Background
Cyclin D1 along with its binding partners CDK 4/6 partially mediate G1 to S-phase transition of the cell cycle through phosphorylation and inactivation of retinoblastoma (Rb) protein with subsequent release of E2F transcription factors [
1‐
3]. The oncogenic activities of the protein have been addressed in numerous studies, [
4‐
7] and many human cancers including breast, colon, and prostate, overexpress cyclin D1 [
8‐
10]. More recently, a number of cyclin D1 studies in breast cancer have focused on functions that are not directly related to cell cycle maintenance. Cyclin D1 can modulate the activity of transcription factors and histone deacetylase [
11], it can activate oestrogen receptor in the absence of oestrogen [
12], and it can bind to the upstream regulatory region of the diverse
Notch1 gene [
13]. Previous work by our group revealed a novel induction of breast cancer cell migration after cyclin D1 silencing, which may account for a worse clinical outcome for patients with low expression of the protein [
14]. Of the genes upregulated following this silencing, Inhibitor of differentiation 1 (Id1), a basic helix-loop helix (bHLH) family member, represents a potential candidate modulating the effect of cyclin D1 on cell migration.
The four Id proteins (termed 1-4) represent the class V subgroup of the bHLH family, however in contrast to other bHLH transcription factors (that modulate gene expression though dimerization and DNA binding of canonical E-box promoter regions in target genes [
15]), the Id proteins lack a DNA binding domain and instead bind to other bHLH family monomers, negatively regulating their activity [
16]. Id1 has been associated with breast cancer progression in a number of studies.
ID1 promoter regulation is lost in aggressive breast cancer cells [
17], Id1 is associated with induction of cell proliferation and invasion [
18], and stable antisense targeting of Id1 represses an aggressive and metastatic phenotype in mammary epithelial cells [
19]. Recent data has also revealed that cyclin D1 binds to the
ID1 promoter region in the mammary gland, and negatively regulates its transcription in mouse retina [
13]. Given the role of Id1 in cell invasion and metastasis, it represents a strong candidate for driving breast cancer cell migration following cyclin D1 silencing.
Increased motility and invasiveness are inherent properties of a mesenchymal phenotype [
20], and the process whereby a non-motile epithelial cell procures these traits is termed epithelial to mesenchymal transition (EMT). Recently, a role for EMT in the process of cancer metastasis has been postulated, and direct evidence of EMT has been demonstrated in a mouse mammary tumour model [
21]. A number of distinct changes occur during the transition to a mesenchymal phenotype, most notably the down-regulation of epithelial markers such as E-cadherin, and an upregulation of mesenchymal markers including Snail, Slug, vimentin, Twist and fibronectin [
22]. In addition, a number of phenotypic changes occur including loss of cell polarity and tight junction regulation, accompanied by cytoskeletal changes [
23,
24] and enhanced cell migration/invasion [
25]. Id1 has previously been implicated with EMT both directly, through suppression of E-cadherin and zonula occludins-1 (ZO-1), in human kidney cells [
26] and indirectly, through loss of Krueppel-like factor 17 (KLF17) in breast cancer cells [
27]. As such, we wished to clarify whether the increase in cell migration following cyclin D1 silencing was due to an Id1-dependent increase in EMT markers.
In this study, we demonstrate that silencing Id1 prevents the cyclin D1 mediated increase in MDA-MB-231 breast cancer cell migration. We have identified that an increase in SNAI2 mRNA expression following cyclin D1 silencing is abolished in cyclin D1/Id1 double knock-down cells. A meta-analysis of primary breast tumours revealed significant associations between CCND1, ID1, CDH1 (E-cadherin) and recurrence-free survival. CCND1 and ID1 gene expression was also correlated with EMT-associated genes including, VIM, SNAI1, SNAI2, and TWIST1. Finally, the recently established claudin-low subtype of breast cancer, which is enriched in EMT markers, was found to have a four-fold greater proportion of CCND1
low/ID1
high tumours compared to other breast cancer subtypes.
