|Home | About | Journals | Submit | Contact Us | Français|
The Polycomb group protein EZH2 is a transcriptional repressor involved in controlling cellular memory and has been linked to aggressive and metastatic breast cancer. Here we report that EZH2 decreased the expression of five RAD51 paralog proteins involved in homologous recombination (HR) repair of DNA double-strand breaks (RAD51B/RAD51L1, RAD51C/RAD51L2, RAD51D/RAD51L3, XRCC2, and XRCC3), but did not affect the levels of DMC1, a gene that only functions in meiosis. EZH2 overexpression impaired the formation of RAD51 repair foci at sites of DNA breaks. Overexpression of EZH2 resulted in decreased cell survival and clonogenic capacity following DNA damage induced independently by etoposide and ionizing radiation. We suggest that EZH2 may contribute to breast tumorigenesis by specific downregulation of RAD51-like proteins and by impairment of HR repair. We provide mechanistic insights into the function of EZH2 in mammalian cells and uncover a link between EZH2, a regulator of homeotic gene expression, and HR DNA repair. Our study paves the way for exploring the blockade of EZH2 overexpression as a novel approach for the prevention and treatment of breast cancer.
Breast cancer is the most common malignancy and second leading cause of cancer-related deaths in women in the Western world . Despite advances in early detection and treatment, once distant metastases develop, the disease is (at present) incurable. There is a need to better understand the molecular events that lead to breast cancer development and progression to improve the design of clinical tests and novel targeted treatments.
Perturbations of the transcriptional memory of a cell may lead to developmental defects and cancer [2,3]. Two groups of proteins have long been found to be involved in the maintenance of heritable transcription patterns: the Polycomb group proteins (PcG) and their counterpart, the Trithorax group proteins (TrxG) . Both maintain spatial patterns of homeotic box gene expression, which occur early during embryonic development of Drosophila. TrxG act as activators and maintain the “on stage” of gene expression, whereas PcG act as repressors and maintain the “off stage” of gene expression. At least two PcG complexes with distinct functions and target genes have been found: Polycomb Repression Complex (PRC)-2, which consists of EZH2/embryonic ectoderm development (EED)  and Su(z)12 , among other proteins, and PRC-1 , which consists of Ring1, Mel18, Mph1, Bmi1, and Mpc2, among other proteins. In human malignancies, PcG and TrxG have been found to be dysregulated first in neoplasms of hematopoietic origin, but also in solid tumors, including breast and prostate carcinomas [7–16].
EZH2 is the human homologue of the PcG Enhancer of Zeste and is involved in gene silencing . EZH2 contains a SET domain, a highly conserved domain found in many chromatin-associated histone methyltransferases. EZH2 and its binding partners EED and Suz(12) interact directly with type 1 histone deacetylases (HDACs), and this has been suggested to be part of the silencing mechanism [17–19]. Furthermore, recent studies have demonstrated that PRC-2 complexes methylate H3-K9 and K27 in vitro, with a strong preference for K27 [20–25]. Methylation of both H3-K9 and H3-K27 is thought to be involved in targeting PRC-1 to specific genetic loci [23,26]. At present, the mechanism of function of EZH2 in the development of malignant tumors is unknown.
The mammalian genome is at constant risk of mutation as a result of damage to DNA. This can occur due to extrinsic or intrinsic insults. Hampered DNA double-strand break (DSB) repair may lead to structural chromosomal abnormalities and aneuploidy, which may result in cell death or neoplastic transformation [27–30]. The two main mechanisms used to repair DSB are homologous recombination (HR), which promotes accurate repair of DSB by copying intact information from an undamaged homologous DNA template, and nonhomologous end joining (NHEJ), which is a homology-independent mechanism that rejoins broken ends, regardless of sequence .
In vertebrates, HR requires the recombinase RAD51 and its five paralogs RAD51B/RAD51L1, RAD51C/RAD51L2, RAD51D/RAD51L3, XRCC2, and XRCC3 . The sixth paralog, DMC1, is thought to be only active in meiotic cells [32,33]. The RAD51 paralogs share up to 20% amino acid identity with the human RAD51 protein and with each other. Null mutations of any of the paralogs prevent RAD51 from assembling at damaged sites, resulting in decreased RAD51 repair foci formation, hampered HR, and increased sensitivity to DNA cross-linking drugs [34–40]. Although the functions and mechanisms of the regulation of the RAD51 paralogs are still elusive, an emerging body of data indicates that the RAD51 paralogs are important in maintaining chromosomal integrity and function in the early and later stages of HR [32,41].
