Joint Requirement for Rac and ERK Activities Underlies the Mid-G1 Phase Induction of Cyclin D1 and S Phase Entry in Both Epithelial and Mesenchymal Cells*

Cyclin D1 gene induction is a key event in G1 phase progression. Our previous studies indicated that signaling to cyclin D1 is cell type-dependent because the timing of cyclin D1 gene expression in MCF10A mammary epithelial cells and mesenchymal cells such as fibroblasts and vascular smooth muscle cells is very different, with epithelial cells first expressing cyclin D1 in early rather than mid-G1 phase. In this report, we induced a mesenchymal phenotype in MCF10A cells by long-term exposure to TGF-β and used the control and transitioned cells to examine cell type specificity of the signaling pathways that regulate cyclin D1 gene expression. We show that early-G1 phase cyclin D1 gene expression in MCF10A cells is under the control of Rac, whereas mid-G1 phase cyclin D1 induction requires parallel signaling from Rac and ERK, both in the control and transitioned cells. This combined requirement for Rac and ERK signaling is associated with an increased requirement for intracellular tension, Rb phosphorylation, and S phase entry. A similar co-regulation of cyclin D1 mRNA by Rac and ERK is seen in primary mesenchymal cells. Overall, our results reveal two mechanistically distinct phases of Rac-dependent cyclin D1 expression and emphasize that the acquisition of Rac/ERK co-dependence is required for the mid-G1 phase induction of cyclin D1 associated with S phase entry.

Cells control their rates of proliferation by sensing external cues in their microenvironment. Cell cycle progression is dependent on these external cues until a time in late G1 phase, called the restriction point, when cells enter the committed and intrinsically controlled portion of the cell cycle (1). The restriction point is thought to occur after activation of the two G1 phase cyclin-dependent kinases (cdks), 2 cyclin D-cdk4 (or its homolog cdk6; hereafter called cdk4/6), and cyclin E-cdk2. A D-type cyclin (mostly cyclin D1 in fibroblastic and epithelial cells) is usually the first of the periodically expressed cyclins to be induced in G1 phase of the cell cycle after reentry from quiescence.
The levels of cyclin D1 mRNA and protein are low in quiescent cells and increase during progression through G1 phase (2)(3)(4). The increase in cyclin D1 mRNA requires cooperative signaling by growth factor receptors and adhesion receptors (5). In mesenchymal cells, activation of integrins by the extracellular matrix (ECM) transmits the adhesion signal to the cyclin D1 gene. Epithelial cell proliferation also requires signals from integrins, but these cells are also strongly influenced by cell-cell adhesion, which is mediated, at least in large part, by E-cadherin (4,6).
The induction of cyclin D1 mRNA requires a sustained activation of extracellular signal-regulated kinases (ERK), ϳ5-6 h after mitogenic stimulation of quiescent cells, at least in fibroblasts (7)(8)(9). ERK activity has also been linked to cyclin D1 gene expression in several other cell types (10 -13). Growth factor receptors and integrins collaboratively regulate the Ras-Raf-MEK-ERK cascade (14), and this allows for sustained ERK activity in cells that have active Rho (3). Activated ERK translocates to the nucleus where it stimulates cyclin D1 transcription in mid-G1 phase. The ERK-dependent immediate-early genes, Jun-B and Fra-1, have been implicated in the ERK-stimulated transcription of the cyclin D1 gene (8,15).
Another Rho-family GTPase, Rac, has also been linked to the induction of cyclin D1 mRNA (16 -20). Rac signaling to cyclin D1 mRNA is thought to be jointly regulated by growth factor receptors and integrins: both receptor systems can support GTP loading of Rac, and integrin-mediated adhesion is required for the coupling of Rac to its effectors (21). In fibroblasts, Rac-dependent induction of cyclin D1 mRNA is not seen unless Rho signaling is inhibited (3). Because Rho inhibition blocks stress fiber formation, Rac signaling to cyclin D1 does not require a high intracellular tensional environment. Rac-dependent cyclin D1 gene expression is readily detected in MCF10A mammary epithelial cells (4), and the absence of welldefined actin stress fibers indicate that these cells do not have the high levels of intracellular tension characteristic of mesenchymal cells.
Rac and ERK signaling have distinguishable effects on the timing of cyclin D1 gene expression within G1 phase, with Rac signaling resulting in an early G1 phase expression of cyclin D1 while ERK signaling results in the mid-G1 phase induction of cyclin D1 (3,4,7). However, most studies on ERK and Rac signaling to cyclin D1 have either been performed in different cell types (epithelial versus mesenchymal) or under conditions of Rac overexpression (3,4,16,17). Thus, little is known about the potential interplay between endogenous ERK and Rac activities as it relates to expression of the cyclin D1 gene within a single cell type.
