TWIST is Expressed in Human Gliomas and Promotes Invasion

Abstract

TWIST is a basic helix-loop-helix (bHLH) transcription factor that regulates mesodermal development, promotes tumor cell metastasis, and, in response to cytotoxic stress, enhances cell survival. Our screen for bHLH gene expression in rat C6 glioma revealed TWIST. To delineate a possible oncogenic role for TWIST in the human central nervous system (CNS), we analyzed TWIST message and protein expression in gliomas and normal brain. TWIST was detected in the large majority of human glioma-derived cell lines and human gliomas examined. Increased TWIST mRNA levels were associated with the highest grade gliomas, and increased TWIST expression accompanied transition from low grade to high grade in vivo, suggesting a role for TWIST in promoting malignant progression. In accord, elevated TWIST mRNA abundance preceded the spontaneous malignant transformation of cultured mouse astrocytes hemizygous for p53. Overexpression of TWIST protein in a human glioma cell line significantly enhanced tumor cell invasion, a hallmark of high-grade gliomas. These findings support roles for TWIST both in early glial tumorigenesis and subsequent malignant progression. TWIST was also expressed in embryonic and fetal human brain, and in neurons, but not glia, of mature brain, indicating that, in gliomas, TWIST may promote the functions also critical for CNS development or normal neuronal physiology.

Introduction

The basic helix-loop-helix (bHLH) family of proteins regulates normal development and differentiation by forming DNA-binding heterodimers composed of tissue-specific class B) and ubiquitously expressed (class A) proteins that direct cell-specific gene expression [1,2]. The class B bHLH protein, TWIST, plays a central role in determining cell fate in mesoderm, primarily during myogenesis [3–6]. For example, TWIST prevents development in normal skeletal muscle by forming inactive heterodimers with other myogenic bHLH proteins (e.g., MyoD and myogenin) that direct the acquisition of a quiescent, differentiated phenotype [7–9]. TWIST also inhibits the induction of p53-mediated apoptosis in rodent fibroblasts in response to genotoxins and prolonged serum deprivation [10]. Both functions of TWIST are likely essential for normal development as evidenced by death in utero at embryonic day 11.5 in TWIST null mice accompanied by a high frequency of apoptotic cells [3].

The antiapoptotic function of TWIST suggests that it could function as an oncogene [10]. In accord, elevated levels of TWIST mRNA have been observed in the mesenchymal tumors, rhabdomyosarcoma and osteosarcoma [10,11]. The recent demonstration of oncogenic cooperation between TWIST and N-myc amplification in neuroblastoma [12] and TWIST expression in a variety of carcinomas [11,13,14] supports an oncogenic function that is not limited to mesenchymal-derived neoplasms. TWIST expression in gastric carcinoma [13] has been associated with the initial phase of the metastatic process whereby tumor cells transition from an epithelial to an invasive mesenchymal phenotype. Indeed, in a mouse breast carcinoma model, TWIST is necessary for cells to metastasize by promoting this epithelial mesenchymal transition (EMT) and inducing cell motility [15]. Notably, elevated levels of TWIST mRNA correlate with invasive breast cancer phenotype [15]. Thus, TWIST promotes increased tumor cell survival and motility—two features characteristic of malignant gliomas. To our knowledge, TWIST expression has not been characterized in human primary brain tumors.

Human adult gliomas arise from mature glia and/or glial progenitors in the brain [16]. Classified histologically in order of increasing malignancy from grades II to IV, gliomas can develop de novo as grade IV neoplasms (glioblastoma multiforme) or can undergo malignant progression from low-grade (grade II) or anaplastic gliomas (grade III) to “secondary” glioblastomas [17,18]. Importantly, as tumors increase in grade, they also demonstrate enhanced cell survival accompanying loss of apoptosis in response to cytotoxic insult, increased tumor cell migration and invasiveness, increased cell proliferation, and induction of tumor angiogenesis [19–21]. As noted above, in other cell types or cancers, the regulation of many hallmark behaviors of gliomas has been attributed to TWIST.

Here we examine TWIST mRNA and protein expression in human glioma-derived cell lines, glioma tissues, and normal developing and mature brains, and also in an in vitro model of glial tumorigenesis using mouse p53-deficient astrocytes [22–24] to further characterize possible associations with gliomagenesis or glial tumor progression. We also examine the contribution of TWIST to the phenotype of human glioma cells by analyzing the effect of TWIST overexpression on invasion in a glioma cell line. Our data support functions for TWIST that contribute to glial tumorigenesis and glioma tumor cell behavior, as well as potential roles in the developing and mature human central nervous system (CNS).

Materials and Methods

Cell Lines and Tissues

Cell lines studied were derived from glioblastoma (C6, SNB19, UW18, UW455, UW456, UW467, U87, U138MG, U373MG, SF763, SF767, A172, and T98G), neuroblastoma (SKNMC and IMR32), medulloblastoma (D283), and colorectal (SW480 and HT29), breast (MDA231 and MDA453), and small cell lung cancer (SCLC; H82 and H209). Lines with the UW designation were established in the authors' laboratories and grown in DMEM/F12 with 10% fetal bovine serum (FBS). The remaining cell lines were either obtained from the Brain Tumor Research Center Tissue Bank (SF767 and SF763; Department of Neurological Surgery, University of California, San Francisco, CA) or purchased from the American Type Culture Collection (ATCC; Manassas, VA), and were cultured as recommended. All human tissues were obtained in accordance with human subjects protocols approved by the University of Washington Institutional Review Board. Tumor and normal brain specimens were obtained with informed consent from adult patients operated for glioma, and epilepsy or benign nonglial tumors, respectively. Clinical and demographic data including pathologic diagnosis and tumor grade (based on clinical neuropathology reports), recurrence status, and prior therapy were noted for each patient. Adult samples were snap-frozen in liquid nitrogen and stored at -80°C. At the time of RNA extraction, a portion of each sample was cut and processed for hematoxylin and eosin (H&E) histology. Fetal tissues, obtained from the University of Washington Birth Defects Research Laboratory, Department of Pediatrics (Seattle, WA), were examined and carefully dissected to include only CNS tissues.

