p21CIP1 Promotes Mammary Cancer–Initiating Cells via Activation of Wnt/TCF1/CyclinD1 Signaling

This article is featured in Highlights of This Issue, p. 1415

Cancer stem cells (CSC) are thought to cause tumor relapse or metastasis due to their ability to indefinitely replenish cancer growth at primary and metastatic sites (1, 2). CSCs constitute a subpopulation of cells that is endowed with self-renewing capability, promoted by asymmetric cell divisions which preserve the stem cell pool while giving rise to differentiated progenitors (3–5). CSCs are often maintained in a quiescent state that prevents hyperproliferation and consequent cellular exhaustion. One molecule known to drive the quiescence of stem cells is p21CIP1 (p21), a cell-cycle inhibitor, that was shown to protect the hematopoietic stem cell pool from depletion (6, 7). Moreover, p21-mediated cell-cycle arrest is necessary for DNA repair that prevents onset of mutations that cause genomic instability and elimination of the CSC pool (8). In further support of this idea, others have shown that inhibitor of differentiation (ID) genes that maintain cancer-initiating cells in an undifferentiated stem cell state, as well as initiate metastatic activity, upregulate p21. p21 is thought to prevent accumulation of DNA damage that results in exhaustion of the CSC pool necessary for replenishing tumor growth (8, 9). Hence, cell-cycle arrest by p21 is essential for maintaining CSC quiescence and genomic integrity, thus preventing onset of mutations that may alter and eradicate the CSC pool.

We previously showed that p21 knockout in the PyMT mammary tumor model suppresses metastasis (10). Interestingly, this effect was associated with mesenchymal-to-epithelial transition that results in differentiation and inhibition of cell migration. These results suggested that p21 supports an undifferentiated mesenchymal state that is conducive of stemness (10). Although studies have shown that p21 loss results in CSC depletion due to excessive proliferation (6, 7, 11), none have addressed the possibility that p21 may also promote stemness via activation of signaling that regulates CSC proliferation and/or differentiation.

Here we demonstrate a novel mechanism whereby p21 promotes a cancer stem–like phenotype. We show that p21 knockout in the PyMT mammary tumor model inhibits tumorsphere formation, Aldefluor activity, and importantly tumor-initiating potential, which were in turn rescued by p21 overexpression in PyMT/p21 knockout cells. Interestingly, p21 ablation led to complete inhibition of Wnt/β-catenin signaling activity, which was due to TCF1 downregulation. TCF1 knockdown in PyMT tumor cells suppressed Wnt signaling and mammosphere formation, which was due to Cyclin D1 attenuation. Thus, p21 acts as a central regulator of the mammary CSC-like phenotype via activation of Wnt/TCF1/Cyclin D1 signaling.

Female FVB mice and athymic nude mice were obtained from Charles River Laboratories. Animal protocols of this study were approved by Institute for Animal Studies at Albert Einstein College of Medicine.

p21CIP1 expression vector was obtained from Dr. Stuart Aaronson (Mount Sinai School of Medicine, New York, NY). Mouse p21CIP1 shRNA and control shRNA were obtained from Santa Cruz Biotechnology Inc. Mouse LEF1, TCF1, Cyclin D1 shRNA clones, and nonsilencing control shRNA in the pLKO.1 lentiviral vector (Open Biosystems) used to knockdown these molecules were obtained from the genomic core facility at Einstein. To generate viruses, lentiviral vectors were transfected into 293T cells with Tat, Rev, Gag/Pol, and VSV-G vectors. TOPFlash, FOPFlash, and Renilla plasmids were obtained from Millipore.

