TLX activates MMP-2, promotes self-renewal of tumor spheres in neuroblastoma and correlates with poor patient survival

Nuclear orphan receptor TLX (Drosophila tailless homolog) is essential for the maintenance of neural stem/progenitor cell self-renewal, but its role in neuroblastoma (NB) is not well understood. Here, we show that TLX is essential for the formation of tumor spheres in three different NB cell lines, when grown in neural stem cell media. We demonstrate that the knock down of TLX in IMR-32 cells diminishes its tumor sphere-forming capacity. In tumor spheres, TLX is coexpressed with the neural progenitor markers Nestin, CD133 and Oct-4. In addition, TLX is coexpressed with the migratory neural progenitor markers CD15 and matrix metalloproteinase-2 (MMP-2) in xenografts of primary NB cells from patients. Subsequently, we show the effect of TLX on the proliferative, invasive and migratory properties of IMR-32 cells. We attribute this to the recruitment of TLX to both MMP-2 and Oct-4 gene promoters, which resulted in the respective gene activation. In support of our findings, we found that TLX expression was high in NB patient tissues when compared with normal peripheral nervous system tissues. Further, the Kaplan–Meier estimator indicated a negative correlation between TLX expression and survival in 88 NB patients. Therefore, our results point at TLX being a crucial player in progression of NB, by promoting self-renewal of NB tumor-initiating cells and altering their migratory and invasive properties.

Neuroblastoma (NB) is the most common extracranial solid tumor found in children, accounting for 8-10% of childhood cancers that likely originates from neural crest-derived sympathoadrenal progenitor cells. NB cells can express a number of neural stem cell and progenitor markers, including CD133, ABCG2 (ATP binding cassette-G2) and Nestin. [1][2][3] As self-renewal and differentiation of neural stem cells is predominantly regulated by a number of stem cell fate determinants such as Notch, Wnt, Hedgehog, PTEN (phosphatase and tensin homolog) and TLX (Drosophila tailless homolog), also named NR2E1, [4][5][6][7][8] it is possible that deregulation of such genes may be responsible for the regulation of tumorigenesis in neural cancers. TLX, an orphan nuclear receptor, is predominantly expressed in the embryonic and adult forebrain, and is a crucial regulator of neurogenesis by regulating neural stem cell self-renewal and maintenance. [8][9][10] Recently, we reported that TLX upon hypoxia stimulates neural stem cell renewal by promoting Oct-4 transcription in adult hippocampal progenitors. 11 However, its role in malignancy in the nervous system is not well understood, even though recent studies suggest a role in the initiation of cancer stem cells of glioma. 13,12 NB of high malignancy acquires the ability to degrade components of extracellular matrix to penetrate the basal membrane of blood vessels to metastasize by activating matrix metalloproteinases (MMPs). NB cells might express these proteins as the normal neural stem cells are regulated by the subfamily, MMP-2 and MMP-9, also called gelatinases. 14 In fact, MMP-2 and MMP-9 have been reported to have an important role in invasion and metastasis of glioma and other cancers. [15][16][17] In this study, we demonstrate that the depletion of TLX in NB cell lines inhibits their sphere-forming capacity and reduces their invasion and migration. We show that the altered migration is a direct function of MMP-2 regulation. On the other hand, under hypoxic conditions, TLX can activate oct-4 gene, promoting self-renewal of tumor spheres. We then correlate TLX levels with patient survival data, pointing at TLX being a crucial player in NB progression.

Results
TLX promotes the proliferation and sphere-forming capacity of NB cells. We first examined the protein levels of TLX in different NB cell lines, including SH-SY5Y, SK-N-SH, SK-N-BE2c, LAN-5 and IMR-32 ( Figure 1a). TLX was expressed at higher levels in SK-N-BE2c, IMR-32 and LAN-5 when compared with the other cell lines. For further studies, we used IMR-32 cells where TLX was stably knocked down using shRNAs. As shRNAs 2 and 3 gave 80% reduction in the protein levels (Figure 1b), further experiments were carried out using clones generated from shRNAs 2 and 3. We next validated the growth characteristics and proliferation potential of TLX-silenced clones and compared them with the wild-type (WT) parental IMR-32 cells. Stably silenced clones were prone to detachment after seeding, but surviving cells showed neurite-like processes (Figure 1c). The doubling time of WT and Sh-control IMR-32 cells was~24 h, whereas those of the TLX-silenced clones were 56-72 h, estimated by MTT cell viability assays (Figure 1d). Interestingly, the relative number of viable cells in each passage in the TLX-depleted cells decreased by twofold as compared with the parental cells ( Figure 1d). Our previous study showed the depletion of TLX in adult hippocampal progenitors increased active caspase-3, indicative of a prosurvival role for TLX in neural progenitors. 11 As Akt is a well-known prosurvival signaling molecule and its activation is a marker for poor outcome and prognosis in NB, 18 the levels of phosphorylated Akt were compared in WT cells before and after transient knockdown of TLX using shRNA. pAkt was significantly reduced upon transient TLX knockdown (Si1 and Si2) as compared with WT and control SiRNAtransfected IMR-32 cells (Figure 1e). These results suggest that TLX mediates survival by maintaining pAkt levels, perhaps via its role as a PTEN transcriptional repressor. 19 TLX is enriched in self-renewing spheres derived from NB cells. NB cell lines and NB cells metastasizing to bone marrow have earlier been demonstrated to harbor tumorinitiating cells (TICs), which can then be isolated by growing them in stem cell media. 1,20 Considering that TLX is essential for maintenance and self-renewal of neural stem cells, we investigated if TLX could have a similar role in maintaining the population of NB-TICs. For this purpose, 1 × 10 5 WT or TLXsilenced IMR-32 cell clones were reseeded in serum-free media containing N2 supplement, basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF), and grown for a period of 21 days with a medium change every third day (Figure 2a, top panel). After 7 days, distinct sphere formation was observed in WT and Sh-control cells, but Sh2 and Sh3 clones showed poor sphere formation ability, even after 21 days, suggesting a requirement of TLX for sphere formation (Figures 2a (bottom panel) and b). To evaluate clonogenic potential, spheres from each of the WT and TLX-silenced cells were dissociated and reseeded at a density of 1000 cells per well and analyzed for secondary sphere formation as an indicator of self-renewal potential. We found that while WT or shRNA-control cells formed 50-60 spheres per well, TLXsilenced stable cells formed only 2-5 spheres per well ( Figure 2c). A strong evidence for the role of TLX in sphere was demonstrated when we found a three-to fourfold increase in TLX protein expression in the same number of cells in primary and secondary spheres compared with the monolayer cells in both SK-N-BE2c and IMR-32 cells (Figure 2d). Further, upon IF analysis we found that the spheres coexpressed TLX and CD133 (Figure 2e, left panel). We also sorted these spheres into CD133-positive and -negative fractions using CD133 Microbead Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) and isolated RNA from these cells. We found that TLX transcript was enriched by sixfold in CD133-positive cells, when normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Figure 2f).
TLX enrichment in spheres correlates with proliferation and markers of neural stemness. To identify if TLX is coexpressed with CD133 in tumor spheres from different cell lines, we assayed the spheres from LAN-5 and SKN-BE2c cells for coexpression of TLX and CD133 using FACS. Both the cell types expressed similar levels of TLX and CD133, comparable to IMR-32 ( Figure 3a). We additionally stained for various proliferation and neural stem cell markers in SK-N-BE-2c cells dispersed from spheres ( Figure 3b). We found that these sphere-forming cells showed an overlap of TLX with Ki67, Nestin, Oct-4, CD133 and hypoxia-inducing factor-2α (HIF-2α). Among them, Nestin and CD133 were detected in the cytoplasm and membrane, respectively. Many TLXpositive cells were proliferating, as shown by nuclear Ki67 staining. Indeed, these spheres could be differentiated by the addition of FBS to express MAP2ab and GFAP (Figure 3c). These results indicate that the tumor spheres have neural stem cell-like characteristics.
TLX is expressed in xenograft tissues of primary NB-TICs derived from patients. To corroborate our findings from cell lines, we examined the coexpression of TLX with neural progenitor marker CD15, which additionally marks migratory neuronal progenitors. For this, we used xenografts from primary human patient-derived NB-TICs ( Figure 4). Paraffin-embedded tissue sections from xenografts recovered from non-obese diabetic/severe-combined immunodeficiency (NOD/SCID) mice grafted with NB-TICs were stained with antibodies for TLX and markers for migratory neural progenitors (CD15). In these xenografts, tightly aggregated TLX-positive cells were surrounded by cells expressing low levels of CD15 or CD15-negative areas. Some of these areas were necrotic with many inflammatory cells. Interestingly, most of the TLX-expressing cells also expressed MMP-2 that is secreted in large amounts, likely to facilitate ECM degradation and tumor dissemination, a hallmark of advanced stages of NB. Several xenografts derived from other primary NB cell lines showed a similar pattern of MMP-2 and CD15 staining with the relation to the TLX staining, although intensity of TLX staining varied among the cell lines (not shown).
TLX increases migratory and invasive properties of NB cells. Staining of NB cell lines and NB-TIC xenograft tissues revealed the co-or juxtalocalization of TLX and MMP-2 and CD15, in particular at the edges of TLX-expressing tumor clusters, suggesting them to be migratory cells. As neural stem cells have a migratory capacity, we asked whether TLX could also promote NB cell migration and invasion. Using a colorimetry-based assay for quantifying migration and invasion separately, we observed that TLX-silenced IMR-32 cells had a two-to threefold reduced migration ability as compared with dispersed sphere-forming WT or control transfected IMR-32 cells (Figure 5a). Similar results were obtained in the invasion assays where the TLX-silenced cells showed a twoto threefold decrease as compared with WT or control cells.
We then asked whether the secretion of MMPs known to be involved in migratory and invasive behavior of cancer cells is altered. Using ELISA, we observed a three-to fourfold reduction of secreted MMP-2 in the TLX-silenced cells ( Figure 5b). These results were verified by blotting for MMP-2 and MMP-9 levels secreted in the conditioned media  Figure 1). 21 We observed a 1.8-fold increase in the pro-MMP-2 level upon TLX overexpression, and simultaneous expression of siMMP-2 and TLX rescued the decrease of MMP-2 level by the silencing effect. This is consistent with TLX being an activator of MMP-2 expression. To confirm the MMP-2-mediated promigratory role of TLX, we silenced MMP-2 with siRNA and after 24 h overexpressed TLX in IMR-32 cells. In the absence of MMP-2, TLX overexpression did not result in an significantly increased migratory activity seen with the control cells, indicating the dependence of TLX on MMP-2 for promoting the migration of NB cells (Figure 5e). In summary, TLX alters the migratory capacity of NB cells by inducing MMP-2 levels.
TLX increases binding to the MMP-2 and Oct-4 promoters in NB cells upon hypoxia. We next examined if TLX regulated the expression of MMP-2 by binding to its promoter. First, we analyzed the effect of TLX overexpression on MMP-2 promoter-luciferase constructs, (a) using a 1780 bp full-length promoter upstream of the transcriptional initiation site and (b) a small promoter construct spanning 177 bp upstream sequence. We identified that while the full-length promoter activity increased by 1.8-fold on TLX overexpression, there was a negligible effect on the shorter construct ( Figure 6a). We then proceeded to analyze the hMMP-2 promoter for putative TLX-binding sites. We identified two 'AAGTCA' sites binding TLX at~1.2 kb upstream of the transcription start site (Figure 6b). Quantitative chromatin immunoprecipitation (ChIP) assays by using chromatin isolated from IMR-32 WT cells revealed a basal level binding of TLX to the MMP-2 promoter, along with RNA polymerase-II (Pol-II) recruitment and acetylated H3K9 (H3K9Ac). Under the same conditions, TLX did not bind to β-actin promoter. As we have previously shown that TLX is a significant contributor to angiogenesis upon hypoxia, we tested if TLX-mediated MMP-2 regulation is affected upon hypoxia. ChIP of IMR-32 cells when grown in a 1% O 2 concentration showed that TLX binding to the MMP-2 promoter increased 2-2.5-fold, which correlated with an increased recruitment of Pol-II and H3K9Ac (Figure 6c). In contrast, no occupancy of TLX was detected at the proximal promoter even in hypoxia. The precipitated DNA was sequenced for confirmation (data not shown). A study from our group has identified binding of TLX on the Oct-4 promoter in neural progenitor cells upon hypoxia, leading to self-renewal of these cells and a further activation of Akt signaling pathways. 11 Recent studies demonstrate the role of Oct-4 in promoting migration and invasion of bladder cancer cells by expression and activation of MMP-2 and MMP-9. 22 In agreement with this, we found that TLX in IMR-32 cells upon hypoxia was recruited to the human Oct-4 core promoter (Figure 6c). The binding to Oct-4 promoter under normoxic conditions was negligible and comparable to preimmune controls (Figure 6a).
Additionally, we tested TLX-specific binding on the MMP-2 promoter consensus element by performing TLX capture using a biotinylated oligonucleotide encompassing the consensus element of TLX-binding site in the MMP-2 promoter. The biotinylated oligonucleotide was incubated with the nuclear lysate containing TLX, which was captured by a TLX-specific antibody, followed by incubation with a secondary antibody conjugated with horseradish peroxidase (HRP). Color was developed by TMB/E substrate and the binding intensity was calculated using absorption at 450/650 nm. Nonspecificity was ruled out using random IgG, and the Occupancy of TLX, Pol-II and H3K9 acetylation across the 1.2 kb upstream regulatory regions of TLX-regulated genes MMP-2 and OCT-4, and control actin promoter was monitored by ChIP analysis upon normoxia (b) or hypoxia (c). Chromatin was isolated from normoxia-or hypoxia-treated cells and ChIP analysis was performed as described in Materials and Methods. Amplicon from each immunoprecipitate is represented as the percentage of input. Each error bar indicates standard deviation calculated from triplicates. (d) Graph represents the binding of TLX to MMP-2 promoter as a function of absorbance at 450/650 nm. Biotin-labeled consensus oligos were used to capture TLX of nuclear lysate from WT IMR-32. A nonspecific capture oligo served as control, and rabbit IgG were used to exclude nonspecific binding. Mutant oligos (Mut1 or Mut2) were used to confirm the specificity of capture. The values obtained are means of three independent experiments along with S.D. as error bars non-biotinylated consensus oligonucleotide was used as a competitor to validate the specific binding. Further mutation of the consensus site at the first two bases (Mut1) or the middle three bases (Mut2) markedly reduced the binding of TLX to the probe. Our results show a 4.5-fold enrichment of TLX binding on the MMP-2 promoter site compared with the preimmune control (Figure 6d).
TLX is expressed in NB tissues derived from patients. We further examined if we could capture an enrichment of TLX expression in patient samples. For this, we screened NB tumor tissue arrays including 10 human cases (ages 5-48 years, two tissues per case) of aggressive NB and two cases of normal peripheral nervous tissues (PNS) for the expression of TLX (Figure 7a). There was an enhanced TLX expression in these tumors compared with normal PNS tissue. We also used the open R2 statistics application (microarray analysis and visualization platform; http://r2.amc. nl) using microarray data from 88 cases of NB-Versteeg-88 MAS5.0-u133p2 (http://hgserver1.amc.nl/cgi-bin/r2/main.cgi). A Kaplan-Meier analysis indicated that the higher expression of TLX (NR2E1) correlates with shorter survival of NB patients, with a cutoff at 8.3, χ 2 = 9.98, d.f. = 1, P = 0.0016 (Figure 7b).

Discussion
It has been recognized that a number of stem cell renewal factors are involved in tumorigenesis. TLX is a neural cellspecific renewal factor, and gene amplification of TLX has been reported to occur in malignant glioma. 13 By expressing TLX, the tumor cells appear to engage neurogenetic niches for their own maintenance. 23 Here we demonstrate that TLX is also highly expressed in the stem cell-like population enriched from NB, originating from the sympathetic nervous system. Some glioma cells are derived from neural stem cells that are normally maintained in neurogenic niches in the brain. 24 However, NB is derived from embryonic neural crest cells, arising from the dorsal aspect of neural tube and migrating to the sympathetic ganglia and the adrenal glands. The high expression of TLX observed in the brain of E13.5 mice 25 indicates the peak of brain neurogenesis. Neural crest cells possess remarkable capacities of migration and multipotency, and begin to migrate around E10.5, detectable in the adrenal glands around E13.5. 26 The HIF-2α-expressing immature neural crest-like NB cells are maintained by perivascular niches 27 we have previously showed TLX to stabilize HIF-2α. 28 We have demonstrated that the expression of TLX increases when the NB cells are cultured in neural stem cell media, resulting in tumor sphere formation. Interestingly, these tumor spheres recapitulate neurospheres in their expression of stem cell markers such as CD133, Nestin, Oct-4 and CD15. Furthermore, TLX is expressed in NB-TICs and induces MMPs, which could activate the remodeling of matrix, migration and, possibly, invasion of NB cells. MMP-2 localizes at the migrating edge of TLX-expressing TIC clusters in the xenograft sections of human NB-TICs, suggesting its importance for migratory activities of cancer cells, which may result in invasiveness leading to metastasis. In this context, it is of interest that CD15 in grafted tumor tissues localizes on the surface of TLX-positive cells. CD15, also known as LeX or SSEA-1, is a set of glycan moieties containing fucosylated N-acetyllactosamine, which is considered to be important for neural stem cell migration. 29 In addition, the sialylated or sulfated forms of CD15 is closely associated with lymphocyte rolling, the first step for cellular extravasation, and cancer metastasis. 31,30 IMR-32 and NB-TICs express MMP in hypoxia, which might be due to a cooperative effect of TLX and its downstream Wnt signaling. In fact, TLX becomes stabilized in hypoxia, 21 and has been shown to induce Wnt7b, which subsequently inhibits GSK3. 9 This leads to stabilization and activation of β-catenin, inducing several target molecules such as Myc. We find that TLX expression correlates with pAkt levels, 11 which could also be a consequence of PTEN repression. 19 Elevated pAkt can also phosphorylate and inhibit GSK3 apart from stabilizing for HIF-1α during hypoxia. 32 HIF-1α also modulates Wnt signaling in hypoxic stem cells and enhances β-catenin activation. Thus, we predict that both TLX and HIF While TLX occupies the MMP-2 promoter endogenously, Oct-4 occupancy occurs in a hypoxic milieu, under which conditions these tumor cells would acquire a more epigenetic and phenotypic resemblance to stem cells. Hypoxia is one of the most important contributing factors in the tumor microenvironment, stimulating tumor dedifferentiation and angiogenesis. 33 In this regard, the expression of HIF-2α has been proposed to be associated with dedifferentiation of NB, which may depend on its angiogenic property rather than cellcycle modulation. 34 TLX is reported to act as a hypoxiainducible proangiogenic switch molecule, strongly expressed in postnatal proangiogenic retinal astrocytes, which secrete vascular endothelial growth factor (VEGF) and fibronectin. Moreover, expression of TLX is rapidly downregulated by contact with blood vessels and a derangement of fibronectin matrix was observed in TLX-null mice. 35 In this context, it is interesting to note that fibronectin fragments from cancer cells can induce the secretion of MMP-2, 36 whereas MMP-2 and MMP-9 have been shown to degrade fibronectin, as the first step of ovarian cancer metastases. 37 Thus, TLX affects not only immediate hypoxia-responsive proteins, that is, HIF-2α and VEGF, but also affects extracellular matrix proteins needed for vascular organization.
Hypoxia is a well-known condition that induces epithelial-tomesenchymal transition (EMT), a hallmark of the morphologic changes of tumor cells leading to metastases by various mechanisms. 38 Interestingly, it has recently been proposed that Oct-4 expression can promote the migration and invasion of glioblastoma cells. 39 It is an obvious possibility that TLX could be a critical factor by virtue of its dual role in matrix remodeling and angiogenesis, along with regulation of cellcycle programs contributing to EMT. Further studies are required to show if TLX is a true contributing factor for cancer metastasis by using NB animal models, and whether the metastatic capacity can be altered by depletion of TLX. In sum, our studies propose that TLX employs multiple pathways, amplifying each other to dedifferentiate NB cells and to maintain the progenitor population in a hypoxic environment. The fact that TLX was identified as one of the significant mRNA responders to EGFR network perturbation when analyzed for a prognostic outcome prediction in glioblastoma multiforme, 40 along with our results, highlight TLX as a crucial candidate for directed cancer therapy.
Animal xenograft. Primary NB cell line (NB273) was cultured as outlined by Hansford et al. 20 Tumors were established using a heterotopic model of tumorigenesis. Briefly, primary NB cells were resuspended in PBS, mixed 1 : 3 with basement membrane extract (Trevigen, Gaithersburg, MD, USA) immediately before injection and injected in a 100 μl volume into the inguinal fat pads of 4-to 5week-old NOD/SCID mice as approved by University Health Network's Animal Ethics Committee (protocols 09-004 and 2178.5). Animals were monitored weekly for evidence of tumor formation and were killed when tumors reached 1.0-1.5 cm 3 . Tumors were recovered and fixed in 10% formalin for 24 h before paraffin embedding. Tumors were characterized as NB tumors by H&E staining and immunohistochemistry for the NB markers MAP2 (1 : 5000; Sigma) and NB84 (1 : 50; Novocastra, Nussloch, Germany), and determined to be free of Epstein-Barr virus contamination by EBER1 in situ hybridization (data not shown).
Invasion and migration assays. Invasion and migration assays were performed using CytoSelectTM 24-Well (Cell Biolabs Inc.; CBA-100-C) according to the manufacturer's instructions. Briefly, equal numbers of cells in suspension (1 × 10 5 ) in DMEM complemented with 2% fetal calf serum (FCS) were added to the upper compartment of the chamber, the bottom of which is a polycarbonate membrane insert with 8-mm pore size in a 24-well plate (insert coated with extracellular matrix membrane for invasion assay and left uncoated for migration assay to use serum as chemoattractant), and kept at 37°C for 16 h. The lower compartment contained DMEM with 10% FCS. At the end of the incubation period, noninvasive or non-migratory cells from the upper surface of the filter were wiped off with a swab. The lower surface of the filter was stained with a dye and extracted. A volume of 100 μl of each sample was transferred to a 96-well plate and the OD at 560 nm was measured. To detect the amount of MMP-2 secreted by control or by TLX knockdown-stable cells, cells were maintained in serum-free media for 30 h at 37°C and conditioned media were collected. Conditioned media were diluted 15fold in diluent buffer from the kit (Invitrogen; KHC3082), added to wells and processed for ELISA along with serially diluted standards and diluent only as negative control. Samples were processed according to the manufacturer's instructions and the results recorded by measuring absorbance at 450 nm. Following collection of conditioned media, cells from different wells were trypsinized and counted to normalize the MMP-2 secreted. Transfection of siMMP-2 (SC29398) was performed as described above. All experiments have been performed as three biological replicates and two technical duplicates. Standard deviation from the mean has been plotted in the graphs.