Over-expression of cyclin a is highly associated with early relapse and reduced survival in patients with primary breast carcinomas

Authors


Abstract

Progression through the mammalian cell cycle is facilitated by cyclin–cyclin-dependent kinase (cdk) complexes, which are activated at specific points during the cell cycle. Alteration in cyclin–cdk complexess may lead to altered cell cycle and tumorigenesis. In this study, we analyzed expression of cyclins A, D1, D3 and E in tumor tissue from 170 patients with primary invasive breast carcinomas. Immunohistochemical methods were used to detect protein expression of these cyclins. We detected positive immunoreactivity in 55 (32%), 22 (13%), 38 (22%) and 37 (21.8%) of the samples for cyclins A, D1, D3 and E, respectively. A highly statistically significant association was observed between expression of cyclin A and early relapse (p = 0.001 univariate analysis, p = 0.006 multivariate analysis) as well as cancer-specific death (p < 0.0001) during the follow-up time. No association was observed between cyclin D1 or cyclin E, respectively, and relapse of disease or survival, while cyclin D3 over-expression was associated with development of metastases during follow-up (p = 0.005 univariate analysis, p = 0.01 multivariate analysis). However, cyclin D3 did not show any statistically significant association when cancer-specific death was examined in a multivariate analysis (Cox regression for survival function). © 2001 Wiley-Liss, Inc.

Cyclins are groups of proteins that play a relevant role in cell-cycle progression. According to their sequence motif, pattern of expression and activity, they exert a regulatory function in the control of transitional stages of the cell cycle1 and are grouped as cyclins of the G1, S and G2/M phases. The phase activity of cyclins contributes to phosphorylation of specific substrates, such as cyclin-dependent kinases and retinoblastoma protein (pRB), required for progression through G1, the beginning of the different phases and the completion of the cell cycle.2 Among the G1 cyclins, ectopic expression of cyclins D1 and E in mammalian fibroblasts shortens G1 and reduces the serum dependence for S-phase entry.3–7 Unlike G1-phase cyclins, A and B family cyclins achieve maximal levels later in the cycle. Cyclin A, which is synthesized during DNA replication and G2/M transition, is involved in cellular activities that promote both replication and transcription.8, 9 The activities of the B cyclins are essential for G2 transition and mitosis.6, 10–13

Loss of restriction point control is a common event in malignancies. Several studies reveal that deregulation of both cyclins and the pRB pathway is involved in human carcinogenesis.14–16 Expression of cyclins and pRB also reportedly represents alternative mechanisms by which tumor cells may escape the G1 restriction point.17 Deregulated expression of cyclins has been found in different types of cancer. Several mechanisms, e.g., chromosomal translocation, amplification of cyclin genes and deregulation by post-transcriptional mechanisms, play a role in deregulated expression in various types of human carcinoma.15 Also, cyclin A may contribute to the regulation of genes involved in cell growth and in the transformed phenotype.11, 14, 18, 19 However, the prognostic value of cyclin A over-expression in clinical material has yet to be studied. Our present aim was to evaluate the prognostic value of expression of cyclins A, D1, D3 and E in tumor tissue from patients with primary invasive breast carcinomas.

MATERIAL AND METHODS

Material was obtained from 170 patients with primary breast carcinoma admitted to the Central Hospital of Akershus and to Ullevål Hospital from 1988 to 1993. Mean age at diagnosis was 60 years (range 28–88). Fifteen (9%) were classified as invasive lobular, 131 as invasive ductal (77%) and 15 (9%) as other types. For 9 patients, no information about tumor classification was available. Lymph node dissection was performed in 166 patients. Of these, 92 (54%) were lymph node-negative and 74 (43%) lymph node-positive. Nine tumors (5%) were classified as histological grade 1, 103 (61%) as grade 2 and 57 (34%) as grade 3. For 1 tumor, no grading was performed. Grading was based on the recommendations of Elston and Ellis.20 All samples were judged after histological evaluation to contain >50% tumor tissue. Mean observation time was 4.9 years (range 0.5–12). Fifty-two patients had distant metastases during follow-up. Eighty patients died during follow-up. Of these, 48 died of breast cancer, 22 died of other causes and for 11 information about cause of death could not be obtained.