Lentiviral particles were generated by transient cotransfection of 293T cells with lentivirus-based vector expressing either the full length clone of desired gene or an shRNA sequence to target-specific RNA of gene. Briefly, a 100-mm dish seeded with 3 × 10⁶ cells were transfected with 0.6 μg of lentiviral packaging gene TAT, RVE, and GAG/POL and 1.2 μg of VSV-G and 12 μg of DNA of interest in lentiviral backbone. Fugene was used to transfect the cells. Following 48 hours of transfection, supernatant was collected, centrifuged at 2,000 rpm for 10 minutes, and filter sterilized.

PyMT and PyMT/p21KO primary mammary tumor cell lines were derived from the PyMT mouse or the PyMT/p21KO mouse as described in detail (10). PyMT and PyMT/p21 KO mice were in a (Bl6/129S/FVB background) mixed background (10). Met-1, is a PyMT mammary tumor cell line that was derived from an MMTV-PyMT mouse in a FVB background. This cell line was serially transplanted into the mammary fat pad of FVB female mice to generate a highly metastatic mammary tumor cell line, known as Met-1 (12). All cell lines were routinely tested for Mycoplasma, and the genetic identity of the cell lines was confirmed by qPCR for expression of the mouse PyMT oncogene. The cell lines were maximally used within 10 passages in culture.

L-cells expressing Wnt3a were a kind gift from Dr. Stuart Aaronson (Mount Sinai School of Medicine, New York, NY). L/Wnt3a cells were cultured in DMEM/10% FBS and allowed to grow for 3 days till they reach confluency. Media was collected, filtered, and mixed 1:1 with DMEM/10% FBS for stimulation of cells.

The antibodies used are against p21 (Millipore), β-actin (Sigma); Slug, LEF1, (Cell Signaling Technology); and Sox9, active-β-catenin, β-catenin, TCF-4 (Millipore). Antibodies against TCF1, Cyclin D1, p-LRP6, and LRP6 were obtained from Santa Cruz Biotechnology.

Cells or tissues were extracted in solubilization buffer (50 mmol/L Tris-HCl pH 7.5, 150 mmol/L NaCl, 0.5 mmol/L MgCl2, 0.2 mmol/L EGTA, 1% Triton X-100) including protease inhibitors. Cells were extracted in solubilization buffer including 1% SDS and sonicated. Thirty-microgram protein were loaded on 7% to 12% SDS-polyacrylamide gels and transferred onto Immobilon membranes. Blots were probed overnight at 4°C with indicated antibodies and developed by chemiluminescence using ECL Detection Reagents (Pierce, Thermo Fisher Scientific).

Cells were seeded at a density of 9 × 10⁴ cells per well of 12-well plates. Next day, 1 μg of polybrene was added to viral solution, mixed well and 250 μL was added dropwise to four different wells. After gently mixing, plates were placed in incubator for 1 hour with rocking for every 15 minutes. After 1 hour, DMEM containing 10 % FBS was added without antibiotics. Incubated at 37°C for 24 hours, cells were collected following trypsinization, pooled, and placed in a 10-cm dish with selective antibiotics.

RNA was isolated using TRizol Reagent (Invitrogen) following the manufacturer's protocol. Murine p21CIP1, Snail2 (Slug), Sox9, TCF4, TCF1, LEF1, Axin2, LRP6, and Cyclin D1 TaqMan primers and RNA-to-CT 1-Step Kit were obtained from Applied Biosystems. Real-time PCR experiments were performed according to the manufacturer's protocol using a StepOnePlus Real-Time PCR Instrument from Applied Biosystems. GAPDH was used as the reference gene and relative mRNA levels were determined using the 2^(−ΔΔCt) method. Three independent experiments were performed. Statistical significance was calculated using the two-tailed t test, P < 0.05.

