SALL1 functions as a tumor suppressor in breast cancer by regulating cancer cell senescence and metastasis through the NuRD complex

We investigated whether SALL1 could function as a tumor suppressor and dissected the molecular mechanism by which it regulates human breast cancer. We first determined SALL1 gene expression levels in breast cancer cell lines using Real-time PCR analyses. In parallel, SALL1 expression in cancer cell lines from other types of cancers (melanoma, prostate cancer, colon cancer and lymphoma), as well as in normal breast primary cell lines, fibroblasts and 293 T cells were also determined. We observed markedly elevated SALL1 gene expression in all melanoma cell lines and moderate gene expression in normal fibroblasts and 293 T cells (Fig. 1a). In contrast, SALL1 gene expression levels in all the tumor cell lines from breast cancer, prostate cancer, colon cancer and lymphoma were significantly down-regulated. In addition, SALL1 expression in normal primary breast cells and cell lines (BN6, BN16, MCF10A and MCF12A) were also relatively low compared with normal fibroblasts and 293 T cells. These results were further confirmed in human breast and melanoma tumor tissues, showing down-regulation of SALL1 gene in breast cancer tissues (Fig. 1b). Clinically breast cancers can be classified into several distinct subtypes based on the histopathology and molecular characteristics which are associated with the therapeutic options and prognostic outcomes. To further investigate the association of SALL1 gene expression with breast cancer subtypes, we utilized The Cancer Genome Atlas (TCGA) normalized log₂ transformed breast cancer agilent microarray expression data sets downloaded from the cBioPortal (http://​www.​cbioportal.​org/) for our studies [27]. Breast cancers have been designated as luminal A, luminal B, epidermal growth factor receptor 2 (HER2) enriched, basal like and normal like, five important categories based on gene expression profiles [28]. We found a substantial decrease of SALL1 expression in the basal like breast cancer compared with that in normal breast tissue (p < 0.001) (Fig. 1c), which is consistent with the previous report [20]. Furthermore, we also compared SALL1 expression in breast cancer patients with different hormone receptors, estrogen receptor (ER) and Progesterone receptor (PR), or HER2 expression statuses (Fig. 1d). The analysis revealed that there was significantly lower expression of SALL1 in the ER⁻ breast cancer than that in the ER⁺ cancer tissues and normal breast tissues (p = 0.006 and p = 0.008, respectively). Similarly the SALL1 expression was lower in PR⁻ breast cancer than that in PR⁺ breast cancer or the normal breast tissues (p < 0.001 and p = 0.0059, respectively). Importantly, we also found a significant lower expression level of SALL1 in triple negative breast cancer tumors (ER⁻, PR⁻, and Her2⁻) compared with that in normal breast tissues (p = 0.0021). To further investigate the SALL1 expression in human breast cancer, we determined the SALL1⁺ cell numbers in cancer tissues from breast and melanoma cancer patients, using immunohistochemical staining analyses. Consistent with the gene expression results (Fig. 1b), we found that melanoma tumor tissues contained larger numbers of SALL1⁺ cells (mean 232/field), while in breast cancer tissues, the SALL1⁺ cells were low (mean 50/field) (Fig. 1e and Additional file 1: Figure S1A). Notably, SALL1⁺ cell numbers in ER⁻ patients were significant lower than those in ER⁺ patients (Fig. 1f). In addition, SALL1⁺ cell population in HER2⁺ patients was much higher than that in HER2⁻ patients (Fig. 1g). To further identify mechanism(s) responsible for the down-regulation of SALL1 gene in breast cancer, we explored the mutations and promoter methylation status of the SALL1 gene via The Catalogue of Somatic Mutations in Cancer (COSMIC) genome browser [27]. However, we did not identify any mutations that could account for SALL1 down-regulation in breast cancer (Additional file 1: Figure S1B). Previous study has shown that the SALL1 promoter was methylated in breast and other epithelial cancers [19]. We selected two probes that indicate a methylation difference of the SALL1 promoter in COSMIC between the primary tumor in breast cancer and solid normal tissues. We observed that the genes were highly methylated in these 2 regions of SALL1 promoter in the breast cancer tissues, which might be partially responsible for the down-regulation of SALL1 in breast cancer (Additional file 1: Figure S1C). Recent studies have shown that SALL4 is involved in leukemogenesis and a marker for hepatoblastoma and non-small cell lung carcinoma [15‐17]. We therefore determined SALL4 gene expression in breast cancer cell lines and primary cancer tissues. In contrast to SALL1 gene expression, we found significantly higher expression of SALL4 in both breast cancer cell lines and primary cancer tissues, suggesting that SALL1 and SALL4 may have distinct functional roles in the regulation of breast cancer (Additional file 1: Figure S1D and E). Collectively, our results suggest that SALL1 expression is significantly down-regulated in human breast cancer cells and in certain types of breast cancer tissues (especially in triple negative breast cancer), which may play a critical role in the pathogenesis of human breast cancer.

