p27Kip1 is an important regulator of the G1 to S transition. While a potent inhibitor of cyclin-dependent-kinase (Cdk)2, p27 is also involved in assembly of cyclin D/Cdk4 complexes. Although rarely mutated, p27 is functionally downregulated in many human cancers by mechanisms involving enhanced degradation, cytoplasmic mislocalization, and/or sequestration by cyclin D/Cdk complexes in response to oncogenic signals. Therefore, low levels and/or cytoplasmic localized p27 have been associated with enhanced malignancy and poor patient prognosis in many neoplasias including breast cancer. Recent data discussed below suggest that a threshold of p27 is required for response to antiestrogens and, conversely, that low levels predict for antiestrogen resistance. These results imply that hormone receptor-positive tumors with low and/or cytosolic p27 respond poorly to antiestrogens and should be considered for alternative therapeutic strategies.
p27 is a member of the Kip family of cyclin-dependent kinase (Cdk) inhibitors, which also includes p21Waf1 and p57Kip1, initially discovered as a Cdk-inhibitory activity induced by extracellular antimitogenic signals (1) . It accumulates in serum-starved and density-arrested cells, and its overexpression can lead to cell cycle arrest in G1 and/or apoptosis. The abundance and cellular localization of p27 are regulated mostly by post-translational modifications but less at the level of gene transcription. Loss of a single allele of p27 confers increased susceptibility to carcinogen-induced tumors in mice (2) . Interestingly, loss of the second p27 allele does not occur in these tumors, consistent with the very infrequent homozygous deletions of p27 in human cancers (reviewed in Ref. 3 ).
We reported recently that loss of one p27 allele accelerates mammary epithelial proliferation and delays post-partum involution (4) . In addition, mammary glands from mouse mammary tumor virus/neu (the mouse homologue of HER2) mice that are p27 haploinsufficient (p27+/−) develop tumors with a 50% shorter latency compared with mouse mammary tumor virus/neu × p27+/+ mice (5) . p27 was still present in the mammary tumors, suggesting lack of a selective pressure in the haploid cancers to lose the remaining p27 allele. Interestingly, the mouse mammary tumor virus/neu × p27−/− tumors were well differentiated, and exhibited a very low proliferation rate and undetectable Cdk4 activity. Nuclear cyclin D1 levels in these tumors were very low. In contrast, mouse mammary tumor virus/neu × p27+/− tumors were poorly differentiated and highly mitotic (5) . The results with the p27-null tumors are consistent with the role of p27 in the assembly of cyclin D1/Cdk4 complexes and in facilitating cyclin D1 translocation to the nucleus (6 , 7) . The acceleration of neu-induced p27 haploid tumors is also consistent with the association of low p27 protein levels with cancers of a worse prognosis (see below).
p27 and Prognosis in Human Breast Cancer.
A large proportion of human cancers downregulate p27 protein levels by mechanisms involving accelerated proteolysis, sequestration by cyclin D-Cdk complexes, and post-translational modifications leading to nuclear export and/or cytoplasmic retention. Several studies have correlated p27 levels with prognostic factors in primary breast cancers. Low p27 levels have been associated with poor disease-free and overall survival (8 , 9) , increased risk of breast cancer relapse (10) , high tumor grade, elevated cyclin E levels (8 , 11) , mutations in BRCA1/2, overexpression of Skp2 and HER-2/neu (12) , and resistance to antiestrogens (reviewed in Ref. 3 ). In a study of 118 breast cancers with BRCA1 mutations, low levels of p27 were associated with a 10-fold higher risk of cancer relapse (13) . Overexpression of HER-2 has been shown to activate mitogen-activated protein kinase-dependent proteasome-mediated degradation of p27 (14 , 15) . HER2 can also upregulate cyclin D1, enhance p27 sequestration by cyclin D/Cdk complexes, and thus facilitate cyclin E/Cdk2 activity (16) . Oncogenic activation of c-myc can also lead to sequestration of p27 through upregulation of cyclin D1 and cyclin D2 levels (17) . Although Ras and mitogen-activated protein/extracellular signal-regulated kinase 1 have been shown to induce p27 degradation in fibroblasts (18 , 19) , it is unclear if this also occurs in breast cancer cells. On the other hand, high levels of p27 have been associated with low histological grade, prolonged survival, and good response to antiestrogen therapy (3) .
