High Cyclin E Staining Index in Blastemal, Stromal or Epithelial Cells Is Correlated with Tumor Aggressiveness in Patients with Nephroblastoma

Purpose Identifying among nephroblastoma those with a high propensity for distant metastases using cell cycle markers: cyclin E as a regulator of progression through the cell cycle and Ki-67 as a tumor proliferation marker, since both are often deregulated in many human malignancies. Methodology/Principal Findings A staining index (SI) was obtained by immunohistochemistry using anti-cyclin E and anti-Ki-67 antibodies in paraffin sections of 54 postchemotherapy nephroblastoma including 42 nephroblastoma without metastasis and 12 with metastases. Median cyclin E and Ki-67 SI were 46% and 33% in blastemal cells, 30% and 10% in stromal cells, 37% and 29.5% in epithelial cells. The highest values were found for anaplastic nephroblastoma. A correlation between cyclin E and Ki-67 SI was found for the blastemal component and for the epithelial component. Univariate analysis showed prognostic significance for metastases with cyclin E SI in stromal cells, epithelial cells and blastemal cells (p = 0.03, p = 0.01 and p = 0.002, respectively) as well as with Ki-67 SI in blastema (p<10−4). The most striking data were that both cyclin E SI and blastemal Ki-67 SI discriminated between patients with metastases and patients without metastasis among intermediate-risk nephroblastoma. Conclusions Our findings show that a high cyclin E SI in all components of nephroblastoma is correlated with tumor aggressiveness and metastases, and that assessment of its expression may have prognostic value in the categorization of nephroblastoma.


Introduction
Nephroblastoma is the most common pediatric tumor of the kidney [1]. It arises from metanephric blastemal cells and recapitulates renal embryogenesis. In Europe, patients are treated according to the International Society of Pediatric Oncology (SIOP) protocol, which consists of preoperative chemotherapy and surgical resection followed by postoperative treatment [2]. This latter step is adjusted on the basis of tumor histology and local tumor stage. Stage I low-risk nephroblastoma receive no postoperative treatment while high-risk tumors (i.e. diffuse anaplasia and blastemal types) are treated with aggressive chemotherapy. In the intermediate-risk tumor group (i.e. epithelial, stromal, mixed, regressive and focal anaplasia types), over 90% of the patients are cured with the SIOP therapeutic strategy, but a small fraction of children will relapse or metastasize. Thus, there is still a need for accurate molecular prognostic markers to identify these intermediate-risk tumors that need more intensive treatment.
A vast amount of prognostic markers in nephroblastoma have been reviewed [3,4] and no biological marker was found that provided consistent predictive information regarding the clinical outcome. Tumor-specific loss of heterozygosity (LOH) for chromosomes 1p or 16q has been shown recently to be associated with a poorer prognosis in favorable-histology Wilms tumor entered in NWTS-5 (National Wilms' Tumor Study 5) [5] and is the only biological marker with immediate implications for treatment in the current Children's Oncology Group (COG) study. However, the prognostic value of these LOH for patients treated with preoperative chemotherapy according to the European SIOP protocol remains to be evaluated.
Some markers, such as Ki-67, may be relevant for assessing proliferative activity [3]. Ki-67, a nuclear antigen associated with cell proliferation, is present throughout the cell cycle and absent in resting cells [6]. High Ki-67 is associated with a more aggressive clinical behavior, and is found to be a significant determinant of distant metastasis and tumor-related death in adult tumors [7].
Cyclin E is the regulatory subunit of the cyclin E-Cdk2 complex, which takes part in the control of progression through G1 phase. Its activity is tightly regulated during normal cell cycle. In neoplastic cells, deregulation is often observed and is thought to play a fundamental role in tumorigenesis [8]. Cyclin E overexpression has been studied and identified as an adverse prognostic marker in a wide variety of human adult cancers [9,10,11,12]. However, to our knowledge, cyclin E levels have never been investigated in solid embryonal tumors, characterized by a high proliferation rate.
The aim of our study was to evaluate cyclin E expression in nephroblastoma using immunohistochemistry. To check if cyclin E overexpression reflects only increased proliferation, levels of the commonly used proliferation marker, Ki-67, were simultaneously assessed. Analysis of the results was carried out taking into account the global SIOP histology.

Ki-67 and cyclin E expression in postchemotherapy nephroblastoma
We determined Ki-67 and cyclin E staining index (SI) in all histological components when present ( Table 1). The median SI for Ki-67 and cyclin E were 33% and 46% respectively in blastemal cells, 10% and 30% in stromal cells, 29.5% and 37% in epithelial cells ( Figure 1).
In normal kidney parenchyma adjacent to the tumor, only scattered Ki-67 positive nuclei were present in the renal tubules, whereas we did not observe positive cyclin E nuclear staining (Figure 2A-C).
We observed no significant difference for Ki-67 and cyclin E blastemal SI in blastemal type nephroblastoma versus mixed type (p = 0.24 for Ki-67 and p = 0.43 for cyclin E). We also did not observe significant difference for Ki-67 and cyclin E stromal SI between mixed and stromal type nephroblastoma (p = 0.23 for Ki-67 and p = 0.61 for cyclin E).