Immunohistochemistry

Sections (4 to 6 μm thick) from formalin-fixed, paraffin-embedded tumor tissue obtained at surgery were applied to coated slides. After antigen retrieval by the microwave technique, immunostaining was performed in an Optimax plus Automated Cell Stainer (model 1.5; BioGenex, San Ramon, CA), following the operating manual. The antibodies for detection of different proteins, their sources and dilutions are shown in Table I. All series included positive and negative controls. Only cells with staining of the nucleus or nucleus and cytoplasm (cyclin D3) were scored as positive. The number of immunopositive cells was estimated semi-quantitatively; grade + corresponded to 15% to 30% positive cells, grade ++ to 30% to 60% positive cells and grade +++ to >60% positive cells. Tumor samples showing immunoreactivity in <15% of tumor cells were scored as grade 0. For every sample, at least 200, usually more than 500, tumor cells were analyzed. Tumor samples were analyzed by 2 investigators (IRKB and JMN) independently, and no discrepancy about immunohistochemical grading was observed between the investigators. For statistical analyses, samples showing immunoreactivity of grade 0 or + were scored as negative, while samples showing immunoreactivity of grade ++ or +++ were scored as positive. (In Figure 2, the relationship between the different grades of cyclin A expression and cancer-specific survival is demonstrated. Elsewhere, the above-described differentiation between positive and negative scores is used.)

Table I. Antibodies and Working Conditions
AntibodyDilutionSourcesPre-treatment
Cyclin A1:50Novocastra (Newcastle-upon-Tyne, UK)2 × 5 min microwave, 1 mM EDTA (pH 8)
Cyclin D11:200Oncogene Research (Manhasset, NY)2 × 5 min microwave, 1 mM EDTA (pH 8)
Cyclin D31:25Dako (Carpinteria, CA)4 × 5 min microwave, 1 mM EDTA (pH 8)
Cyclin E1:100Santa Cruz Biotechnology (Santa Cruz, CA)4 × 5 min microwave, 2 mM citrate buffer (pH 6)

Statistical methods

Follow-up data were taken from the time of the last clinical appointment or the date of death. When cancer-specific deaths were analyzed, patients who died from unrelated causes were censored. All statistical calculations were performed using the SPSS (Chicago, IL) Data Analysis program (SPSS for Windows). For univariate analysis of survival and metastasis, the χ2 test was used to assess differences in categorical variables (Fisher's exact 2-tailed test). Multivariate survival outcomes were assessed using Cox regression analysis for comparison of co-variates. The association of different variables with metastasis was examined by multivariate analysis using binary logistic regression.

RESULTS

For cyclins A, D1 and E, immunostaining was detected only in the tumor cell nuclei, while cyclin D3 staining was also, to some degree, detected in the cytoplasm of tumor cells as well as tumor cell nuclei. Fifty-five samples were scored as positive to cyclin A immunoreactivity (++/+++), while 115 samples were scored as negative (0/+). Twenty-two tumor samples were scored as positive (++/+++) for cyclin D1, while 148 samples were scored as negative. When immunoreactivity for cyclin D3 was evaluated, 38/170 tumors were scored as positive. Only 37 tumor samples were scored as positive for cyclin E. The distribution of different grades for different cyclins is shown in Table II. Immunostaining examples for cyclins A, D3 and E are shown in Figure 1. A photograph of cyclin D1 immunostaining has previously been published.21

Table II. Number of Samples in Different Immunoreactivity Grades for Different Cyclins
CyclinGrade
0++++++
A53623718
D1110381210
D3113191919
E105282215
Figure 1.

Tumour cell nuclei showing immunoreactivity to cyclin A (a), cyclin D3 (b) and cyclin E (c). For cyclin D3, some cytoplasmic staining was also detected together with nuclear staining.

Multiple-comparison post hoc tests according to both Bonferroni and Tukey showed a significant difference between the 0 and + score groups, on the one hand, and the ++ and +++ score groups, on the other hand, for cyclin A with regard to metastasis and survival. To make the statistical tests more robust, cyclin scores were recoded into 2 groups (0 and + = 0, ++ and +++ = 1).