When MCF10A cells are treated with TGF-␤ for 3 days, they acquire a mesenchymal phenotype in a process that resembles an epithelial-mesenchymal transition (22). We exploited this approach to study the interplay between endogenous ERK and Rac signaling to cyclin D1 mRNA in a cell type-specific manner yet within cells of the same origin. Our results identify two effects of endogenous Rac signaling on cyclin D1 gene expression, which can be distinguished by cell type (epithelial versus mesenchymal), the dependence or independence from ERK activity, a requirement for intracellular tension, and the ability to support Rb phosphorylation and S phase entry.
In some experiments, trypsinized MCF10A (2-3 ϫ 10 5 cells/ ml) or VSMCs (2-3 ϫ 10 4 cells/ml) were suspended in starvation media and preincubated in suspension (30 min at 37°C) with vehicle (DMSO) or 50 M U0126 (Promega). MCF10A cells were also preincubated for a 30-min suspension with 1 M cytochalasin D (CCD; EMD Biosciences). The treated cells were then reseeded with the appropriate mitogens in the continued presence of the inhibitors.
RNA Interference and Adenoviral Infection-siRNA-mediated knockdown was performed as previously described (23). The siRNA oligonucleotide sequences for human E-cadherin, human Rac1, and mouse Rac1 have been previously reported (23), and those for p21 were: No. 1: AACAUACUGGCCUG-GACUGtt and No. 2: AUCGUCCAGCGACCUUCCUtt. All siRNA oligonucleotides were used at 150 nM. Adenoviruses were titered and used as described (23). Rac N17 adenovirus was a generous gift from Anne Ridley (Ludwig Institute for Cancer Research, University College London, UK).
Induction of a Mesenchymal Phenotype in MCF10A Cells-MCF10A cells in maintenance medium were seeded at low confluence (3 ϫ 10 5 and 6 ϫ 10 5 cells for control and TGF-␤treated cells, respectively) in 100-mm dishes containing autoclaved glass coverslips. A mesenchymal phenotype was induced similar to Maeda et al. (22) by incubating the cells for 3 days in maintenance medium with 3 ng/ml human recombinant TGF-␤1 (R&D Systems). The transitioned cells were serumstarved for 2 days in the continued presence of TGF-␤. Duplicate cultures were incubated in parallel without TGF-␤. Each dish of quiescent control and TGF-␤-treated cells was then trypsinized, reseeded into two 100-mm dishes coated with 1.27 g/cm 2 collagen (containing autoclaved coverslips), and stimulated with 10% FBS and growth factor mixture in DMEM:F12 in the continued absence or presence of TGF-␤.
Rho and Rac GTPase Activity Assays-MCF10A cells were seeded in 6-well plates at 10 4 cells/well (control cells) or 2 ϫ 10 4 cells/well (TGF-␤-treated cells). Cells were treated for 3 days Ϯ 3 ng/ml TGF-␤ and starved for 2 days. The cells were stimulated with 10% FBS and the growth factor mixture. Protein was collected and quantified, and active Rho and Rac GTPase levels were measured using G-LISA small G-protein activation assay kits (Cytoskeleton, Inc.) according to the manufacturer's directions.
Other Methods-BrdU incorporation and E-cadherin surface expression was determined by epifluorescence microscopy performed as described (23). F-actin staining was performed using fluorescein isothiocyanate-or rhodamine-phalloidin (Invitrogen).
Total RNA was prepared from collected cells and analyzed by quantitative real-time RT-PCR (QPCR) as described (23). Taqman primer and probe sequences for human cyclin D1 mRNA and 18S rRNA have been listed previously (23). Immediate-early gene induction was analyzed by SYBR Green QPCR using the following primers (forward and reverse pairs, respectively): human Fra-1 CAGGCGGAGACTGACAAACT and CTTCCAGCACCAGC-TCTAGG and mouse Fra-1 ACCGAAGAAAGGAGCTGACA and CTGCTTCTGCAGCTCTTCAA. The expression of mesenchymal markers was determined using Taqman Assay-On-Demand primers and probes (Applied Biosystems): fibronectin, Hs00415006_m1; vimentin, Hs00185584_m1; snail, Hs00195591_m1; and slug, Hs00161904_m1. RNA expression was quantified by standard curve. QPCR results were normalized to 18S rRNA levels and show the mean Ϯ S.D. of duplicate PCR reactions.