Correlation between Ki-67 and cyclin E expression
A correlation between Ki-67 and cyclin E SI was observed in blastemal (r = 0.68, p,10 24 ) and epithelial (r = 0.51, p = 0.01) components. In contrast, no correlation was observed for the stromal component (r = 0.15, p = 0.44).

Correlation between Ki-67 and cyclin E expression and metastases
Univariate analysis showed prognostic significance of blastemal Ki-67 SI for metastases (p,10 24 ). Epithelial and stromal Ki-67 SI did not have any prognostic value (p = 0.29 and p = 0.90 respectively). In contrast, cyclin E SI showed prognostic significance for metastases in blastemal, stromal and epithelial cells (p = 0.002, p = 0.03 and p = 0.01 respectively). Multivariate analysis could not be performed because of the low frequency of events. Table 2 summarizes the main data in patients with and without metastasis.

Discussion
Our study showed that a high cyclin E SI in all components of nephroblastoma was correlated with tumor aggressiveness and metastases. The three histological components of nephroblastoma (blastemal, epithelial and stromal) have different proliferating potential. Furthermore, each component may have different proliferating activity in each different histological type of Numerous studies have been reported which analyze the expression of genes at the RNA or protein level in order to detect correlations with adverse outcome [3,4,13,14,15,16,17]. Potential markers included proliferation markers, apoptosis-associated molecules, growth factors and cell adhesion molecules. For some of these molecules, i.e. the antiapoptotic factor Bcl-2 [18], TGFa [19], the angiogenic growth factor VEGF and its receptor Flt-1 [20], IGFIR (type I insulin-like growth factor receptor) [14], blastemal staining was indicative of clinical progression but epithelial staining has no prognostic value. To date, among the predictive markers identified, no immunohistochemistry staining has proven to be relevant in the three histological components of nephroblastoma.
It is well established that characterization of clinical relevant prognostic markers will require studies of combination of markers. However, difficulties in validating molecules previously reported [4] highlight the fact that the first step that consists of identifying single reliable markers, which could be further included in large association studies, must go on.
In this study, we analyzed two cell-cycle markers, Ki-67 and cyclin E, using immunohistochemistry in different components of postchemotherapy nephroblastoma histologically classified according to the SIOP classification. We determined Ki-67 SI in the three components of nephroblastoma and found the highest values in blastemal and epithelial components, as previously reported [21]. The clinical value of proliferation markers such as Ki-67 in nephroblastoma is still subject to debate. Nagoshi et al. [22] did not find any relationship between proliferation rates and patient survival whereas other studies [21,23] showed that blastemal Ki-67 is a prognostic factor in pretreated nephroblastoma. Our study demonstrates that Ki-67 SI in blastemal cells is a highly significant indicator of metastases whereas it has no prognostic value in epithelial nor in stromal cells. When focusing only on intermediate-risk nephroblastoma, we still found higher values of blastemal Ki-67 SI in nephroblastoma with metastases in comparison with those without metastasis.
Cyclin E, a G1-cyclin, is a marker of cell-cycle progression. Overexpression of cyclin E has been demonstrated to be an indicator of poor prognosis in many cancers [9,10,11,12]. In the cyclin E promoter, several WT1 binding sites have been identified and WT1, which is the tumor suppressor gene defective in some nephroblastoma, has a repressive effect on cyclin E expression [24]. But, to our knowledge, cyclin E expression has never been studied in nephroblastoma. Faussillon et al. reported overexpression of another cyclin, cyclin D2 (CCND2), in 86% of Wilms' tumor, supporting the idea that alteration at the G1/S cell cycle control point is of biological significance in nephroblastoma [25]. However they found no association between relapse and cyclin D2 overexpression.
We used an antibody that can recognize both full-length and lower molecular weight isoforms of cyclin E. The latter are of particular interest because recent data demonstrate that they are specific to tumor cells [26,27]. We found the highest values for cyclin E SI in blastemal cells with prognostic significance for distant metastases. Cyclin E is also largely expressed in epithelial and stromal cells where it has a prognostic value too. Our study demonstrates that increased cyclin E expression is associated with distant metastases for whichever histological component of nephroblastoma it is assessed in. Moreover when focusing on intermediaterisk nephroblastoma, the ones with metastases have higher cyclin E SI in the three components than those without metastasis.
We observed a correlation between Ki-67 and cyclin E levels in blastemal and epithelial components. These data suggest that high cyclin E SI, at least in these two components, may reflect increased proliferative activity. However, we found a decreased blastemal proliferation, as assessed by Ki-67 after chemotherapy, in accordance with another study [21] whereas no variation of blastemal cyclin E SI was found, suggesting that cyclin E is not only a proliferation marker. Some studies assessing Ki-67 and cyclin E activities in breast cancer also concluded that cyclin E provides additional information in that it represents a marker for both proliferation and oncogenesis [28,29]. Porter et al. [28] reported a subset of tumors with discordant cyclin E expression and proliferation index (i.e., high cyclin E and low Ki-67 levels) showing a strong association with mortality. Moreover, cyclin E staining detects tumor cells committed to cell division (late G1 and beyond). Such improvement in tumor phenotyping may enable prediction of responsiveness to chemotherapy targeted at cells in S and M phases [29]. Even if the basis for cyclin E prognostic significance is not known, a major hypothesis is that cyclin E overexpression reflects alterations in any part of the p16-cyclinD-Rb-E2F pathway, which is mutated in most cancers. Thus, cyclin E expression may be a single convenient marker of alterations occurring in a very complex pathway [8].
Using a CART procedure, we determined cyclin E and Ki-67 cutoff scores for metastasis prediction. A high specificity was found for Ki-67 and cyclin E, but sensitivity was higher for blastemal Ki-67 than for cyclin E. Optimal determination of these cutoffs needs a larger cohort of patients and a centralized immunohistochemical study to avoid technical variations.
As a summary, blastemal, epithelial and stromal cyclin E SI and blastemal Ki-67 SI are significant determinants of distant metastases. Accurate postoperative chemotherapy is based on histology and staging. Risk histological stratification could also include immunohistochemical analyses to better explore aggressiveness in the different components of nephroblastoma. Cooperative larger studies focusing on intermediate-risk nephroblastoma are needed to determine if these two markers should be considered for stratification in future trials.

