p27Kip1 is a predictive factor for tamoxifen treatment response but not a prognostic marker in premenopausal breast cancer patients

Authors

  • Maria Stendahl,

    1. Center for Molecular Pathology, Lund University, Malmö University Hospital, Malmö, Sweden
    Search for more papers by this author
  • Sofie Nilsson,

    1. Center for Molecular Pathology, Lund University, Malmö University Hospital, Malmö, Sweden
    2. Breakthrough Breast Cancer Research Unit, School of Cancer, Enabling Sciences and Technology, University of Manchester, Manchester Academic Health Science Centre, Paterson Institute for Cancer Research, The Christie NHS Foundation Trust, Manchester, United Kingdom
    Search for more papers by this author
  • Caroline Wigerup,

    1. Center for Molecular Pathology, Lund University, Malmö University Hospital, Malmö, Sweden
    Search for more papers by this author
  • Karin Jirström,

    1. Center for Molecular Pathology, Lund University, Malmö University Hospital, Malmö, Sweden
    Search for more papers by this author
  • Per Ebbe Jönsson,

    1. Department of Surgery, Helsingborgs Lasarett, Helsingborg, Sweden
    Search for more papers by this author
  • Olle Stål,

    1. Division of Oncology, Department of Clinical and Experimental Medicine, Linköping University, Linköping, Sweden
    Search for more papers by this author
  • Göran Landberg

    Corresponding author
    1. Center for Molecular Pathology, Lund University, Malmö University Hospital, Malmö, Sweden
    2. Breakthrough Breast Cancer Research Unit, School of Cancer, Enabling Sciences and Technology, University of Manchester, Manchester Academic Health Science Centre, Paterson Institute for Cancer Research, The Christie NHS Foundation Trust, Manchester, United Kingdom
    • Breakthrough Breast Cancer Research Unit, School of Cancer, Enabling Sciences and Technology, University of Manchester, Manchester Academic Health Science Centre, Paterson Institute for Cancer Research, The Christie NHS Foundation Trust, Wilmslow Road, Manchester, M20 4BX, UK
    Search for more papers by this author
    • Tel.: +44 1619187041


Abstract

The cell-cycle regulating protein p27Kip1 (p27) has dual roles by acting as both a cdk inhibitor and as an assembly factor for different cdk complexes. Loss of p27 has been linked to malignant features in tumours; however, the exact role of p27 deregulation in breast cancer regarding prognostic and treatment predictive information has not been fully clarified. We have evaluated p27 expression in 328 primary, Stage II breast cancers from premenopausal patients who had been randomised to either tamoxifen treatment or no adjuvant treatment after surgery. p27 was associated with the oestrogen receptor and cyclin D1, and p27 downregulation was associated with high proliferation. There was no association between recurrence-free survival (RFS) and p27 (HR = 0.800, 95% CI 0.523–1.222, p = 0.300), indicating that p27 is not a prognostic marker. The predictive value of p27 was analysed by comparing RFS in tamoxifen-treated and untreated patients in subgroups of low and high p27 expression (HR = 0.747, 95% CI 0.335–1.664, p = 0.474 and HR = 0.401, 95% CI 0.240–0.670, p < 0.001, respectively). Only patients with p27-high tumours benefited from tamoxifen (multivariate interaction analysis p = 0.034). Our study suggests that p27 downregulation is associated with tamoxifen resistance in premenopausal breast cancer but is not linked to impaired prognosis.

Progression through the cell cycle is a complex process regulated by cyclins and cyclin-dependent kinases that together coordinate certain transitions during cell cycling. Cdk4 and cdk6 both associate with cyclin D and act in middle to late G1 phase, whereas cdk2 partners with cyclin E and cyclin A in the G1 to S-phase transition and in S phase, respectively.1 Cdk inhibitors are divided into 2 groups on the basis of their structure and cdk targets; the cdk4/6 inhibitors p15, p16, p18 and p19 cause cyclin D dissociation, whereas p21, p27 and p57 inhibit cdk2 complexes. By acting as negative growth regulators, cdk inhibitors function as tumour suppressors,2 but p21 and p27 also mediate the assembly and nuclear import of cyclin D-cdk4/6 complexes, thereby both promoting and inhibiting cell cycle progression.3–5