Posttranslational Modifications and Cellular Localization.
p27 is subjected to post-translational modifications that regulate its stability, protein/protein associations, and cellular localization (reviewed in Ref. 1 ). Using HER-2-overexpressing BT-474 cells, we reported recently that Akt phosphorylates p27 at T157 (20) . This threonine is contained within a classical Akt consensus RXRRXT157D in the nuclear localization motif of p27. Inhibition of HER2 in these cells blocks Akt-mediated phosphorylation of p27 and redirects p27 to the nucleus. Myristoylated Akt phosphorylates wild-type p27 in vivo but was unable to phosphorylate a T157A p27 mutant. Wild-type p27 localizes in cytosol and nucleus, whereas T157A p27 localizes exclusively in the nucleus. T157A p27 is more effective than wild-type p27 in inhibiting cyclin E/Cdk2 and cell proliferation, and these effects are not rescued by myristoylated Akt. Expression of (active) S473 P-Akt in primary breast cancers correlates with expression of p27 in tumor cell cytosol. In 32/33 P-Akt-negative specimens, p27 was exclusively detected in tumor cell nuclei (20) . These data suggest that oncogene-activated Akt contributes to tumor cell proliferation by phosphorylation and cytosolic retention of p27, thus relieving Cdk2 from p27-mediated inhibition and subsequent growth arrest. Similar data were simultaneously reported by two other groups. Viglietto et al. (21) reported that cytoplasmic but not nuclear p27 isolated from primary breast tumors reacted with a T157 phosphospecific p27 polyclonal antibody. In the second study by Liang et al. (22) , cytoplasmic p27 was correlated with poor histological differentiation, high tumor grade, and poor patient prognosis. Indeed, the tumors with the worse outcome in this study were those with low as well as cytosolic p27.
The data summarized above suggest several possibilities: (a) that in cells with aberrant HER-2 signaling (and therefore activated Akt), p27 is phosphorylated at T157; (b) that T157 P-p27 can be construed as a marker of aberrant HER-2 signaling; and (c) that inhibition of HER-2 or phosphatidylinositol 3′-kinase/Akt signaling would: (a) induce loss or reduction of T157 P-p27, and (b) redirect p27 to the nucleus. Supporting these possibilities, treatment of HER-2 gene-amplified cells with the phosphatidylinositol 3′-kinase inhibitor LY294002 results in loss of T157 P-p27 staining, whereas using antibodies against total p27 produces a redirection of p27 to the nucleus. In this same study, treatment with the anti-HER-2 antibody trastuzumab or LY294002 eliminated both P-Akt and T157 P-p27 simultaneously (20) .
In response to mitogenic stimulation, p27 is phosphorylated in Ser10. This modification results in binding of p27 to the exportin CRM1 and nuclear export in late G1 (23 , 24) . The human kinase interacting stathmin is a nuclear protein that binds p27, phosphorylates it on Ser10, and induces its export to the cytoplasm (24) . Cytoplasmic redistribution of p27 has been reported in primary human tumors and cell lines (20 , 25, 26, 27) . Interestingly, introduction of a transducible p27 protein into HepG2 cells of p27-null fibroblasts results in rearrangement of the actin cytoskeleton and cell motility (28) , suggesting “gain-of-function” for cytosolic p27. Overexpression of HER-2 has been shown to mediate mitogen-activated protein kinase-dependent nuclear exclusion of p27 and cellular transformation (15) . Taken together, these data suggest that Akt and human kinase interacting stathmin are kinases potentially involved in the cytoplasmic mislocalization of p27 in human cancers and, thus, worthy of additional investigation.
p27 and Response to Antiestrogens.