To test whether SALL1 is a tumor suppressor [20], we determined the effect on cancer cell growth and function by SALL1. SALL1 was transfected into human and murine breast cancer cell lines MDA-MB-231 (TNBC), MCF7 and E0771 (basal-like) (all with no or minor expression of SALL1) for the gain-of-function studies. Prostate cancer cell line PC-3 (low SALL1 expression) and melanoma cell line B16F0 (high SALL1 expression), as well as normal breast cell line MCF12A cells were included as controls. Tumor cell growth and proliferation were determined by analyzing cell growth curves and [³H]-thymidine incorporation assays. We found that transfection of SALL1, but not SALL4 or vector in MCF-7, MDA-MB-231, E0771 and PC-3 tumor cells significantly inhibited cell growth and proliferation. However, over-expression of SALL1 in B16F0 melanoma cells, and normal breast cell MCF12A did not affect cell growth and proliferation (Fig. 2a and b, and Additional file 1: Figure S2A and B). These results further suggest that SALL1 directly inhibits breast cancer cell growth, but it may have different functions in the other types of cancers.

Suppression of tumor cell proliferation and growth mediated by SALL1 expression could be due to the induction of apoptosis or cytolysis in the tumor cells. We therefore measured apoptosis and cell death in SALL1-transfeced breast tumor cells. We found that breast tumor cells MCF7, MDA-MB-231 and E0771 in medium alone or transfected with control vector contained some apoptotic cells (around 10% in MCF-7, 4% in MDA, and 20% in E0771). However, overexpression of SALL1 in cancer cells did not induce increased apoptosis or cell death in breast cancer cell lines (Fig. 2c and Additional file 1: Figure S2C). In parallel, we studied the cell cycle distribution of the breast cancer cells transfected with SALL1. SALL1 transfection in MCF-7, MDA and E0771 cells significantly induced cancer cells to arrest in S phase and decrease in G0/G1 phase (Fig. 2d). Notably, transfection of SALL1 in melanoma B16F0 cells induced neither cell apoptosis nor cell cycle arrest. To further identify the potential mechanism responsible for the SALL1-mediated breast cancer cell arrest in S phase, we determined the cell cycle regulation gene expressions in MDA-MB-231 cells using Real-time PCR analysis, including Cyclin A2, B1, D1 and E1, as well as CDK2, 4 and 6. We observed that transfection with SALL1 significantly increased the gene expressions of Cyclin A2, Cyclin B1, Cyclin E1, CDK2 and CDK4 in breast cancer MDA cells, which are important for checkpoint regulation in G1-S transition and S phases (Additional file 1: Figure S3). These data suggest that over-expression of SALL1 in breast cancer strongly suppresses tumor growth and proliferation, as well as induces cell cycle arrest, which is mechanistically independent of apoptosis or cytolysis in tumor cells.

To further confirm the functional role of SALL1 in regulating breast cancer cell growth, we also utilized a loss-of-function strategy to knockdown SALL1 gene expression with lentivirus-based shRNA in breast cancer cell lines and then determined its effect on tumor growth and proliferation. As expected, silencing of SALL1 expression in MCF-7, MDA-MB-231 and E0771 breast cancer cells dramatically promoted tumor growth and increased cell proliferation compared with lentivirus carrying control shRNA infected breast cancer cells (Fig. 3a and b). Furthermore, the numbers and sizes of tumor cell colonies were also significantly increased in three breast tumor cell lines after knockdown of SALL1 in a colony formation assay (Fig. 3c and d). In contrast, over-expression of SALL1 in breast cancer cells dramatically inhibited tumor cell grown and proliferation, as well as decreased colony numbers, which were consistent with the results shown in Fig. 2. In addition, we did not observe any effects of SALL1 overexpression or down-regulation on cell growth and proliferation in normal breast MCF12A cells. These results further suggest that SALL1 expression in breast cancer cells directly controlled cell growth and function.