RESULTS
Differential Expression of Cyclin D1 in MCF10A Cells before and after Acquisition of a Mesenchymal Phenotype-Our previous studies in MCF10A cells and fibroblasts indicated that the timing of cyclin D1 gene induction might be cell type-specific with an early-G1 phase induction in epithelial cells and a mid-G1 phase induction in mesenchymal cells (3,4,7). To address this issue without using distinct cell lines, we induced a mesenchymal phenotype in MCF10A mammary epithelial cells (MCF10A cells) by a 3-day treatment with TGF-␤. TGF-␤ efficiently induced a mesenchymal phenotype as assessed by loss of cell-cell contacts, appearance of actin stress fibers, and loss of membrane-localized E-cadherin (Fig. 1A). Consistent with previous reports (22,24,25), the transcription factors snail and slug, as well as their downstream targets, fibronectin and vimentin, were also induced (Fig. 1B). Thus, this system allows us to explore cell type-specific signaling to cyclin D1 mRNA without use of overexpression or cells of different origins.
We serum-starved control and TGF-␤-treated MCF10A cells and monitored the expression of cyclin D1 mRNA during a 12-h period (the time when cyclin D1 appears in fibroblastic cells). Interestingly, the early-G1 phase induction of cyclin D1 mRNA seen in control MCF10A cells ( Fig. 2A, Control, 3 h) was lost after TGF-␤ treatment: cyclin D1 mRNA was now induced   NOVEMBER 7, 2008 • VOLUME 283 • NUMBER 45 in mid-G1 phase ( Fig. 2A; TGF-␤, 9 -12 h) as it is in fibroblasts. A similar shift in the kinetics of cyclin D1 protein expression was seen before and after TGF-␤ treatment (Fig. 2B). When control cells were treated with TGF-␤ for 12 h (which was an insufficient time to induce the mesenchymal phenotype; not shown), the early-G1 phase induction of cyclin D1 was retained (Fig. 2C). Thus, the change in cyclin D1 expression kinetics is associated with the transition to a mesenchymal phenotype rather than simple exposure to TGF-␤.

Joint Regulation of Cyclin D1 by Rac and ERK
Loss of E-cadherin and the appearance of actin stress fibers are two hallmarks of an epithelial-mesenchymal transition (22,24,25). To determine if the change in cyclin D1 expression kinetics in TGF-␤-treated MCF10A cells was due to loss of E-cadherin, we transfected control MCF10A cells with E-cadherin siRNA. E-cadherin depletion led to the loss of cell-cell adhesions (Fig. 3A), and consistent with our previous report showing that E-cadherin-mediated adhesion contributes to cyclin D1 mRNA expression (4), MCF10A cells treated with E-cadherin siRNA had lower levels of cyclin D1 (Fig. 3C). However, E-cadherin knock-down was not sufficient to induce mesenchymal markers (Fig. 3B), nor did it generate mesenchymal expression kinetics of cyclin D1: near constitutive expression persisted between 3 and 9 h after mitogenic stimulation (Fig. 3C).

Cyclin D1 Gene Expression in MCF10A Cells Occurs in Two Distinct Phases with Differing Requirements for Cytoskeletal
Integrity-We then tested the importance of actin polymerization and intracellular tension on early-and mid-G1 phase cyclin D1 gene expression by treating control and transitioned MCF10A cells with CCD. CCD treatment blocked mid-G1 phase cyclin D1 mRNA expression both before and after TGF-␤ treatment (Fig. 4A, 9 h). In contrast, the early G1 phase induction of cyclin D1 seen in control MCF10A cells was relatively resistant to CCD (Fig. 4A, 3 h).
Because our previous studies indicated that early-G1 phase cyclin D1 gene expression is mediated by Rac and repressed by Rho in fibroblasts (3), it was possible that our ability to detect early-G1 phase cyclin D1 mRNA in control MCF10A cells was due to an increased level of Rac signaling or a reduced level of Rho signaling relative to that found after TGF-␤ treatment. This potential mechanism would also be consistent with the increase in actin stress fibers seen after TGF-␤ treatment. However, we found that Rac and Rho GTP levels were similar in control and TGF-␤-treated MCF10A cells (Fig. 4, B and C, respectively). Collectively, these data indicate that there are two distinct phases of cyclin D1 expression in MCF10A cells (earlyand mid-G1 phase), which are resolvable on the basis of their requirements for an organized actin cytoskeleton. If changes in Rac and Rho signaling differentially regulate these two phases, these changes must be imposed downstream of GTP loading.