Cells were plated in 24-well plates in duplicates. Reporter plasmid TOP-Flash or FOP-Flash, which contain three optimal copies of the TCF/LEF-binding site (TOPFlash), or mutated copies of the TCF/LEF-binding site (FOPFlash) upstream of a minimal thymidine kinase promoter directing transcription of a luciferase gene were used. (TOPFlash) or (FOPFlash; 0.5 μg) together with a Renilla luciferase plasmid (0.1 μg) were transfected using Lipofectamine LTX (Invitrogen) according to the manufacturer's protocol. To activate the Wnt signaling following 24 hours of transfection, Wnt3a-conditioned media with or without 10 μmol/L ICG001 (Selleck Chemicals) was added. Cell lysates were obtained using lysis buffer provided in Dual Luciferase Assay Kit (Promega). Firefly and Renilla luciferase readings were recorded 48 hours posttransfection using Luminometer (Promega). The firefly luciferase activity is then normalized to the Renilla luciferase activity, and fold increase in TOP-Flash activity compared with FOP-Flash was plotted as mean ± SEM of triplicate tests and validated by t test, P < 0.05.

The Aldefluor Kit (Stemcell Technologies) was used to measure the cell population with high ALDH enzymatic activity. Dissociated single cells were suspended in Aldefluor assay buffer containing the ALDH substrate, Bodipy-aminoacetaldehyde (BAAA) at 1.5 mmol/L and incubated for 45 minutes at 37°C. To distinguish between ALDH-positive and ALDH-negative cells, for each sample of cells an aliquot was incubated under identical condition in the presence of a 10-fold molar excess of the ALDH inhibitor, diethylaminobenzaldehyde (DEAB). This results in a significant decrease in the fluorescence intensity of ALDH-positive cells and was used to calibrate the flow cytometer. After incubation, cells are resuspended in 0.5 mL of ice-cold Aldefluor assay buffer for flow cytometry analysis using the BD LSR II Flow Cytometer in the 488/Green fluorescent channel.

Cells were digested by 0.25% Trypsin-EDTA, resuspended in HBSS plus 2% FBS at 2 × 10⁵ cells/100 mL, and incubated with antibodies against mouse CD49 (PE-Cy7-conjugated, 1:100, from BD Biosciences) and CD24 (FITC-conjugated, 1:200, from BD Biosciences) in cold room for 30 minutes. Flow cytometry analysis was performed using the BD LSR II Flow Cytometer at the Albert Einstein Flow Cytometry Core facility.

Single-cell suspensions were plated on a 60-mm ultralow attachment tissue culture dish (Corning) at a density of 1 × 10⁵/dish in DMEM/F12 containing 10 ng/mL FGF-2, 4 μg/mL heparin, 20 ng/mL EGF, 5 μg/mL insulin, and 5 μg/mL hydrocortisone for 7 days. Inhibition of mammospheres was tested with ICG001 (Selleck Chemicals), which was added at 5 or 10 or with 20 μmol/L of Tankyrase inhibitor XAV 939 (Tocris Bioscience) in the medium for 7 days. For secondary sphere formation, primary spheres were dissociated in 1:1 trypsin/DMEM at 37°C, and mechanically dispersed by passing through a 23-gauge needle. Single cells were replated at 5 × 10³/dish, and incubated in 37°C 5% CO₂ for 7 days. At the end of the treatment, cells were transferred to a 35-mm MatTek dish and spheres, as well as total cells (including mammospheres, single cells, and clusters) per microscopic field over five fields were counted. Mammospheres were expressed as percentage using the average number of spheres per 100 cells.

PyMT, PyMT/p21KO (p21KO), p21KO+p21 (p21 rescue) cells, or Met1 control and Met1p21shRNA cells were each resuspended in 25% Matrigel PBS at indicated cell numbers and injected into two bilateral sites of mammary fat pads from female athymic nude mice or FVB immunocompetent mice. Tumor onset was determined by palpation and aggregate mammary tumor growth from two sites per each mouse, was measured by calipers to determine tumor volume as an aggregate of two tumors per mouse (10).

All of the data are presented as the mean ± SEM. A Student t test was used, unless otherwise specified, to calculate the P; P < 0.05 is considered significant.

