|Home | About | Journals | Submit | Contact Us | Français|
Medulloblastoma (MB) is the most common pediatric brain cancer. Several important developmental pathways have been implicated in MB formation, but fewer therapeutic targets have been identified. To locate frequently overexpressed genes, we performed a comprehensive gene expression survey of MB. Our comparison of 20 primary tumors to normal cerebellum identified neuronatin (NNAT) as the most frequently overexpressed gene in our analysis. NNAT is a neural-specific developmental gene with α and β splice forms. Functional evaluation revealed that RNA interference knockdown of NNAT causes a significant decrease in proliferation. Conversely, coexpression of both splice forms in NNAT-negative MB cell lines increased proliferation, caused a significant shift from G1 to G2/M, and increased soft agar colony formation and size. When expressed individually, each NNAT splice form had much less effect on these in vitro oncogenic predictors. In an in vivo model, the coexpression of both splice forms conferred the ability of xenograft formation to human MB cells that do not normally form xenografts, whereas a control gene had no effect. Our findings suggest that the frequently observed over-expression of both NNAT splice forms in MB enhances growth in this cancer.
Medulloblastomas (MBs) are the most common pediatric solid tumor malignancy; however, the current therapeutic regimen involves radiation therapy, which can damage the developing brain. Therapies that target the genes involved in tumor formation might provide the means to enhance therapy for MBs without radiation, and perhaps increase survival while decreasing the debilitating side effects of treatment. Genetic alterations known to give rise to MBs include MYC and MYCN amplification, as well as PTCH mutations,1 which occur in 5%–15% of patients. Recently, genomic amplification of NOTCH2 and OTX2 in MBs has been reported.2,3 Despite these very important advances in understanding the pathological basis of MB growth, the known oncogenes are generally expressed in only a small percentage of MBs. This presents difficulties in developing a molecular target for most MBs. To identify commonly overexpressed MB genes and to gain insight into MB growth, we performed large-scale gene expression analyses and functionally evaluated candidate MB-related genes. Evidence, by in vitro functional assays and an in vivo model, supports our main hypothesis presented here that neuronatin (NNAT) stimulates malignant growth when both of its splice forms are expressed in MB.
Two snap-frozen MBs were purchased from the Duke University Brain Tumor Bank. Two autopsy specimens of normal cerebella from a 78-year-old male and a 21-year-old female with diabetes mellitus were obtained from the Brain Bank of the Alzheimer Research Center at Duke University. Genomic DNA from MBs was isolated by CsCl centrifugation or the DNAeasy kit (Qiagen), Valencia, CA, USA. Total RNA was isolated by CsCl centrifugation or the RNAgents Total RNA Isolation kit (Promega, Madison, WI, USA). The MB cell lines used in this study were D283Med, D341Med, D425Med, D581Med, D721Med, Med-mhh1, MCD-1, and UW228.
Serial Analysis of Gene Expression (SAGE) and Micro-SAGE were performed as previously described.4,5 SAGE was performed using 1–100 μg of total RNA from two MBs, two normal cerebella, and four MB cell lines (D341Med, Med-mhh1, UW228-pLAPSN, and UW228-NNATα+β) to obtain transcript profiles for various comparisons using a p-value, 0.01. An additional 18 MBs were made using the same procedure and have been previously reported by our laboratory.6 The various comparisons of the transcript profiles for these SAGE libraries were performed using either the SAGE2000 software, version 4.5 (available upon request from the B. Vogelstein/K. Kinzler lab, Johns Hopkins University) or the Digital Gene Expression Displayer tool on the National Cancer Institute’s SAGE Genie Web site (http://cgap.nci.nih.gov/SAGE). All expression profiles are posted and can be downloaded from SAGE Genie.
The following dicer RNA interference (RNAi) duplexes were obtained from Integrated DNA Technologies (Coralville, IA, USA): HSS.RNAIN005386.3.3, which targets the NNATα splice form only; ACC U31767 and HSC.RNAI.N005386.3.5, which target both NNATα and NNATβ (NNATα+β) splice forms; and a randomized small interfering RNA (siRNA) control. One million D283Med cells were electroporated at 200 V for 25 ms (GenePulser, Bio-Rad, Hercules, CA, USA) with 5 μg of RNAi duplex.
