anchoring enzyme, as described in 49 with a centrifugation of cell lysates to discard melanin. One thousand clones from each library were sequenced. Sequencing reactions were performed by MWG (Martinsried, Germany).

SAGE tags were extracted from sequence files and processed to remove duplicate ditags, linker sequences and repetitive tags using the SAGE 2002 version 4.5 software 6. Statistical significance was determined using Monte Carlo simulation analysis included in the SAGE software. A P value of less than 0.05 was considered significant. Tags were identified using the mammalian Genbank database 50 analysed by the SAGE software, or the 51. Pig RACK1 partial cDNA corresponding to the nucleotide sequence -70 to 900 bp from the ATG start codon was subcloned into pCR4TOPO plasmid (Invitrogen). The resulting plasmid was linearized with NotI or PmeI to obtain sense or antisense RNA probes, respectively. In vitro transcription was performed as described 52.

In situ hybridisation was performed as described 53 with modifications to bleach the sections. Briefly, deparaffinized 5 μm sections were treated for 15 minutes with 0.075% KMnO₄ and discoloured for 1 minute in 5% oxalic acid with brief rinses between and after treatments. Sections were fixed for 20 minutes in 4% PFA, rinsed, dehydrated and air-dried. Sense or antisense radiolabelled riboprobes at about 15 × 10⁶ cpm/ml were hybridised as described 52. Slides were exposed to Biomax MR films (Kodak, France) for 3 days, then dipped in Kodak NTB2 emulsion and exposed for 4 weeks.

Human melanoma tissues were obtained at the Curie Institute (Paris, France) from patients undergoing lymphadenectomy (n = 13), hepatectomy (n = 5) or epidermal cutaneous resection (n = 19). Cutaneous melanoma specimens consisted of 2 superficial spreading melanoma stages IV (Breslow depth of 2,35 and 2,5 cm) and 3 recurrent dermal cutaneous melanoma. Samples from 37 patients, 25 women and 12 men, were examined. Normal skin from breast plastic surgery was used as control (n= 4). Tissues were fixed either in 4% PFA (n = 38) or in Bouin fixative (n = 2) and embedded in paraffin

Mouse monoclonal antibodies were anti-MITF (1:50) (Zymed, Clinisciences, Montrouge, France) and anti-RACK1 (1:150) (Transduction Laboratories, BD Biosciences, Le Pont de Claix, France). Rabbit PEP-8 anti-DCT (1:1000) was obtained from Dr. Hearing and anti-phospho PKCαβ (1:50) was polyclonal (Cell Signaling, Ozyme, France). Cross reaction and specificity of the antibodies to pig tissue were checked by Western blot.

Antigen retrieval was performed by microwaving deparaffinized sections in citrate buffer, pH 6, for pig sections, or in Tris-EDTA, pH 9, for human sections. For immunohistochemistry, primary antibodies were reacted with the avidin-biotin complex (ABC Elite, Vector, Biovalley, France). For double immunofluorescence, antibodies applied overnight at 4°C were revealed with anti-mouse isotype or anti-rabbit antibodies, one labelled with Alexa Fluor 555, the other coupled to biotin and revealed with Alexa Fluor 488-labelled streptavidin (Molecular Probes, Invitrogen, France). Nuclear counterstaining was achieved with Topro 3 (Molecular probes). Sections were observed with a Leica laser TCS SP2 scanning confocal microscope producing 0.7 μm-thick optical sections. Controls without the first antibodies showed no unspecific labelling. Confocal images were processed with the computer program Leica Lite or Zeiss LSM Image Browser for orthogonal projections. All images shown are individual sections of z series, plus the orthogonal projections of the stack when indicated. Final Figures were assembled with Adobe Photoshop (Adobe Systems, USA).