Human embryonic stem cell microenvironment suppresses the tumorigenic phenotype of aggressive cancer cells

To better characterize the role of Nodal in the maintenance of pluripotency and tumorigenicity, we examined the expression of key components of the Nodal signaling pathway in human normal, neoplastic and stem cell types. Western blot analyses revealed that in a manner similar to hESCs (MEL-2, H1, and H9), metastatic melanoma (C8161) and breast carcinoma (MDA-MB-231) cells express Nodal protein at ≈48 kDa (Fig. 1A). This is in contrast to corresponding normal cell types [melanocytes, myoepithelial cells (Hs 578 Bst), and primary human mammary epithelial cells (HMEpC)], in which Nodal was not detected. Using RT-PCR, we determined that Nodal is expressed in additional melanoma (WM278), and breast carcinoma (T47D, MDA-MB-468, MDA-MB-330, and ZR75–30) cell lines [supporting information (SI) Fig. 5A]. Nodal propagates its signal by binding to a receptor complex comprised of Cripto and a heterodimer of type I (ALK 4/7) and type II (ActRIIB) activin-like kinase receptors. Genetic studies in zebrafish and mice have determined that Cripto directly associates with ALK 4 and Nodal and that these associations facilitate the ability of Nodal to propagate its signal (15, 16). Using Western blot analysis and immunofluorescence microscopy, we determined that hESCs uniformly express high levels of Cripto at ≈35 kDa; however, only a low level of Cripto was heterogeneously expressed in C8161 and MDA-MB-231 cells (Fig. 1 A and B). Moreover, only one additional melanoma line (WM278) and two additional breast carcinoma lines (T47D and ZR75–30) examined in this study expressed detectable levels of Cripto transcript (SI Fig. 5A).

The Lefty proteins (Lefty-A and Lefty-B), divergent members of the TGF-β superfamily, spatially and temporally antagonize Nodal in embryological systems (17). Moreover, the Lefty genes are downstream targets of Nodal signaling, thereby providing a powerful negative-feedback loop for this pathway (17). By using Western blot analysis we determined that, in accordance with previously published accounts of Lefty protein processing patterns, hESCs express Lefty protein at ≈42, 34 and 28 kDa (17, 18). In contrast, Lefty is not expressed by metastatic breast carcinoma and melanoma cells or by corresponding normal somatic cell types (Fig. 1A and SI Fig. 5A). To visualize the localization of Lefty and Nodal in hESC colonies and their microenvironment, we performed immunofluorescence localization with confocal microscopy. Using this method, we determined that Lefty protein localizes to the areas where hESCs contact the underlying Matrigel matrix, and that hESC-derived Lefty permeates into this microenvironment (Fig. 1C). This is in contrast to Nodal protein, which localizes to the surface of hESC colonies, and is secreted into the media. These results were confirmed with Western blot analyses, which demonstrated that Lefty protein can be detected in Matrigel conditioned by H9 hESCs (H9 CMTX; Fig. 1C). Furthermore, neither Nodal nor Lefty is found in unconditioned control Matrigel alone (data not shown).

We next sought to elucidate the expression of Nodal, Lefty, and Cripto in other human stem cell types and in first trimester human cytotrophoblast cells (HTR-8/SVneo). Western blot analyses revealed that umbilical-cord-derived mesenchymal stem cells (MSC; SC00125) and adult MSCs do not express Nodal and Cripto, and that although amniotic-fluid-derived stem cells (GM00473, GM00957A) and cytotrophoblast cells express Cripto, only the latter expresses appreciable levels of Nodal (SI Fig. 5B). Of note, in contrast to hESCs, none of the other stem cell lines examined expressed an appreciable level of Lefty protein (SI Fig. 5B). To confirm the apparent exclusivity of Lefty expression to hESCs, we performed real-time RT-PCR. Using this method, we determined that relative to hESCs, metastatic cancer cells, normal somatic cells and trophoblast cells do not express appreciable levels of Lefty mRNA (SI Fig. 5C). Moreover, the other stem cell types used in this study expressed a comparably insignificant level of Lefty (SI Fig. 5C). We previously demonstrated that Nodal expression is positively correlated with melanoma progression, such that Nodal protein is not expressed in normal melanocytes or radial growth phase melanomas, but is present in vertical growth phase and metastatic lesions (4). The current study characterized Nodal localization (previously unexplored) in breast tissues: Using immunohistochemical analysis of a human breast tissue microarray (TMA) we observed that Nodal protein is absent in normal breast tissue, and that its expression is positively correlated with breast cancer progression (P < 0.05; SI Fig. 5D and SI Table 1). Given the significant observation that like hESCs, cancer cells express Nodal, although unlike hESCs, they do not express Lefty, we hypothesized that hESC-derived Lefty and possibly other tumor-suppressive factors found in hESC-conditioned matrices (CMTX), may inhibit Nodal signaling in cancer cells (Fig. 1D). We further proposed that by neutralizing Nodal in cancer cells, hESC microenvironments will reprogram these tumor cell types toward a less tumorigenic phenotype.