High levels of p27 are expressed in normal human mammary epithelium, but loss of p27 is frequent in breast cancer and has been demonstrated to have prognostic implications.6–9 Porter et al.10 showed that patients with tumours expressing low levels of p27 were associated with worse prognosis, and this was especially pronounced in hormone-receptor–positive tumours. HER2-overexpressing primary breast cancers often reveal reduced levels of p27.11 Low levels of p27 have also been shown to be a negative prognostic factor in colon, gastric, lung and prostate carcinomas.12–15 p27 is important for cellular differentiation and development, and p27 null mice exhibit multiorgan hyperplasia and are prone to develop tumours.16, 17

To assess a breast cancer patient's risk of relapse and death, a number of well-known clinicopathological parameters are assessed, such as tumour size, histopathological grade, number of axillary lymph nodes involved, steroid receptor status, patient age and menopausal status.6, 18 Despite the availability of all of these parameters, it is not possible to predict which tumours will recur. Therefore, it is of great importance to find new prognostic markers that can help to determine the natural course of the disease and even more crucial to find predictive tools that can be used to indicate potential response to therapy and avoid unnecessary side effects. The most widely used adjuvant therapy in breast cancer is varying forms of endocrine treatment with the selective oestrogen receptor modulator tamoxifen as the most common and best evaluated drug available. Pohl et al.19 demonstrated that only patients with tumours expressing high levels of p27 benefited from a combined treatment of tamoxifen and goserelin.

Altogether, various studies have suggested that p27 levels have both prognostic and endocrine treatment predictive values in breast cancer.7, 10, 19 The exact predictive value of p27 regarding tamoxifen response has not been convincingly addressed. Hence, our aim was to evaluate the prognostic and predictive value of p27 in a well-characterised material of premenopausal Stage II breast cancer patients that had been randomised to 2 years of tamoxifen treatment or no adjuvant treatment after surgery. We further wanted to delineate associations among p27, oestrogen receptor (ER), cyclin D1 and proliferation in different subgroups of breast cancer to better understand the role of p27 as an inhibitor or an assembly factor in various molecular contexts.

Material and Methods

Between 1986 and 1991, 564 premenopausal breast cancer patients were enrolled in a clinical trial comparing tamoxifen treatment (20 or 40 mg daily) to no adjuvant endocrine treatment. Nine patients (1.6%) received adjuvant polychemotherapy with cyclophosphoamide, methotrexate and fluorouracil (CMF) or groserelin. All patients were premenopausal or aged <50 years (mean and median age 44 and 45 years, respectively) and had been diagnosed with Stage II invasive (pT2 N0 M0, pT1 N1 M0 or pT2 N1 M0) breast cancer. No stratification for tumour size or nodal status was made, and patients were included irrespective of hormone receptor status. All patients underwent standard surgical treatment with, either modified radical mastectomy or breast-conserving surgery, axillary lymph node dissection. We constructed tissue microarrays with paraffin-embedded specimens (n = 500) from these patients. The mean follow-up for patients without breast cancer events was 13.9 years, equal in both arms. The primary outcome was recurrence-free survival (RFS) and included local, regional and distant recurrences and breast cancer-specific death but not contralateral breast cancer as primary events. Details of the study design have been reported elsewhere.18

We used immunohistological staining for ER, progesterone receptor and p27 on 4-μm sections of the tissue microarrays

The fractions of p27-positive tumour cell nuclei were subgrouped as 0 (0–10%), 1 (11–50%), 2 (51–75%) and 3 (76–100%). Tumours were considered ER/progesterone receptor (PR)-positive if >10% of the nuclei were positive, which is in accordance with clinical practice. All evaluations were made by a pathologist blinded to patient information. The analysis of the p27 as a prognostic marker was performed in the group of patients not receiving any adjuvant treatment, irrespective of ER/PR status, whereas the predictive value of p27 for tamoxifen response was delineated in ER- and ER/PR-positive tumours only.