Degenerate Reverse Transcription Polymerase Chain Reaction (RT-PCR) Cloning

Degenerate RT-PCR cloning was utilized to identify novel class B bHLH genes expressed in the rat C6 glioma cell line. The design of the degenerate primers was based on conserved class B-specific sequences within the bHLH region as previously described [25]. PCR products were cloned into pCR2.1-TOPO vector with the TOPO-TA system according to the manufacturer's instructions (Invitrogen, Carlsbad, CA), and individual clones were sequenced to identify bHLH genes. A probe generated from the insert of the TWIST clone was used for Northern blot confirmation of TWIST expression in the C6 glioma cell line according to methods described below.

Northern Blot Analysis

Twenty microgram of total RNA was analyzed for steady-state levels of TWIST mRNA by Northern blot analysis using previously published methods [26]. TWIST probe was generated from a cloned 556-bp fragment from the 3′ end of the gene using the Megaprime DNA labeling system (Amersham Pharmacia, Piscataway, NJ) with ³²P-labeled dCTP. Images of ethidium-stained 18S and 28S RNA bands were obtained before and after transfer to confirm loading, RNA integrity, and complete transfer.

TWIST Expression Constructs and Cell Transfection

The human TWIST cDNA clone (GenBank ID BC036704) was purchased from ATCC. Open reading frame (ORF) was amplified with PCR primers: 5′-cgccaccatgatgcaggacgtgtccagctcg-3′ and 5′-actagtgggacgcggacatggaccag-3′. The upstream primer was designed with Kozak sequence [27] for improved protein translation. The resulting PCR product was subcloned into pCR2.1-TOPO vector (Invitrogen) and verified by sequencing. The EcoRI fragment was then subcloned into the EcoRI site of the pLXSN retroviral vector (BD Biosciences, San Jose, CA). Following sequence verification, the TWIST expression construct or empty vector was transfected into packaging Phoenix cells in serum-free DMEM/F12 using Lipofectamine Plus reagent (Invitrogen) according to the manufacturer's recommendations. Twelve hours later, the medium was replaced with medium containing 10% FBS. Following an additional 24 hours of incubation, the medium containing virus was collected, filtered through a PVDF filter, and used for infection of SF767 cells in the presence of 4 µg/ml Polybrene (Sigma, St. Louis, MO). Selection of infected cells was begun after 48 hours by culturing cells in the presence of G418 (Sigma). Pools of G418-resistant SF767 cells were collected and expanded, and endogenous expression of TWIST protein was confirmed by Western blot analyses. Cells overexpressing TWIST are designated SF767Tw and empty vector controls SF767 LXSN.

TWIST Antibody

Twenty one amino acids (SSGGGSPQSTEELQTQRVMA) specific for TWIST-1 protein near the loop region of the HLH domain were chemically synthesized and conjugated with a carrier, Keyhole limpet hemocyanin. After high-performance liquid chromatography (HPLC) analysis to determine Keyhole limpet hemocyanin binding efficiency and peptide purification, the protein antigen was injected into three rabbits for immunization. The specificity of the peptide sequence for TWIST and the lack of homology to other related bHLH or other proteins were confirmed by performing a BLAST search of the National Center for Biotechnology Information protein sequence database. Polyclonal serum was prepared from whole blood three times after primary and booster immunizations (Beckman Research Institute, Duarte, CA). The specificity of the antibody was tested in Western blots using cell lines with undetectable, low, and high levels of TWIST mRNA and by immunohistochemical analysis in tissue sections comparing antibody staining patterns with patterns of TWIST message detected by in situ hybridization.

Western Blot Analysis

Cells were harvested by scraping into cold PBS and lysed in M-PER protein extraction reagent (Pierce, Rockford, IL) supplemented with protease inhibitors. Nuclear extracts were prepared as previously described [28]. Protein concentrations were determined using the BCA Protein Assay Reagent Kit (Pierce). Proteins were separated in 4% to 15% Tris-HCl gradient polyacrylamide SDS-PAGE gels (Bio-Rad, Hercules, CA) and transferred by electroblotting to an Immun-Blot PVDF membrane (Bio-Rad) in Tris-glycine buffer with 20% methanol. Membranes were blocked with 5% nonfat dry milk (NFDM) in TBST (10 mM Tris, pH 8.0, 150 mM NaCl, and 0.02% Tween-20) for 30 minutes and incubated overnight at 4°C with an antihuman TWIST rabbit polyclonal antibody diluted 1:1000 in TBST containing 2% bovine serum albumin (A3059 Fraction V; Sigma). After several washes with TBST, membranes were incubated for 2 hours at room temperature with horseradish peroxidase (HRP)-conjugated goat antirabbit secondary antibodies (Pierce) diluted 1:5000 in TBST containing 3% NFDM. After washes with TBST, antibody binding was visualized on X-OMAT Blue XB-1 film (Kodak, Rochester, NY) by chemiluminescence. For protein loading control, the membranes were stripped and incubated in 5% NFDM/TBST with either anti-β-actin goat polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA), diluted 1:1000 for total proteins, or with INI1 (H-300) rabbit polyclonal antibody (Santa Cruz Biotechnology), diluted 1:400 for nuclear proteins. After washes with TBST, membranes were incubated in 3% NFDM/TBST for 45 minutes with secondary antibodies diluted 1:5000: HRP-conjugated chicken antigoat antibody (Santa Cruz Biotechnology) or HRP-conjugated goat antirabbit antibody (Pierce).

RT-PCR and Southern Blot Analysis

cDNA was synthesized by RT using as a template DNase I-treated total RNA that was extracted from cultured cells and tissues using Trizol (Invitrogen). Controls for genomic DNA contamination were performed without reverse transcriptase on matching samples. The sequences of TWIST and GAPDH primers and of internal oligonucleotide probes used for Southern blot analysis were either derived from published sources, or designed using the MIT Primer 3 program (Table 1). PCR conditions were optimized to demonstrate linear product formation at the cycle numbers used for TWIST detection. Products were identified by Southern hybridization with gene-specific internal oligonucleotide probes. In addition, products from randomly selected RT-PCR reactions for each gene were directly sequenced to confirm their identity. Band density images of Southern hybridizations or ethidium bromide-stained gels were quantified by using Image J shareware (Scion Corporation, Frederick, MD; http://rsb.info.nih.gov/ij/) and normalized to corresponding GAPDH readings for the same sample. Band density was determined on a representative Southern hybridization for tumor samples, and in triplicate for mouse astrocyte experiments. Between-group differences were tested for significance by using the Kruskal-Wallis test and post hoc Mann-Whitney U tests.