We have previously reported that overexpression of the PcG EZH2 plays an important role in breast cancer by promoting the growth and invasion of breast epithelial cells . Moreover, EZH2 is a promising novel biomarker of aggressive breast cancer, being elevated in invasive and metastatic tumors, when compared to normal breast tissues. Indeed, EZH2 is an independent predictor of breast cancer recurrence and death . In breast tissues, EZH2 protein expression increases steadily from normal epithelium to epithelial hyperplasia, ductal carcinoma in situ, invasive carcinoma, and distant metastasis, supporting the finding that EZH2 plays a central role in neoplastic transformation . In this study, we show that EZH2 overexpression downregulates all five RAD51 paralogs, which are necessary for HR in breast epithelial cells. This downregulation is associated with an impaired survival capacity of breast epithelial cells amidst the DNA-damaging effects of etoposide and ionizing radiation. Taken together, these data uncover a link between regulators of homeotic gene expression and the DNA repair pathway, and they propose a novel mechanism for EZH2 function in breast tumorigenesis.
Spontaneously immortalized human mammary epithelial MCF10A cells (ATCC, Manassas, VA) were grown in DMEM/F-12 (Cellgro) supplemented with nonessential amino acids, penicillin/streptomycin, l-glutamine, Fungizone, 5% horse serum, 20 ng/ml embryonic growth factor, 100 ng/ml cholera toxin, 10 μg/ml insulin, and 500 ng/ml hydrocortisone. Cells were kept at 37°C and 10% CO2. For detaching, cells were treated with 0.25% trypsin for 5 minutes. MCF7 breast cancer cells (ATCC) were grown in a DMEM/F12 mixture supplemented with 5% FBS. SUM102 human breast cancer cells were grown in F12 supplemented with 5% FBS, 10 μg/ml insulin, and 500 ng/ml hydrocortisone.
Adenoviral constructs were generated by in vitro recombination. Full-length EZH2 (av-EZH2) was inserted into an adenoviral shuttle [pACCMVpLpA(−) loxP-SSP] provided by the University of Michigan (Ann Arbor, MI) vector core. The virus was complemented in a 293 helper cell line, propagated in 911 cells, and purified on a CsCl gradient. The virus was stored in 10 mM Tris–HCl (pH 7.4), 137 mM NaCl, 5 mM KCl, and 1 mM MgCl2 in 10% (vol/vol) glycerol. As control, adenovirus containing no transgene was used. As an additional control, the full-length luciferase gene, an unrelated gene, was inserted in an adenovirus (av-luciferase).
RNA was harvested from MCF10A cells infected with vector or av-EZH2 constructs using TRIZOL (Invitrogen, Carlsbad, CA), according to the manufacturer's instructions. One microgram of RNA was reverse-transcribed using Superscript II (Invitrogen) following the manufacturer's instructions and reconstituted in 20 μl of RNAse-free water. One microliter of cDNA product was used for quantitative real-time PCR analysis. Reaction mixes were made by combining 12.5 μl of 2× SYBR green RT-PCR master mix, 0.5 μl of RNAse inhibitor (Promega, Madison, WI), primer, template, and H2O to a final volume of 25 μl into Smart Cycler Tubes (Cepheid, Sunnyvale, CA). Primers were used at a final concentration of 0.25 μM:
Primers were selected using LocusLink software and spanned an intron–exon junction to prevent amplification of genomic DNA. To confirm the absence of nonspecific amplification, a no-template control was included. Forty cycles were programmed as follows: 15 seconds of denaturation at 94°C, followed by 1 minute of extension at 60°C. Results were quantified against a β-actin standard set in relation to the vector control using Excel.