The Two Phases of Cyclin D1 Gene Expression Require Rac Activity and Are Distinguished by Their Requirement for ERK Activity-We asked whether ERK activity is required for cyclin D1 induction in MCF10A cells both before and after transition to a mesenchymal phenotype by incubating control and TGF-␤-treated MCF10A cells with the MEK inhibitor, U0126. U0126 either had little effect (not shown) or increased (Fig. 5A, 3 h) early-G1 phase cyclin D1 expression in MCF10A cells. In contrast, U0126 strongly inhibited cyclin D1 mRNA induction in mid-G1 phase in MCF10A cells both before (Fig. 5A, 12 h) and after (Fig. 5B, 12 h) TGF-␤ treatment. Intermediate time points showed an intermediate dependence on ERK activity (Fig. 5A,  9 h), likely reflecting the transition from ERK-independent to ERK-dependent cyclin D1 gene expression.
As opposed to ERK activity, we found that inhibition of Rac using either a dominant negative mutant (N17) or Rac1 siRNA efficiently blocked both the early (3 h) and mid (9 -12 h) G1 phase cyclin D1 mRNA expression (Fig. 6, A and B). The Rac requirement was retained in MCF10A cells that had undergone TGF-␤ treatment (Fig. 6A). Thus, the early-and mid-G1 phase expression of cyclin D1 are both dependent on Rac but can be distinguished by their requirement for ERK activity.
We considered the possibility that Rac was required for cyclin D1 gene expression through an interaction with the ERK pathway mediated by the stimulatory effect of PAK on Raf and/or MEK (26). However, Western blotting with antibodies specific to dually phosphorylated (activated) ERK showed that Rac inhibition attenuated the expression of cyclin D1 mRNA without affecting ERK activity in control MCF10A cells (Fig. 6C and Ref. 4).
One previous study concluded that Rac activity is necessary for the translocation of ERK from the cytoplasm to the nucleus FIGURE 4. Two phases of cyclin D1 gene expression distinguished by their dependence on cytoskeletal tension. A, starved control and transitioned MCF10A cells were treated with DMSO (vehicle) or 1 M CCD prior to reseeding on collagen-coated dishes and stimulation with 10% FBS and growth factor mixture. Total RNA was collected at indicated times and analyzed by QPCR. B and C, starved control and TGF-␤-treated MCF10A cells were stimulated (without reseeding) with 10% FBS and growth factor mixture. Lysates were prepared at the indicated times, and Rac (B) or Rho (C) GTPase activities were measured using G-LISA kits as described under "Experimental Procedures." GTPase activities were normalized to serum-starved MCF10A cells.

FIGURE 5. Role of ERK activity in cyclin D1 expression in MCF10A cells.
A and B, starved control (A) and TGF-␤-treated (B) MCF10A cells were pretreated with DMSO (control) or 50 M U0126 for 30 min prior to seeding on collagen-coated dishes and stimulation with 10% FBS and growth factor mixture in the continued absence or presence of TGF-␤. Total RNA was collected at the indicated times and cyclin D1 mRNA levels were measured by QPCR. (27). In this scenario, Rac inhibition would prevent the mid-G1 phase induction of cyclin D1 mRNA by blocking ERK-dependent transcription. To examine this possibility, we asked whether Rac was required for the induction of Fra-1, an immediately-early gene that is strongly regulated by ERK activity and linked to the expression of cyclin D1 mRNA (8,15). Consistent with previous reports (15), the induction of Fra-1 in MCF10A cells was strongly blocked by MEK/ERK inhibition with U0126 ( Fig. 6D) but not by the knock-down of Rac1 (Fig. 6E). Thus, Rac and ERK signal through parallel pathways to stimulate the mid-G1 phase induction of cyclin D1 mRNA.
This joint Rac/ERK requirement for mid-G1 phase cyclin D1 appears to be a conserved feature of cells. Inhibition of either Rac or ERK prevented the mid-G1 phase induction of cyclin D1 mRNA in primary mouse vascular smooth muscle cells (Fig.  7A) and MEFs (not shown). As seen in MCF10A cells, Rac knock-down affected neither ERK activation (Fig. 7B) nor ERKdependent Fra-1 gene induction (Fig. 7, C and D) in mouse vascular smooth muscle cells.