Table 1.

RT-PCR Primer Sequences and Conditions for Human and Mouse Genes.

	Gene	Forward Sequence	Reverse Sequence	Internal Oligo	TA	Cycle #	Product Size
Human	Twist	GAGTCCGCAGTCTTACGAGG	CTGCCCGTCTGGGAATCACT	TTAGAAGAGCAAAATCCAAA(1)	59	35	555
				CTACGCCTTCTCGGTCTGGA(2)
	GAPDH	ACGGATTTGGTCGTATTGGG	TGATTTTGGAGGGATCTCGC	ATGGCACCGTCAAGGCTGAG	60	28	231
Mouse	Twist	CACGCTGCCCTCGGACAA	GGGACGCGGACATGGACC	GCTGAGCAAGATTCAGACCC	60	30	197
	GAPDH	ATTGTCAGCAATGCATCCTGCA	AGACAACCTGGTCCTCAGTGTA	AACTTTGGCATTGTGGAAGG	59	25	412

In Situ Hybridization

Sense and antisense TWIST riboprobes were synthesized by in vitro transcription from a 298-bp fragment from the 3′UTR of human TWIST cloned into pBluescript SK⁺ (Stratagene, La Jolla, CA) using T7/T3 polymerases. Twelve-micrometer sections previously fixed in 4% paraformaldehyde were freshly cut from tissue stored at -80°C. TWIST in situ hybridization was performed using previously published methods with minor modifications [26] (full protocol available on request). The slides were either washed in water and mounted in aqueous mounting media, or counterstained in methyl green, dehydrated, cleared, and permanently mounted. Adjacent sections were also stained with H&E.

Immunohistochemistry

Tumor or non-neoplastic samples were either fixed fresh frozen with 4% paraformaldehyde or in formalin with paraffin embedding and processed for H&E histology and immunostaining. Frozen samples were fixed in 4% paraformaldehyde, processed for histology, and stored frozen in OCT compound (Sakura Finetek USA, Torrance, CA) at -80°C. Both frozen and formalin-fixed paraffin-embedded samples were sectioned at 6 µm thickness and processed for antigen retrieval by heating in a microwave oven for 4 minutes at 100% power, then 10 minutes at 10% power in 10 mM citrate buffer, pH 6.0. Sections were allowed to cool to room temperature for 30 minutes before staining. For TWIST immunostaining, sections were incubated at 4°C with TWIST-specific rabbit polyclonal antibody described above (1:800) with detection of bound primary by the ABC method (Vectastain, Vector Laboratories, Burlingame, CA) using DAB as chromogen. Cell nuclei were visualized by hematoxylin counterstaining. For immunoflourescent detection of TWIST protein in paraffin sections, the same protocol was followed, except that TWIST primary antibody was detected with a donkey CY3-conjugated antirabbit IgG (711-165-152, 1:400; Jackson ImmunoResearch Laboratories, West Grove, PA). Double immunofluorescence was performed by sequential staining for TWIST as above, followed by incubation with mouse monoclonal primary antibodies against either NeuN (MAB377, 1:1000; Chemicon, Temecula, CA), GFAP (G3893, 1:500; Sigma), CD45 (M0701, 1:1000; Dako, Carpentiera, CA), or CD68 (M8014, 1:200; Dako), followed by incubation with FITC-conjugated antimouse IgG secondary antibodies (715-095-151, 1:100; Jackson ImmunoResearch Laboratories). Nuclei were counterstained with DAPI (D-1306; Molecular Probes, Eugene, OR), and coverslips were mounted with fluoromount (SouthernBiotech, Birmingham, AL) and light-protected prior to visualization with fluorescent microscopy (Zeiss Axioplan 2, Carl Zeiss USA, Thornwood, NY). Controls omitting primary antibody or using an isotype IgG were performed on all samples. Digitally captured images (Slidebook software; III) were analyzed for double-labeled TWIST immunopositive cells using Image J software (Scion Corporation) by counting at least four x20 fields from two separate brain samples that included both gray and white matter.

Cell Growth and Saturation Density

For cell growth, 1 x 10⁵ cells in DMEM/F12 supplemented with 10% FBS were plated in 60-mm plates in triplicate. Cells were counted everyday for 4 days using a Coulter counter (Beckman-Coulter, Fullerton, CA). For saturation density assays, 2 x 10⁵ cells were seeded in six-well plates in triplicate. Cells were counted every second day for 7 days.

Matrigel Invasion Assays

The invasion assay was performed using 24-well BD BioCoat Matrigel invasion chambers (BD Biosciences) and QCM cell invasion chambers (Chemicon). Chambers consist of a cell culture insert with an 8-µm pore size membrane coated with a thin layer of BD Matrigel Matrix or Chemicon ECMatrix. SF767 LXSN or SF767Tw cells were harvested, resuspended in a serum-free DMEM, and loaded into the insert (BD Biosciences: 5 x 10⁴ cells/500 µl; Chemicon: 3 x 10⁵ cells/300 µl) according to the manufacturer's recommendations. DMEM/F-12 with 10% FBS was then added to the lower chamber (750 µl for BD Biosciences; 500 µl for Chemicon). Cells were incubated at 37°C for 70 hours (BD Biosciences) or 48 hours (Chemicon) before noninvading cells and matrix gel from the insert were gently removed with a cotton swab. For both assays, invasive cells that penetrated through pores and migrated to the underside of the membrane were stained with the staining solution provided by the manufacturer (Chemicon) for 20 minutes and photographed. For quantitation in both assays, stained cells were dissolved in 200 µl of 10% acetic acid and 100 µl was used for colorimetric reading of OD at 560 nm.

In Vitro p53-Deficient Model of Astrocyte Transformation

Mouse astrocytes were cultured from postnatal day 1 mice from wild-type (p53^+/+), p53 hemizygous (p53^+/-), or p53 homozygous knockout (p53^-/-) animals as previously described [22,29]. Frozen cell pellets of p53^-/- astrocytes and selected samples of p53^+/-, p53^+/+, and p53^-/- astrocytes harvested at varying passage numbers were used for the isolation of total RNA and RT-PCR for TWIST mRNA.