MCF10A, MCF7, and SUM102 cells were infected with vector or av-EZH2. After 2 days, cells were lysed in 10% glycerol, 150 mM KCl, 10 mM MgCl2 25 mM HEPES, 2 mM DTT, and one Complete Mini protease inhibitor tablet (Roche, Nutley, NJ) and sonicated 4 × 30 seconds at medium speed. Samples were separated on a precast gel (10% Tris–glycine; Cambrex, Rockville, MD) and transferred to nitrocellulose membrane using a Transblot Semidry Transfer Cell (Bio-Rad, Hercules, CA). Membranes were blocked in 5% (wt/vol) milk at room temperature for 1 hour. After three washes in Tris-buffered saline–Tween (TBS-T), primary antibodies against RAD51L2 (Abcam, Cambridge, MA), RAD51L3 (Novus Biologicals, Littleton, CO), RAD51L3, XRCC2, and XRCC3 (Abcam), at a dilution of 1/1000, were added in PBS-T [PBS containing 0.1% (vol/vol) Triton X-100] containing 3% BSA. The membranes were incubated overnight at 4°C, washed three times for 15 minutes in TBS-T, and incubated with anti–mouse horseradish peroxidase (Amersham, Buckinghamshire, UK) at a dilution of 1/2500. After incubation for 1 hour and three more washes, immunoreactive proteins were visualized using ECL reagents (Amersham), following the manufacturer's instructions.
MCF10A cells were plated onto chamber slides and infected with av-EZH2 or av-vector. On the next day, the cells were treated with 0.3 mM etoposide for 24 hours. Following incubation, the cells were washed three times with PBS and fixed in 3% (wt/vol) paraformaldehyde in PBS-T and 0.15% (wt/vol) BSA for 20 minutes. Prior to incubation with antibodies against RAD51 (Santa Cruz Biotechnology, Santa Cruz, CA) or Myc (Cell Signaling, Beverly, MA) at a dilution of 1/1000 for 16 hours, the slides were rinsed twice in PBS-T. After incubation, the slides were washed 4 × 15 minutes in PBS-T. The slides were incubated for 4 hours with Alexa Fluor 488 or Texas Red antibody (Molecular Probes, Carlsbad, CA) at a dilution of 1/500, washed 4 × 15 minutes in PBS-T, and mounted with ProLong antifade reagent with DAPI (Molecular Probes). Images were obtained with an Olympus F500 inverted confocal microscope using ×10, ×40, and ×60 oil immersion objectives. The frequencies of cells containing RAD51 foci were determined by counting at least 250 nuclei in each slide in three independent experiments. Following strict published criteria, only nuclei containing ≥10 RAD51 foci were classified as RAD51-positive .
For colony outgrowth assay, 500 MCF10A, 500 MCF7, and 500 SUM102 cells were plated onto a Petri dish 4 hours before infection. Cells were infected with vector or av-EZH2 constructs. An av-luciferase was used as additional control. One day after infection, the culture medium was changed and cells were continuously treated with increasing concentrations of etoposide (0.1, 0.3, and 0.5 mM). After 7 to 9 days, when colonies had been observed, the cells were washed twice in PBS, and colonies were fixed and stained using methylene blue (Merck, Whitehouse Station, NJ), following the manufacturers' instructions. Colonies with more than 50 cells were counted as positive.
For clonogenic survival assay, 500 MCF10A, 500 MCF7, and 500 SUM102 cells were plated 4 hours before infection. Cells were infected with av-vector, av-EZH2, or av-luciferase constructs. One day after infection, the culture medium was changed and cells were treated for 24 hours with increasing concentrations of etoposide (0.1, 0.3, and 0.5 mM). After 7 to 9 days, colonies were already visible. Cells were washed twice with PBS, fixed, and stained using methylene blue (Merck), following the manufacturer's instructions. Colonies exceeding 50 cells were counted.
For ionizing radiation experiments, 4000 MCF10A, 4000 MCF7, and 4000 SUM102 cells were plated and infected with av-vector, av-EZH2, or av-luciferase. Two days after infection, the cells were irradiated with doses of 0, 300, 600, 900, and 1200 rad using a 137Cs Mark I irradiator. Seven days postinfection, the colonies were stained with crystal violet and counted as described in the above paragraph. For all assays, transient gene expression of EZH2 in MCF10A, MCF7, and SUM102 cells was validated by Western blot analysis.