Antisense-mediated inhibition of p27 protein levels has been shown to block antiestrogen-mediated cell cycle arrest. This is associated with reduction in cyclin E/Cdk2-bound p27, high cyclin E/Cdk2 activity, and entrance into S phase (29) . Donovan et al. (30) reported two antiestrogen-resistant breast cancer cell lines with increased mitogen-activated protein kinase activity. Inhibition of mitogen-activated protein kinase restored p27 inhibitory function and sensitivity to antiestrogens. Pohl et al. (31) reported recently a retrospective study in 1034 premenopausal stage I and II patients with estrogen receptor + and/or progesterone receptor + breast cancers. Patients were treated with goserelin (3 years) and tamoxifen (5 years) versus six cycles of CMF (cytoxan, methotrexate, 5-fluorouracil). Patient outcome was evaluated as a function of low p27 (<50% tumor cells) versus high p27 (≥50% tumor cells) as measured by immunohistochemistry of tumor sections. Relapse-free survival in patients treated with hormonal therapy was significantly better in patients with high p27 compared with patients with low p27 protein levels. Interestingly, the relapse-free survival in patients with high p27 was statistically better in those treated with tamoxifen plus goserelin versus those treated with chemotherapy, suggesting that high p27 levels might be useful for the selection of premenopausal patients with steroid receptor-positive breast cancer for adjuvant hormonal therapies.
The results with the high p27 tumors suggest, as indicated by the retrospective studies above, that these cancers might be those low S phase, low grade, hormone-dependent tumors in which antiestrogen therapy is particularly effective. The poor outcome of the low p27 tumors suggests that a threshold level of p27 is necessary to achieve antiestrogen-induced inhibition of Cdk activity. Although the study by Pohl et al. (31) did not report cytosolic expression of p27, it is likely that in many tumors with low levels the Cdk inhibitor, p27, was localized in tumor cell cytosol. Although this study and others have used 50% of p27-positive tumor cells as a cut-off, the optimal threshold below which nuclear and/or total p27 content are most informative as a negative prognostic factor, clearly requires additional investigation. Of note, two prostate cancer studies have used a lower threshold of p27 expression (<25%) with an apparent better predictive power (32 , 33) . We propose that the low levels and/or cytoplasmic mislocalization of p27 is a potential surrogate marker of overexpressed kinases, i.e., HER-2, Erk, and Akt, which directly or indirectly alter estrogen receptor transcription and enhance tumor cell survival, thus contributing to escape from hormonal dependence and/or antiestrogen-induced growth inhibition. In addition, the low levels of p27 compromise the inhibition of Cdk2, further contributing to the dysregulated cell proliferation and the poor outcome of these tumors. These mechanisms would explain the poor response to both hormonal therapy and chemotherapy, and suggest that patients with low levels or mislocalized p27 should be considered for alternative and novel antisignaling therapeutic strategies.
Dr. Daniel Medina: In the human tumors, to what extent is that lower amount of p27 due to a transcriptional versus a posttranslational regulation?
Dr. Arteaga: All the data seem to indicate a predominance of post-translational regulation. However, the p27 promoter has binding sites for several transcription factors including Sp1, AFX, Myb, and CRE. AFX is a forkhead transcription factor that activates p27 transcription, but this effect appears to be modest. Akt phosphorylated AFX and excludes it from the nucleus, thus blocking p27 transcription. p27 can be also be regulated by increased proteolysis. I should add that some phosphorylation events retain p27 in the cytosol where it does not inhibit Cdk2.
Dr. Medina: How much of that posttranslational modification can be attributed to alterations in Skp-2 levels of function?
Dr. Arteaga: In order to be degraded, p27 phosphorylated in T187 binds an SCF complex consisting of Skp1, Cul1, the F box protein, and Skp2. Therefore, overexpression of Skp2 can lead to enhanced degradation of p27. Indeed, transgenic mice overexpressing Skp2 in the mammary gland develop ductal hyperplasias in part because p27 is downregulated. There are clinical papers that show that high Skp-2 correlates with low p27 in several human cancers. However, there are no papers showing this association in human breast cancer, although this has not been studied extensively.
