Expression levels of vimentin and p21 correlate in NB cell lines and tumor samples with low p21 expression being associated with worse prognosis in high risk patients
We initially observed that levels of vimentin and p21 were correlated at the protein level in S-type-like NB cell lines [4]. To test whether this was also true at the mRNA level, ten NB cell lines were tested by real-time quantitative PCR (Q-PCR). As seen in Figure 1A, the three S-type-like cell lines expressed high levels of vimentin and p21 mRNAs. Although more heterogeneous, expression levels were consistently lower in the six N-type-like cell lines tested.
We then measured mRNA levels for both vimentin and p21 in NB tumor samples obtained at diagnosis from 77 children (all stages; see Additional file 1: Table S1). We found a highly significant correlation between the mRNA levels of the two genes (Spearman r = 0.5787, 95% confidence interval (CI) = 0.4019 to 0.7139, p < 0.0001) (Figure 1B). Although weaker, this correlation held true when the 40 stage 4 NB tumor samples were considered separately (Spearman r = 0.3719, 95% CI = 0.0587 to 0.6184, p = 0.0181) (Figure 1C, left panel). We thus conclude that vimentin and p21 mRNA levels are correlated in both NB cell lines and tumor samples.
Loss of p21 expression has been reported to be a poor prognostic factor in colorectal carcinomas and small-cell lung cancers [11]. In the presently used NB cohort, no statistically significant association was identified between p21 mRNA levels in tumor cells and any available clinical data (age at diagnosis, INSS stage, MYCN amplification status) (data not shown). Many reports have indicated that clinical and genomic tumor criteria allow a prognostic classification of NB [12, 13]. However, no stratifying prognostic factor has been proposed so far for high risk NB (MYCN-amplified and stage 4 NB of patients > 12 months). We here analyzed p21 tumor expression levels in relation with the overall survival (OS) of the 40 high risk NB patients. Despite the limited size of the cohort, two groups of patients were identified using the Log-rank Kaplan-Meier estimation (Figure 1C, right panel) that appeared to exhibit different survival curves according to the "high" or "low" p21 mRNA expression levels in the corresponding tumors (p = 0.2141, Hazard Ratio (HR) = 1.598, 95% CI = 0.7610 to 3.384), the median tumor p21 mRNA level of the cohort being chosen as the cut-off between high (> median) and low (< median) levels of expression. The patients with low p21 tumor expression at diagnosis had a 24.6 months median survival compared with 37.2 in high p21 tumor expressors.
We analyzed the relationship between p21 mRNA expression in stage 4 NB tumors and overall survival using microarray gene expression data from the R2 repository http://hgserver1.amc.nl/cgi-bin/r2/main.cgi. While the p21 probe set #202284 did not give significant results, the p21 probe set #1555186 revealed that low levels of p21 mRNA are highly significantly associated with a reduced OS (data not shown), consistent with the tendency we observed.
Altogether, these results suggest that our finding would be of great interest to orientate therapeutic decisions, if confirmed in larger groups of high risk NB patients.
Vimentin participates in p21 transcriptional regulation
To further explore the relationship between expression of vimentin and p21, we then asked whether vimentin could regulate p21 expression. A pool of siRNAs was used to inhibit vimentin expression in the SK-N-SH S-type-like NB cell line (Figure 2A). First of all, it is worth noting that vimentin depletion was not accompanied by any cellular morphological changes (data not shown). Q-PCR analysis (upper panel) revealed that p21 and vimentin mRNA levels decreased concomitantly. Whereas p21 was strongly induced in vimentin expressing control cells whose DNA was damaged following NCS treatment, cells inhibited for vimentin expression using siRNAs proved unresponsive to NCS with hardly any increase in p21 expression (lower panel). We also generated SK-N-SH clones stably expressing a shRNA directed against a vimentin mRNA sequence different from that targeted by the siRNAs. Four SK-N-SH shRNA-Vim clones were selected with various protein expression levels of vimentin. Strikingly, p21 protein levels were found to strictly correlate with those of vimentin (r = 0.97, see Additional file 2, Figure S1). Altogether, these results suggest that vimentin may directly or indirectly regulate p21 expression.