In a univariate analysis (χ2 test), a statistically significant association was observed between expression of cyclin D3 (p = 0.005) and of cyclin A (p = 0.001) and development of metastases during follow-up as well as positive lymph node status at the time of surgery (p = 0.018). In a multivariate analysis (binary logistic regression) comparing the effect of the cyclins (A, D1, D3 and E), lymph node metastases, tumor type, estrogen receptor status and histological grade, expression of cyclins D3 and A as well as positive lymph node status were independently associated with development of distant metastases (cyclin D3 p = 0.01, cyclin A p = 0.006, lymph node metastases p = 0.05). Tumor type, histological grade, estrogen receptor status, cyclin D1 and cyclin E did not show any association with development of metastases in either univariate or multivariate analysis (Table III). Histological grade, however, showed a borderline association (p = 0.07) with development of distant metastases in a multivariate analysis model (Table III).

Table III. Relationship Between Expression of Cyclins, Other Clinicopathological Parameters and Development of Metastases in Patients with Primary Breast Cancer
 Metastases
YesNop1
UnivariateMultivariate
  • 1

    Univariate analyses were performed using the χ2 test, while multivariate analyses were performed using binary logistic regression. Multivariate analyses included lymph node status, estrogen receptor status, tumor type, histological grade and the cyclins (A, D1, D3 and E). Cyclin scores were recoded after post hoc multiple-comparison tests according to Bonferroni and Tukey. For cyclin scores, 0 and + were recoded as 0 and ++ and +++ were recoded as 1.

  • 2

    Only patients with known receptor status are included.

  • 3

    Only patients with histological grade 2 or 3 are included.

  • 4

    Only patients with either invasive ductal or invasive lobular are included.

Cyclin A
 Positive2629
 Negative26890.0010.006
Cyclin D1
 Positive715
 Negative441001.00.95
Cyclin D3
 Positive1919
 Negative33990.0050.01
Cyclin E
 Positive1225
 Negative40930.840.64
Lymph node metastases
 Positive3044
 Negative21710.0180.05
Estrogen receptor2
 Positive2946
 Negative23710.070.11
Histological grade3
 23370
 318391.00.07
Tumor type4
 Invasive lobular510
 Invasive ductal41901.00.97

Both expression of cyclin A and the number of positive lymph nodes were significantly associated with cancer-specific death when analyzed individually (univariate; cyclin A p < 0.0001, positive lymph node status p = 0.024). A statistically significant association between lymph node metastases and cancer-specific death was seen only in patients who had more than 4 positive lymph nodes at the time of surgery (Table IV). Also in a multivariate analysis of prognostic factors on survival (Cox regression), including estrogen receptor status, tumor type, histological grade, age at surgery, lymph node metastases and the cyclins (A, D1, D3 and E), only expression of cyclin A (p < 0.0001) showed an independent influence on cancer-specific death (Fig. 2, Table IV), while positive lymph node status showed a borderline association (p = 0.06). Expression of cyclins D1, D3 and E as well as tumour type, histological grade (data not shown) and estrogen receptor status did not appear to be independent prognostic factors with regard to survival (Table IV). The Cox regression model hazard function showed linearity before and after recoding the cyclin data. However, to ensure increased robustness to the linearity of the hazard function, all statistical analyses represented in Table IV were performed on recoded cyclin data. Figure 2 demonstrates the effect of increasing cyclin A over-expression on cancer-specific survival before recoding. As is clearly shown, the main difference is between cyclin A immunohistochemistry scores + and ++ (p < 0.0001). The odds ratio for cancer-specific death related to over-expression of cyclin A was 5.2 (95% confidence interval 2.6–10.2).

Table IV. Relationship Between Different Clinico-Pathological Parameters and Cancer Death1
 Cancer deathp2
YesNoUnivariateMultivariate*
  • 1

    Patients who died from causes unrelated to breast cancer and where death cause is unknown are not included.