We showed that p21CIP1 gene knockout in the mouse MMTV-PyMT mammary tumor model inhibits lung metastasis and that p21 overexpression into PyMT/p21 knockout cells restores metastatic colonization (10). To determine whether metastasis promotion by p21 was associated with stem cell activity, we tested the effect of p21 knockout on CSC properties, including tumorsphere formation, aldehyde dehydrogenase 1 (ALDH1) activity, and importantly, tumor-initiating potential (Fig. 1). We used primary tumor cell lines derived from PyMT and PyMT/p21KO mouse mammary tumors (10). Compared with approximately 60% of PyMT mammary tumor cell lines (PyMT1 and PyMT2) which formed robust tumorspheres, PyMT/p21KO cell lines (p21KO1 and p21KO2) were devoid of sphere-forming ability (Fig. 1A). Replating single cells from first-generation PyMT1 and p21KO1 tumorspheres, showed consistent inhibition of secondary sphere formation by p21 loss (Fig. 1B), thus suggesting a role of p21 in CSC/progenitor cell proliferation. Moreover, the percentage of tumor cells with ALDH1 activity, inherent to tumor stem/progenitor cells, was reduced from 73% in PyMT1 to 19% in p21KO1 cells (Fig. 1C). In addition, FACS sorting for CD49/CD24 expression, known to be medium-high in basal1 mammary stem cells (11), showed that the fraction of CD49/CD24-positive cells was reduced from 99% and 98% in PyMT (PyMT1 and PyMT2) cell lines to 27% and 3% in p21KO (p21KO1 and p21KO2) cell lines (Fig. 1D).

Furthermore, tumor-initiating potential, a gold standard for testing stemness activity, was measured by mammary implantation of limiting dilutions of tumor cells into two sites at the mammary fat pads of female athymic nude mice. Compared with PyMT1 cells which formed exponentially growing tumors at high (1 × 10⁶) and low (5 × 10⁴) cell densities, p21KO1 cells were unable of forming sizeable tumors at both dilutions (Fig. 2A). Interestingly, rescue of p21 expression in p21 KO1 cells (Fig. 2B) induced tumor forming ability as compared with p21KO1 cells at high (1 × 10⁶) and low (5 × 10⁴) cell densities (Fig. 2B and C). Of note, p21-rescued cells (p21KO1+p21) grew tumors with delayed onset as compared with PyMT cells (24 days vs. 4 days; Fig. 2A–C). This may be either due to a partial rescue of p21 expression, which was at most noted in 60% of p21-infected p21KO1 cells, or delayed initial growth, resulting from p21 reexpression in p21 KO1 cells. Despite a gap in tumor onset, p21KO1+p21 cells were able of forming similar tumor sizes as PyMT controls, when compared 15 days post their respective onset time (Fig. 2A–C). In contrast, p21KO1 tumors were consistently reduced in size throughout the time course of tumor growth (Fig. 2B and C). In support of these data, p21 overexpression in p21KO1 cells was also able of restoring sphere formation to a similar level as PyMT1 cells (Fig. 2D), and caused a 4.5-fold increase in the fraction of ALDH1-positive cells (Fig. 2E). Consistent with partial rescue of p21, ALDH1 activity was noted in 30% of p21KO+p21 cells as compared with 73% of PyMT1 cells (73%; see Fig. 1C). Thus, these data demonstrate that p21 promotes a powerful cancer stem–like phenotype that is consistent with its prometastatic activity (10).