Rb Phosphorylation and S Phase Entry Linked to Co-regulation of Cyclin D1 Gene Expression by Rac and ERK-Inhibition of ERK activity with U0126 efficiently blocked S phase entry in MCF10A cells (Fig. 8A, NS) despite the fact that Rac-dependent cyclin D1 expression persists for 9 h (refer to Fig. 5A). This result implies either that the effects of ERK activity extend beyond cyclin D1 or that continued expression of cyclin D1 is required for mitogen-dependent S phase entry. Other studies have indicated that ERK activity can regulate the cip/kip family of cdk inhibitors (20,28), and indeed we found that U0126 strongly increased p21 cip1 (but not p27 kip1 ) levels (Fig. 8B). While the suppressive effect of ERK activity on p21 levels likely contributes to G1 phase arrest in U0126-treated cells, S phase entry was effectively blocked by U0126 even when p21 cip1 was knocked-down with siRNA (Fig. 8A). Consistent with this data, we found that despite expressing cyclin D1 in early-G1 phase, Rb was not phosphorylated on cyclin D1-cdk4/6 sites in MCF10As until ϳ12-15 h after mitogenic stimulation (Fig. 8C), the time at which cyclin D1 gene expression is strongly dependent on ERK activity (Fig. 5A). These results indicate that joint Rac/ERKdependent expression of cyclin D1 throughout mid-G1 phase is required for Rb phosphorylation and S phase entry.

DISCUSSION
Our previous studies have shown that cyclin D1 gene expression in mammary epithelial cells is regulated by an endogenous Rac signaling pathway (4), whereas most cell types rely on sustained ERK activity to induce cyclin D1 mRNA (29). We now reconcile these two results by showing that Rac-dependent, G1 phase cyclin D1 gene expression in mammary epithelial cells can be resolved into two phases on the basis of their requirements for an organized actin cytoskeleton and ERK activity. Early-G1 phase expression of cyclin D1 in MCF10A cells proceeds in the absence of an organized actin cytoskeleton, is independent of ERK activity, and is insufficient to support Rb phosphorylation or S phase entry (Fig. 9). This phase is either absent or repressed in mesenchymal cells such as fibroblasts (3), smooth muscle, or MCF10A cells after 3 days of exposure to TGF-␤. In contrast, mid-G1 phase expression of cyclin D1 mRNA is jointly regulated by Rac and ERK activities, and this co-regulation is conserved among these cell types (Fig. 9). We did not observe this joint requirement in our previous studies on cyclin D1 gene expression using ␣5-overexpressing NIH-3T3 cells (3), mostly likely because the overexpression of ␣ 5 ␤ 1 integrin in those cells (30) may have overcome the need for co-regulation of cyclin D1 mRNA by Rac and ERK.
The two phases of cyclin D1 expression (early-and mid-G1 phase) in MCF10A cells are reminiscent of the two phases of FIGURE 6. Rac activity is required for cyclin D1 gene expression throughout G1 phase. A, control and TGF-␤-treated MCF10A cells were starved, infected with 300 MOI of lacZ (control) or Rac N17 adenovirus, and reseeded in 10% FBS with growth factor mixture in the continued absence or presence of TGF-␤. Total RNA was collected at the indicated times, and cyclin D1 mRNA levels were measured by QPCR. B, C, and E, MCF10A cells were transfected with either control (nonspecific, NS) siRNA or one of two Rac1 siRNAs, starved, and reseeded on collagen-coated dishes with 10% FBS and growth factor mixture. B, total RNA was collected at the indicated times, and cyclin D1 mRNA levels were measured by QPCR. The results are normalized to the serum-starved cells within each treatment. C, cell lysates were collected 0 and 9 h after mitogenic stimulation and Westernblotted with the indicated antibodies. D, MCF10A cells were starved, trypsinized, preincubated in suspension for 30 min with DMSO (control) or 50 M U0126, and reseeded with 10% FBS and growth factor mixture as described under "Experimental Procedures." Total RNA was collected 0 and 3 h after mitogenic stimulation and Fra-1 mRNA levels were measured by QPCR. E, total RNA was collected at the indicated times, and Fra-1 mRNA levels were measured by QPCR.
ERK activity seen during G1 phase progression. In both cases, the early-G1 phase (also referred to as "transient activity" for ERK) component is dispensable for S phase entry. Moreover, early-G1 phase ERK activity is dispensable for cyclin D1 gene expression, and early-G1 phase cyclin D1 expression is dispensable for Rb phosphorylation. In fact, the timing of Rb phosphorylation in MCF10A cells is similar to that seen in mesenchymal cells, which do not express cyclin D1 until mid-G1 phase (2,3).