MCF10A cells were grown to 20% confluency and infected with av-EZH2 or vector control. Two days postinfection, 0.004 μg/ml KaryoMax Colcemid (Gibco, Carlsbad, CA) was added. Sixteen hours later, the cells were trypsinized and washed in PBS. After centrifugation, the cells were resuspended in 250 μl of PBS. Eight milliliters of 0.06 M KCl was carefully added while gently shaking the cells at 37°C in a water bath. Cells were incubated at 37°C for 30 minutes and then fixed with 2 ml of methanol:glacial acetic acid (3:1). After 5 minutes of centrifugation at 1000 rpm, the cells were resuspended in 8 ml of fixative, centrifuged again, and finally resuspended in 1 ml of fixative.
Cells were dropped from a pipette onto a wet frosted slide (Fisher, Pittsburg, PA) from a height of about 50 cm. The slides were dried for 15 minutes at 37°C. ProLong Gold (Molecular Probes) antifade reagent with DAPI was used to fix the cover slip. Slides were dried overnight and then analyzed using Spectral Imaging SkyView software (Applied Spectral Imaging, Vista, CA). The number of chromosomes in 75 to 100 cells in metaphase was counted independently by three investigators (M.Z., D.O.F., and C.G.K.). Experiments were performed in triplicate.
In our previous studies, we demonstrated that EZH2 promoted growth and invasion in human mammary epithelial cells, which resulted in an aggressive and metastatic breast cancer phenotype . Patients with tumors with high EZH2 expression had a worse disease-free and overall survival than those with low levels of EZH2 . EZH2 has been reported to function as a transcriptional repressor in mammalian cells [14,19,43]. Preliminary data from our laboratory of a cDNA microarray comparing EZH2-overexpressing breast epithelial cells to vector controls prompted us to investigate the possible influence of EZH2 on the expression of the RAD51 paralogs (data not shown). Notably, quantitative real-time PCR on av-EZH2– and av-vector–infected human mammary epithelial cells revealed that EZH2 overexpression led to a drastic decrease in the transcripts of the five RAD51 paralogs. As shown in Figure 1A, RAD51L1, RAD51L2, RAD51L3, XRCC2, and XRCC3 transcripts are significantly decreased in EZH2-overexpressing MCF10A cells. These five RAD51 paralogs have crucial nonredundant roles in the HR DNA repair pathway. DMC1, a RAD51 paralog that is active only in meiosis, is not affected (Figure 1A). Immunoblot analyses of MCF10A, MCF7, and SUM102 cells showed that EZH2 overexpression was associated with decreased protein levels of RAD51L1, RAD51L2, RAD51L3, XRCC2, and XRCC3 (Figure 1B).
RAD51L1, RAD51L2, RAD51L3, XRCC2, and XRCC3 are required for the formation of DNA damage–induced RAD51 repair foci. This is a crucial event in HR repair as studies have demonstrated that RAD51 foci formation is significantly attenuated in mutant cells defective in any of the RAD51 paralogs [34–37,40,41,44,45]. EZH2 overexpression in MCF10A cells resulted in a marked decrease in RAD51 foci formation after induction of DSB by etoposide (Figure 2). To investigate the effect of EZH2 overexpression on the repair foci formation ability of breast cells, we determined the number of cells with ≥10 RAD51 foci in EZH2-overexpressing and control cells at 0, 6, 12, and 24 hours following a 24-hour treatment with etoposide. Figure 2B shows that, immediately after removal of etoposide, EZH2-overexpressing cells displayed only about 50% of repair foci compared to the controls. These cells could not compensate for the hampered foci formation by sustaining a low level of repair over a longer period of time. Thus, 6 hours after etoposide withdrawal, the number of cells with RAD51 foci was significantly decreased in all cells, but the number of foci in EZH2-overexpressing cells was still significantly lower than that in the respective controls. After 12 hours, EZH2-overexpressing cells and controls had about the same number of cells with RAD51 repair foci. Twenty-four hours after etoposide withdrawal, RAD51 foci formation ceased almost completely in EZH2-overexpressing cells and controls. These data suggest that HR repair in EZH2-overexpressing cells is less effective than that in the controls.