To further confirm the effect of p21 gene knockout on CSC properties, we used shRNA to knockdown p21 expression in a mammary PyMT tumor cell line, known as Met-1, which was derived from an independent PyMT model in the FVB background (12). This cell line was serially transplanted into the mammary fat pad of FVB female mice to generate a highly metastatic cell line that can be tested in an immunocompetent FVB mouse (12). p21 shRNA–mediated knockdown in Met-1 cells attenuated sphere forming efficiency from 30% to 12% (Fig. 3A) and reduced the Aldefluor-positive fraction of Met-1 cells from 20% to 7% (Fig. 3B). Moreover, p21 knockdown in Met-1 cells reduced tumor incidence following mammary fat pad inoculation of 5 × 10⁵ and 5 × 10⁴ tumor cells into syngeneic FVB female mice (Fig. 3C). The number of mice developing tumors at 14 weeks (end points) post fat pad injection of 5 × 10⁵ cells was higher for Met-1 control cells (5/6 mice) than for Met-1/p21shRNA cells (2/6 mice; Fig. 3C). At the lower cell density (5 × 10⁴), Met1 cells produced tumors in 3 of 6 mice, whereas Met-1/p21shRNA cells in 2 of 6 mice. Of note, Met-1 control cells produced substantially larger tumors than Met-1/p21shRNA cells at both cell densities. These data confirmed that ablation of p21 expression by shRNA recapitulates the effects of p21 gene knockout in PyMT cells by attenuating CSC-like properties.

It has been reported that Slug and Sox9 are two transcription factors that cooperatively drive the mammary stem cell state, as well as tumorigenic- and metastasis-seeding abilities (13, 14). Indeed, p21 knockdown in Met-1 cells (Fig. 3D), inhibits Slug, and to a lesser extent, also Sox9 expression (Fig. 3D). Interestingly, Slug rescue into Met-1/p21shRNA cells was able of restoring sphere formation (Fig. 3E). In comparison, Slug and Sox9 were both substantially reduced in four of the p21KO cell lines as compared with PyMT control cell lines (Fig. 3F); however, reexpression of Slug in p21KO1 cells (Fig. 3G) did not rescue sphere formation (Fig. 3H). This might be due to the dramatically lower levels of Sox9 in p21 KO1 cell lines relative to Met-1/p21 shRNA cells. However, reexpression of Sox9 and Slug in p21KO1 cells was also unable of rescuing sphere formation (not shown), which could be due to insufficient levels of expression. Alternatively, PyMT cells may be deficient in other stemness factors that may be present in Met-1 cells such as Sox10 (15, 16). These data remain nevertheless consistent with that Slug and Sox9 can cooperate in creating a mammary stem cell phenotype.

The Wnt signaling pathway is known to promote stemness by increasing levels of transcriptionally active β-catenin induced by Wnt ligands, resulting in gene transcription leading to stem cell renewal (17). Wnt/β-catenin is therefore thought to promote self-renewal and suppress differentiation of stem cells (17). Comparison of four p21KO to four PyMT mammary tumor cell lines, revealed that p21KO cell lines expressed significantly reduced levels of unphosphorylated or stabilized β-catenin (active-β-catenin) as compared with PyMT control cell lines, whereas total β-catenin expression was unchanged (Fig. 4A).

To test whether p21 promotes canonical Wnt signaling, we assayed for β-catenin activation of the T-cell factor (TCF)/lymphoid enhancer factor (LEF) transcription factors, using a TCF/LEF reporter plasmid, which consists of three TCF-binding sites fused to luciferase (18). β-catenin–mediated TCF/LEF transcriptional activation was measured as fold induction of a consensus TCF-β-catenin reporter (TOP-FLASH) with respect to FOP-FLASH (a scrambled consensus reporter), normalized for transfection efficiency. Comparison of PyMT with p21KO cell lines, showed an overall significant decrease in Wnt3a-stimulated TOP/FOP FLASH reporter activation in p21KO (p21KO1–5) cell lines relative to control PyMT (PyMT 1–3) cell lines, which overall expressed higher, yet variable levels of TOP/FOP FLASH, likely due to tumor cell heterogeneity (Fig. 4B). Moreover, p21 rescue in p21KO1 cells enhanced TOP/FOP FLASH reporter activity by 2-fold in ligand-untreated cells, and by 4-fold in Wnt3a-stimulated cells (Fig. 4C), implying a stimulatory effect of p21 on Wnt signaling, which occurs even in the absence of Wnt ligand. Collectively, these data confirm an active role for p21 in potentiating Wnt/β-catenin/TCF signaling.

