Lymph node status is the most important prognostic factor in patients with VC . Routine pathological staging is limited to the microscopic evaluation of H&E-stained LN sections. Unfortunately, small tumour foci may occasionally be missed by the pathologist in this examination . As stated by Regauer , sentinel LNs of VC patients with even single tumour cells should be regarded as positive. Thus, to improve detection of LN metastases, pathological ultrastaging should ideally be performed [19, 20]. This histological work-up of LNs involves serial sectioning and immunohistochemical analysis; the procedure is labour-intensive and time-consuming. Real-time RT-PCR technology is more sensitive, as it allows to detect a very small number of cells in larger volumes of previously pulverized (or homogenized) tissues. However, this method carries some disadvantages, for example it makes morphologic evaluation of the analyzed samples impossible. Therefore, a careful marker selection for RT-PCR analysis is a fundamental issue.
Since there are no molecular markers for diagnosing VC dissemination, we aimed to identify the potential mRNA markers of VC dissemination into LNs, and to verify their prognostic value by correlating their levels with time to groin recurrence. We used oligonucleotide microarrays which measure the expression level of over 47,000 transcripts and variants of human genes.
At first, the PUMA package, enabling a comparison between small-sample gene expression data  was employed, to analyze two samples obtained from one patient, who underwent a rapid progression of the disease during the follow-up time, with local recurrence, and groin recurrence in the same groin where the LN(+), and died due to cancer progression. The PUMA analysis produced a list of genes differentially expressed in LN(+) and LN(−). The most significantly over-represented gene ontology (GO) classes in the list included epidermal molecules. Thus, some of the differentially expressed genes are those specifically expressed in the primary tumour site. This finding testifies the methodology used for the comparison of the single microarray experiments. Similar results were obtained for the three additional pairs of LNs obtained from three patients.
Interestingly, gene expression signature distinguishing between the four pairs of histologically positive and negative LNs comprised no genes down-regulated in the metastatic LNs of the same patients. A possible explanation of this observation is that only transcripts that were highly up-regulated in the metastatic cells could be distinguished from the background of the numerous non-cancer cells as an “added value” while down-regulated transcripts, i.e. transcripts’ deprivation, would not be noticed in this background environment.
The consecutive steps of narrowing-down of the list of the differentially expressed in the four LNs obtained from the four VC patients genes led us to the choice of seven genes, namely PERP
JUP and CSTA, to be validated by the real time RT-PCR. Gene expression data of 27 non-cancer vulvar tissue samples deposited in the Genevestigator V3 system [22, 23] have shown that these seven genes are regularly up-regulated in vulvar tissue. In addition, Hsiao et al. , who created a compendium of tissue-selective genes based on microarray data, described five of these genes as “vulva-selective.” This strongly supports tissue specificity of the selected genes for the primary tumour site.
To our knowledge, this is the first time that these genes have been linked to VC dissemination. PERP, as a component of intercellular desmosome junctions, plays a role in epithelial integrity and cell-cell adhesion  and constitutes a proapoptotic transcriptional target of TP53 . PERP-deficiency promotes cancer by enhancing cell survival, desmosome loss, and inflammation . S100 calcium binding protein A8, S100A8, belongs to S100 proteins, a family of EF-hand signalling proteins . The complex of S100A8 and S100A9 called calprotectin induces a proinflammatory and thrombogenic response, and is involved in danger signalling; what is more, its stimulation results in a loss of cell–cell contacts [28, 29]. S100A8/S100A9 expression is increased in patients with various tumours, being involved in invasion and migration processes [30, 31]. S100A8/9 expression is minimal in normal epidermis and is elevated in skin diseases . However, Dell'oste et al.  showed that in HPV-immortalized keratinocytes S100A8/9 expression was downregulated. Fatty acid binding protein 5 (psoriasis-associated, FABP5), found in epidermal cells, was first identified to be up-regulated in psoriasis keratinocytes . In cancer cells, FABP5 increases cell proliferation and invasiveness, as recently demonstrated in oral squamous cell carcinoma  where its expression may be HPV-related . In non-cancer keratinocytes FABP5 elevation may be necessary for the activation of cell motility during epidermal wound healing . Stratifin, SFN (14-3-3 sigma), plays a role in various cellular processes being the most cancer-associated 14-3-3 isoform . 14-3-3 sigma expression is lost in numerous tumours , including VC , while its increased expression could be associated with a loss of TP53 function [40, 41]. However, Wang et al.  found high levels of SFN detected immunohistochemically in over 70% of analyzed VC tumors, significantly correlating to large tumor diameter and deep invasion. Importantly, 14-3-3 sigma protein is present in various normal epithelia and absent in LNs [43, 44], and lymphocytes may express its mRNA at relatively low levels . Carbonic anhydrase XII is one of the tumour-associated carbonic anhydrases. The enzyme is overexpressed under hypoxic conditions and constitutes a possible target for anticancer therapy [46, 47]. Junction plakoglobin, JUP, a member of the catenin family, is present both in desmosomes and in intermediate junctions . The majority of studies suggest that plakoglobin has a tumour suppressor role . Cystatin A (stefin A), CSTA, functions as a cysteine protease inhibitor and plays a role in epidermal development and maintenance. Serum level of CSTA has been proposed as a tool predicting nodal stage and poor prognosis in nasopharyngeal carcinoma .