Senescent human cells have permanent growth arrest, which could occur due to telomere shortening or a DNA damage response [29, 30]. We therefore reasoned that induction of senescence might be the mechanism involved in the suppressed cell growth and proliferation mediated by SALL1 overexpression in breast cancer cells. In addition to cell cycle arrest and morphologic characteristics, SA-β-Gal is the first biomarker used to identify senescent human cells [31, 32]. We observed that transfection of SALL1 in MCF-7, MDA, and E0771 tumor cells significantly increased the number of SA-β-Gal⁺ cells, indicating the induction of tumor cell senescence (Fig. 4a and b, and Additional file 1: Figure S4A). In contrast, transfection with SALL4 or control vector in these tumor cells did not induce increased SA-β-Gal expression. In addition, we did not observe increased senescent cell populations in melanoma B16F0 cells after over-expression of SALL1.

The induction of DNA damage is the key molecular process in senescent cells, which could be induced by telomere erosion and/or other forms of stress. The nuclear kinase ataxia-telangiectasia mutated protein (ATM) is the chief inducer of the DNA-damage response. We thus determined whether induction of ATM-associated DNA damage is the main trigger for SALL1-induced senescence in breast tumor cells [33, 34]. Over-expression of SALL1 significantly induced active, phosphorylated ATM in MCF-7, MDA and E0771 cancer cells (Fig. 4c and Additional file 1: Figure S4B). In addition, we further investigated the other key DNA damage response proteins involved in the induction of senescence due to the DNA damage response. These proteins include ATM substrates H2AX and 53BP1, as well as the downstream target checkpoint kinase 2 (CHK2) [30, 34]. We observed that transfection of SALL1, but not SALL4 or control vector also significantly induced phosphorylation of H2AX, 53BP1 and CHK2 in MCF-7, MDA and E0771 cells (Data not shown). To confirm the involvement of ATM-associated DNA damage response in SALL1-mediated breast cancer cell senescence, we next determined whether we can prevent the SALL1-mediated senescence in breast cancer cells through the functional blockade of ATM-induced DNA damage using loss-of-function approaches with the specific pharmacological ATM inhibitor KU55933 and shRNA against ATM. As shown in Fig. 4d, treatment of MCF-7, MDA and E0771 breast cancer cells with KU55933 dramatically suppressed the phosphorylation of ATM in SALL1-transfected tumor cells and prevented induction of senescence in tumor cells. In addition, knockdown of ATM expression with shRNA significantly decreased the senescent cell populations in SALL1-transfected breast tumor cells, further confirming the involvement of the ATM-associated DNA damage response in SALL1-induced tumor cell senescence (Fig. 4e). These data provide the first evidence that suppression of breast cancer growth and proliferation mediated by SALL1 expression is due to the induction of tumor cell senescence.

Our previous studies have shown that endogenous SALL1 binds to the NuRD complex to regulate gene transcription and specific developmental processes [4, 6, 21‐23]. We further identified a highly conserved 12-amino acid motif in the SALL1 protein that is sufficient for the recruitment of NuRD [22]. Importantly, increasing evidence suggests that the NuRD complex plays an essential role in regulating oncogenesis and metastasis programs of breast cancer [24‐26]. We therefore hypothesized that SALL1-mediated suppression of breast cancer growth and proliferation, and induction of tumor cell senescence may also act through the recruitment of the NuRD complex. To test this hypothesis, we first transfected a mutated SALL1 (mSALL1) encoding a protein in which the conserved 12-amino acid peptide motif that specifically binds to NuRD was deleted, into breast cancer cells and determined the capacity for senescence induction (Additional file 1: Figure S5). Consistent with the above results, transfection of full-length SALL1 into MCF-7 and E0771 breast cancer cells significantly induced tumor cell senescence (around 40%) and promoted cell cycle arrest in S phase in 3 days (Fig. 5a and b). However, transfection of the mutated SALL1 in the breast cancer cells did not have any effects on cell senescence and cell cycle, similar to that of control vector, suggesting that breast cancer cell senescence mediated by SALL1 depends on NuRD recruitment.