Discussion
In this study we demonstrate that the increase in MDA-MB-231 cell migration following cyclin D1 silencing is dependent on the upregulation of Id1. Previous studies have found both similarities and differences to our experimental model. Caldon
et al. showed an increase in Id1 protein in mouse mammary epithelial cells isolated from cyclin D1
-/- mice compared to wild type, in line with our observations. Moreover, they also established the inability of Id1 to promote proliferation of mammary acini in the absence of cyclin D1 [
43]. Swarbrick
et al. revealed a decrease in cyclin D1 expression 48 h after Id1 silencing in MCF7 cells [
44], and others report the same effect in both MCF7 and MDA-MB-231 cells [
45]. We did not observe this decrease in cyclin D1 protein expression in MDA-MB-231 cells after 24 h in our study. However, qPCR analysis showed a similar decrease in cyclin D1 mRNA levels which may become more apparent on the protein level after 48 h. Bienvenu
et al. demonstrated binding of cyclin D1 to the promoter region of
ID1 in mouse retinal cells, and when comparing wildtype to
CCND1
-/- mice found an 8-fold enrichment of
ID1. We have also observed occupancy of the Id1 promoter region by cyclin D1 in MDA-MB-231 cells, where it may repress Id1 expression. These data demonstrate the complex relationship between cyclin D1 and Id1. It is important to note that here we are only proposing this mechanism in MDA-MB-231 cells and in a distinct subset of representative breast tumours. We observed this complexity during the course of our work, where despite an increase in ZR75-1 migration following cyclin D1 silencing, Id1 protein levels were so low as to not substantially contribute to this effect. We postulate that in ZR75-1 cells other known transcription regulators of Id1 such as TGF-beta may be responsible for repressing expression of the protein. Importantly, TGF-beta and other known Id1 regulators (KLF17, Src) were unchanged in our MDA-MB-231 microarray following cyclin D1 silencing, indicating they do not contribute to the upregulation of Id1 or migration in our analysis.
It is pertinent to highlight that the increase in migration we have observed is occurring in an already highly invasive, mesenchymal-like cell line. This may account for a lessened migratory response to cyclin D1 silencing. Further evidence of this concept is shown in the more epithelial-like, less typically invasive ZR75-1 cells, where the increase in cell migration is more pronounced (1.89 fold vs. 1.3 fold in MDA-MB-231) following cyclin D1 knock-down. In addition, cyclin D1 is known to be expressed at variable levels across cell lines and subtypes of breast cancer thus, silencing of cyclin D1 is unlikely to increase migration uniformly in all cell types.
A common feature in our MDA-MB-231 and ZR75-1 cells was an increase in
SNAI2 expression 24 h after cyclin D1 knock-down, which coincided with an increase in cell migration. In MDA-MB-231 cells, silencing of Id1 reversed this and
SNAI2 expression was decreased, as was cell migration. Moreover, silencing of Slug- the
SNAI2 protein, significantly decreased MDA-MB-231 migration, and cyclin D1 silencing was unable to rescue this effect. These migratory observations for
SNAI2 are in line with previous experimental data, indicating that Slug expression induces a migratory phenotype and can represses E-cadherin, inducing an EMT in epithelial cells [
46]. Moreover, siRNA against Slug decreases MDA-MB-231 cell migration [
47], and Slug and Snail are overexpressed invasive ductal carcinoma [
48]- a form of breast cancer hallmarked by cell migration. In our experimental model, Slug would appear a likely candidate mediating the observed migratory effects, however it is entirely plausible that it does so in conjunction with other EMT factors. We also found statistically significant changes in
TWIST1 and
CDH11 (the positive EMT-regulator also known as OB-cadherin) following cyclin D1 silencing, both of which have been implicated with enhanced cell motility [
49,
50]. The changes in our EMT markers are in the order of 1.13 to 1.19 fold of control by expression array analysis (Figure
2A). We note that these figures are more meaningful when taken in the context of the most increased gene in our expression array, which was only upregulated 1.8 fold [
14]. As may be expected from treatment with siRNA, many more genes were downregulated in the array analysis than upregulated, again highlighting the importance of the increases in our mesenchymal markers. It is likely that all of these factors work in concert to promote a migratory and EMT-like phenotype, and that small gains in expression of a number of EMT genes can contribute to a greater overall effect.
The relationship between cyclin D1 expression and patient outcome remains a controversial area, with studies reporting both positive and negative associations.
CCND1 gene amplification has been related to poor disease outcome in ER-positive patients [
51,
52], but others correlate cyclin D1 protein expression with both better [
53,
54] and worse [
55] prognosis. It has been proposed that subgroup analysis with small numbers of patients [
56] and splice variants of the gene have contributed to these contrasting results. In agreement with others [
57], we found an association between high
CCND1 expression and poor prognosis (Figure
3A). However, when examining
ID1 high tumours, both the highest and lowest expression quartiles of
CCND1 were correlated to reduced RFS/DFS but only in the ER-positive subgroup (Figure
4C). A similar trend was noted for
ID1, where in all patients low expression of the gene was associated with a shortest RFS (Figure
3B), but in the
CCND1 low ER-positive subgroup of tumours, a positive correlation was found (Figure
4B).