Results

TWIST mRNA and Protein Are Frequently Expressed in Glioma Cell Lines

Examination of rat C6 glioma cells for class B bHLH genes by a degenerate RT-PCR cloning strategy revealed the presence of TWIST mRNA. TWIST expression in C6 was confirmed by Northern blot analysis (Figure 1A), suggesting that TWIST may play a role in glial tumorigenesis. To evaluate the relevance of our finding in human gliomas, we examined TWIST expression by Northern blot analysis in 11 human cell lines, including six gliomas. Five of six glioma lines expressed variable levels of the documented major 1.4-kb TWIST transcript, together with minor transcripts at 2.2 and 3.6 kb (Figure 1A) [30,31]. TWIST message was also detected in human neuroblastoma and medulloblastoma, but not in colon, breast, or SCLC cell lines examined.

Figure 1.

Expression of TWIST mRNA and protein in rodent glioma and human glioblastoma cell lines. (A) Northern blot analysis showing the expression of TWIST in rat C6 glioma cells (far left lane), confirming the results of a degenerate RT-PCR screen for class B bHLH genes. TWIST mRNA expression was detected in five of six human glioma cell lines (A172, T98G, SNB19, UW455, UW18, and SF767), as well as in lines derived from a neuroblastoma (SKNMC) and a medulloblastoma (D283). Lines derived from colon (SW480), breast (MDA453), and small cell lung cancer (H82) had no detectable expression. The gel stained with ethidium bromide (lower panel) demonstrates equivalent loading and integrity of total RNA. (B) Southern hybridization of TWIST-specific probes to RT-PCR products from a larger panel of human tumor cell lines. Blots of GAPDH controls show that comparable amounts of total RNA were analyzed for each sample. Differential levels of TWIST expression were detected in 11 of 12 human gliomas and in two neuoblastomas and one medulloblastoma, whereas no detectable expression was noted in non-neural tumors. TWIST mRNA was not detected by RT-PCR or by Northern blot analysis in the glioblastoma line, UW455, or in the non-neural cell lines examined. (C) Western blot analysis of 30 µg of whole cell lysates of UW455 and SF767, transfected either with an empty expression vector (LXSN) or a TWIST overexpression construct (Tw), demonstrates the presence of a band at the expected molecular mass of 26 kDa. UW455, which lacks detectable TWIST mRNA and SF767LXSN with low basal message levels, shows no detectable protein. (D) Nuclear extracts of the SF767LXSN, SNB19LXSN, and T98G cell lines (50 µg each) demonstrate endogenous expression of a 26-kDa TWIST protein comparable to that detected in nuclear extracts from SF767Tw (10 µg) at levels commensurate with mRNA levels (panel A). The nuclear protein, INI1, is used to compare the levels of nuclear-specific protein loaded from each preparation.

The findings in Northern blot analysis were confirmed and expanded on using RT-PCR to assay TWIST expression in a larger panel of 21 human tumor cell lines including those examined by Northern blot analysis. As shown in Figure 1B, 11 of 12 glioblastoma (i.e., grade IV) cell lines expressed TWIST; RT-PCR confirmed the lack of TWIST message revealed by Northern blot analysis of UW455. TWIST was also detected by RT-PCR in cell lines derived from neuroblastoma and medulloblastoma—primitive neuroectodermal tumors that arise in the peripheral nervous system and CNS, respectively [32,33]. To our knowledge, this is the first demonstration of TWIST expression in medulloblastoma cell lines, whereas expression of TWIST in neuroblastoma has recently been described in association with N-myc amplification [12]. As observed in the Northern blot analysis (Figure 1A), TWIST was not detected in the two colorectal, small cell lung, and breast carcinoma-derived cell lines analyzed by RT-PCR. Our results demonstrate that TWIST mRNA is frequently expressed in human glioblastoma-derived cell lines and indicate that TWIST expression is a hallmark of glial and nonglial tumors of neuroectodermal lineage.

To analyze TWIST protein in cell lines, first we confirmed the specificity of polyclonal anti-TWIST antibody using SF767 cells transduced with TWIST-expressing retrovirus. We chose SF767 cells because they express low basal levels of message by Northern blot analysis (Figure 1A). UW455 cells, which lack detectable TWIST message (Figure 1A), served as a negative control. Figure 1C shows that in 30 µg of whole cell extracts of glioma cells overexpressing TWIST (SF767Tw), the antibody detected a single protein band of approximately 26 kDa. The 26-kDa molecular weight is consistent with the previously reported size for the TWIST protein [30]. Then we tested endogenous levels of TWIST expression in glioblastoma cell lines using cell nuclear extracts. Because endogenous levels of TWIST protein are low, we reduced the amount of nuclear proteins from positive control cells (SF767Tw) loaded on the gel to 10 µg, whereas 50 µg of nuclear extracts from other cell lines was loaded alongside (Figure 1D). Levels of endogenous TWIST protein of the same molecular weight were detected and correlated well with Northern blot analysis of RNA from the same cell lines. Therefore, we conclude that the polyclonal TWIST antibody interacts specifically with TWIST protein, and that TWIST protein present in human glioma cell lines reflects steady-state levels of TWIST mRNA.

TWIST mRNA Content Is Associated with Grade in Human Gliomas

We examined 10 grade II, 6 grade III, and 12 grade IV gliomas of differing histologies to determine if TWIST mRNA is expressed in vivo in human gliomas. Nineteen tumors were newly operated, whereas nine had recurred after either previous surgery (n = 1), surgery followed by radiotherapy (n = 3), or surgery followed by radiotherapy and chemotherapy (n = 5). Using semiquantitative RT-PCR, TWIST mRNA was detected in 89% (25 of 28) of gliomas (Figure 2, A and B). TWIST abundance did not differ discernibly between newly operated and recurrent tumors (Figure 2A). As shown in Figure 2, B and C, mean TWIST transcript band intensity, expressed as a percentage of that for GAPDH, was approximately three-fold greater in grade IV than in grade II tumors (29.6 ± 24.5 vs 10.0 ± 8.9; P = .015) and in combined grade II and III tumors (10.5 ± 7.4; P = .01). Mean normalized TWIST expression in grade III gliomas (11.2 ± 4.5) was markedly less than grade IV, but the difference was not statistically significant (P = .09). These results indicate that overall TWIST mRNA levels reflect degree of malignancy (Figure 2C), a conclusion supported by the observation that the three tumors with no detectable TWIST were grade II. Nonetheless, TWIST levels varied considerably between and within diagnostic subtypes, as evidenced by the ∼70-fold range of normalized TWIST content for grade IV (Figure 2B). Our data indicate that TWIST mRNA expression commonly accompanies glial tumorigenesis, and that higher expression is associated with grade IV tumors.

