Tissue samples from 50 patients with primary breast carcinomas that had been operated on more than 6 years before the beginning of the study were collected for immunohistochemical studies, after we had obtained the approval of the institution's Human Investigation Committee. Patients presenting with metastatic disease were excluded from this study. Clinical and histological data were available for all of these patients. Long-term follow-up data were provided by patients' medical records and the Israel Cancer Registry.

Human breast cancer cell lines (MCF-7, T47D, and MDA-MB-231) were generously provided by Dr H Degani (Weizmann Institute of Science, Rehovot, Israel). Because Skp2 levels change during the cell cycle (being normally highest in the S phase and lowest in the G1 phase) we cultured the cells in different media under conditions of similar proliferation rate in all cell lines. MDA-MB-231 cells were grown in RPMI medium (Biological Industries, Beth Ha'emek, Israel) supplemented with 10% fetal calf serum, 100 Units penicillin, and 100 μg streptomycin per ml and 1 mM sodium pyruvate. MCF-7 breast carcinoma cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Biological Industries) supplemented with 10% fetal bovine serum, 4.5 g/l glucose, antibiotics as described above, and 4 mM glutamine. The T47D cell line was cultured in RPMI medium supplemented with 10% fetal calf serum, 100 Units/ml penicillin, 100 μg/ml streptomycin, and 10 μg/ml insulin. All cell lines were cultured at 37°C in 5% CO₂. Under these conditions, the proliferation rate of all cell lines was similar (21 to 22 hours doubling time).

Samples containing 30 μg of protein were resolved by electrophoresis on a 12.5% SDS-polyacrylamide gel and were transferred to nitrocellulose membranes. The membranes were probed with affinity-purified rabbit polyclonal antibody directed against a 12-amino-acid synthetic peptide from the extreme C terminus of human Cks1 (gift of Dr A Hershko, Technion, Haifa, Israel). This antibody does not cross-react with Cks2. The antibody was diluted 1:200 in TBST (Tris-buffered saline and 0.1% Tween 20) containing 10% (w/v) nonfat dry milk. Membranes were also probed against the mouse monoclonal antibody directed against Skp2 (Zymed Laboratories Inc, South San Francisco, CA, USA) diluted 1:500, or with a rabbit polyclonal antibody directed against p27^Kip1 (Zymed) diluted at 1:250. Membranes were also probed with a mouse monoclonal antibody directed against Skp1 (Transduction Laboratories, Lexington, KY, USA; 1:1000 dilution). Since levels of Skp1 do not change in the cell cycle, this protein served as an internal control to normalize for loading of cellular protein. After being washed with TBST, the immunoreactive proteins were visualized with horseradish-conjugated IgG (Pierce, Rockford, IL, USA) diluted 1:10,000 and an enhanced chemiluminescence system (SuperSignal West Pico; Pierce). All immunoblot analyses were done at least twice.

Immunohistochemical studies were performed on formalin-fixed tissue sections embedded in paraffin wax. Five-micron sections were deparaffinized with xylene and rehydrated in a series of ethanols.

For epitope retrieval, slides were heated in 1 mM ethylenediaminetetraacetic acid buffer (pH 8), either in a microwave oven at 92°C for 20 min (p27^Kip1) or in an Antigen Retrieval Processor (Milestone Inc, Sorsiole, Italy) at 120°C for 8 min (Skp2 and Cks1). After they had cooled, the slides were washed in distilled water.

Skp2 staining was carried out in the NexES IHC Immunostainer (Ventana Medical Systems, Tucson, AZ, USA), in accordance with the manufacturer's instructions, using a monoclonal antibody (Zymed) diluted 1:100.

Slides for p27^Kip1and Cks1 staining were treated for 10 min with 3% H₂O₂ in methanol to block endogenous peroxidase and for 30 min with 10% nonimmune rabbit serum to block nonspecific protein binding. The slides were then washed in water and soaked in washing buffer (pH 7.4, Optimax; Biogenex, San Ramon, CA, USA) for 5 min. For p27^Kip1 staining, slides were incubated overnight at 4°C with the monoclonal p27^Kip1 antibody diluted 1:500. Staining was completed with a Histostain-plus kit (Zymed) in accordance with the manufacturer's instructions. For Cks1 staining, slides were incubated with an affinity-purified rabbit polyclonal antibody directed against a 12-amino-acid synthetic peptide from the extreme C terminus of human Cks1 (described above) diluted 1:40 for 2 hours at room temperature. Staining was completed using a Super-sensitive Multilink Immunodetection kit (Biogenex) in accordance with the manufacturer's instructions. The color reaction product was developed with aminoethyl carbazole (AEC). All sections were counterstained with hematoxylin.

For negative controls, slide sections that were positive for staining were treated with 10% nonimmune rabbit serum (Zymed) instead of the primary antibody. No staining was observed in any of these controls.

The immunohistochemical slides were scored according to the percentage of tumor cells exhibiting nuclear staining. To define high and low protein expression, we used a cutoff of 50% p27^Kip1 and 10% for Skp2, which is the cutoff used in most previous studies 1415. The cutoff for Cks1 was 10%, similar to the level used for Skp2, as in previous studies. When stained, cells displayed similar intensities of nuclear staining regardless of the percentage of cells stained, and therefore the intensity of staining was not included in the score. The specificity of immunohistochemistry staining procedures for Skp2, Cks1, and p27^Kip1 was previously verified by comparing the protein levels as determined by immunohistochemistry with the protein levels determined by immunoblot analysis from the same tumor specimens 1523.

Statistical data analyses were performed using SPSS 11.0 statistical software package (SPSS Inc, Chicago, IL, USA). First, the relations between Skp2, Cks1, and p27^Kip1 and those between protein levels and various clinical and pathological features were studied using cross tabulation and Pearson's χ². Survival curves were constructed using the Kaplan–Meier method and multivariate analysis by Cox regression; P values less than 0.05 were considered significant.