To examine the effect of hESC-derived Lefty on tumor cell phenotype, we used Dynabeads covalently coupled to anti-Lefty antibody to isolate Lefty from hESCs cultured on a feeder-free Matrigel matrix. The specificity and utility of the anti-Lefty antibody used to recover Lefty protein was confirmed by Western blot analysis and corresponding Coomassie stained SDS/PAGE of known rLefty-A and -B standards whose sequences were subsequently confirmed by liquid chromatography/mass spectrometry (LC/MS; SI Fig. 6A). Purified hESC-derived Lefty was seeded into Matrigel and the effects of the “Lefty-containing” matrix on cancer cell phenotype were determined. Western blot analysis revealed that hESC-derived Lefty significantly diminished Nodal protein expression in C8161 and MDA-MB-231 cells (Fig. 2A).

To further substantiate the specificity of these results, the cancer cells were also exposed to Matrigel conditioned by hESCs in which Lefty protein expression was knocked down with FITC-tagged Morpholinos specific for Lefty-A and Lefty-B (MO^LEFTY). [The fluorescently tagged Morpholinos could be detected microscopically in >75% of the hESC colonies treated (SI Fig. 6B Upper), and Western blot analysis confirmed the efficient knock down of Lefty protein in hESCs for up to 3 days (SI Fig. 6B Lower)]. The expression of Oct-3/4 and Nanog, representative of pluripotency, was not affected during this time, and morphology of the hESC colonies was not altered (SI Fig. 6C). Thus, although MO^LEFTY knocked down Lefty protein in the hESCs, it did not induce stem cell differentiation during this window of observation. Western blot analysis revealed that C8161 and MDA-MB-231 cells up-regulate Nodal protein expression in response to this “H9 Lefty-deficient” hESC matrix (Fig. 2B). C8161 and MDA-MB-231 cells treated with rNodal (100 ng/ml) similarly experienced an up-regulation of Nodal mRNA expression (SI Fig. 6D), and hESCs treated with MO^LEFTY up-regulated Nodal protein expression (Fig. 2B). These findings are in accordance with the positive feed-back loop that characterizes the Nodal signaling pathway. In contrast to rNodal, which has biological activity, previous studies have reported that recombinant Lefty protein (rLefty) is unable to inhibit Nodal signaling (17). We confirmed these results and extended them to discover that rLefty-B can inhibit Nodal protein in C8161 cells, but at a nonphysiological dose (1,000 ng/ml) (SI Fig. 6E). In an effort to understand the disparate results between hESC-derived Lefty and rLefty on Nodal signaling, we analyzed glycoprotein content in rLefty-B, rLefty-A and a lysate from the H9 hESCs plus their CMTX. We determined that in an apparent contrast to the rLefty proteins, H9-derived Lefty is glycosylated (SI Fig. 6F).