We used Spearman's rank-order correlation coefficient to examine the statistical significance between p27 expression and other variables. RFS was estimated according to the Kaplan-Meier method, and the log-rank test was used to compare survival in different treatment groups. A Cox regression hazards model was used for the estimation of the relative risk in the multivariate analyses together with the interaction variable for tamoxifen treatment and p27 status. All statistical tests were two sided, and the calculations were done in Statistical Package for the Social Sciences version 17.0 (SPSS, Chicago, IL).

Results

Immunohistochemical evaluation of p27 status was possible in 328 of 500 cases (66%), ER status was available for 321 (98%) and PR status for 314 (96%) (Fig. 1). The majority of the excluded cases lacked malignant cells in the biopsies or were lost in the staining process. Thirty-one tumours (9%) had 0–10% p27-positive nuclei, 57 (17%) had 11–50% positive nuclei and 45 (14%) had 51–75% positive nuclei. A majority of the tumours, 195 (60%) showed more than 75% positive cell nuclei. We next analysed associations between p27 nuclear staining fraction and clinicopathological parameters as illustrated in Table 1. There was a significant inverse association between p27 nuclear fraction and tumour grade (NHG); a majority of tumours expressing low levels of p27 had high NHG scores, whereas tumours with high levels of p27 had lower scores. Although all analysed patients were premenopausal or aged <50 years, high p27 expression was more common in older patients. When analyzing tumour type, low p27 levels were correlated to medullary tumours, whereas high p27 was more common in ductal and lobular tumours. Interestingly, low p27 levels were observed in larger tumours, whereas high levels were more common in patients with lymph node metastases. p27 expression was significantly associated with both ER and cyclin D1, and low levels of p27 were associated with high Ki-67 levels and many mitoses (Table 2). By separating ER-negative from ER-positive tumours, we observed that p27 nuclear staining was only inversely associated to number of mitoses in the ER-negative group (Table 3). We further observed that in ER-positive tumours, there was a significant association between cyclin D1 and the proliferation marker Ki67 (Table 4).

Figure 1.

p27 staining of primary breast tumours. Tissue microarrays were stained with the p27 antibody (brown) and scored according to number of stained tumour cell nuclei. a = >75%, b = 11–50% and c = 0–10%.

Table 1. Nuclear p27 staining in relation to clinicopathological parameters
inline image
Table 2. Nuclear p27 staining in relation to cyclin D1, ER and proliferation markers
inline image
Table 3. Associations between p27 levels and cyclin D1 and proliferation markers according to ER status
inline image
Table 4. Associations between cyclin D1 and proliferation marker Ki67 and ERα in ER-positive tumours according to p27 levels
inline image

To evaluate the prognostic value of p27, we analysed the patients that had been randomised to no adjuvant treatment, irrespective of ER/PR status. Interestingly, there was no significant difference in RFS for patients with low or high levels of p27 (Fig. 2a, HR = 0.800, 95% CI 0.523–1.222, p = 0.300). Moreover, there was no trend toward impaired survival for patients having p27-low tumours, and notably, patients with low p27 even had a significantly better survival compared to patients with high p27 when analysing the subgroup of ER-negative tumours separately (Fig. 2c, HR = 0.316, 95% CI 0.146–0.682, p = 0.002). In the ER-positive subgroup, there was no association between p27 and RFS (Fig. 2b, HR = 1.243, 95% CI 0.724–2.133, p = 0.430). Taken together, these results support that p27 does not confer any prognostic information in premenopausal Stage II breast cancer.

Figure 2.

Recurrence-free survival in untreated patients according to p27 status, irrespective of ER status (a) (HR = 0.800, 95% CI 0.523–1.222, p = 0.300), in ER-positive patients (b) (HR = 1.243, 95% CI 0.724–2.133, p = 0.430) and in ER-negative patients (c) (HR = 0.316, 95% CI 0.146–0.682, p = 0.002).