Figure 2.

TWIST expression in human gliomas. The abundance of mRNA was analyzed by Southern hybridization of TWIST-and GAPDH-specific probes to RT-PCR products. (A) Glial tumors, grouped by grade, include grade II fibrillary astrocytomas (1–3), oligodendroglioma (4–9), and mixed oligoastrocytoma (10); grade III anaplastic mixed oligoastrocytomas (11–13) and anaplastic astrocytoma (14–16); and grade IV glioblastoma (17–26) and gliosarcoma (27 and 28). The nine tumors recurrent after prior treatment are indicated by a caret beneath the corresponding lane. No statistical difference in gene expression levels between primary or recurrent samples was noted (data not shown). Representative RT-PCR controls without RT or cDNA are shown at the far right. (B) Histogram of relative TWIST expression for gliomas, grouped by grade. The signal density of TWIST is expressed as a percentage of the corresponding GAPDH signal. (C) Mean relative TWIST expression with standard errors of the mean by glioma grade. Significant differences (P ≤ .05) indicated by asterisks were demonstrated between grades II and IV (**) and combined grade II plus grade III and IV tumors (*).

TWIST Expression Is Localized to Tumor Cells

To elucidate the intratumoral localization of TWIST mRNA in gliomas, we examined expression by in situ hybridization in two grade II, one grade III, and three grade IV gliomas; representative results are shown in Figure 3. Importantly, TWIST-specific hybridization reflected TWIST mRNA content detected by RT-PCR as evidenced by greater hybridization associated with increasing tumor grade. Thus, a low-grade glioma with barely detectable TWIST expression by RT-PCR (sample 7; Figure 2A) showed only faintly discernible hybridization (Figure 3E), whereas two grade IV gliomas with high levels of TWIST mRNA (samples 24 and 28 in Figure 2A) showed intense hybridization in a majority of cells (Figure 3, F and G). Notably, the grade IV tumors displayed in situ hybridization signal intensity comparable to that of rhabdomyosarcoma (Figure 3H), a highly malignant mesenchymal neoplasm known to overexpress TWIST [10]. Although the overall intensity of TWIST hybridization corresponded to tumor grade, hybridization intensity varied markedly among cells within a glioma (Figure 3, E–H). The variation in TWIST message abundance in these samples was corroborated by immunostaining of the same specimens (Figure 3, I–L). Although only sparse immunoreactivity was detected in the low-grade tumor (Figure 3I), markedly elevated TWIST immunoreactivity was demonstrated in tumor nuclei of samples from the grade IV gliomas (Figure 3, J and K) and rhabdomyosarcomas (Figure 3L).

Figure 3.

TWIST expression patterns in low-grade and high-grade gliomas by in situ hybridization and immunostaining. H&E histology (A–D), TWIST in situ hybridization (E–H), and immunostaining (I–L) are shown from adjacent sections of sample 7, a grade II glioma with low levels of TWIST by RT-PCR (A, E, and I); sample 24, a grade IV glioblastoma (B, F, and J); sample 28, a grade IV gliosarcoma (C, G, and K) with strong TWIST expression; and a rhabdomyosarcoma (D, H, and L), previously reported to express TWIST [10]. H&E and in situ hybridization photomicrographs were taken with x20 objective, whereas iummunostained images were taken with a x40 objective. Sense control hybridization reactions for each sample are shown as insets in the right lower hand corner (E–H). In situ hybridization sections were counterstained with methyl green to facilitate visualization of tissue architecture. Immunostaining demonstrates occasional faint staining in the low-grade sample (I), whereas abundant pronounced nuclear staining is present in all of the high-grade tumors (J–L). Nuclear staining is not apparent in isotype IHC controls (lower right corner insets), indicating the specificity of the antibody. (The 50-µm scale bar in (A) applies to (A)–(H), and the 25-µm scale bar in (I) applies to (I)–(L).)

TWIST Expression and Malignant Progression

The increased levels of mRNA expression in grade IV tumors suggest that elevation of TWIST accompanies the malignant progression of gliomas. To further characterize the association of TWIST with malignant progression in gliomas, we examined changes in mRNA and protein content that accompanied the recurrence of a low-grade oligodendroglioma as a grade IV glioblastoma in the same patient. As shown in Figure 4, TWIST protein content detected by immunostaining (Figure 4, A and B) and TWIST mRNA abundance (Figure 4D) was markedly increased in the high-grade recurrence. Importantly, the increase in TWIST protein content colocalized with malignant histologic hallmarks (increased cellularity, nuclear and cellular atypia, endothelial proliferation, and necrosis) in the recurrent tumor (Figure 4C). Areas histologically similar to the original low-grade glioma either lacked or exhibited low levels of TWIST protein (data not shown). This correlation between the presence of TWIST and anaplastic histologic features suggests that, in this case, elevated levels of TWIST protein expression accompanied malignant glial progression. This observation supports a role for TWIST in malignant progression.

Figure 4.

Increased TWIST protein and mRNA content accompanies malignant progression. Immunostaining revealed increased abundance of TWIST protein accompanying the recurrence of a grade II oligodendroglioma (A) to a grade IV glioblastoma (B). The recurrent tumor demonstrates increased cellularity characteristic of glioblastoma, together with pronounced TWIST immunopositivity in tumor cell nuclei. High levels of TWIST protein (C) were particularly evident in areas of the recurrent tumor with malignant histologic features of increased cellularity, necrosis (asterisk), and microvascular proliferation (arrow). In accord with the elevated level of TWIST protein evident by immunostaining, the recurrent sample demonstrates increased abundance of TWIST mRNA by RT-PCR (D). (The 40-µm scale bar in (A) applies to (A)–(C).)