In our efforts to identify the relationship and direct interactions between SALL1 and NuRD, we have demonstrated that protein kinase C phosphorylates serine 2 of the SALL1 repression motif and regulates the association with NuRD [21]. Furthermore, we showed that substitution of the serine with a glutamic acid (SALL1-S2E, phosphomimetic) significantly abolished the effect on NuRD recruitment and repression activity; whereas mutating the serine to an alanine (SALL1-S2A) modestly increased the transcriptional repression [21, 22] (Additional file 1: Figure S5). We next utilized SALL1 constructs with these two separation-of-function mutations to test its effects on senescence induction in breast cancer cells. As we expected, transfection of SALL1-S2E into MCF-7 and E0771 breast cancer cells lost the ability to induce tumor cell senescence. In contrast, transfection of SALL1-S2A into tumor cells significantly augmented senescence induction in both cell lines compared with that of wild type SALL1-transfected tumor cells (Fig. 5c and d). To confirm the physical interaction of SALL1 with the NuRD complex in cancer cells, we transfected MCF-7 breast cancer cells with GST fusions of wild type SALL1, or SALL1-S2A and SALL1-S2E. GST-SALL1 fusion proteins were isolated on glutathione-Sepharose beads. Western-blot assays were then performed to determine the expression of components of the NuRD complex, including MTA2, RbAp46/48, HDAC1 and MBD3, after SALL1 and GST pulldown [21, 22]. Transfection of the three fusion constructs equivalently expressed SALL1 protein (Fig. 5e). Expression of wild type SALL1 and SALL1-S2A in MCF-7 tumor cells recruited the endogenous NuRD complex components. However, GST-SALL-S2E did not pulldown NuRD components even though it is expressed at a level comparable to wild type SALL1. Collectively, these results clearly indicate that SALL1 recruits NuRD in breast cancer cells resulting in suppression of tumor growth and proliferation, and induction of tumor cell senescence.

SALL proteins heterodimerize and in some contexts can cross-regulate their respective expression. We thus determined whether SALL1 could alter the expression of other members in the SALL family mediating breast cancer growth suppression [10]. Our results showed that transfection of SALL1 in both MCF-7 and E0771 cells did not change the gene expression levels of SALL2, SALL3 and SALL4 at different time points using Real-time PCR analyses (Additional file 1: Figure S6).

MAPK signaling pathways play a major role in regulating cell cycle re-entry and oncogenic ras-induced senescence [35]. It has been reported that ERK1/2 and p38 activation can induce p21-dependent G1 cell cycle arrest [36]. Our recent studies further demonstrated that MAPK ERK1/2 and p38 signaling controls the molecular process of human CD4⁺CD25^hiFoxP3⁺ Treg-induced responder T cell senescence [32, 34]. We therefore explored whether SALL1-induced breast cancer cell cycle S phase arrest and conversion of cancer cells into senescent cells involved MAPK signaling modulation. We first determined the activation and phosphorylation of MAPKs, including ERK1/2, p38 and JNK in breast cancer cells transfected with SALL1 using western blot analyses. We found that transfection of SALL1 but not mutated SALL1 selectively activated ERK1/2 and p38, but not JNK, resulting in significantly enhanced phosphorylation of ERK1/2 and p38 in both MCF-7 and E0771 breast cancer cells (Fig. 6a). To further determine the role of ERK1/2 and p38 signaling in controlling the molecular process of SALL1-induced senescence in breast cancer cells, we utilized loss-of-function strategies with specific pharmacological inhibitors and lentivirus-based shRNAs to block ERK1/2 and p38 activities in breast cancer cells, as we previously described [32]. The optimal concentrations (10 μM) of different inhibitors used for our experiments, including SB203580 (p38 inhibitor) and U0126 (ERK1/2 inhibitor), were selected based on their toxic effects on tumor cell viability and proliferation. As shown in Fig. 6b, we observed that inhibitors U0126 and SB203580 significantly reduced SALL1-induced senescent cell populations in both MCF-7 and E0771 breast tumor cells. We then used shRNAs to specifically knock down p38 and ERK1/2 genes in MCF-7 and E0771 cells, and measured the effects on SALL1-induced tumor cell senescence. Consistent with the results obtained in inhibitor experiments described above, knockdown of p38 and ERK1/2 in MCF-7 and E0771 cells significantly decreased the senescence induction in SALL1-transfected tumor cells (Fig. 6c). These results suggest that SALL1 expression in breast cancer cells induces selective modulation of specific MAPK p38 and ERK1/2 signaling pathways in tumor cells that control the molecular process of SALL1-induced tumor cell senescence and growth suppression. Moreover, this process depends on an intact NuRD-recruitment motif in SALL1.

In addition to MAPK signaling, mTOR kinase signaling activation is important for tumor cell proliferation and senescence induction [37‐41]. We next investigated whether mTOR signaling is also involved in the SALL1-induced breast cancer growth inhibition and senescence induction. We determined the activation of mTOR and its downstream substrates p70S6K and 4E-BP1, in breast tumor cells after transfection with SALL1 [42]. Transfection of SALL1 but not mutated SALL1 in both MCF-7 and E0771 breast tumor cells significantly induced the phosphorylation of mTOR, p70S6K, and 4E-BP1, further confirming the activation of mTOR signaling in tumor cells after SALL1 expression (Fig. 6d). Using loss-of-function strategies, we demonstrated that the mTOR inhibitor rapamycin and shRNA to specifically knock down mTOR gene expression in breast cancer cells dramatically prevented induction of senescence in tumor cells mediated by SALL1 expression (Fig. 6e and f). These results suggest that the mTOR signaling pathway is also critical in regulating breast cancer cell senescence mediated by SALL1 expression. Furthermore, our studies indicate that SALL1 expression in breast cancer selectively utilizes both MAPK and mTOR signaling pathways controlling tumor cell fate and functions.