The above experiments demonstrated that EZH2 overexpression results in a marked downregulation of the RAD51 paralogs involved in HR and an attenuated formation of RAD51 foci after induced DSB. We next investigated the effects of EZH2 overexpression on cell survival after induction of DSB in the spontaneously immortalized human mammary epithelial cell line MCF10A and two breast cancer cell lines MCF7 and SUM102. In addition, to control for possible effects of protein overexpression, we infected these cells with av-luciferase, a gene unrelated to EZH2 that is not involved in HR repair. To induce DSB, we chose to independently use etoposide and ionizing radiation—both previously shown to cause DNA DSBs that are repaired by the HR pathway . As presented in Figure 3, EZH2 overexpression significantly decreased the survival, clonogenic capacity, and colony-forming ability of the three breast cell lines after exposure to ionizing radiation and etoposide. Overall, our experiments suggest that overexpression of EZH2 hampers HR repair of DSB, at least in part, by downregulation of the five RAD51 paralogs RAD51L1, RAD51L2, RAD51L3, XRCC2, and XRCC3.
Knockout of the RAD51 paralogs has been shown to cause multiple forms of genetic instability, including chromosome aberrations, centrosome fragmentation, and missegregation of chromosomes [32,34–37,40,41,44,45]. In particular, a defect in chromosomal segregation leading to aneuploidy appears to be specific for HR repair defects, as it has not been found for cells deficient in alternative repair pathways (for instance, simple end joining of DNA breaks) [32,46,47]. Cytogenetic analysis of EZH2-overexpressing MCF10A cells and vector controls showed that EZH2 overexpression led to a shift in the number of chromosomes. Table 1 shows that, although the control cells show the characteristic chromosomal changes of immortal but untransformed MCF10A cells, EZH2-overexpressing cells have an increased number of chromosomes, with over 65% of cells having 49 to 54 chromosomes (t-test, P < .0001). Adenovirus infection did not affect the number and integrity of the chromosomes, as adenovirus-infected and uninfected MCF10A cells had the same low-level (baseline) aneuploidy characteristic of MCF10A cells [48–53].
In this study, we provide evidence for a hitherto unknown mechanism of action for EZH2 in the development of breast cancer. We report that EZH2 overexpression results in specific downregulation of five RAD51 paralogs RAD51L1, RAD51L2, RAD51L3, XRCC2, and XRCC3, which are crucial for the normal function of the HR pathway of DNA DSB repair in mammalian cells [36,37,40]. Coupled with this effect on gene expression, EZH2 overexpression attenuates RAD51 foci formation at sites of DNA damage, impairing the ability of human breast epithelial cells to repair DNA by HR. Our results show that EZH2 overexpression leads to a lower survival after DNA-damaging insults. We suggest that suppression of the RAD51 paralogs by EZH2 is a likely mechanism of EZH2-driven malignant transformation of breast epithelial cells.
EZH2 is a member of the PcG of early-onset gene repressors involved in maintaining heritable gene expression profiles, thus regulating cell type identity [22–25]. There is compelling evidence that EZH2 overexpression leads to cancer. EZH2 is involved in the pathogenesis of myeloid leukemia, Hodgkin's disease, B-cell lymphoma, and multiple myeloma [9,10,12]. Our group and other investigators have reported that EZH2 promotes the development and progression of breast cancer, prostate cancer, and other solid tumors [11,13–15]. By performing an in silico analysis of published cDNA datasets using ONCOMINE , we found that, in addition to breast cancer, EZH2 mRNA is overexpressed in lung, liver, prostate, bladder, and adrenal carcinomas when compared to their normal epithelial counterparts (data not shown). These data suggest that misexpression of EZH2 may be a fundamental event in the transition from normal state to cancer in several organs. At present, the mechanism of action of EZH2 in cancer is unknown.
There are several lines of evidence supporting a role for the RAD51 gene family in cancer development, especially in light of the functional links between RAD51 and the BRCA genes . In breast cancer, it has been shown that BRCA2 mutation carriers are at greater risk for breast cancer when they also carry the RAD51 variant G135CD [55–57]. XRCC2 codon 188 variant may carry an increased risk for breast cancer . The variant Glu233Gly of RAD51L3 may increase the risk of breast cancer in families without BRCA gene mutations . The emerging association between the RAD51 gene family and cancer is not surprising given the critical role of RAD51 and its paralogs in HR repair, which is essential for the maintenance of genomic stability and tumor avoidance. Cell lines deficient in XRCC2 and XRCC3 have a high frequency of chromosomal instability, with especially high frequencies of chromosome exchange and aneuploidy [32,46,47,60]. XRCC2- and XRCC3-deficient hamster cells have been shown to have increased missegregation of chromosomes resulting in aneuploidy, likely due to a centrosome defect [46,47]. It has been suggested that loss of RAD51 paralog genes leads, in particular, to a chromosomal segregation defect that results in aneuploidy [32,46,47]. Studies have shown that aneuploidy causes imbalances in groups of proteins involved in chromosome segregation, synthesis, and repair, possibly leading to malignant transformation of the cell [27–30].