Patients
This retrospective study included 54 tumors removed from 53 patients (23 males and 30 females). Eight of these patients were syndromic: one WAGR syndrome, one Denys-Drash syndrome, one Perlman syndrome and five Beckwith-Wiedemann syndrome or hemihypertrophy patients. The patients were seven to 163 months of age; the median age at surgery was 32 months.
All patients were treated by neoadjuvant chemotherapy before nephrectomy. All the slides were reviewed by the French National referent for SIOP-2001 Nephroblastoma Study (LBG) and local pathologists. Exclusion criteria were: low-risk tumors and regressive type nephroblastoma with more than 90% of chemotherapy-induced changes.
Thirteen high-risk nephroblastoma (5 anaplastic and 8 postchemotherapy blastemal types) and 41 intermediate-risk nephroblastoma (9 stromal, 3 regressive, 27 mixed and 2 epithelial types) were studied. Pathological local staging was: stage I (n = 20), stage II (n = 21), stage III (n = 13). Three blastemal type and 2 mixed type from primary nephrectomy cases (4 under seven months of age and one with a highly cystic tumor) were also separately analyzed for comparison.
Forty-six tumors were removed from patients with a localized disease at diagnosis (i.e. unilateral or bilateral nephroblastoma, without metastasis), and 8 tumors from patients with metastases at diagnosis (clinical stage IV). Among the 46 patients with localized disease, 4 developed metastases after nephrectomy. The median follow-up was 57 months.

Specimen characteristics
This retrospective study was performed on samples obtained from the SIOP-2001 nephroblastoma protocol and conducted

Antibodies
Primary antibodies were: mouse monoclonal anti-human Ki-67 antigen (clone MIB-1, DakoCytomation, Glostrup, Denmark) and rabbit polyclonal affinity purified antibody against a peptide mapping at the C-terminus of the human cyclin E (C-19, Santa Cruz Biotechnology, Santa Cruz, CA).

Immunohistochemical staining
After paraffin removal, sections were heated in a water bath in sodium citrate buffer 0.1 M, pH 6.0 for 40 min for antigen retrieval. Endogenous peroxidase activity was quenched with a 3% hydrogen peroxide solution. The slides were then incubated with primary antibody: anti-MIB-1 (1:50) for 30 min at room temperature or anti-cyclin E antibody (1:80) overnight at 4uC. For MIB-1 primary antibody detection, incubation with biotinylated sheep anti-mouse immunoglobulin for 30 min followed by incubation with horseradish peroxidase coupled streptavidin for 30 min was performed. For anti-cyclin E primary antibody detection, the DAKO EnVision system (DakoCytomation) was used (30 min). In both cases, 3,39-diaminobenzidine was used as a substrate. Sections were counterstained with hematoxylin.

Quantification of labeled cells
Immunostaining was assessed by two independent observers blinded to clinical outcome and local staging. Staining was evaluated in tumor cells and adjacent normal kidney. Only nuclear staining was considered as positive. The intensity of staining was not recorded. For anti-cyclin E antibody, the positive control was a breast adenocarcinoma [28]. We recorded at high magnification the number of cyclin E and Ki-67 positive nuclei in at least 500 cells in representative fields. The average percentage of stained tumor cells was calculated and assigned to a staining index (SI).

Statistical analysis
Results are expressed as median and extreme values for quantitative variables, and numbers (percentages) for qualitative variables. Relationships between metastasis and age, Ki-67 SI, cyclin E SI were studied by non-parametric Wilcoxon Rank Sum test. Association between Ki-67 and cyclin E SI was studied by Spearman correlation statistics.
We developed cutoff scores to predict metastasis with Ki-67 and cyclin E in the different components of nephroblastoma by the CART (Classification and Regression Trees) procedure [30]. All tests were two tailed. Statistical analyses were performed using the SAS 9.1 (SAS Inc, Cary, NC) software package, with p,0.05 considered as statistically significant.