Elevation of TWIST mRNA Precedes In Vitro Transformation of Mouse Astrocytes

To gain additional insights into the association of TWIST with glial tumorigenesis, we analyzed its expression in an in vitro model of astrocyte transformation. Cultured neonatal mouse astrocytes hemizygous for p53 spontaneously undergo a well-defined sequence of genetic and physiological changes culminating in malignant transformation [22–34]. We examined by RT-PCR the abundance of TWIST mRNA in serially passaged p53^+/- mouse astrocytes derived from postnatal day 1 animals. As shown in Figure 5, TWIST mRNA content normalized to GAPDH was approximately 10- and 7-fold greater in p53^+/- than in p53^+/+ cells at passages 1 and 4, respectively. By passage 8, when these cells are known to lose the remaining wild-type copy of p53 and develop aneuploidy [22], TWIST expression increased an additional 30%. Similar changes in TWIST expression were observed in homozygous p53^-/- astrocytes (data not shown). TWIST content remained elevated through passage 50, at which time the cells displayed serum-independent growth and formed tumors in vivo [22]. In contrast, TWIST abundance was unchanged in wild-type astrocytes through passage 4, the approximate limit of their proliferative potential in vitro, and was comparable to the level in adult wild-type mouse brain. Wild-type astrocytes senesce by passage 6 or 7 and additional comparison of TWIST levels with p53-deficient astrocytes at subsequent serial passage numbers are not possible. Our data suggest that elevated TWIST expression accompanies loss of p53 function and the resulting increase in genomic instability associated with malignant transformation.

Figure 5.

TWIST mRNA abundance is elevated by p53 deficiency in mouse astrocytes. (A) Comparison of TWIST mRNA detected by Southern hybridization of RT-PCR reaction products from cultured p53 wild-type (+/+) and hemizygous (+/-) astrocyte cultures derived from neonatal mice. Also shown are levels in normal adult mouse brain (NAB) and a control lacking RT. (B) Density of TWIST mRNA signal normalized to that of GAPDH of samples shown in (A). Values are the mean ± SD of three independent determinations. The appearance of features accompanying malignant transformation at specific passage numbers is indicated as follows: DNA ploidy abnormalities (*), serum-independent growth, and tumor formation in nude mice (^). The marked increase of TWIST expression in p53^+/- astrocytes relative to normal brain and wild-type astrocytes at passages 1 and 4 is maintained through serial passages coinciding with progressive acquisition of malignant phenotype in vitro, including complete malignant transformation (growth of solid tumors in vivo) by passage 50.

Overexpression of TWIST Promotes Glioma Cell Invasion

To delineate potential roles for TWIST in glioma biology, we examined the effect of TWIST overexpression in SF767 (SF767Tw) on cell proliferation, saturation density, and invasion through an artificial basement membrane matrix. As shown in Figure 6A, SF767Tw cells displayed slightly slower growth rates that led to reduced saturation density (Figure 6B) compared to SF767 containing vector alone. Interestingly, suppression of TWIST expression in 4T1 mouse mammary carcinoma cells resulted in a small increase in growth rate [15]. TWIST overexpression has been reported to alter cell morphology in MDCK kidney epithelial cells, immortalized human mammary epithelial cells [15], and osteosarcoma cells [30], but similar morphologic changes in TWIST-overexpressing SF767 cells were not observed (data not shown). The report that TWIST expression is necessary for mammary carcinoma metastasis in rodents and is elevated in invasive human breast-carcinomas led us to hypothesize that TWIST promotes invasion by human gliomas. In accord with our hypothesis, SF767Tw displayed greater infiltration through two different Matrigel extracellular matrices than control cells (Figure 6, C and D). These findings suggest that, by promoting invasiveness, TWIST contributes to a hallmark characteristic of human gliomas.

Figure 6.

Overexpression of TWIST promotes invasion but not cell proliferation in SF767 glioma cells. (A) The impact of TWIST overexpression on cell growth rate in culture was examined by plating 1 x 10⁵ SF767Tw or SF767 LXSN vector control cells in 60-mm plates in triplicate. Cells were trypsinized and counted each day for 4 days using a Coulter counter, and the mean of three wells from each time point is shown. (B) TWIST effects on cell saturation density were assayed by plating 2 x 10⁵ SF767Tw or SF767 LXSN vector control cells in six-well plates in triplicate. Triplicate wells were trypsinized and counted every second day for 7 days using a Coulter counter and plotted. Calculated standard errors were small and cannot be visualized at this scale. (C, D) Invasion through two extracellular matrices is increased by overexpression of TWIST in SF767 assayed by two separate invasion assays (C, Chemicon; D, BD Biosciences). Data were analyzed after 70 hours for BD Biosciences and 48 hours for the Chemicon assay. Photomicrographs of stained cells on the undersurface of matrix-coated 8-µm pore size filter demonstrate differential invasion of SF767 LXSN and SF767 TW cells. Cells from triplicate assays were stained and quantified by colorimetric reading at OD of 560 nm. Invasion by SF767TW was calculated as a percentage of the SF767 LXSN vector control values. Data shown are mean ± SE.

TWIST Is Expressed in Normal Fetal and Adult Human Brain

Expression of TWIST in developing or adult mammalian brain has not been unambiguously delineated [31,35]. To better clarify the potential role of TWIST expression in gliomagenesis, we examined TWIST mRNA content in human developing and mature brain by RT-PCR. As shown in Figure 7A, abundant levels of TWIST were detected in all samples of embryonic and fetal brain that ranged from day 52/53 to day 132 postconception, suggesting that TWIST participates early in the development of the human brain. Notably, the level of expression in developing brain was comparable to that in mesenchyme-derived human embryonic bone (Figure 7A). TWIST was also detected by RT-PCR at levels comparable to that of adult temporalis muscle in seven non-neoplastic brain samples obtained from adults operated either for epilepsy (four males and one female, 36–46 years old) or benign skull base tumors (two males, 32 and 48 years old). TWIST message was detected by in situ hybridization (data not shown) in adult muscle in satellite cells, consistent with previous observations that TWIST is expressed in adult myoblasts [36]. In brain samples that were macroscopically dissected from fresh brain tissue, TWIST expression was exclusively detected in the gray matter in three of four specimens with only trace expression in one of four white matter specimens. Representative results are shown in Figure 7A. In situ hybridization performed on two brain specimens corroborated these findings and revealed pronounced staining of cells with morphologic characteristics of neurons (Figure 7, B and C), with rare staining evident in white matter, which did not localize to morphologically identifiable cell types (Figure 7, F and G). Immunostaining of normal gray matter (Figure 7D) and white matter (Figure 7H) or gray-white junction (Figure 7, E and I) samples further corroborated the findings of RT-PCR and in situ hybridization, which suggested that in normal brain, TWIST protein and mRNA are almost exclusively localized to neurons in gray matter.