To confirm that Wnt/β-catenin activity regulates CSC-like behavior in PyMT cells, we used a selective Wnt/β-catenin signaling inhibitor, ICG001, which antagonizes β-catenin/TCF/CBP-mediated transcription (19). Treatment of PyMT cell lines with ICG001, which abrogates TOP/FOP FLASH reporter activation by Wnt3a (Supplementary Fig. S1A), dramatically suppressed sphere forming activity (Fig. 4D and E). Importantly, ICG001 prevented rescue of sphere formation by p21 overexpression in p21KO1 cells (Fig. 4D and E). Similarly, a tankyrase inhibitor, XAV939, which directly inhibits Wnt/β-catenin signaling by causing Axin stabilization (20), was also efficient in blocking sphere formation by PyMT1 cells, as well as by p21-rescued p21KO1 cells (p21KO1+p21; Fig. 4F and G). These data underscore an active role of p21 in promoting CSC activity via Wnt/β-catenin/TCF transactivation.

We tested whether p21 stimulation of TCF/LEF transcriptional activity was due to increases in expression of Wnt signal transducers including LEF1 and TCF4 (21, 22). There was no difference in TCF4 mRNA expression between PyMT1 and p21KO1 cells before or after Wnt3a treatment (Fig. 5A and B). LEF1 expression was also unchanged between PyMT1 and p21KO1 cells that were untreated with Wnt3a; however, Wnt3a stimulated a dramatic increase in LEF1 mRNA levels in PyMT1 cells, but not in p21KO1 cells (Fig. 5A and B). At the contrary, Wnt3a reduced LEF1 expression in p21KO1 cells below levels found in Wnt-untreated p21KO1 cells, implying a synergy between p21 and Wnt3a in regulating LEF1 expression (Fig. 5A and B).

To determine whether p21 ablation in PyMT cells affects the expression of Wnt transducing effectors that are upstream of β-cat/TCF/LEF transactivation, we measured the expression of Axin2, a central node in the Wnt signaling cascade, as well as LRP6, a Frizzled coreceptor, which regulates Wnt ligand activation (22, 23). Wnt3a stimulated Axin2 mRNA expression in PyMT1 cells, and to a lesser degree also in p21KO1 cells, (Fig. 5C). LRP6 mRNA (Fig. 5C) or protein (Fig. 5D) expression was not significantly affected by p21 knockout. Similarly, LRP6 phosphorylation, which regulates binding of Wnt to Frizzled (23), was not significantly changed (Fig. 5D). These data were confirmed by densitometry of immunoblots showing that p-LRP6 or LRP6 expression was not significantly changed in p21 KO cells relative to PyMT cells (Supplementary Fig. S1B and S1C). Frizzled 7 expression could not be assessed by Western blot analysis due to variability in detection levels by different antibodies. In sum, our data demonstrate that p21 ablation in PyMT cells readily suppresses the expression of LEF1, a main Wnt signal effector as well as a target gene (21, 22).

To further examine the role of p21-mediated LEF1 upregulation in Wnt signal activation, we knocked down LEF1 in PyMT cells by two independent shRNAs. This led to dramatic suppression of LEF1 mRNA expression, in Wnt3a-treated and -untreated cells (Supplementary Fig. S2A). LEF1 knockdown, however, did not inhibit Wnt3a-mediated TOP/FOP FLASH reporter activation as compared with marked inhibition by the Wnt signal antagonist ICG001 (Supplementary Fig. S2B). In support of this finding, LEF1 shRNA knockdown by two targeting sequences in PyMT cells did not affect sphere formation (Supplementary Fig. S2C and S2D).