As a functional correlate, we determined that exposure of C8161 and MDA-MB-231 cells to H9 Lefty-derived matrix significantly reduces anchorage-independent growth, and that this inhibition of in vitro clonogenicity can be rescued by the inclusion of recombinant Nodal (rNodal; 100 ng/ml) (Fig. 2C). Furthermore, exposure of C8161 and MDA-MB-231 cells to H9 CMTX reduced their ability to undergo anchorage-independent growth but this inhibition of in vitro clonogenicity was only partially rescued by the inclusion of rNodal (100 ng/ml) (Fig. 2D). In addition, hESC supernatant inhibited anchorage-independent growth, even when Lefty was decreased via the aforementioned Dynabead isolation, although not to the same level as purified Lefty. Collectively, these findings suggest that hESC-derived Lefty has important anti-tumorigenic potential, but that it is not the only tumor suppressive factor within this unique embryonic microenvironment.

We next determined the effects of H9 CMTX on Nodal expression in C8161 and MDA-MB-231 cells. Real-time RT-PCR showed that exposure to H9 CMTX down-regulates Nodal mRNA expression in both C8161 and MDA-MB-231 cells, and that this effect is reversible over time (Fig. 3A). Exposure to H9 CMTX similarly decreases Nodal protein expression in these cells (SI Fig. 7A). Exposure of C8161 cells to matrices conditioned by adult bone-marrow-derived stem cells (MSC), amniotic-fluid-derived stem cells (GM00473, GM00957A), or cytotrophoblast cells (HTR-8/SVneo) did not inhibit Nodal mRNA or protein expression (Fig. 3B and SI Fig. 7B), thus illuminating the exclusivity of the epigenetic influence of the hESC microenvironment.

We implemented orthotopic mouse models to examine the effects of the hESC microenvironment on the in vivo tumorigenicity of melanoma and breast carcinoma cells. Exposure of C8161 and MDA-MB-231 cells to H9 CMTX resulted in a significant reduction in tumorigenicity compared with cells exposed to unconditioned Matrigel (Fig. 4 A and B). To establish a mechanism for the reduction in tumorigenicity, we examined the effects of this treatment on in vivo tumor cell proliferation and apoptosis. Using immunohistochemical staining for Ki67 as a measure of proliferation, and terminal deoxynucleotidyl transferase biotin-dUTP nick-end labeling (TUNEL) as a measure of apoptosis, we found that exposure to hESC CMTX decreased proliferation and increased apoptosis in C8161 and MDA-MB-231 cells (Fig. 4C and SI Fig. 8). These in vivo data confirm the tumor suppressive effects of the hESC microenvironment on aggressive melanoma and breast cancer cells and implicate the potential involvement of apoptotic pathways.

Human cutaneous melanoma (C8161, WM-278) and breast carcinoma (MDA-MB-231, MCF-7, T47D, MDA-MB-468, MDA-MB-330 and ZR75–30) lines were maintained as described (29–31). Normal human neonatal epidermal melanocytes (HEMn-LP; Cascade Biologics), myoepithelial cells [Hs 578 Bst; American Type Culture Collection (ATCC)] and primary mammary epithelial cells (HMEpC; Cell Applications) were maintained as per distributor instructions. Live umbilical cord blood stem cells (SC00125; New Jersey Stem Cell Resource at Coriell Institute for Medical Research), amniotic-fluid-derived stem cells (GM00473, GM00957A) and adult bone-marrow-derived MSCs (Stem Cell Technologies) were maintained under the recommended conditions. HTR-8/SVneo is a well characterized human extravillous cytotrophoblast cell line, and was maintained as described in ref. 32. H1 and H9 hESCs (WiCell, Madison WI) and MEL-2 hESCs (Millipore) were maintained as described in ref. 14. For conditioned matrix experiments, hESCs were maintained in stem cell medium preconditioned on irradiated mouse embryonic fibroblasts (CF-1, ATCC) as described in ref. 14. Recombinant Nodal and Lefty (R&D Systems) were diluted as per manufacturer suggestions.

Conditioned matrices were prepared by using hESCs, melanocytes, myoepithelial cells, amniotic-fluid-derived stem cells, or trophoblast cells on growth factor-reduced Matrigel (14 mg/ml; BD Biosciences) as described in ref. 14. Alternatively, hESC-derived Lefty protein was seeded into Matrigel before polymerization (for complete methods outlining the purification of Lefty, see SI Methods). C8161 or MDA-MB-231 cells were exposed to specific matrices preconditioned by these cells for 3 to 4 days.