A rabbit polyclonal antibody was designed against the C-terminus of human NNAT (Zymed/Invitrogen, Carlsbad, CA, USA), used to stain 5-μm sections of formalin-fixed, paraffin-embedded MBs, normal cerebella, and normal white matter at a dilution of 1:100, and visualized by the chromagen diaminobenzidine.8
Cytoplasmic extracts were made using the NE-PER Nuclear and Cytoplasmic Extraction Reagent Kit (Pierce, Rockford, IL, USA) following the manufacturer’s instructions. Twenty micrograms of protein were heated for 5 min with 2.5% β-mercaptoethanol/lithium dodecyl sulfate sample buffer (Invitrogen, Carlsbad, CA, USA) and loaded onto a 4%–12% NuPAGE Bis-Tris gel (Invitrogen). The gel was transferred to a polyvinyl difluoride membrane and probed according to standard procedures using the anti-NNAT antibody described above at a dilution of 1:400. Signals were visualized by the Pico SuperSignal chemiluminescent system (Pierce).
The two splice forms of NNAT, α (Mammalian Gene Collection clone; National Institutes of Health, http://mgc.nci.nih.gov/) and β (cloned by PCR), were each cloned into the retroviral vector pLNCX2 (Clontech, Mountain View, CA, USA). Briefly, both constructs and a control vector containing alkaline phosphatase (pLAPSN) were transfected separately into the ecotropic packaging cell line QNX. Supernatants from these cultures were then used to infect the amphotrophic packaging cell line PT-67. Subsequent supernatants were collected from PT-67+NNATα, PT-67+NNATβ, and PT-67+pLAPSN to infect MCD-1 and UW228 cells, resulting in four daughter cell lines (pLAPSN, NNATα, NNATβ, and NNATα+β) for each cell line. All infected cultures of PT-67, MCD-1, and UW228 cells were subsequently placed under selection by G418 antibiotic (1 μg/ml) to select stable virus producers. Expression of NNAT transcript and protein was confirmed by reverse transcriptase (RT)-PCR and Western blotting, respectively (data not shown).
To determine the effects of the different splice forms on the cell cycle, flow cytometry was performed. For each of the four daughter cell lines of MCD-1 and UW228 (described above), 1 × 106 cells were counted, washed twice with phosphate-buffered saline (PBS), fixed overnight in 70% ethanol at 4°C, stained with 1 ml propidium iodide (PI) mix (5 μg/ml PI in PBS, 20 μg/ml RNase A, 0.1% Triton X-100),9 and then analyzed by flow cytometry. For continuous monitoring of D283Med cells postelectroporation, alamarBlue (Invitrogen) was used according to the manufacturer’s instructions. Proliferation rates for the various infections were analyzed by Cell Counting Kit 8 (CCK-8; Alexis Biochemical Corp., Lausen, Switzerland) according to the manufacturer’s instructions.
Anchorage-independent growth was assayed by plating 1 × 103 cells/ml of each of the daughter cell lines of MCD-1 and UW228 in 0.4% agar on a 0.8% agar base. Six weeks after the initial seeding, colonies were counted using a scored eyepiece.
Four million UW228-pLAPSN, UW228-NNATα+β, MCD1-pLAPSN, or MCD1-NNATα+β cells were injected in an equal volume of reduced-growth-factor Matrigel (BD Biosciences, San Jose, CA, USA) in a total volume of 200 μl subcutaneously in the flanks of 8-week-old female athymic nude mice (National Cancer Institute, Frederick, MD, USA). Eight weeks after injection, animals were sacrificed and any tumorlike tissue at the injection site was collected, fixed in 10% formalin for 24 h followed by 70% ethanol, and embedded in paraffin. Samples were sectioned at 5 μm, stained with hematoxylin and eosin per standard histopathological technique, and analyzed by a neuropathologist (C.G.E.).