Because TCF1 might be redundant to LEF1 in recruiting β-catenin to derepress Wnt target gene transactivation (21–24), we tested whether p21 knockout inhibits TCF1 expression. Of note, TCF1 mRNA was significantly attenuated in p21KO1 relative to PyMT1 cells under basal or Wnt3a-treated conditions (Fig. 6A), suggesting p21 supports TCF1 expression. Knockdown of TCF1 in PyMT1 cells by two independent shRNAs led to approximately 60% reduction in TCF1 mRNA expression before or after Wnt3a treatment (Fig. 6B). Consistently, both TCF1shRNAs reduced Wnt3a-stimulated TOP/FOP FLASH activation by 50% as compared with control cells (Fig. 6C). Interestingly, ICG001 was able of further attenuating Wnt reporter activity in TCF1shRNA cells as compared with control cells, underscoring the dependence of Wnt signaling on TCF1 (Fig. 6C). Moreover, TCF1 knockdown was highly efficient in suppressing sphere formation (Fig. 6D). Thus, p21 knockout attenuates TCF1 expression, thereby attenuating canonical Wnt signaling and sphere formation.

Among the TCF/LEF target genes, Cyclin D1 was shown to regulate mammary CSC renewal (25, 26). We tested whether TCF1 knockdown in PyMT cells affects Cyclin D1 expression. Indeed, TCF1 knockdown in PyMT1 cells, led to significant attenuation of Cyclin D1 mRNA expression (Fig. 6E). In support of this idea, knockdown of Cyclin D1 by two shRNA sequences in PyMT1 cells, markedly reduced sphere formation (Fig. 6F).

Importantly, most of these findings were recapitulated in the metastatic PyMT/Met-1 cell line (Fig. 7). p21 knockdown by shRNA in Met-1 cells, caused a dramatic reduction in LEF1 expression, but had no effect on TCF4 or LRP6 mRNA expression (Fig. 7A–C). p21 knockdown, however, increased Axin 2 mRNA in untreated cells, but did not meaningfully change Axin 2 mRNA expression in Wnt3a-treated cells (Fig. 7C).

Consistent with the results in PyMT cells, Wnt3a stimulation caused a dramatic increase in LEF1 and TCF1 protein expression that was suppressed by p21 shRNA in Met-1 cells (Fig. 7D). Furthermore, TCF1 knockdown in Met-1 cells (Fig. 7E) reduced sphere formation (Fig. 7F). Interestingly, TCF1 knockdown in Met-1 cells also reduced Cyclin D1 expression, similar to the effect of p21 shRNA (Fig. 7G). Consistent with a role of TCF1/Cyclin D1 in promoting stem-like properties, Cyclin D1 attenuation by siRNA in Met-1 cells (Fig. 7H), caused a significant decrease in sphere formation (Fig. 7I).

Thus, the cumulative data demonstrate that p21 regulates sphere formation or CSC/progenitor proliferation by upregulating TCF1 and consequently Cyclin D1, thereby promoting a CSC-like phenotype.

p21CIP1 is a cell-cycle inhibitor that acts as a tumor suppressor by halting cell proliferation and promoting DNA repair, both involving cell-cycle arrest. However, increasing evidence suggests that p21 may also act as a malignancy promoter (8). Wnt/β-catenin signaling is known to regulate stem cells by promoting dedifferentiation and self-renewal (17). It also promotes quiescence via p21 upregulation to protect cells from excessive proliferation that leads to CSC exhaustion (27). p21 was also shown to act downstream of genes that regulate cancer-initiating cells and metastasis such as “ID” genes, thereby preventing excess DNA damage by stress or chemotherapy which results in CSC exhaustion (9).