Figure 7.

Expression of TWIST in human developing and adult brain. (A) Southern hybridization of RT-PCR reactions from embryonic, fetal, and adult tissues including control bone tissue (79 days) known to express TWIST and a panel of embryonic and fetal brain (Br) samples from 52/3 to 132 days showing TWIST expression throughout this developmental time frame. TWIST expression in temporalis muscle (Ad Muscle) and rhabdomyosarcoma (Rhabdo) is consistent with the known expression of TWIST in adult muscle progenitor cells [36] and the mesenchymal-derived tumor [10]. TWIST is differentially expressed in adult brain, with expression detected in gray matter (GM) but at low or undetectable levels in white matter (WM). (B and F) H&E histologic sections (x20) of normal gray matter (B) and white matter (F) are compared with ISH (x20) for TWIST expression from adjacent sections of gray (C) and white matter (G). Insets in the right lower corner (C and G) are sense controls. TWIST IHC (x20) for normal gray matter (D) and white (H) matter corresponds to expression patterns by ISH. Inset: x40 images in (D) and (H) demonstrate nuclear staining pattern in cells with neuronal morphology. (E and I) The demarcation of TWIST protein expression between normal gray and white matter is demonstrated in low-power (x10) photomicrographs of a normal brain sample with a diagonally oriented junction between gray (upper right) and white matter (lower left) by H&E histology (E) and TWIST IHC (I). These data indicate that TWIST message and protein are almost exclusively detected in gray matter. (The 50-µm scale bar in (B) applies to (B)–(D) and (F)–(H); the 25-µm scale bar in the (D) inset applies to insets of (D) and (H); and the 100-µm scale bar in (E) applies to (E) and (I).)

TWIST Expression in Adult Brain Localizes to Neurons

The identity of TWIST-expressing cells in the normal brain is of central importance to establishing the importance of TWIST in human gliomagenesis. To confirm the results of in situ hybridization, which suggested that TWIST mRNA expression is confined to neurons, we performed double-label immunofluorescent staining experiments using two different normal brain samples and found that of 315 TWIST-expressing cells, 305 (or 97%) coexpressed the neuronal marker NeuN (Figure 8, A–C). Conversely, we were not able to unambiguously identify any TWISTexpressing cells that coexpressed GFAP (Figure 8, D–F), indicating that TWIST protein expression is almost exclusively restricted to neuronal cells in the non-neoplastic human brain. TWIST protein was not detectable in white matter (Figures 7, H and I and 8F), in agreement with in situ hybridization and RT-PCR results. Our analysis also indicated that TWIST is not expressed in lymphocytes or microglia (data not shown). To our knowledge, these results are the first evidence of expression of TWIST mRNA and protein in human developing brain and mature neurons. Our findings suggest that TWIST participates in the development of the CNS, and may have as yet undetermined roles in the biology of mature neurons. The restriction of expression to neurons indicates that TWIST in human glioma cells is a consequence of oncogenic transformation, rather than persistent expression of a gene present in normal glia.

Figure 8.

TWIST is expressed almost exclusively in neurons of the adult brain. (A–C) Photomicrographs of indirect immunofluorescent staining with TWIST antibody (Cy3) of normal human gray matter (A), the neuronal marker NeuN (FITC) (B), and the merged image (C) show colocalization of TWIST with NeuN in almost every TWIST immunoreactive cell (arrow). Sections were also stained with DAPI and indicate that NeuN-negative cells do not express TWIST. (D–F) A section of normal human brain containing gray and white matter stained for TWIST (Cy3) (D) and GFAP (FITC) (E), and the merged image with DAPI nuclear stain (F) demonstrate that TWIST is found exclusively in gray matter and that TWIST-expressing nuclei (arrowhead) do not colocalize with GFAP-immunopositive cells (arrow). The junction between gray and white matter is indicated by the dotted line. These results demonstrate that TWIST is almost exclusively found in neurons. (The 50-µm scale bar in (A) applies to (A)–(F).)

Discussion

Primary brain tumors remain among the most lethal of human cancers despite more than three decades of intensive effort to develop more effective surgical techniques, and radiation-based and chemotherapy-based adjuvant therapies [37,38]. The persistent dismal prognosis for the most common brain malignancies in adults and children has stimulated the search for novel therapies. The recognition that neural precursor cells in the adult brain are the likely targets for malignant transformation [39–41] has stimulated investigation of the recapitulation of developmental gene expression during glial tumorigenesis in the hope of uncovering new targets for therapeutic intervention. Here we show that human gliomas frequently express TWIST, an essential regulator of mesodermal development [42]. Expression was observed in all major glioma diagnoses as well as in medulloblastoma-derived and neuroblastoma-derived cell lines, indicating that TWIST participates in the genesis of a wide range of neuroectodermal tumors. Expression was also observed in embryonic and fetal brain and, unexpectedly, in mature neurons, suggesting that TWIST promotes gliomagenesis by the aberrant activation of a previously unrecognized pathway(s) normally involved in the development and function of the human CNS. Importantly, the demonstration that TWIST promotes glioma cell invasion supports functions for TWIST in human gliomas similar to those demonstrated for other non-neural cancers [15].

Current evidence indicates that TWIST functions as an oncogene by enhancing cell survival through inhibition of p53-mediated apoptosis [10], preventing growth arrest that accompanies developmental maturation [36,43–46] and promotion of tumor cell invasiveness [15]. The ability of TWIST to foster cell survival would enhance tumor progression by increasing resistance to cytotoxic therapies as well as endogenous cytotoxic stresses such as reactive oxygen free radicals generated in association with increased cellular proliferation [47,48]. Indeed, elevated levels of TWIST in other transformed or normal cells are associated with resistance or reduced sensitivity to cytotoxic drugs [14,35]. The antiapoptotic function of TWIST may be a primary defense against endogenous insults in the sizeable minority of gliomas that possess functional p53 and may also provide a secondary or complementary mechanism that promotes survival in the majority of gliomas in which p53 function is suppressed or inactivated [49–51]. In accord with this hypothesis is the observation that coexpression of c-myc and TWIST inhibits p53-mediated apoptosis and promotes transformation in embryonic mouse fibroblasts [10].