As vimentin has been reported to bind nucleic acids [7, 14], we then used confocal immunofluorescence microscopy to determine the cellular distribution of vimentin in NB cell lines. The two cell lines strongly expressing vimentin, SK-N-SH and GI-M-EN, displayed much stronger staining than the SJN-B1 and Kelly cell lines, which produce little or no vimentin, respectively (see Additional file 2, Figure S2). In addition to the classical dense cytoplasmic staining for filamentous vimentin, we observed weak but nonetheless evident staining for vimentin in the nucleus of SK-N-SH and GI-M-EN cells (see Additional file 2, Figure S2). We also observed some interesting colocalization of vimentin with the nuclear matrix lamin A/C protein on the inner side of the nuclear membrane (see Additional file 3, Figure S3). We then used a biochemical approach for the detection of nuclear and cytoplasmic soluble proteins by immunoblotting. Soluble vimentin was detected in nuclear extracts in all three cell lines strongly expressing vimentin (SK-N-SH, CHP-212, GI-M-EN), and, to a lower extent, in the SJN-B1 cell line, which contained smaller amounts of vimentin, but no soluble vimentin was detected in the nucleus of the vimentin-negative Kelly cell line (Figure 2B). However, we cannot formally exclude the potential binding of vimentin to the outside of the nucleus that could produce its inclusion in the nuclear fraction.
Being potentially partly localized in the nucleus, vimentin may well behave like a bona fide transcription factor in these NB cells. To test this hypothesis, the SK-N-SH, CHP-212 or GI-M-EN cell lines which express high levels of p21 were transfected with plasmid vectors containing the luciferase reporter gene downstream from either 2400 bp of the p21 gene promoter (pWaf1-luc; [10]) or a minimal gene promoter used as a control (pGL3-luc). In all three cell lines, p21 promoter activity depended upon the expression of vimentin as demonstrated using a vimentin siRNAs pool in comparison with control siRNAs (Figure 2C, upper panel). Conversely, co-transfecting the luciferase reporter vectors along with a plasmid directing vimentin expression demonstrated that the transient overexpression of vimentin induced a significant, but moderate, increase in p21 promoter activity in vimentin-negative or low-expressing cell lines Kelly and SJN-B1, respectively. The inducing effect was even stronger in the already high vimentin-expressing cell line SK-N-SH (Figure 2C, lower panel). Transfection efficiency was similar (approximately 35% of cells transfected, data not shown) in all three cell lines as revealed by flow cytometry analysis of cells co-transfected with a GFP expression vector and the pcDNA3.1-VIM plasmid. That vimentin was indeed overexpressed in all three cell lines, albeit to different extents, was demonstrated by immunoblotting analysis shown in Figure 2D (upper panel). Finally, the vimentin negative Kelly cell line and the SK-N-SH cell line which displayed high levels of vimentin were transiently transfected with an empty vector or with a construct encoding vimentin. We then fractionated the cells into cytoplasmic soluble, nuclear soluble and cytoskeletal/matrix-associated insoluble fractions. In all cases where vimentin was expressed (at the basal level or after transfection), most of it was detected in the insoluble fraction, as expected for the cytoplasmic intermediate filament form of the protein (Figure 2D, lower panel), but some vimentin was also detected in the nuclear soluble fraction.
From our results, we conclude that vimentin regulates p21 gene transcription, directly or indirectly, through promoter activation. Indeed, it remains unclear whether vimentin regulates p21 gene transcription through direct or indirect effects on its promoter, and further experiments are required to address this issue. Vimentin could have an indirect effect by modifying the chromatin organization of the promoter, as suggested by its specific binding to chromatin-modifying proteins, such as the HBO1 histone acetyltransferase or the SETDB1 histone methyltransferase, by acting as transcription cofactor, as suggested by its binding to the PIASγ transcription regulator [15], or by modifying the global organization of the nucleus, as we observed colocalization with the lamin A/C proteins known to regulate nuclear organization [16].
Cytoskeletal proteins thus appear to be endowed with dual functions, both in maintaining cell architecture and in regulating gene expression. Actin has been known for some years to regulate transcription [17]. More recently, some lamins have been reported to perform similar function [18] but to our knowledge, there has been no previous demonstration of a gene regulating function for vimentin.
In the context of neuroblastoma oncogenesis, the expression of the vimentin and p21 genes is regulated during neuronal differentiation. The vimentin gene is expressed early in neural crest cells development [19] and is involved in initiating neurite outgrowth [20, 21], and p21 is also known to be induced during neural development [22]. Strong vimentin and p21 expression in some tumors may indicate that oncogenesis occurred at a specific time point in neural crest cell development. Low or no vimentin and p21 in other tumors may indicate that oncogenesis took place at a different stage of differentiation of the tumor progenitors.