We next evaluated p27 as a predictive factor for tamoxifen response by comparing RFS between tamoxifen-treated patients and untreated patients in subgroups defined by p27 status. In the analyses, we only included ER-positive or ER/PR-positive patients that theoretically and also in our previous study responded to tamoxifen.20 A total of 314 tumours were evaluable for both p27, ER and PR. Of these, 212 (68%) had >10% ER-positive cells, 207 (66%) had >10% PR-positive cells and 202 (64%) were both ER- and PR-positive. Interestingly, in ER-positive patients, there was no effect of tamoxifen treatment in the subgroup with p27-low tumours (Fig. 3a, HR = 0.747, 95% CI 0.335–1.664, p = 0.474), whereas there was a beneficial effect of tamoxifen treatment in the subgroup with p27-high tumours (Fig. 3b, HR = 0.401, 95% CI 0.240–0.670, p < 0.001). The difference in tamoxifen treatment effect between low and high p27 was even more pronounced in ER- and PR-positive tumours (Fig. 3c, HR = 0.907, 95% CI 0.392–2.099, p = 0.820, and Fig. 3d, HR = 0.393, 95% CI 0.232–0.664, p < 0.001). The difference in tamoxifen treatment response in relation to p27 status was significant in a multivariate interaction analysis (Table 5, HR = 3.07, 95% CI 1.09–8.68, p = 0.034).

Figure 3.

Recurrence-free survival after tamoxifen treatment compared to no treatment in ER-positive/p27-low patients (a) (HR = 0.747, 95% CI 0.335–1.664, p = 0.474), ER-positive/p27-high patients (b) (HR = 0.401, 95% CI 0.240–0.670, p < 0.001), ER- and PR-positive/p27-low patients (c) (HR = 0.907, 95% CI 0.392–2.099, p = 0.820) and ER- and PR-positive/p27-high patients (d) (HR = 0.393, 95% CI 0.232–0.664, p < 0.001).

Table 5. A multivariate Cox regression hazards model for p27 nuclear staining and treatment interaction based on ER- and ER/PR-positive patients
inline image

Discussion

The cdk inhibitor p27 is a tumour suppressor that inhibits cell cycle progression, and the gene product is downregulated in many malignancies. Low p27 protein levels in cancer samples are often linked to aggressive behaviour and a less favourable outcome for the patient even though the exact role for p27 as a prognostic and/or treatment predictive marker in breast cancer has not been fully elucidated.6, 7, 11, 21 By analysing a tissue microarray material containing 500 breast cancer samples from premenopausal Stage II breast cancer patients randomised to tamoxifen or no adjuvant treatment, we were able to detail prognostic and potential treatment properties of p27. This material had previously been characterised with regards to predictive and prognostic properties of ER, PR and cyclin D1 content.20, 22 Notably, we could not confirm p27 as a prognostic factor but did observe that downregulation of p27 was strongly associated to an impaired tamoxifen response in ER- and ER/PR-positive tumours.

Several studies have suggested p27 to be a prognostic factor in breast cancer as thoroughly reviewed by Alkarain et al.21 These studies were carried out in various patient materials; consecutive and unselected, node-negative patients, node-positive patients, small tumours, young patients and other types of patient material. In general, the patients had also received various adjuvant treatments but not always tamoxifen, making it difficult to assess the prognostic value of the studies and to draw comparisons with our work. It should be emphasised that only by analyzing an untreated patient material, it is possible to assess correct prognostic information, and the majority of earlier studies have analysed a mixture of untreated patients as well as patients receiving different types of adjuvant treatment. It is also difficult to find comparable predictive studies regarding p27 and tamoxifen response, but some studies have suggested a better response to tamoxifen in tumours with high levels of p27, which is in line with our results.19 In this study of Stage II premenopausal breast cancer patients, we could not confirm p27 as a prognostic factor in an untreated patient cohort, but we could however show that p27 had predictive values for tamoxifen treatment response; p27-high tumours being more susceptible to this endocrine treatment, further suggesting a link between steroid receptors and p27. The results were even more striking in the ER/PR-positive subgroup that we have previously shown to be highly susceptible to tamoxifen treatment.20 One potential explanation to the discrepancy between our findings and earlier publications regarding prognostic features for p27 could be that we only analysed premenopausal breast cancer patients of Stage II, whereas previous studies usually included patients of all stages and ages.