We previously showed that p21 knockout in the PyMT mammary tumor model inhibits lung metastasis (10). We speculate that p21 loss may inhibit metastasis by suppressing CSC behavior. Indeed, we show that p21 knockout in PyMT mammary cancer cells dramatically suppresses CSC properties, involving tumorsphere formation, AlDH1 activity, and tumor-initiating potential. The stem promoting effect of p21 was corroborated by genetic rescue of p21 in knockout cells, resulting in reactivation of CSC properties, including tumor regenerative potential.

Interestingly, p21 appears to stimulate stemness properties via activation of canonical Wnt signaling, a key pathway implicated in stem and CSC self-renewal (17). Consistent with this idea, p21 knockout in PyMT cells suppressed Wnt3a-stimulated β-catenin/TCF transactivation, and importantly, p21 overexpression in p21 knockout cells was able of rescuing Wnt reporter activity. We show that p21 activates canonical Wnt signaling via positive regulation of TCF1 and Cyclin D1, two main effectors as well as target genes of Wnt signaling (21). Our data demonstrate that p21 strongly supports the expression of LEF1 and TCF1; however, only TCF1, and not LEF1, knockdown in PyMT cells was able of attenuating Wnt reporter activation and sphere formation. This implied that TCF1 is the likely protein recruiting β-catenin to the transcriptional complex to derepress Wnt signaling PyMT cells, which is consistent with the higher abundance of TCF1 relative to LEF1 protein in PyMT cells. We believe that TCF1 regulates Cyclin D1, a bona fide Wnt target gene, which in turn promotes CSC/progenitor proliferation (25, 26). In support of this idea, we found that TCF1 knockdown in PyMT cells attenuates Cyclin D1 expression and Wnt/TCF TOPFLASH reporter activation, as well as sphere formation, implying a role for Cyclin D1 upregulation by p21 in driving CSC/progenitor self-renewal (25). In addition to promoting dedifferentiation and self-renewal, p21 also activates EMT and stemness factors such as Slug and Sox9, which may also be activated by Wnt signaling (24). These data are partly consistent with our observation that Slug rescue in p21-knockdown mammary tumor cells restores sphere formation.

At last remains the question whether p21 loss inhibits CSCs by enhancing cell proliferation, which might lead to CSC exhaustion (6, 7). We previously showed that p21 loss in PyMT mammary tumor cells causes increased cell growth in vitro, but not in vivo, where it surprisingly reduced mammary tumor growth, an effect which was associated with striking epithelial differentiation (10). We speculate that the effect of p21 loss on proliferation is offset by its effect on differentiation, which in turn stops cell proliferation. This effect was mostly observed in vivo, but not in vitro, likely due to abundance of growth factors in the culture media that tilt the balance toward proliferation over differentiation. These data suggest that p21 loss causes a switch from proliferation to differentiation, which in turn inhibits EMT, migration, and ultimately stemness, a matter that warrants further investigation.

Our cumulative data point to an interesting and novel function of p21 in potentiating CSCs via activation of canonical Wnt signaling due to TCF1/Cyclin D1 upregulation, resulting in promotion of self-renewal, and leading to proliferation of CSC/progenitor cells that fuels tumor growth and also metastasis. Further investigation on the precise molecular mechanism whereby p21 activates canonical Wnt signaling is warranted.

No potential conflicts of interest were disclosed.

Conception and design: X. Qian, L. Norton, R.B. Hazan

Development of methodology: X. Qian, Z. Ren

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): O. Benard, X. Qian, H. Liang, K. Suyama

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): O. Benard, X. Qian, H. Liang, K. Suyama, L. Norton

Writing, review, and/or revision of the manuscript: O. Benard, X. Qian, L. Norton, R.B. Hazan

Study supervision: R.B. Hazan

This work was supported by grants from the Avon foundation, Breast Cancer Research Foundation, and Cure Breast Cancer Foundation to (R.B Hazan and L. Norton). This work was also supported by the Albert Einstein Cancer Center Support Grant of the National Institutes of Health, P30CA013330.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.