These in vitro studies provided us with important information regarding the mechanism and molecular signaling of SALL1 in suppressing breast cancer tumor cell growth and metastasis. We next performed complementary in vivo studies, using murine breast cancer E0771 cells in humanized NOD-scid IL2Rγ^null (NSG) mouse models, and explored whether SALL1 functions as a tumor suppressor for the tumorigenesis and metastasis of breast cancer in vivo. The E0771 cell line is a basal-like and derived from medullary breast adenocarcinoma cells, representing a good model for spontaneously developed breast cancer [43, 44]. We also included murine B16F0 melanoma cells as a tumor control for this study. We first performed xenograft models to investigate whether over-expression of SALL1 in breast cancer cells can affect tumor growth and development. Mouse E0771 and B16F0 tumor cells infected with Lentivirus carrying SALL1, mSALL1 or vector, were subcutaneously injected into NSG mice. Tumor growth was evaluated. At the end of the experiments, tumors were isolated from the sacrificed mice and weighed. As shown in Fig. 7a, we observed that E0771 tumor cells alone or transfected with mutated SALL1 gene or vector control, grew progressively in NSG mice. However, over-expression of wild type SALL1 in E0771 cells dramatically inhibited breast tumor progression and growth. Furthermore, tumor sizes collected from the E0771-SALL1 group on day 21 post inoculation were significantly smaller than those in the control groups of E0771 tumor alone, E0771 cells transfected with mSALL1 or vector (Fig. 7b). In addition, the average tumor weights obtained from the E0771-SALL1 group were much lower than those of the three control groups (Fig. 7c). Notably, tumor growth, tumor sizes and weights were very similar among the three control groups of E0771 tumor alone, E0771 cells transfected with mSALL1 or vector. In contrast to the E0771 breast tumor model, there were no differences of tumor growth, progression and sizes among the experimental groups of B16F0 cells transfected with SALL1, mSALL1 or vector, in the B16F0 melanoma model (Fig. 7a-c). These results were consistent with our in vitro studies showing that SALL1 had different effects and functions in breast and melanoma tumor cells. Furthermore, we confirmed, using histochemical staining of SA-β-Gal expression on sections from embedded tumor tissues, that a high amount of senescent tumor cells were observed in tumor tissues obtained from the E0771-SALL1 group, but not from the control groups of E0771 tumor alone, E0771 cells transfected with mSALL1 or vector (Fig. 7d). These results clearly suggest that SALL1 expression in breast tumor cells directly controls tumor growth and tumorigenesis.

We next investigated whether over-expression of the SALL1 inhibited breast tumor metastasis. In vitro wound closure assays were utilized to determine the capacity of tumor cell migration after transfection with or without SALL1. We observed that over-expression of SALL1 in E0771 and MCF7 breast tumor cells markedly inhibited the migration of tumor cells compared with the tumor cells alone, or tumor cells transfected with mSALL1 or vector (Fig. 8a and Additional file 1: Figure S7A). We then investigated whether over-expression of SALL1 affected tumor metastasis using previously established adoptive transfer tumor models with a live imaging system [45]. Mouse E0771 tumor cells infected with or without lentivirus carrying SALL1 or control mutated SALL1 gene were labeled with VivoTag®680 XL and then injected intravenously into the tail vein of NSG mice. Tumor cell distribution and metastasis in mice were imaged at dorsal, right lateral and ventral positions with an In Vivo Spectrum Imaging System (IVIS) at different time points post injection. As shown in Fig. 8b and Additional file 1: Figure S7B, at early time points (before 3 days) post tumor cell adoptive transfer, tumor cells randomly migrated into different organs, including spleen, liver and lung with similar signal densities among different groups. The signal density of tumor cells significantly increased and accumulated in these sites in the late time points (after 3 days) post tumor cell transfer among the groups of E0771 tumor cells only, or transfected with mSALL1 or vector. This tumor cell accumulation with strong signal density continued to persist throughout the whole observation period (19 days), indicating accumulated colonization of tumor cells into the lung and liver. However, the signal density from the tumor cells transfected with SALL1 markedly decreased in the late time points (after 3 days) post tumor cell injection, suggesting the lower capacity of tumor metastasis and/or colonization compared with that of the other groups. To further confirm the live imaging results, livers and lungs from the four groups were harvested at day 19 post tumor injection and macro-metastatic tumors were evaluated. As expected, we observed that transferred E0771 tumor cells alone or transfected with mutated SALL1 gene or vector control, grew significantly with metastasis/colonization in lungs and livers (Fig. 8c and d). In contrast, transfection of SALL1 in E0771 cells dramatically decreased tumor macro-metastatic numbers both in lung and in liver surfaces. Furthermore, we confirmed, using histological staining on sections from embedded liver and lung tissues that a high number of tumor cells infiltrated into livers and lungs obtained from control groups of mSALL1 and vector-transfected E0771, but not from the SALL1-transfected tumor group (Fig. 8e and Additional file 1: Figure S7C). Our studies clearly demonstrate that SALL1 is a tumor suppressor in breast cancer and plays a critical role in directing tumorigenesis and metastasis.