Immunohistochemistry has been frequently used in previous studies analysing p27 in breast cancer materials, and our data and others' support that p27 is mainly localised in the nucleus in normal tissues and that downregulation of nuclear p27 in a fraction of the tumours can be detected by an immunohistochemical approach.2, 6, 23 We have used <75% positive nuclei as a cutoff for defining p27-downregulated tumours, which has been used in other studies even though a somewhat lower cutoff also has been applied.6, 24 A range of 40% p27-downregulated tumours is higher than observed previously and could either be due to the cutoff used or that we analysed a premenopausal breast cancer material containing a higher fraction of p27-low tumours. There were strong inverse associations between p27 and proliferation as well as tumour grade, similar to the observations of others.6, 11, 14

It has been suggested that p27 has dual roles in primary breast cancer, either acting as a cdk inhibitor or as a cyclin D1-cdk assembly factor, making it difficult to delineate its main purpose and biological function. When functioning as a cdk inhibitor, the outcome will be that of a tumour suppressor, promoting a controlled cell cycle progression and hindering tumour transformation. Our results confirm this by suggesting an inverse correlation between p27 and proliferation markers such as Ki-67 and number of mitoses. We also show that loss of p27 in ER-negative tumours was associated with increased proliferation, again supporting its tumour suppressor role. The impact of p27 as an assembly factor for the cyclin D/cdk4 and 6 complex is less straightforward. Cyclin D1 has been shown to be essential for normal mammary gland development,25 but high levels can be crucial for patient outcome. Our results suggest that p27 might be essential in ER-positive tumours for cyclin D1 to adequately regulate proliferation as it was only in the ER-positive tumours with high levels of p27 that cyclin D1 was associated with Ki67. The lack of correlation between cyclin D1 and Ki67 in ER-positive tumours expressing low levels of p27 might indicate that these tumours are reliant on another proliferative pathway, independent of cyclin D1 and oestrogen. This in turn might explain why these tumours do not respond to tamoxifen to the same extent as ER-positive tumours expressing high levels of p27. One possibility might be that the cancer cells adjust to a cyclin E-dependent proliferation when levels of p27 are low. The access to a well-characterised material has enabled us to evaluate p27 in different strata, and its functions as a tumour suppressor and an assembly factor, respectively, appear to be connected to different subsets of tumours.

Earlier studies have shown that cyclin D1-high tumours do not respond to tamoxifen treatment,22 and in this study, we now show that p27-low tumours also have an impaired tamoxifen response. The fact that we and others have observed a strong association between the expressions of p27 and cyclin D1 might appear contradictory when referring to tamoxifen resistance. However, as showed in Figure 4, there is a large fraction of ER-positive, p27-high tumours with high cyclin D1 expression, illustrating that the two groups with an impaired tamoxifen response, i.e., p27-low tumours and p27/cyclin D1-high tumours, are different.

Figure 4.

Distribution of cyclin D1 staining intensity levels within the 2 cohorts dependent on p27 status in ER-positive patients. Arrow pointing to the group of tumours with high p27 levels and high cyclin D1 levels.

The mechanism of tamoxifen resistance in p27-low tumours is not clear; it is possible that the tamoxifen effect is reliant on a p27-mediated process to arrest the cell cycle. It is also likely that there are several mechanisms of tamoxifen resistance in breast cancer. We and others have suggested that HER2, Pak1 and cyclin D1 are some of the important interplayers in tamoxifen resistance in breast cancer, and we can now add an additional protein to the list of potential treatment predictive markers in breast cancer.26–28 In addition to these markers, including p27, having the potential to be used to predict treatment response in the clinical setting, they can also be used as targets for novel treatment approaches, potentially increasing the effect of endocrine treatment. p27 seems to have several key functions in the cell cycle and is further linked to ER status and also to tamoxifen response, one of the most commonly used endocrine treatments for breast cancer.

Acknowledgements

The authors thank Ms. Elise Nilsson for excellent technical assistance. They also thank the South Swedish and South-East Swedish Breast Cancer Groups.

Ancillary