To identify pathways potentially regulated by NNAT overexpression, we used two different programs, Web-Gestalt10 (Vanderbilt University; http://bioinfo.vanderbilt.edu/webgestalt) and Bibliosphere Pathway Analysis (Genomatix, Munich, Germany) to analyze the SAGE data. Both programs organize genes based on Gene Ontology categories, Biocarta pathways, or Kyoto Encyclopedia of Genes and Genomes (KEGG) biochemical pathways. The WebGestalt program performs a ratio of enrichment (R) calculation as well as the p-value calculation associated with this ratio when the input list is compared to a reference human set. The version of the Bibliosphere Pathway Analysis program we used identifies pathways and networks of pathways by literature mining.
To find new genes or pathways activated in MB pathogenesis, we first compared the SAGE4 transcript profiles of two MBs to those of normal cerebella, the site of origin for MBs. This revealed that these MBs express the neural developmental gene NNAT at very high levels. NNAT had been previously identified as up-regulated in MBs by subtractive hybridization screening as well.11 We then compared the SAGE transcript profiles from an additional 18 MBs to those of the two cerebella and found that only NNAT transcription was up-regulated more than 10-fold in a majority (90%) of MBs, based on Bayesian statistics, with a p-value of 0.01. In this comparison, NNAT was the gene most induced compared to the normal controls. The NNAT transcript counts are shown in Table 1. Seventeen of 20 MBs had expression of NNAT greater than 1 in 10,000 transcripts, indicating that NNAT overexpression in MBs is a common event.
To confirm our transcript expression analysis results in a larger panel of MBs, we performed conventional PCR (Fig. 1A) and real-time PCR (data not shown) to determine the frequency and level of NNATα+β overexpression in MBs. Both RT-PCR and real-time PCR of the MB panel and the 20 SAGE libraries showed that both NNAT splice forms are overexpressed at the RNA level in 90% (36 of 40) of MBs. We also confirmed NNAT protein expression in MBs by immunostaining MB sections and tissue arrays and found that 80% (57 of 72) of MBs were immunopositive for NNAT; although most of the immunoreactivity is cytoplasmic or membranous, in some tumor cells nuclear staining is present (Fig. 1B).
The PCR results indicated that the NNATα splice form constituted the majority of NNAT expression in the tumors. In order to ascertain whether NNATα was preferentially reexpressed in MBs, we performed semi-quantitative PCR. We found that the higher the total NNAT expression as measured by SAGE transcripts, the more NNATα is proportionally overexpressed (Fig. 1C).
Since NNAT, a neural development gene, was over-expressed in a high percentage of MBs at both the transcript and protein level, we next investigated whether it had a functional role in this cancer. To initially define the role of NNAT in MBs, we performed RNAi-mediated knockdown of either the NNATα splice form only or NNATα+ β splice forms (Fig. 2), both of which resulted in a significant decrease (40%–50%) in growth compared to the scrambled RNAi negative control at 48 h after electroporation.
We then sought to determine whether NNATα and/or NNATβ was capable of enhancing tumorigenicity when expressed in MB. We chose to use MB cell lines instead of cell lines such as NIH3T3 as the in vitro model because we were going to perform gene expression analysis of these cell lines and believed that in order to interpret the results in the context of MB biology, it would be more appropriate to use MB cell lines. Additionally, the MB cell lines we used do not normally form in vivo xenografts, which meant that tumor formation might indicate tumor-promoting ability of NNAT. We created cell lines from parental MB cell lines MCD-1 and UW228, both of which do not express NNAT. Lines were made that express (1) NNATα+β, (2) each splice form individually, and (3) pLAPSN (an alkaline phosphatase control gene). PI staining and flow cytometric analysis of these modified cell lines showed that NNATα or NNATβ individually caused a small shift from G1 to G2/M, increasing the G2/M. However, the presence of NNATa1b caused a larger shift from G1 to G2/M of approximately 2-fold compared to control (Fig. 3A). In vitro growth of NNATα, NNATβ, NNATα+β, and control was measured over 4 days. Again, the presence of NNATα+β caused the largest increase in the rate of proliferation (~2-fold that of control), whereas either NNATα or NNATβ alone exhibited only a modest increase in proliferation compared to the control pLAPSN (Fig. 3B). Similar effects of increased proliferation with both NNAT splice forms were observed in both the MCD-1 and UW228 background cell lines.