Tumor samples were obtained from breast cancer patients treated at the Department of Surgery, Saint Louis University from 2004 to 2015 who have given informed consents for enrollment in a prospective tumor procurement protocol approved by the Saint Louis University Institutional Review Board. Paired fresh tumor tissues and normal breast tissues were obtained perioperatively and snap frozen in liquid nitrogen. In addition, fresh-frozen metastatic cutaneous melanoma tumor tissues were also collected as controls for this study. Breast tumor cell lines (human MDA-MB-231, MCF7, BC80, 31, 30, 29, 16, 12, 10, and murine 4 T1 and E0771), Melanoma cell lines (Mel1938, Mel1586, Mel1860, Mel1363, Mel1526 and Mel1628, and murine B16F0), prostate cell line PC3 and DU145, colon cancer cell line SW480 and lymphoma L428 and L504, as well as normal breast cells and fibroblast cells, were either obtained from the American Tissue Culture Collection (ATCC) or established by our group, and maintained in RPMI 1640 medium containing 10% fetal calf serum (FCS) and penicillin-streptomycin (Invitrogen, Inc. San Diego, CA).

Full length flu-tagged SALL1 wild type and mutant constructs cloned into pcDNA3.1 were prepared as previously described [21]. Point mutants were created by PCR-mediated site directed mutagenesis using QuikChange (Stratagene). The amplified PCR products were cloned into lentivirus vector pCDH-CMV-EF1-GFP. The nucleotide sequences of all constructs were verified by DNA sequencing.

The cell populations of SALL1⁺ cells in cancer and normal tissues (frozen sections) were determined using immunohistochemical staining with the Histostain®-Plus 3rd Gen IHC Detection Kit (Invitrogen, CA), as we described previously [31]. Immunohistochemical reactions were performed using either mouse monoclonal or rabbit polyclonal antibodies against SALL1 at dilution of 1:1500. Controls were performed by incubating slides with the isotype control antibody instead of primary antibodies, or second antibody alone. SALL1⁺ cells in tissues were evaluated manually using a computerized image system composed of a Leica ICC50 camera system equipped on a Leica DM750 microscope (North Central Instruments, Minneapolis, MN). Photographs were obtained from 20 randomly selected areas within the tumor tissues of 10 cancer nest areas and 10 cancer stroma areas at a high-power magnification (400 ×). Both cancer nest and stroma areas were counted and summed, and the means of positive cell numbers per field reported.

Total RNA was extracted from tumor or normal tissues and cell lines using Trizol reagent (Invitrogen), and cDNA was transcribed using a SuperScript II RT kit (Invitrogen), both according to manufacturers’ instructions. mRNA expressions of each gene were determined by reverse-transcription PCR using specific primers, and mRNA levels in each samples were normalized to the relative quantity of Glyceraldehyde-3-phosphate dehydrogenase (GAPDH). All samples were run in triplicate. The primers for each gene used were as following:

Tumor cell lines were plated at 2 × 10⁴/well in 24 wells and transfected with one of the following plasmids: pcDNA3.1-SALL1, pcDNA3.1-SALL4, and pcDNA3.1. Cell growth was evaluated at different time points by counting cell numbers. Proliferation assays were performed as previously described [60]. In brief, different numbers of tumor cells (2 × 10⁴, 5 × 10⁴, or 1 × 10⁵) transfected with or without the related genes were cultured in 96-well plates in cell assay medium containing 2% FCS. After 56 h of culture, [³H]-thymidine was added at a final concentration of 1 μCi/well, followed by an additional 16 h of culture. The incorporation of [³H]-thymidine was measured with a liquid scintillation counter.