Whilst this may appear contrasting to our in vitro data, we reason that cyclin D1 low, ER-positive tumours best represent our cell line model. We chose two cell lines (MDA-MB-231 and ZR75-1) based on their high expression of cyclin D1 (regardless of oestrogen receptor status). We then reduced these high levels using siRNA and noted an increase in cell migration and EMT markers. As ER-negative tumours are consistently cyclin D1 low, these are less representative our in vitro experiments. ER-positive tumours however are typically cyclin D1 high, thus by choosing tumours that are cyclin D1 low in this subgroup, we are more correctly mimicking our in vitro setting, where expression of cyclin D1 may have been lost. This yields the interesting observation that ER-positive tumours with low cyclin D1 appear to behave similarly to ER-negative tumours with regards to their relationship to EMT markers and the claudin-low subtype. Thus, should ER-positive tumours that have lost expression of cyclin D1 be considered more ER-negative-like? Whilst the answer to this question is far beyond the scope of this study, what is clear is that the effect we are observing is centred on loss of cyclin D1 and not on the oestrogen receptor status of our testing material.
Interestingly, the
CCND1
low/
ID1
high and
CCND1
high/
ID1
high tumours both displayed increased expression of EMT-related genes (Figure
4A, yellow and green bars respectively). This suggests that in the context of these subgroups,
ID1 is vital for increased EMT gene expression and when
CCND1 is low it enhances the EMT phenotype.
We did not observe any meaningful impact of EMT genes in individual Kaplin-meier analysis on patient survival in our dataset. There has been an explosion of EMT related data in recent years in the breast cancer field. Central to many of these publications has been the ability of EMT to putatively enhance stem cell-related features and promote the metastatic process [
58,
59]. Of particular note, the idea of cells that have undergone EMT residing at the leading edge of an invasive tumour and promoting metastasis at the tumour- stroma interface has garnered much attention [
60]. This hypothesis may be one explanation as to why EMT markers such as
SNAI1,
SNAI2,
TWIST1 and
VIM do not show any prognostic significance in our model- if the cells that have undergone EMT reside at the leading edge of the tumour, strong expression of their genes could easily be lost amongst the entirety of the tumour body. In these circumstances, any strong links to prognosis would also be diluted.
A second, more straightforward explanation as to why we have not observed prognostic significance of EMT- related genes centers upon a keystone principal. Upregulation of one EMT gene, e.g.
SNAI1, is not enough to induce a transition to mesenchymal phenotype. This is supported by the board range of expression values of EMT genes across all breast cancer tumours and subtypes in our study (Figure
5D). Induction of EMT requires a reduction in
CDH1 expression and upregulation of the potent
SNAI1, SNAI2 and
TWIST1 genes (amongst others). In order to examine the effect of EMT in our cohort, we would have to combine all tumours with these gene properties- giving us a 'claudin-low' subgroup. Unfortunately, we have too few cases in our claudin-low dataset to give any relevant prognostic information. In order to explore this further a cohort consisting of a large representation of claudin-low tumours, preferably with micro-dissection of the tumour-stroma interface would be required.
Much like
CCND1, some controversy surrounds expression patterns of
ID1, and despite numerous links to invasion and migration in breast cancer [
43,
44] some groups report an absence of the protein in the normal mammary gland [
61]. Perk
et al. assessed Id1 protein expression in mammary carcinomas [
62] and found nuclear expression of Id1 in a rare subtype of breast cancer, metaplastic mammary tumours. Metaplastic cancers have a unique genetic profile that is notably, most closely related to the claudin-low subtype of breast cancer [
41,
63,
64] and are very poorly differentiated. Given the poor outcome associated with metaplastic cancer, it may indicate why high
ID1 expression in
CCND1 low tumours gave the shortest RFS.
Adding further weight to our analysis, we found the greatest proportion of
CCND1
low/
ID1
high cell lines and tumours in the claudin-low subgroup, which have a poor prognosis [
65], associations with EMT and chemotherapy resistance [
66] and has stem-cell tumour initiating features [
42]. A number of these properties are reflected in both the cell lines and patient material used within this study, potentially indicating a central role for cyclin D1 and Id1 in this subgroup.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
NT performed the experimental work and the statistical analyses as well as drafting the manuscript. KL helped with the experimental work, and statistical analysis as well as drafting the manuscript. SL performed migration assays and western blots. AHS performed the statistical analyses and the bioinformatics related experimental work, as well as drafting the manuscript. GL was the principal investigator of the study and participated in the study design and interpretation of the data and helped to draft the manuscript. All authors read and approved the final manuscript.