We then assayed the ability of these cell lines to grow in an anchorage-independent fashion, which is another indicator of the tumorigenic potential of transformed cells. Colonies formed in soft agar at a high rate only when NNATα+β was present (19-fold that of pLAPSN), suggesting that both splice forms are necessary for anchorage-independent growth (Fig. 3C). In contrast, the presence of NNATα or NNATβ separately did not induce highly significant soft agar colony formation, although cell lines expressing NNATα did form more and larger colonies (~4-fold more colonies) than did the NNATβ-or pLAPSN-expressing cell lines (Fig. 3C). Not only did cell lines expressing NNATα+β form significantly more soft agar colonies, but also these colonies were much larger, supporting the results of the proliferation assays (Fig. 3B). To ascertain the effect of NNAT on invasion and migration, we performed Matrigel invasion chamber assays and scratch assays, respectively, and the presence of NNAT did not affect tumor cell motility in either (data not shown). Our data so far are consistent with the hypothesis that expression of both NNAT splice forms promotes tumor growth by altering the cell cycle. NNAT may also help the tumor establish in certain microenvironments.
Increased colony formation in soft agar suggested that we should evaluate the ability of NNATα+β to enhance tumor formation in an in vivo model. Neither of the MB cell lines MCD-1 and UW228 ordinarily grows as a xenograft when injected into nude mice.9 In our studies, the UW228 cell line expressing the pLAPSN control also failed to form flank tumors in three of three mice, whereas the MCD-1 cell line failed to form tumors in mice regardless of whether it expressed pLAPSN or NNATα+β. However, the UW228 cell line expressing NNATα+β was able to form small subcutaneous tumors in three of three mice (Fig. 4A). Histologically, these tumors were consistent with the pathological definition of MBs; they were composed of embryonal cell that had pleomorphic nuclei and showed increased mitotic activity (Fig. 4B, arrows).
Given this evidence supporting a functional role in tumor formation, we sought to gain preliminary insight into the oncogenic signal promoted by NNAT overexpression. To find the pathway(s) potentially regulated by NNAT, as reflected by mRNA levels, we made SAGE libraries of UW228-pLAPSN and UW228-NNATα+β. SAGE data analysis resulted in a list of 414 genes that were overexpressed in UW228-NNATα+β compared to UW228-pLAPSN, using the parameters of p-values, 0.01 and increased expression by at least 3-fold. We then used this list of differentially expressed genes as the input genes for the WebGestalt pathway analysis software suite and for the Bibliosphere Pathway Analysis software (Table 2). Both software programs use Gene Ontology, Biocarta pathways, and KEGG biochemical pathways. We found several oncogenic pathways potentially associated with NNAT overexpression according to the analyses performed with the two different programs. These pathways are the phosphoinositide 3-kinase (PI3K) pathway, the mitogen-activated protein kinase (MAPK) pathway, the calcium signaling pathway, and various metabolic pathways. We also performed additional SAGE library comparisons of two NNAT-positive primary MBs versus one NNAT-negative MB and one low-NNAT-expressing primary MB. After identifying genes that were significantly (p-value, 0.05, at least 3-fold greater) overexpressed in the NNAT-positive population, we performed the same bioinformatic analyses using the WebGestalt program. Some of the pathways we identified as being significantly up-regulated in our NNAT-overexpressing in vitro model (p-value, 0.01) were also significantly up-regulated in the NNAT-overexpressing primary MBs with endogenous NNAT overexpression (Table 2), suggesting that these pathways are indeed regulated by NNAT. Pathway analysis by the WebGestalt program also found that the MAPK signaling pathway and the signaling pathway downstream of mTOR, a serine/threonine kinase that is the mammalian target of rapamycin, were both significantly up-regulated in our in vitro system, although not in the NNAT- positive MBs. This may perhaps be due to the high degree of heterogeneity of MBs and the lack of truly NNAT-negative primary MBs. There was an overlap in many of the pathways identified by both the WebGestalt program and the Bibliosphere Pathway Analysis program, although the latter identified additional pathways that may also be associated with NNAT overexpression. One of these included the insulin-signaling pathway, which is of particular interest because previous studies had demonstrated the role of NNAT in insulin secretion by the pancreas.12