Transfected cells were cultured for 72 h and apoptosis was analyzed after staining with PE-labeled Annexin V and 7-AAD (BD Biosciences, San Diego, CA). For cell cycle analysis, transfected cells were fixed with 70% ethanol overnight, washed with PBS and incubated with propidium iodide (10 μg/ml) and RNase A (100 μg/ml). Untransfected cells served as controls. All the stained cells were analyzed on a FACSCalibur (BD Bioscience) and the data were analyzed with FlowJo software (Tree Star, Ashland, OR).

Five thousand per well of tumor cells infected with lentivirus carrying shRNA against SALL1 or control scramble shRNA, or transfected with SALL1, were seeded in 6-well plates with 0.4% agar for culture. Cell colonies were stained with crystal violet and counted after 2–3 weeks of culture.

Senescence associated β-Galactosidase (SA-β-Gal) activity in tumor cells was detected as we previously described [31, 32]. Briefly, tumor cell lines were transfected with or without plasmids and cultured for 3 or 5 days. Cells were washed in PBS (pH 7.2), fixed in 3% formaldehyde, and followed to incubate overnight at 37 °C with freshly prepared SA-β-Gal staining solution (1 mg/ml X-gal, 5 mM K₃Fe[CN]₆, 5 mM K₄Fe[CN]₆, 2 mM MgCl₂ in PBS at pH 6.0). The stained cells were washed with H₂O and examined with a microscope. For some experiments, SA-β-Gal⁺ populations were determined in the transfected tumor cells after exposure to various inhibitors or combined transfection with shRNAs: ATM inhibitor KU55933 (20 μM, Tocris Bioscience); mTOR inhibitor Rapmycin (5 μM, Sigma); MAPK inhibitors U0126 (10 μM), SB203580 (10 μM) and SP600125 (10 μM), or PI3 Kinase inhibitor Wortminnin (10 μM) (Calbiochemistry), or transfection with shRNAs against p38, ERK and mTOR, for 3 or 5 days. The treated tumor cells were then detected for SA-β-Gal expression.

Breast cancer cells transfected with or without plasmids pcDNA3.1-SALL1 or pcDNA3.1-mSALL1, were cultured for 0, 24 h, 48 h and 72 h. Whole cell lysates were prepared for western blotting. The antibodies used in western blotting are as follows: anti-SALL1, anti-ERK, anti-phospho-ERK, anti-p38, anti-phospho-p38, anti-JNK, anti-phospho-JNK, anti-phospho-p53 (ser15), anti-mTOR, anti-phospho-mTOR; anti-P70S6K, anti-phospho-P70S6K; anti-4E-BP1, anti-phospho-4E-BP1, anti-PTEN, anti-phospho-PTEN and anti-GAPDH rabbit polyclonal antibodies (Cell Signaling Technology, Danvers, MA).

Protein interaction analysis of NuRD complex members with SALL1 (2–137) was performed as previously described [21, 22]. In brief, MCF-7 breast tumor cells were transfected with plasmids pEBG-SALL1, pEBG-SALL1-S2A, and pEBG-SALL1-S2E, and allowed to expression GST-SALL1 fusion proteins for 48–72 h. Cells were incubated for 1 h on ice in lysis buffer (1% Triton X-100, 200 mM sucrose, 50 mM Tris pH 7.4, plus protease cokatail) and the cell suspension was disrupted by sonication. GST-SALL1 fusions and associated protein complexes were isolated by precipitation of 50 μg of total protein with glutathione-Sepahrose beads (Amersham Biosciences) for 2 h at 4 °C. Protein pulldowns were separated by SDS-PAGE and proteins detected by western blot. Primary antibodies were all used at 1:1000 dilution and included: rabbit anti-SALL1, rabbit anti-Mta2 (Abcam, ab 8106), mouse anti-RbAp48 (GeneTex, GTX 70237), rabbit anti-Hdac1 (Abcam, ab 19,845), rabbit anti-Mbd3 (Abcam, ab 157,464). Secondary antibodies were used at 1:10,000 included: goat anti-rabbit (Sigma, A0545) and rabbit anti-mouse (Jackson immune Research 315–035-048).

The expression of DNA damage response markers on tumor cells were determined by FACS analysis after staining with anti-human specific antibodies conjugated with either PE or FITC. These human antibodies included: anti-phosphorylated H2Ax, anti-phosphorylated p53bp, and anti-phosphorylated ATM, which were purchased from Cell Signaling Technology or BD Biosciences. All stained cells were analyzed on a FACSCalibur flow cytometer (BD Bioscience) and data analyzed with FlowJo software (Tree Star).