Dr. Myles Brown: The genetic changes that can drive p27 downregulation seem to be HER2 amplification. What about cyclin D amplification—does that alter the available p27 levels by the sequestration model?
Dr. Arteaga: No, not that I’m aware. Theoretically, cyclin D1 amplification should subtract p27 from Cdk2, counteract p27-mediated inhibition of Cdk2, and thus contribute to increased cell cycle progression. Of note, a paper by Fredersdorf et al. [Proc. Natl. Acad. Sci. USA 94: 6380, 1997] showed a high correlation between high cyclin D1 and high p27 levels but colocalization and/or association of these two proteins was not shown in this report.
Dr. Brown: Is Skp-2-mediated degradation of p27 a reasonable drug target?
Dr. Arteaga: Theoretically, if we inhibit Skp-2, p27 will be stabilized and this could lead to cell cycle growth arrest. It is possible that very high levels of p27 may induce apoptosis. There are a couple of papers showing that high overexpression of adenoviral p27 can lead to apoptosis. Those results could be an artifact of adenovirus transduction or they could reflect that at very high levels p27 becomes an apoptotic signal.
Dr. Richard Santen: You have a system involving multiple signals including cyclin D, cyclin E, and at least six inhibitors. You are specifically looking at one of them. Perhaps the downstream marker that tells you how the whole process is interacting would be E2F1. That leads to the question whether there would be an inverse relationship between the p27 in the nucleus and the levels of free E2F1 in the cells. This putative relationship is based upon the concept that E2F1 provides the integration of all of the upstream signals. Have you looked at E2F1?
Dr. Arteaga: No, we have not. In principle, however, with high levels of p27 in the nucleus and inhibition of Cdk2’s catalytic activity, Rb should remain bound to E2F1 and free E2F1 might not be readily detectable.
Dr. Bissell: Does p27 change as a function of cell cycle, whether or not it retains nuclear localization?
Dr. Arteaga: In nontumor cells, p27 is highest and predominantly nuclear during G1 and decreases during S phase. There is probably some degradation during S phase as a function of Cdk2-mediated phosphorylation in T187 and by Akt in T157. During S phase, there is also translocation of p27 to the cytosol. So, both levels and cellular compartmentalization of p27 change during the cell cycle.
Dr. Bissell: So the tumors that have high level of cytosol p27, they are probably rapidly growing tumors?
Dr. Arteaga: High levels of p27 in the cytosol probably reflect the presence of hyperactive kinases, i.e., Akt, that modify and retain p27 in the cytosol and thus contribute to cell cycle progression by relieving Cdk2. It is possible that these signals per se contribute to dysregulated tumor cell proliferation. It is also possible that cytosolic p27 may have gain-of-function effects. This is based on the recent observation that a mutant form of p27 incapable of localizing in the nucleus was able to induce cell motility via binding to F-actin [Mol. Cell. Biol., 23: 216, 2003]. There is a paper by Ruoslahti et al. [Oncogene, 16: 2575, 1998] showing cytoplasmic displacement of high levels of p27 protein in a number of cancer cell lines. This cytoplasmic displacement has also been observed in some human cancers and appears to correlate with high tumor grade. Whether that cytoplasmic p27 is contributing to tumor progression, either through proliferation, motility, and/or another effect, remains to be studied.
Presented at the Third International Conference on Recent Advances and Future Directions in Endocrine Manipulation of Breast Cancer, July 21–22, 2003, Cambridge, MA.
Grant support: R01 CA80195, Breast Cancer Specialized Program of Research Excellence (SPORE) Grant P50 CA98131, and Vanderbilt-Ingram cancer Center Support Grant P30 CA68485.
Requests for reprints: Carlos L. Arteaga, Division of Oncology, Vanderbilt University Medical Center, 2220 Pierce Avenue, 777 PRB, Nashville, TN 37232-6307. Phone: (615) 936-3524; E-mail: