ZEB1 Promotes Invasion in Human Fetal Neural Stem Cells and Hypoxic Glioma Neurospheres

Cell culture, hypoxic and digoxin treatment

The GBM neurosphere (gliomasphere) lines HSR‐GBM1 (GBM1) and 040622 (622) were generously provided by Dr Angelo Vescovi (Milan, Italy), and the pediatric glioma cell line KNS42 54 was obtained from the Japan Cancer Research Resources Cell Bank. Both were cultured in serum‐free media as neurospheres as previously described 23. Their identity was confirmed by analysis of nine tandem repeats plus a gender‐determining marker, Amelogenin using the StemElite kit (Promega, http://www.promega.com/) in the John Hopkins Core Facility for DNA fragment analyses (http://grcf.med.jhu.edu/) as part of their standard STR fingerprinting service for cell lines. The polymerase chain reaction (PCR) product was electrophoresed on an ABI Prism^® 3730xl Genetic Analyzer using the Internal Lane Standard 600 (Promega). Data were analyzed using GeneMapper v 4.0 software (Applied Biosystems, http://www.appliedbiosystems.com). The profiles of the used cell material are attached in Supporting Information Table S2.

Human fetal neural stem and progenitor cells (NSCs) were obtained from first trimester human fetal autopsy specimens in concordance with German law and Ethics Board evaluation (University Medical Center Freiburg, Germany) and cultured as neurospheres as previously described 30. The study was also approved by the Johns Hopkins Institutional Review Board. The pancreatic cancer cell line PANC1 was a gift from William Matsui from The Sidney Kimmel Comprehensive Cancer, Johns Hopkins Hospital, Baltimore, MD, USA.

Hypoxic growth (1% O₂, 5% CO₂ and 94% N₂) was maintained in a special chamber (COY laboratory equipment, Grass Lake, MI, USA). For all analyses, it was ensured that cells resided in exponential growth phase. Digoxin (Baxter, Deerfield, IL, USA) was applied in a working concentration of 100 nM dissolved in Dulbecco's modified Eagle's medium (DMSO) (Sigma‐Aldrich, St. Louis, MO, USA) for 24 h. Selection for resistance against the antibiotic puromycin [short hairpin RNAs (shRNAs)] was achieved at a concentration of 2 μg/mL (for tumor cultures) and 0.2 μg/mL (for fetal NCSs) in standard growth culture media.

mRNA and protein analyses

The abundance of mRNA was assessed by conversion into complementary DNA and subsequent quantitative real‐time PCR measurements using SYBR green‐based fluorescence (Bio‐Rad, Hercules, CA, USA). To ensure purity of RNA, on‐column genomic DNA digestion was applied prior to transcription into RNA. Relative quantification to housekeeping gene β‐ACTIN was assessed with ddCt‐method. Primer sequences can be found in Supporting Information Table S1. Western blotting was performed as described before 34; antibodies were used as per the manufacturer's instructions (Supporting Information Table S1).

DNA constructs and viral infection

We used the following lentiviral constructs: pLKO.1 scramble shRNA (Addgene plasmid 1864) 45, and ZEB1 I shRNA and ZEB1 V shRNA (both from The RNA Consortium, Broad Institute, Clone ID TRCN0000017563 and TRCN0000017567). Lentiviral particles were produced by transfecting 293T cells with VSVG envelope plasmid, delta 8.9 gag/pol plasmid and the plasmid of interest, as described previously 23 using Lipofectamine2000 (Life Technologies, http://www.lifetechnologies.com) as per the manufacturer's instructions. After 24 h, the medium was changed to FGF/EGF neurosphere media [DMEM and F12 from Life Technologies, hEGF and b‐FGF from Peprotech, http://www.peprotech.com, Heparin from Sigma‐Aldrich 23 ] and supernatants were collected at 48 and 72 h. These supernatants were pooled, passed through a 0.45‐μm filter, concentrated using polyethylene glycol and then frozen at −80°C until needed.

Immunohistochemistry

Tissue microarrays (TMAs) were constructed using tumor samples obtained from 1977 to 2009. These as well as normal tissues from autopsy specimens were retrieved from the archives of the Johns Hopkins Hospital Department of Pathology following institutional review board approval. TMAs were constructed and sectioned using standard procedures at the Johns Hopkins TMA core facility. For each tumor, four cores measuring 0.6 mm in diameter were included per array. ZEB1 staining was performed using sodium citrate (pH = 6) antigen retrieval, anti‐ZEB1‐antibody incubation overnight, anti‐rabbit antibody detection using biotinylated secondary‐antibody (Vectastain, #PK6101) and 3,3′‐diaminobenzidine (DAB)‐based stain. A pancreatic cell line (PANC1) cell block was used as a positive control, while negative controls included omission of primary antibody and the analysis of multiple non‐neoplastic control tissues included on the TMAs 56. A neuropathologist (CGE) blinded to diagnosis scored average ZEB1 staining intensity in tumor nuclei as negative (0), weak (1), moderate (2) or strong (3). The average percent of cell staining rounded to the nearest decile (ie, 0%, 10%, 20%) was also scored for each tumor. These two scores were then multiplied to generate Z‐scores between 0 and 300.

I n vitro invasion assay

Invasion of GBM‐derived cells was analyzed using the modified Boyden Chamber assay. A 24‐well system was used with each well containing a permeable Transwell insert (Becton Dickinson Biosciences, https://www.bdbiosciences.com/) with 8‐μm pores. To mimic three‐dimensional extracellular environment and to investigate infiltrative potential of the cells, the inserts were coated with growth factor reduced Matrigel (Becton Dickinson Biosciences) and incubated for 1 h at 37°C. Subsequently, cells suspended in 500 μL of DMEM were placed on top of each insert membrane. The bottom was filled with 700 μL of DMEM media containing 10% fetal calf serum (FCS). For GBM1, 75 000 cells, for KNS42, 25 000 cells, and for the fetal NSCs, 15 000 cells were used for the assays.

All Boyden chamber assays were analyzed 14 h after exposure to hypoxia or normoxia. The upper side of the membrane was then wiped carefully with a cotton swab to remove the rest of the plated cells. The membrane was then fixed in ice‐cold 90% ethanol for 25 minutes and stained with hematoxylin. The migration/invasion of the GBM cells was evaluated by counting the cell nuclei on the lower side of the membrane under light microscope and 20× magnification (4–6 view fields).

Statistical analyses

Significance was tested with two‐sided Student's t‐test performed in Statistica v5.1. (Statsoft, Tulsa, OK, USA). The Mann–Whitney test was used for immunohistochemical analyses. Bar plots present the mean and standard error of the mean.

Increased ZEB1 protein in higher grade and more invasive brain tumors

We screened a range of pediatric and adult brain tumors for ZEB1 expression in order to determine if this protein was associated with a particular grade or type of neoplasm. These included relatively compact tumors such as World Health Organization (WHO) grade I pilocytic astrocytoma (PA, n = 65) and WHO grades I to III ependymoma (EP, n = 71), as well as infiltrating lesions including WHO grade II diffuse astrocytoma (A, n = 15), WHO grade III anaplastic astrocytoma (AA, n = 16), pediatric (n = 32) and adult (n = 61) WHO grade IV glioblastoma (pGBM and aGBM), WHO grade II oligodendroglioma (O, n = 14) and medulloblastoma (MB, n = 21). ZEB1 immunostaining was almost entirely nuclear, and varied in intensity and percentage of neoplastic cell stained. Almost no staining was noted in entrapped non‐neoplastic neuronal or glial elements within tumor specimens. Among adult and pediatric GBM represented on the tissue microarray, both extremely cellular regions presumable from the tumor core and less cellular ones most likely from the periphery could show high levels of ZEB1 protein.

Representative images from multiple tumor types with varying expression levels are shown in Figure 1A–H, and mean expression for all groups of lesions is depicted in Figure 1I. ZEB1 expression was significantly higher in infiltrating astrocytic, oligodendroglial and embryonal brain tumors as compared to the more compact ependymoma and pilocytic astrocytoma (Figure 1J, P < 0.0001). Within the infiltrative group, low‐grade (WHO grade II) tumors had significantly lower amounts of ZEB1 as compared to high‐grade (WHO grade III and grade IV) tumors (Figure 1K, P < 0.0001). Finally, pediatric GBM had significantly higher ZEB1 protein levels than adult GBM (P < 0.002).

Hypoxia increases invasion of glioma cultures and expression of EMT factors

To test the effect of reduced oxygen on glioma cell dispersion, we compared the ability of adult GBM1 cells and the pediatric GBM line KNS42, both grown in serum‐free media as neurospheres, to invade when cultured in 1% or 21% oxygen. In both the adult and the pediatric GBM lines, we observed a significant increase in invasive behavior over 14 h in hypoxic cells, with a two‐ to fourfold increase in the number of hypoxic cells moving through pores in a modified Boyden chamber assay as compared to normoxic cells (P < 0.001, Figure 2A). Representative light‐field images of GBM1 cells, which have traversed the membrane under normoxia and hypoxia, are shown in the lower panels of Figure 2A.

Next, we aimed to identify possible molecular effectors of the hypoxia‐augmented motile behavior of glioma neurospheres. Hypoxia resulted in elevated mRNA expression of the three tested EMT‐associated transcription factors (TWIST1, SNAI1 and ZEB1; Figure 2B) compared with normoxic conditions. These transcription factors are known to play key roles in dissemination of many types of cancer 4, 50 including GBMs 24, but their roles in hypoxia are not known in glial tumors. Because relatively little is known about the functional role of ZEB1 in gliomas, and given its association with infiltrating and higher grade brain tumors, we focused our further investigations on that transcription factor. Western blot analysis confirmed that GBM cells grown in 1% oxygen contain higher levels of ZEB1 protein compared with normoxic ones (Figure 2C). A histopathological hallmark of GBM is the presence of pseudopalisades around hypoxic and necrotic cores, and it has been suggested that these hypoxic cells are particularly motile 6. Interestingly, ZEB1 staining was prominent in these regions of hypoxia in GBM (Figure 2D).

Inhibition of HIF1α can reduce expression of ZEB1 under hypoxia

Hypothesizing that the classical hypoxic response drives the expression of ZEB1 in our glioma models, we tested if pharmacological reduction of a core regulator of hypoxia, HIF‐1alpha (HIF1α), affects the ZEB1 levels. The cardiac glycoside digoxin has been shown to lower HIF1α levels in cancer 60, including glioma neurospheres 3. We confirmed that Digoxin treatment caused robust reduction of HIF1α protein in hypoxic GBM1 cells (Figure 3A). Consistent with the notion that HIF1α regulates ZEB1, we observed a marked decrease in ZEB1 expression following digoxin treatment in the GBM neurospheres (Figure 3B). Digoxin‐mediated reduction of HIF1α and ZEB1 also resulted in impairment of in vitro GBM1 invasion under hypoxia (Figure 3C).

ZEB1 knockdown inhibits invasion in both normoxia and hypoxia

Prior reports have implicated ZEB1 in the invasion of glioma cells in the normoxic periphery of tumors 49. To directly examine the role of ZEB1 in GBM invasion in the hypoxic core of tumors as well, we tested the effects of targeted ZEB1 knockdown in vitro. We identified two shRNA species able to reduce ZEB1 protein levels in glioma neurospheres by 63%–100% as quantified using NIH ImageJ following lentiviral delivery compared with scrambled control (Figure 4). This decrease in ZEB1 levels had a profound effect on the invasive capacity of glioma cells, with a significant reduction in the number of tumor cells extending through pores in Transwell assays under both normoxic and hypoxic conditions. Similar effects were observed following ZEB1 knockdown in the pediatric GBM line KNS42 (Figure 4). These data suggest that ZEB1 is an important factor promoting invasion in both normoxia and hypoxia.

ZEB1 is expressed in human fetal stem and progenitor cells and promotes their migration

Many brain tumors are thought to arise from neural stem and progenitor cells, and we found particularly high ZEB1 expression levels in pediatric brain tumors. In fetal brain development, the majority of stem cells are found around the ventricles in a “ventricular zone,” with more differentiated progenitors forming a dense germinal matrix then migrating outward to generate neurons and astrocytes. We stained three fetal human brains from the late first trimester or early second trimester for ZEB1, and saw moderate to strong expression in the cells nearest the ventricles (Figure 5A, red arrowheads), as well as moderate expression in many of the neural progenitors deeper in the germinal matrix (Figure 5A, black arrowheads). Scattered cells in the developing fetal deep gray matter structures, white matter and cortical ribbon also expressed ZEB1, but these were less frequent and often showed weaker staining than those in the periventricular zones.

We were also able to detect ZEB1 protein in human neural stem and progenitor cells isolated from the brain of first trimester fetuses and grown as neurospheres. Exposure to low oxygen increased the level of ZEB1 proteins in a manner similar to that seen in GBM neurospheres, suggesting that this EMT‐associated transcription factor may also respond to physiological hypoxia during normal development (Figure 5B). Growth in hypoxia resulted in a non‐significant increase in cellular invasion of those human fetal neurospheres compared with normoxic cells in Boyden chamber assays (data not shown). We were able to reduce ZEB1 protein expression in fetal NCSs by over 80% using two shRNAs (Figure 5C), and this significantly inhibited the invasion in normoxia (Figure 5D). We were not able to assess invasive capacity of the neurospheres in hypoxia because of the limited availability of cells following ZEB1 knockdown. Nevertheless, these studies support the concept that the EMT regulator ZEB1 promotes human NCS motility.

Further support for the expression of ZEB1 in neural stem cells and migrating progenitors comes from analysis of the publicly available Allen Brain Atlas database (http://www.brain‐map.org/). Their in situ hybridization studies in the developing mouse brain show highest ZEB1 abundance in neurogenic regions around the ventricles, in cerebellar progenitors and in the rostral migratory stream (Supporting Information Figure S1A). Moreover, Affymetrix chip‐based mRNA expression analyses of human brains revealed its highest expression in very young fetal samples (weeks 8–9), as highlighted by the red signature in the heat map shown in Supporting Information Figure S1B.