  • 2

    Univariate analyses were performed using the χ2 test, while multivariate analyses were performed using the Cox regression for survival function. Multivariate analyses included age at surgery, lymph node status, estrogen receptor status, tumor type, histological grade and the cyclins (A, D1, D3 and E). Cyclin scores were recoded after post hoc multiple-comparison tests according to Bonferroni and Tukey. For cyclins scores, 0 and + were recoded as 0 and ++ and +++ were recoded as 1.

  • 3

    Only patients with known receptor status are included.

  • 4

    Only patients with 0 lymph nodes with metastases are considered negative.

  • 5

    Only patients with >4 positive lymph nodes are scored as positive.

Cyclin A
 Negative1770
 Positive3119<0.0001<0.0001
Cyclin D3
 Negative3372
 Positive15170.1390.15
Cyclin E
 Negative3966
 Positive9230.400.57
Cyclin D1
 Negative4274
 Positive6111.00.53
Estrogen receptor3
 Negative2254
 Positive26340.10.16
Lymph node metastases4
 Negative2150
 Positive25380.270.16
Lymph node metastases5
 Negative3479
 Positive1290.0240.037
Figure 2.

Survival function plot showing association between immunoreactivity grade of cyclin A and cancer-specific survival. 0, grade 0; 1, grade +; 2, grade ++; 3, grade +++. Multivariate analyses were performed using Cox regression for survival function and included age at surgery, lymph node status, estrogen receptor status, tumor type (invasive ductal, invasive lobular), histological grade (2, 3) and the cyclins (A, D1, D3 and E).

DISCUSSION

In the present study, we analyzed the relationship between expression of cyclins A, D1, D3 and E and the development of metastases and survival in patients with primary invasive breast carcinomas. We found a highly significant association between over-expression of cyclin A and development of metastases as well as reduced survival.

The exact mechanism behind how cyclin A leads to more aggressive disease and thereby reduced survival and early relapse is not known. Expressed discontinuously during the cell cycle, cyclins promote the phosphorylation of specific substrates and allow cells to proceed through the sequential phases of the cell cycle.22, 23 Cyclin deregulation may lead to oncogenesis.14, 18, 24 Biochemical studies on the function of cyclin A have shown that it is required to initiate DNA replication.25 Collecchi et al.26 showed a positive association between cyclin A expression and tumor proliferation in breast carcinomas. Since the presence of cyclin A protein leads to initiation of DNA replication, a sustained abnormal elevation of cyclin A in cancer cells may lead to more rapid enlargement of the tumor, both at the primary site and in metastases, causing early relapse and reduced survival. It is interesting to observe that cyclin A expression had a statistically stronger association with metastasis and survival than with lymph node metastasis in a multivariate model (binary logistic regression and Cox regression).

We also analyzed the expression and prognostic value of cyclins D1, D3 and E. Expression of cyclin E did not show any association with development of metastases and survival. These results are in contrast with results from other groups.27, 28 The exact cause of this discrepancy is difficult to explain, but it may be due to a different composition of patient groups, regarding clinicopathological parameters. Cyclin D3 was highly associated with the development of metastases, in both univariate and multivariate analyses. Cyclin D3 expression did not influence survival.

Over-expression of cyclin A was highly associated with reduced overall survival as well as cancer-specific death. The association with overall survival may be explained by there being only a few patients in this series for whom cause of death was something other than breast cancer. It is also possible that many of the patients for whom information about cause of death was not available died from their cancer. One indication that this may be the case is the observation that lymph node metastases were also associated with reduced overall survival (data not shown).

Histological grade was not associated with the development of distant metastases in univariate analysis (p = 1.0), but when analyzed in a multivariate model, it showed a borderline association (p = 0.07). At the same time, the highly significant association between lymph node metastases and distant metastases was reduced from p = 0.018 to p = 0.05 when histological grade was taken into account. The same was true when the association between cancer-specific death and lymph node status was analyzed (univariate p = 0.024, multivariate p = 0.06) (Table IV). This may indicate that there is co-variance between histological grade and lymph node metastases.

In summary, the present study provides data about a highly statistically significant association between over-expression of cyclin A and relapse as well as reduced survival in patients with primary invasive breast carcinomas.

Acknowledgements

The excellent assistance of Ms. E. Hellesylt, Mr. P. Blom and Ms. G. Pedersen are gratefully acknowledged.

Ancillary