Breast cancer E0771 and MCF-7 tumor cells transfected with Lenti-SALL1, Lenti-mSALL1 or vector, were plated in 6-well plates and grown to confluence. A wound area was generated by scraping cells with a 200 μl pipette tip across the entire diameter of the dish and extensively rinsed with the medium to remove all cellular debris. Low-serum RPMI 1640 with mitomycin (2 μg/ml) was then added to inhibit cell proliferation during the experiment and the closing of the wound was observed at different time points.

The methods for design and construction of shRNA specific for ERK1, ERK2, P38α, JNK1 and mTOR or scrambled lenti-shRNAs, and generation of recombinant lentivirus carrying GFP and shRNA, have been described previously [60]. shRNAs specific for ATM (TRCN0000194861, TRCN0000039951 and TRCN0000360327), mouse SALL1 (TRCN0000238153, TRCN0000238154 and TRCN0000238155) and human SALL1 (TRCN0000003956, TRCN0000003957 and TRCN0000003958), were purchased from Sigma Aldrich. For lentivirus infection, concentrated lentiviral supernatant with a multiplicity of infection (MOI) of 5–10 in a total volume of 0.5 ml culture medium was added to the tumor cells growing in 24 well plates containing 8 μg/ml polybrene (Sigma), and then centrifuged at 1000 x g for 1 h at room temperature. The infected tumor cells were then transfected with or without pcDNA3.1-SALL-1, and induction of senescence was determined.

NOD-scid IL2Rγ^null (NSG, 6–8 weeks) immunodeficient mice were purchased from The Jackson Laboratory and maintained in the institutional animal facility. All animal studies have been approved by the Institutional Animal Care Committee. For tumorigenesis studies, mouse E0771 (2 × 10⁵/mouse) and B16F0 (1 × 10⁵/mouse) tumor cells infected with lentivirus carrying SALL1, mSALL-1 or vector, were subcutaneously injected into NSG mice. Five mice were included in each group. Tumor size was measured with calipers every 2–3 days. Tumor volume was calculated on the basis of two-dimensional measurements. At the end of experiments, the mice were sacrificed and tumors were isolated and weighted. Furthermore, tumor tissues were embedded into OCT and prepared for cryostat sections (4~ 8 μm), and SA-β-Gal expression was assayed, as described above.

For tumor metastasis studies, lentivirus-transfected E0771 tumor cells were incubated with 100 μg/ml of VivoTag®680 XL (PerkinElmer) for 30 min. Stained tumor cells were washed and then injected intravenously into the tail vein (5 × 10⁴/mouse in 200 μl of buffered saline) into NSG mice. Five mice were included in each group. Mice were imaged with an In Vivo Spectrum Imaging System (IVIS) (Caliper Life Science) at 120 min, and 1, 3, 5, 7, 10, 14, 17 and 19 days post injection. The appropriate filter set for VivoTag®680 XL imaging of 665 nm excitation and 688 nm emission was used. Mice were imaged in the dorsal, right lateral and ventral positions at all the time points. Livers and lungs were harvested at 19 days post injection and stained with 15% black India ink. Visible lung and liver surface macro-metastatic appeared as white spots and were counted using a dissecting microscope. Lungs were collected and fixed in 10% formalin. For tissue morphology and metastasis evaluation, liver and lung tissues were embedded into OCT and frozen sections (4~ 8 μm) were prepared and stained with hematoxylin and eosin (H & E).

Statistical analysis was performed with GraphPad Prism5 software. Unless indicated otherwise, data are expressed as mean ± standard deviation (SD). For public database analysis, The Cancer Genome Atlas (TCGA) normalized log₂ transformed breast cancer Argilent microarray expression data sets and methylation database were downloaded from the cBioPortal (http://​www.​cbioportal.​org/) [27] and used to compare SALL1 mRNA expression among the different breast cancer subtypes and solid normal tissues, as well as analyze methylation difference of SALL1 promoter in COSMIC between the primary tumor in breast cancer and solid normal tissues. The Mann-Whitney U test was utilized for statistical analysis of association between SALL1 expression and breast cancer subtypes. P valve < 0.05 was considered statistically significant. For multiple group comparison in vivo studies, the one-way analysis of variance (ANOVA) was used, followed by the Dunnett’s test for comparing experimental groups against a single control. For single comparison between two groups, paired Student’s t test was used. Nonparametric t-test was chosen if the sample size was too small and did not fit a Gaussian distribution.