In the present study, we found that EZH2 overexpression in human mammary epithelial cells resulted in a drastic downregulation of the transcripts of the five RAD51 paralogs involved in HR repair (RAD51L1, RAD51L2, RAD51L3, XRCC2, and XRCC3). Notably, the expression of the sixth paralog, DMC1, and of RAD51 (data not shown) was not affected by EZH2. In light of these data, it is likely that EZH2 may play a role in the transcriptional regulation of the RAD51 paralogs and thus modulate the function of HR DNA repair. In Drosophila, PcG exert their function through a DNA motif called Polycomb response element (PRE) [61,62]. Although a mammalian PRE has not been found yet, a recent study found that the EZH2 complex is able to associate with the hDAB2IP promoter in prostate cells and is able to recruit HDAC1 to the promoter region . Whether the EZH2 complex also associates with the promoters of the RAD51 paralogs is intriguing and warrants investigation.
Consistent with the observed decrease in RAD51L1, RAD51L2, RAD51L3, XRCC2, and XRCC3, EZH2-overexpressing mammary epithelial cells had an attenuated formation of RAD51 repair foci after treatment with etoposide. EZH2 overexpression resulted in a marked decrease in the number of cells with repair foci when compared to controls. The attenuated RAD51 foci formation most likely led to unrepaired chromosomal breaks. Although most cells would succumb to the damage, a subset of cells may survive and acquire additional genetic alterations, which may lead to neoplastic transformation.
We next tested the hypothesis that attenuated DNA repair foci formation may lead to defective HR. Indeed, EZH2-overexpressing cells exhibited decreased survival ability following etoposide and ionizing radiation, both of which cause DNA DSBs that necessitate HR repair [42,63–67]. EZH2 overexpression not only affected HR repair in the spontaneously immortalized human mammary epithelial cell line MCF10A, but had a similar effect on two breast cancer cell lines MCF7 and SUM102. Furthermore, EZH2-overexpressing MCF10A cells exhibited an increase in the number of chromosomes when compared to controls. These results are not surprising given the important role of HR not only in DNA repair, but also in the resolution of crossover during cell divisions . Taken together, our data suggest that EZH2 overexpression occurs early in neoplastic transformation and may be important for the development of prevention strategies for breast cancer based on blockade of EZH2. Our data open the way to future investigations on the detection of EZH2 overexpression in histologically normal breast epithelial cells as a harbinger of carcinoma. This could result in the development of a new clinically applicable tissue-based test to determine the individual risk of breast cancer before histologic atypia is evident.
Our data are in agreement with previous studies demonstrating that aneuploidy is particularly seen in association with defects in HR, but not in association with defects in other DNA repair mechanisms such as NHEJ repair [32,46,47]. These data suggest further investigations to elucidate whether EZH2 overexpression induces random or specific chromosomal alterations. Our data strengthen the emerging link between decreased expression of the RAD51 paralogs, aneuploidization, and cancer development.
In summary, we have identified and delineated a new functional role for EZH2 in the HR mechanism of DNA repair, which may cause aneuploidy in breast epithelial cells. We provide a mechanistic basis for our initial observations implicating EZH2 in promoting breast cancer development and also uncover a role for EZH2 in the regulation of a central pathway in DNA repair and in maintaining the stability of the genome.
We thank Diane Roulston (University of Michigan Medical School) for critical reading of the manuscript and suggestions, Lei Ding for assistance with ONCOMINE, as well as other members of the Kleer laboratory for their constructive suggestions throughout the execution of this project, and Karilynn Schneider and Robin Kunkel for artwork.
1This work was supported, in part, by National Institutes of Health grants RO1CA107469 (C.G.K.), CA090876K08 (C.G.K.), and RO1CA77612 (A.M.C); Army grants DAMD17-02-1-0490 and DAMD17-02-1-491 (C.G.K.); the Burroughs Wellcome Research Fund (S.D.M.); and the Breast Cancer Research Foundation (S.D.M).