NNAT is an alternatively spliced gene whose two splice forms, NNATα and NNATβ, are expressed during mouse development with a period of strong overlapping expression.13,14 NNATα appears first, followed later by NNATβ, but by birth NNATα expression has been largely abated, leaving NNATβ as the predominant splice form.14 The two splice forms differ by the absence of a small transmembrane domain in NNATβ, which in normal children and adults is present at only low levels.15–17 The NNAT gene is located in the intron of a biallelically expressed, antisense transcript, bladder cancer-associated protein (BLCAP),16 which we concluded was not directly involved in MB because it is not overexpressed in our MB SAGE libraries. It has also been shown that BLCAP is not involved in antisense regulation of NNAT.16
During development, NNAT expression is regulated by epigenetic modifications.18–21 By virtue of its expression pattern during development, NNAT has been postulated to participate in the maintenance of segment identity in the mammalian hindbrain.13 The two NNAT isoforms, α and β, have distinct patterns of expression during development,14 suggesting that there is differential regulation between the two isoforms and that there are possibly different roles played by the two isoforms during development. Interestingly, NNAT is a known target of NeuroD1 (neurogenic differentiation 1),12 which itself has been shown to be involved in granule cell precursor differentiation,22 and is overexpressed in 15 of the 21 MB SAGE libraries constructed in our lab (posted at http://cgap.nci.nih.gov/SAGE). NNAT is expressed in the granule cell layer of newborn mouse cerebellum,19 as is NeuroD1.22 So far, there is no direct evidence that the NNAT is either directly or indirectly activated by NeuroD1 in MBs, but these findings taken together suggest this is a hypothesis worth testing.
NNAT had a dramatic increase in transcription compared to adult brains in our expression survey. It also has an active role in neural development, which is consistent with known oncogenes and tumor suppressors for MB. We found that high levels of total NNAT expression in MBs are due mostly to increased expression of NNATα, but is also associated with some increased expression of NNATβ (Fig. 1C). This finding, and the developmental pattern of expression of the NNAT splice forms, suggests that both splice forms might be needed for the tumorigenic potential of NNAT. This was supported by our demonstration of the significant increase in anchorage-independent growth and a shift from G1 to G2M only with the presence of both NNAT splice forms. Additionally, our findings that NNATα is the predominant splice form in MBs and that knockdown of endogenous NNATα alone or with endogenous NNATβ simultaneously causes a decrease in proliferation, suggests that it is specifically the presence of NNATα that is primarily responsible for the tumor-promoting behavior of this gene.
Pathway analysis of the isogenic cell line model we created revealed that known cancer-related pathways such as the PI3K and MAPK pathways are up-regulated in NNATα+β-expressing cells, further supporting the potential role of NNAT in promoting tumorigenicity. Previous studies have shown that some of these pathways are up-regulated in MBs, including the PI3K/Akt pathway.23 Interestingly, known oncogenes such as H-RAS and MAPK1 were significantly up-regulated in UW228-NNATα+β cells (p-value, 0.01; 3.5-fold and 18.7-fold, respectively). From data mining of transcript profiles on the National Cancer Institute’s SAGE Genie database, we observed that NNAT is up-regulated in MB cells that are resistant to 4-hydroperoxycyclophos-phamide, which is commonly used to treat MBs. NNAT has previously been linked to tamoxifen resistance in mammary carcinoma xenografts;24 both of these findings suggest that NNAT may have a role in chemotherapeutic drug resistance. Unfortunately, the gene expression analysis we performed does not reveal how NNAT might promote drug resistance, but functional studies of the genes induced by NNAT might resolve the question of whether NNAT is directly involved in promoting chemoresistance.