Protein lysates were prepared and quantified as described in ref. 4. Equal amounts of protein were separated by SDS/polyacrylamide gel electrophoresis (PAGE) under reducing conditions, and resolved proteins were transferred onto Immobilon-P membranes (Millipore). Membranes were blocked, incubated with primary antibody (SI Table 2), washed, and incubated with the appropriate horseradish peroxidase-labeled secondary antibody. Secondary antibodies were detected by enhanced chemiluminescence (Super Signal; Pierce) with exposure to autoradiography film (Midwest Scientific).

RNA was isolated by using TRIzol (Invitrogen, Carlsbad, CA) and 1 μg of was reverse transcribed as described in ref. 4. Real-time PCR was performed as described in ref. 4 by using TaqMan gene expression human primer/probe sets for Lefty1/B (Hs00764128_s1), Nodal (Hs00250630_s1), Nanog (Hs02387400_g1), and OCT 3/4 (Hs00999632_g1). Gene expression was normalized to 18S rRNA (Hs99999901_s1). Data were analyzed by using the Applied Biosystems Sequence Detection Software (Version 1.2.3). Semiquantitative RT-PCR was performed for Lefty, Nodal, Cripto, and HPRT1 as outlined in SI Methods.

Cells were fixed with 4% paraformaldehyde, made permeable with 20 mM Hepes, 0.5% Triton X-100 and blocked with serum-free protein block (DAKO). Primary antibodies were diluted in antibody dilutent (DAKO) to concentrations outlined in SI Table 2, and appropriate fluorochrome-conjugated secondary antibodies were used according to manufacturer recommendations. Nuclei were stained with DAPI (0.1 mg/ml; Invitrogen/Molecular Probes), and images were obtained by using confocal microscopy (Zeiss 510 META; Carl Zeiss).

Assays were conducted as described in ref. 4.

Protein lysates underwent SDS/PAGE and transfer and were stained for glycoproteins by using the Pro-Q Emerald 300 staining kit (Molecular Probes). After drying the blot, glycoproteins were visualized by using an UV transilluminator and an image of the green fluorescing proteins were captured by using a color CCD camera (Toshiba) equipped with a deep yellow #15 filter. The blot was then rehydrated as per manufacturer's instructions and Lefty protein was detected with immunoblotting. See SI Methods for more in depth details.

Five-week-old mice were injected s.c. with 250,000 C8161 cells in 50 μl of complete RPMI; or 500,000 MDA-MB-231 cells in 50 μl of complete RPMI were injected into the mammary fat pad of 6- to 8-week-old mice. When tumors became palpable measurements were taken twice per week. All experiments involving animals were approved by the Institutional Animal Care and Use Committee at the Children's Memorial Research Center.

Immunohistochemical staining for Nodal in a breast carcinoma progression tissue microarray (CBL-TMA-029; Creative Biolabs) was performed as described in ref. 4. In addition, tissues from the orthotopic tumor models were formalin-fixed and paraffin-embedded and immunohistochemical staining on this tissue was conducted by using a Ki67-specific antibody (SI Table 2) or ChromPure Goat IgG (Jackson Laboratories) as described in ref. 4. TUNEL assays to measure apoptosis were conducted as per manufacturer's instructions (Millipore).

Fluorescein (FITC)-conjugated control (5′-CCTCTTACCTCAGTTACAATTTATA-3′), Lefty-A (5′-GCCACATGGTGCTGCCCTGGG-3′) and Lefty-B (5′CTGCATGGTGCTGCCCTGGAGGA-3′) Morpholinos (20 μM) (Gene Tools) were delivered by using the scrape method (4). Cancer cells were sorted for FITC and were allowed to recover for 1 day before experimentation.

For tumor formation studies, we determined statistical significance by using the Kruskal–Wallis one-way ANOVA on ranks, followed by Dunn's method or a one-way ANOVA followed by the Student–Newman–Keuls method for pairwise multiple comparisons. For the clonogenic assays we determined statistical significance by using ANOVA followed by the Student–Newman–Keuls method for pairwise multiple comparisons. For the correlation of breast cancer stage and Nodal expression, a Spearman rank order correlation was used. In all cases, differences were statistically significant at P < 0.05.