Recently, it has been shown in neuroblastoma that oncogenic cooperation between TWIST and N-myc promotes malignant transformation and cell survival in response to cytotoxic stress by inhibiting p53 function [12]. In genetically engineered mice, c-myc overexpression targeted to GFAP-expressing cells produces tumors analogous to human glioblastoma [52]. Notably, grade III and IV human gliomas demonstrate marked elevations of c-myc expression relative to normal brain and low-grade tumors [53]. Thus, these data and our demonstration of TWIST mRNA and protein in human gliomas suggest that, similar to neuroblastoma, oncogenic cooperation between TWIST and other oncogenes such as c-myc may contribute to glioma treatment resistance by promoting cell survival, in part, through abrogation of p53-mediated apoptosis.

Another function of TWIST relevant to human gliomas is promotion of EMT. In epithelial-derived solid tumors, EMT is a critical component of the metastatic cascade whereby transformed epithelial cells lose normal cell-cell interaction and adopt a mesenchymal phenotype characterized by invasion through basement membrane [54]. TWIST expression has been associated with EMT in gastric carcinoma [13] and is essential in mouse mammary carcinoma models for EMT and metastasis, in part, by promoting cell motility [15]. Of note, elevated levels of TWIST are found in the most invasive types of breast carcinoma cell lines and tumors [15]. In the present study, the highest levels of TWIST expression were present in grade IV tumors and overexpression of TWIST promoted glioma invasion in vitro. Tumor cell migration and invasion into surrounding brain are two features of human gliomas that increase with tumor grade. Thus, the promotion of invasion appears to be a function of TWIST shared by both breast and glial cancers. The limitations to surgical resection and effective delivery of systemic and local therapies posed by the diffusely invasive behavior of glioma tumor cells underscore the importance of characterizing TWIST function in primary brain tumors.

Although direct evidence is currently lacking, TWIST function may also be linked to the activation of receptor tyrosine kinases (RTKs) overexpressed in human gliomas that control cell survival and invasion. In other cell types, TWIST expression is elevated by the RTKs IGFR1 and c-met, whose ligands are IGF-1 and HGF, respectively [35,36]. Antisense-mediated suppression of TWIST reduces the ability of IGF-1 to rescue NIH 3T3 fibroblasts from etoposide-mediated cell death [35]. In human gliomas, both RTKs are overexpressed and their signaling pathways are activated, in direct correlation with tumor grade [35,55–58]. Importantly, activation of the IGF-1/IGFR1 and HGF/c-met signaling pathways promotes survival of glioma cells exposed to cytotoxic agents [58–60] and enhances tumor cell migration and invasiveness [61–64]. Of particular interest, IGF-1/IGFR1 and HGF/c-met signaling activates AKT [58,65], which mediates increased resistance to apoptosis demonstrated by migratory glioma cells [66]. These observations and the demonstrated role for TWIST in regulating glioma cell invasion and survival in other cancers warrant further investigation of TWIST interactions with RTK signaling and activation of AKT.

Our examination of embryonic, fetal, and mature human brain, the most comprehensive to date, unambiguously indicates that TWIST expression is associated with early development of the human CNS. Conceivably, the antiapoptotic and invasion-promoting activities of TWIST facilitate the migration and expansion of neural progenitor populations during CNS development. We note that our findings stand in contrast to an earlier report of an absence of TWIST in human fetal astrocytes and adult human brain by Northern blot analysis [31]. Our analyses, however, support the expression of TWIST in fetal brain, and in neurons in adult brain. The low levels of TWIST expression detected in mouse neonatal astrocytes and adult human white matter, and the relative differences in sensitivity between Northern blot analysis and RT-PCR analysis may account for these differences. TWIST promotes resistance to p53-mediated apoptosis and cell proliferation, and inhibits differentiation in experimental models [10,46]. Accordingly, in normal tissues, its expression has predominately been localized to proliferative and immature cells of mesodermal origin [67–69], although expression in some mature cell types has been described [31,67]. Thus, finding TWIST mRNA in mature neurons was unexpected. It is conceivable that the antiapoptotic function of TWIST operates in quiescent neurons as well as in proliferating cells. Neuronal activity is associated with high rates of oxygen utilization that elevate the rate of endogenous generation of reactive oxygen free radicals. TWIST would likely promote neuronal survival by inhibiting the induction of apoptosis by these damaging agents.

Our results, together with previous findings in other human tumors, suggest the following model in which TWIST, by promoting survival, malignant progression, and invasion, facilitates glial tumorigenesis and glioma clinical behavior. TWIST in low-grade gliomas may promote resistance to endogenous and exogenous cytotoxic stress that would otherwise lead to apoptotic cell death. By promoting survival, TWIST could facilitate the accumulation and propagation of mutations that are associated with glial tumor progression. TWIST could then interact synergistically with mutation-induced oncogene activation, in similar fashion to the oncogenic cooperation observed between N-myc and TWIST in neuroblastoma [12], to promote and maintain a more malignant phenotype. Candidate target oncogenes for interaction with TWIST include RTKs as well as c-myc. This scenario outlines the potential functions of TWIST related to the genesis of low-grade gliomas and their malignant progression to “secondary” glioblastomas. In the case of “primary” glioblastoma, which infrequently demonstrates p53 mutations [70], TWIST may function alone or in combination with other oncogenes to promote tumor growth and treatment resistance, in part by inhibiting p53-mediated apoptosis. Importantly, TWIST may promote glioma cell invasion of surrounding normal brain, a major factor contributing to the dismal prognosis for glial tumors. The mechanisms by which TWIST contributes to gliomagenesis and glioma physiology and normal CNS development remain to be more fully characterized. Elucidating the pathways that elicit aberrant TWIST expression in gliomas, as well as characterizing the effectors of TWIST activity, are likely to provide new insights into the genesis of gliomas and to identify new targets for therapeutic intervention.

Abbreviations

bHLH

basic helix-loop-helix

CNS

central nervous system

FBS

fetal bovine serum

GBM

glioblastoma

GS

gliosarcoma

H&E

hematoxylin and eosin

HGF

hepatocyte growth factor

IGF

insulin-like growth factor