We have accumulated evidence that NNAT expression promotes tumorigenic growth of MB cells, but we have not identified any genomic alterations that pathologically trigger its expression in cancer. A screen of the coding region of NNAT in 20 primary MBs failed to identify any mutations, suggesting that mutations in NNAT are rare. We cannot exclude the possibility that increased NNAT expression is triggered by upstream signaling, promoter mutations, epigenetic changes, or any combination of these. A report of imprinting at the NNAT locus prompted us to investigate methylation changes at the presumed NNAT promoter, but this failed to yield conclusive evidence of this potentially important means of expression reactivation.
Although our studies are limited to NNAT’s role in MB, its involvement in some other cancers cannot be excluded. Using the SAGE Genie database, we found little or no expression of NNAT in 23 classes of normal tissue. However, we found that NNAT is also overexpressed, with levels of more than 3 per 10,000 transcripts, in ependymomas, a mesothelioma, and a rhabdomyosarcoma. Other groups have demonstrated a correlation between NNAT overexpression and poor prognosis of non-small-cell lung cancer.25 A recent report indicates that NNAT is expressed as a result of acquired tamoxifen resistance in breast cancer xenografts.24
To the best of our knowledge, this is the first report characterizing NNAT as having a potential functional role in cancer. Our in vitro and in vivo results indicate that the increased expression of the NNATα splice form is pathological and that its coexpression with NNATβ causes increased proliferation of MBs. The high expression and proportion of NNATα in total NNAT expression in MBs correlates with that seen during neural development (Fig. 1A).14 Previous rat developmental studies have shown that the coexpression of both splice forms, with NNATα predominantly expressed relative to NNATβ, coincides with a time in gestation of neuroepithelial proliferation and differentiation,14 suggesting that the NNAT splice forms interact during development. This is consistent with our findings that the combined presence of NNATα+β caused an increase in proliferation, anchorage-independent growth, and tumor formation in vivo. Furthermore, our expression analysis suggests that coexpression of NNATα+β may regulate both known cancer-related pathways as well as pathways known to be regulated by NNAT but not previously described in MBs or other cancers. Up-regulated metabolic pathways in response to NNAT also suggest additional approaches to evaluate for MB therapy.
NNAT is approximately 50% homologous to the class of proteolipids that are regulatory subunits of ion channels.17 If NNAT does indeed act to regulate ion channels, one mechanism of NNAT action may involve the formation of a heterodimer or multimer comprising the NNATα+β splice forms in order to exert this regulation in MBs. We have demonstrated that knockdown of NNATα results in decreased proliferation and that only expression of NNATα+β simultaneously results in the highest increase in proliferation; these findings suggest that the two NNAT splice forms may interact, although physical interaction of the two splice forms has not been addressed by these experiments.
We conclude that NNAT is one of the most commonly overexpressed genes in MBs, that it acts in an oncogenic fashion, and that the coexpression of both NNAT splice forms is necessary for significant increase in proliferation. The ability of coexpressed NNATα+β to promote tumor formation in cells devoid of tumor-forming capacity also hints at a critical role for this gene in MB progression.
We thank Stephen Baylin, J. Keith Killian, and James G. Herman for helpful discussion and insights about methylation; Ashani Weeraratna, Vinay Prabhu, and Kevin Becker for helpful discussion regarding bioinformatic pathway analysis of the SAGE data; John Silber, William Freed, and Torsten Pietsch for the UW228, MCD-1, and Med-mhh1 cell lines, respectively; and Brenda J. Raymond for her assistance in the preparation of the manuscript. This work was supported by funds from National Institutes of Health grant NS052507, the American Brain Tumor Association, and the Virginia and D.K. Ludwig Fund for Cancer Research. G.J.R. is the recipient of the Irving J. Sherman M.D. Research Professorship.