One of the ways we approached the issue of cell type-specific signaling was to exploit the finding that long-term exposure to TGF-␤ allows MCF10A mammary epithelial cells to acquire a mesenchymal-like phenotype (22,24,25). Control MCF10A cells have the characteristic epithelial cobblestone appearance at confluence which is the result, at least in part, of robust cellcell adhesions mediated by E-cadherin. Following TGF-␤ treatment, surface E-cadherin is lost, the cells dissociate, and the resulting single cells acquire mesenchymal features such as fibronectin and vimentin expression and prominent actin stress fibers. We found that post-TGF-␤ treatment, the timing of cyclin D1 expression was delayed to mid-G1 phase just as we have seen in fibroblasts and smooth muscle cells (3,31). This cell type-specific change in cyclin D1 kinetics was not associated with changes in the activity profiles of Rac or Rho, but was associated with appearance of actin stress fibers as is typical for mesenchymal cells in culture. FIGURE 7. Joint Rac-and ERK-mediated induction of mid-G1 phase cyclin D1 is a conserved feature of mesenchymal cells. A, starved VSMCs were either transfected with nonspecific siRNA (control, NS) or one of two different Rac1 siRNAs or infected with Rac N17 adenovirus (AdV; 300 MOI). Some of the cells were treated with DMSO (vehicle) or 50 M U0126 prior to reseeding and stimulation with 10% FBS. Total RNA was collected 0 and 24 h after stimulation, and cyclin D1 mRNA levels were measured by QPCR. B and D, VSMCs were transfected with either control (nonspecific, NS) siRNA or one of two Rac1 siRNAs, starved, and stimulated with 10% FBS. B, cell lysates were prepared 0 and 3 h after mitogenic stimulation and Western-blotted with the indicated antibodies. C, VSMCs were starved, trypsinized, preincubated in suspension for 30 min with DMSO (control) or 50 M U0126, and reseeded with 10% FBS as described under "Experimental Procedures." Total RNA was collected 0 and 3 h after mitogenic stimulation, and Fra-1 mRNA levels were determined by QPCR. D, total RNA was collected at 0 and 3 h, and Fra-1 mRNA levels were measured by QPCR. FIGURE 8. Early-G1 phase cyclin D1 is not sufficient to drive Rb phosphorylation or S phase entry. A, MCF10A cells were transfected with either nonspecific (NS) siRNA or one of two p21 cip1 siRNAs. Following serum-starvation, the cells were trypsinized, preincubated in suspension for 30 min with either DMSO or 50 M U0126, and reseeded with 10% FBS, growth factor mixture, and BrdU. Cell lysates were collected 3 h after stimulation and analyzed by Western blotting to confirm p21 knockdown. Cells were fixed 24 h after stimulation, and BrdU incorporation was measured by immunofluorescence microscopy. Results show mean Ϯ S.D. of two independent experiments. B, MCF10A cells were starved, preincubated with either DMSO or 50 M U0126, and reseeded with 10% FBS and growth factor mixture. Cell lysates were collected at indicated times and analyzed by Western blotting. C, experiment in B was repeated in the absence of U0126.
Both E-cadherin and integrins can support Rac signaling to cyclin D1 (4). The results described here indicate that the early-G1 phase cyclin D1 expression seen in MCF10A cells is due, at least in part, to E-cadherin, because that phase of cyclin D1 expression is lost coincident with down-regulation of E-cadherin during TGF-␤ treatment. The absence of strong cadherin-mediated adhesion likely cooperates with Rho-mediated repression (3) to ensure efficient suppression of early-G1 phase cyclin D1 gene expression in mesenchymal cells. We do not yet know if the Rac-dependent signaling pathway(s) that leads to early-G1 phase cyclin D1 mRNA and mid-G1 phase cyclin D1 gene expression are the same or different. Evidence that they are different includes the observation that Rac-dependent induction of cyclin D1 in early-G1 phase requires concomitant signaling by NF-B, whereas its induction of cyclin D1 in mid-G1 phase is independent of NF-B (17).
Collectively, our data indicate that Rac and ERK are mutually co-dependent for functionally productive expression of cyclin D1 mRNA and mitogenesis. As a conserved feature of epithelial and mesenchymal cells, the transcriptional mechanism underlying this co-regulation will be an important matter for further study, and its elucidation should provide interesting insights into extracellular control of the cell cycle.