The comparison of LNs(+) with LNs(−) confirmed that LNs(+) express significantly more transcripts of PERP, S100A8, FABP5, SFN, CA12, JUP and CSTA. The results of PCR analysis showed particularly increased expression of S100A8 and SFN in the metastatic LNs, where there was over 10-fold change of the mean expression level. Higher expression of PERP, FABP5, CA12, JUP and CSTA were also found, but the differences were lower.
The significance of gene expression levels in LN samples for time to groin recurrence (TTR) was statistically analyzed using Cox's proportional hazard model. A coefficient computed by this method for each variable expression of “predictor” genes indicates the direction and degree of flexing that the predictor has on the TTR curve. A positive value of coefficients indicated that larger values of the expression were associated with greater groin recurrence rates. Two genes, namely SFN and CA12, were found significantly predictive of such recurrence. High expression levels of the other genes of the set of five in LNs were also associated with shorter TTRs, but the trend did not reach statistical significance.
The log-rank test used for testing significance of the TTR functions in Kaplan-Meier analysis yielded significant P values for separating the two groups of VC patients with different TTR according to their LN gene expression levels of SFN, CA12, JUP and CSTA. Patients with low expression of these genes had a superior TTR values of not reached versus 6.9 to 8.7 months.
It is worth noting that our study had some important limitations. Firstly, only seven of the most differentially expressed genes in the microarray experiment were analyzed by the qRT-PCR. Notwithstanding, the remaining genes (see Additional file: 2 Table S2) are candidates for validation in further studies. Secondly, LN assessment was based on routine pathological evaluation with H&E staining, while the state-of-the-art would be to perform ultrastaging on the excised LNs. Lastly, Kaplan-Meier analysis of TTR that was applied to SFN, CA12, JUP and CSTA expression, should be considered as just an illustration of potential clinical use of the identified marker genes. Future studies should include more patients to assure a better assessment of the novel markers’ performance. Such studies should also provide better cut-off values for patients’ stratification than the arbitrarily chosen median expression levels in LNs, and enable the ROC curve analysis. Importantly, our microarray data were obtained for HPV-negative patients in order to exclude the interference of the infection with the expression results and to focus on metastasis-associated genes’ selection. However, patients with HPV infection should also be enrolled in future studies to assess the influence of HPV on the markers’ performance.
Groin recurrences from VC are often fatal. Unfortunately, “the optimum mode of treatment and predictive factors for patients with groin recurrences are unknown” . The size of LN metastases correlates with survival in patients with early  and advanced  stage VC. On the other hand, “inguinofemoral lymphadenectomy can be avoided when the sentinel node is negative for disease” . However, although Oonk and colleagues  reported that the risk of non-sentinel LN involvement increases with size of sentinel LN metastases, they were unable to determine the cut-off size below which the risk of non-sentinel LN metastases would be inconsiderable. Therefore, the authors concluded that all patients with sentinel LN metastases - regardless of their size - should undergo additional groin treatment. The risk of additional metastases when only isolated tumour cells are present in the sentinel LN is 4% . Still, approximately 12% of early-stage VC patients with a negative sentinel LN develop local recurrence” [54, 56].
For the reasons discussed above, novel means to better stratify early-stage VC patients in order to optimize treatment modality, i.e. a tool to decide who should not undergo lymphadenectomy vs who should receive additional treatment to the groin would be highly beneficial. Further validation of markers and development of molecular tests is indispensable for the reliable up- or down-staging of LNs. Our results may ideally be used to develop a test that could be used during surgery to decide whether to remove inguinal LNs, similarly to the RT-PCR based, FDA-approved GeneSearch™ Breast Lymph Node (BLN) Assay (Veridex, LLC, Warren, NJ), used for the rapid intra-operative detection of sentinel LN metastases in breast cancer. Such tests may reduce the need for a second surgery for the axillary LN dissection [57–59]. As sentinel node procedure was confirmed to be safe in the early-stage VC patients , tests similar to GeneSearch™ should also be of clinical utility for VC patients.