We studied the expression of Cks1, Skp2, and p27^Kip1 by immunohistochemistry in 50 tumor samples obtained from patients with breast cancer. The quality of immunostaining was good, with minimal background reactivity. Cks1 levels were very low or absent in all normal breast tissues adjacent to tumors. However, Cks1 levels were not uniformly high in all tumor specimens, being very high in some but virtually absent in others (Table 1). Given the important biochemical link between Cks1, Skp2, and p27^Kip1, we examined the correlation between these proteins. Regression analysis of all 50 cases showed an inverse correlation between Skp2 and p27^Kip1 levels (r = -0.395; P = 0.005). Cks1 expression was strongly related to Skp2 expression (r = 0.477; P = 0.001) and inversely related to p27^Kip1 levels (r = -0.726; P < 0.001). Thus, strong relations were found between the expressions of all three proteins in 74% of the patients. High levels of Cks1 and Skp2 with low levels of p27^Kip1 were observed in 15 patients, and low levels of Cks1 and Skp2 with high levels of p27^Kip1 were observed in 22 patients. A typical representative immunohistochemical sample is shown in Fig. 1. In this tumor sample, cancer cells show strong nuclear staining for Cks1 and Skp2 and weak staining for p27^Kip1, but high p27^Kip1 staining and low Cks1 and Skp2 staining in the normal surrounding breast tissue. In 13 of the 50 cases, the above relation was not found. In eight cases an inverse relation was found between Cks1 and p27^Kip1 expression but not between Skp2 and p27^Kip1 expression; in three cases, levels of both p27^Kip1 and Skp2 were high whereas Cks1 expression was low; and in another five cases, p27^Kip1 levels were low despite high Cks1 levels and low Skp2 levels. In another five cases, the differences between the three proteins were not significant.

To examine the relation between Cks1 expression and common parameters associated with aggressive tumor behavior, we compared Cks1 levels with various clinicopathological features such as age, tumor size, pathological grade, lymph node status, and ER and PR expression (Table 1). A significant direct correlation was found between high Cks1 levels and loss of tumor differentiation (r = -0.421; P = 0.012), negative ER expression (r = -0.356; P = 0.013), and negative PR expression (r = -0.404; P = 0.006). We also found a strong relation between Cks1 overexpression and patient age less than than 50 years (r = 0.327; P = 0.049), but did not observe a significant correlation between Cks1 levels and lymph node status (r = 0.029; P = 0.508) or tumor size (r = 0.192; P = 0.333).

Examination of the association between Skp2 or p27^Kip1 and patients' clinicopathological characteristics showed a strong association between Skp2 expression and loss of tumor differentiation (r = -0.400; P = 0.005), young age (r = -0.337; P = 0.022), and negative ER or negative PR status (r = -0.589; P = 0.001 and r = -0.674; P = 0.001), respectively. Loss of p27^Kip1 was associated with loss of tumor differentiation (r = -0.323; P = 0.032) and, respectively, loss of ER and PR (r = 0.298, P = 0.048; r = -0.316, P = 0.030 (Table 1)).

Because of the strong association between Cks1, Skp2, and p27^Kip1 levels and the ER status of the tumors, we next examined the expression of these proteins in estrogen-dependent (MCF-7 and T47D) and estrogen-independent (MDA-MB-231) breast carcinoma cell lines, and the possible effects of estrogen modulation on the regulation of Skp2 and Cks1 expression. Basal levels of Cks1 and Skp2 were higher in estrogen-independent cells than in estrogen-dependent cells, whereas p27^Kip1 levels were lower (Fig. 2a). Interestingly, between the two estrogen-dependent cell lines, levels of Cks1 and Skp2 were higher in T47D cells, which phenotypicallly have poorer cellular differentiation (Fig. 2b). Treatment of T47D cells with estradiol increased Cks1 and Skp2 levels, whereas treatment with tamoxifen down regulated Cks1 and Skp2, resulting in an upregulation of p27^Kip1 levels (Fig. 2b).

To determine the association between Cks1 expression and prognosis, we plotted Kaplan–Meier curves for disease-free survival and overall survival. Complete 80-month follow-up data were available for all the patients. Patients presenting with high tumor Cks1 levels had significantly shorter disease-free survival rates than patients with low Cks1 levels (54 months vs 76 months, respectively; P = 0.0007; Fig. 3a). There was only one case of disease recurrence in patients with low Cks1 levels, whereas in patients with high Cks1 levels, disease recurrence was observed in 11 (42% of patients). Similarly, Skp2 and p27^Kip1 were also good predictors for disease-free survival (P = 0.0014 and P = 0.0191, respectively; data not shown).

We also examined the prognostic role of Cks1 in predicting local or distant relapse disease-free survival (Fig. 3b,c, respectively). Patients with high Cks1 levels had both higher local recurrence rates (P = 0.0352) and more distant relapse rates (P = 0.0006) than patients with low Cks1 levels. Skp2 was also found to be a good predictor for local (P = 0.0051) and distal (P = 0.003) relapse rates, while p27^Kip1 expression was a predictor for distant relapse (P = 0.031) but not for local recurrence (P = 0.136).

There was also a strong association between Cks1 expression and overall survival (Fig. 4a). The mean survival rate for patients with low Cks1 expression was 79 months, significantly higher than that (72 months) of patients with high Cks1 levels (P = 0.0415). Similarly, Skp2 expression was also found to be a strong predictor of overall survival (69 months vs 79 months P = 0.0224) (Fig. 4b), but p27^Kip1 expression was not (P = 0.0831; Fig. 4c).