To investigate if PIR deregulation was a common feature in human malignancies other than AML, we studied PIR protein expression in a collection of normal and transformed human tissues by immunohistochemistry (IHC) using a rabbit polyclonal antibody raised against human PIR. In accordance to previous data [1], PIR protein was expressed at very low levels in most tissues (data not shown). The sole exception was represented by a subgroup of samples deriving from nevi, primary and metastatic melanomas. We, therefore, analyzed by IHC a melanoma-specific tissue microarray (TMA), containing surgical samples from 117 primary melanomas and 57 metastatic melanomas. Strikingly, PIR localization was not homogeneous across all samples. Previous studies described PIR as a nuclear protein, predominantly localized within sub-nuclear dot-like structures [1]. Concordantly, normal melanocytes from healthy tissue sections showed PIR expression exclusively in the nucleus (Figure 1a), whereas melanoma samples expressing PIR showed a varied pattern of protein localization, ranging from nuclear to prevalently cytoplasmic (Figure 1b-e). Interestingly, cytoplasmic PIR appeared to correlate with melanoma progression (Figure 1f): in Radial Growth Phase (RGP) melanoma, which represents an early step of melanoma progression, PIR is localized in the nucleus in 50% of the cases, and shifts to the cytoplasm in 40% of cases, the remaining 10% having both nuclear and cytoplasmic expression. In more infiltrating Vertical Growth Phase (VGP) melanomas, the percentage of cases with nuclear PIR is of 11,5%, whereas cytoplasmic localization is detectable in 84,6% of cases and only a minor fraction (3,8%) has both nuclear and cytoplasmic staining. Finally, all tested metastatic melanomas express PIR in the cytoplasm, with some degree of nuclear staining in 12,5% of the cases only.
In summary, we found a shift of PIR sub-cellular localization from the nucleus to the cytoplasm in a relevant subset of melanoma samples; furthermore, the proportion of cases displaying cytoplasmic PIR increases with melanoma progression. The cytoplasmic localization is less frequent in RGP than in advanced VGP melanoma and, in the latter, the percentage of cases bearing cytoplasmic PIR is comparable to that of metastatic melanoma, where there is virtually no PIR in the nucleus. These data suggest that the shift of PIR localization from nucleus to cytoplasm may be a marker of melanoma progression.
Data obtained by IHC analysis of primary tumours are highly informative and reliable as to protein expression and localization in vivo. However, to further explore the relevance of PIR delocalization during progression of the malignancy, we decided to switch to primary cultures and cell lines derived from melanomas. In order to assess if PIR localization in in vitro experimental model systems is representative of the pattern identified in vivo, we analyzed PIR expression in 17 primary or established melanoma cell lines by means of a slightly modified ICA method [18, 19] to perform simultaneous analysis of several cell lines with different antibodies.
The cell lines included in this study represent primary melanomas with different growth patterns: RGP (WM35, WM1552C, WM1575), VGP (WM278, WM793B, WM902B, IGR39, WM115) and metastatic melanomas derived from different sites (skin, lymph nodes, visceral). Apart from two established cell lines, IGR37 and WM266-4, all other metastatic melanoma cells were of primary origin, isolated by mincing and dissociating surgical biopsies and cultured in complete medium. The melanoma origin of the cultures was proven by positivity to the S-100, HMB-45 and MART-1 staining (data not shown).
Cells were seeded on gelatin-coated multiwell-slides at comparable densities (50-60% confluence), and different antibodies were simultaneously hybridized to multiple spots on the array (Figure 2). An affinity-purified polyclonal anti-PIR antibody was used to assess PIR expression. A polyclonal antibody recognizing the C-terminus of the ShcA adaptor protein was used as control for exclusive cytoplasmic localization, while anti-PIR total serum and pre-immune serum were included in the experiment as further controls. Signals were revealed with Cy3 labelled goat anti-mouse or anti-rabbit secondary antibodies, and nuclei were stained with Diamino-2-phenylindole (DAPI, Figure 3a, b). Image analysis was performed using a specific freeware software (CellProfiler) [20] to obtain fluorescence values corresponding to nuclear and cytoplasmic staining for each single cell. The average values of fluorescence for each antibody were then calculated.
For each cell line, the PIR fluorescent signal was quantified cell-by-cell in the nucleus and in the cytoplasm. The relative fraction of nuclear versus cytoplasmic signal varied among the cell lines. On the contrary, the ShcA signal, as expected, had a unique cytoplasmic localization in all cell lines, thus providing a reliable reference in the subsequent analysis. The cell lines included in the array differ for morphology, size and level of PIR expression (Figure 3a); in order to eliminate the influence of these factors on the analysis of PIR distribution, for each cell we normalized PIR and ShcA cytoplasmic fluorescence to the total cellular fluorescence. Figure 3c shows the distribution of PIR and ShcA normalized cytoplasmic signals for all cell lines (higher values on the x-axis correspond to higher intensities of cytoplasmic fluorescence). With the exception of PIR signal in the WM1552 cell line, the normalized cytoplasmic fluorescence signals are symmetrically distributed across the populations. The cytoplasmic signals for ShcA are homogeneous among all cell lines (both primary and metastatic cells) and are centred on a mean value of 0.6, which we considered as a reference value for cytoplasmic protein localization. The distribution of PIR cytoplasmic signals are, instead, centred on lower mean values: 0,4 in primary melanoma cell lines and 0,5 in metastatic melanoma cell lines, suggesting that there is on average a higher amount of cytoplasmic PIR in metastatic melanoma cell lines, as previously observed on TMA samples using IHC. In the case of the WM1552 cell line, we observed a bimodal distribution of PIR, reflecting the presence of two subpopulations, one predominantly cytoplasmic and one with a higher level of nuclear localization. Comparable results were obtained using total anti-PIR serum, compared to pre-immune serum as a negative control (data not shown).
We next performed classical immunofluorescence (IF) experiments on five cell lines included in the array, as a validation of ICA results. PIR displayed a distinct nuclear localization in two primary melanoma cell lines tested, IGR39 and WM115 (Figure 4a, b respectively). Among metastatic melanoma cell lines, AdMa1955 and AnSe1935 displayed PIR staining predominantly in the cytoplasm (Figure 4d, e respectively), while CaCi1962 presented mixed nuclear and cytoplasmic PIR staining (Figure 4c). These results are in perfect agreement with those obtained by ICA for the same cell lines and confirm the technical reliability of ICA technology. Therefore, both techniques indicated a variable localization of PIR protein in melanoma cell lines, and, in accordance with IHC, suggest that a higher proportion of cytoplasmic PIR is characteristic of cell lines derived from more advanced stages of melanoma progression. The number of RGP and VGP samples analyzed by ICA is not sufficient for the identification of significant differences between the two categories; however, similarly to IHC, the ICA approach revealed progressive shift of PIR localization towards the cytoplasm associated to disease progression. Among the cell lines of metastatic origin, we did not find differences in localization associated to the different sites of metastasis, suggesting that cytoplasmic PIR may be a general characteristic of metastatic melanomas. This is in agreement with data obtained on surgical samples analyzed by IHC.