Suppression of Glioma Progression by Egln3

Grade IV astrocytoma or glioblastoma has a poor clinical outcome that can be linked to hypoxia, invasiveness and active vascular remodeling. It has recently been suggested that hypoxia-inducible factors, Hifs, increase glioma growth and aggressiveness [1], [2], [3]. Here, we tested the hypothesis that Egl 9 homolog 3 (Egln3), a prolyl-hydroxylase that promotes Hif degradation, suppresses tumor progression of human and rodent glioma models. Through intracranial tumorigenesis and in vitro assays, we demonstrate for the first time that Egln3 was sufficient to decrease the kinetics of tumor progression and increase survival. We also find that Klf5, a transcription factor important to vascular remodeling, was regulated by hypoxia in glioma. An analysis of the tumor vasculature revealed that elevated Egln3 normalized glioma capillary architecture, consistent with a role for Egln3 in eliciting decreases in the production of Hif-regulated, angiogenic factors. We also find that the hydroxylase-deficient mutant, Egln3H196A partially maintained tumor suppressive activity. These results highlight a bifurcation of Egln3 signaling and suggest that Egln3 has a non-hydroxylase-dependent function in glioma. We conclude that Egln3 is a critical determinant of glioma formation and tumor vascular functionality.


Introduction
Glioblastoma is a highly invasive, fast-growing cancer, classified by hypoxia, necrosis and the active formation of intra-tumor blood vessels [4]. Collectively, the pathological features of glioblastoma render this malignancy extremely refractory to surgical resection, chemotherapy, radiotherapy and anti-angiogenic treatments, leading to a median patient survival of 12-15 months after diagnosis [5,6,7].
Many of the cellular and systemic adaptations to hypoxia (,8% O 2 ) are mediated by hypoxia-inducible factors (Hifs) such as Hif-1a and Hif-2a, basic helix-loop-helix PAS domain transcription factors that regulate the expression of genes involved in angiogenesis, cell proliferation and metabolism [8,9,10]. For a variety of tumor types, high levels of Hif-1a and Hif-2a are tightly correlated with malignancy, invasiveness, metastasis and vascular density [10,11,12,13,14,15,16]. While Hif-1a is ubiquitously expressed under hypoxia, Hif-2a exhibits a relatively restricted, cell type-specific pattern of expression [8,17]. Hif-2a promotes tumor-initiation, the up-regulation of pro-angiogenic factors such as Vegf and expression of the embryonic stem cell gene, Oct4 [2,18,19,20,21,22]. These data highlight Hifs as potential targets for dismantling the initiation and vascularization potential of glioma and raises the possibility that endogenously occurring Hif inhibitors could be employed to antagonize tumor progression.
Hifs are primed for degradation by Egln1, 2, and 3 hydroxylases (also termed prolyl-hydroxylase domain enzymes, PHD) that act upon specific proline residues in an O 2 -dependent manner and thereby target Hif-a subunits for ubiquitination by von Hippel-Lindau tumor suppressor and proteasomal-degradation [8,23,24].
Eglns manifest substrate biases in a number of cell contexts, with Egln1 prone to recognize Hif-1a and Egln3 preferentially hydroxylating Hif-2a [25]. Nonetheless, the roles of Egln1-3 in cancer biology are poorly understood, in some cases characterized as tumor suppressors and in others implicated in tumor aggressiveness [26,27,28,29,30,31,32]. Indeed, Egln proteins are heterogeneously expressed in various glioma cell lines in vitro [26,30], making it difficult to correlate Egln expression with tumorigenicity. These ambiguities could be attributed to non-Hif related functions, as alternate Egln targets have been identified [26,33,34].
In this study, we report for the first time the effects of Egln3 expression upon glioma progression in vivo. When glioma-forming cells were engrafted intracranially and induced to express Egln3, tumor aggressiveness decreased as evidenced by markedly increased survival of injected mice. Using a catalytically inactive Egln3 mutant, we demonstrate that these effects are hydroxylase activity-dependent and independent, indicative of Hif and non-Hif functions of Egln3.

Reduced Egln3 expression correlates with the upregulation of Hifs in glioma cells
Hifs participate in maintaining the transcription of Oct4, a gene widely expressed within glioma cells, and are causal to aggressive glioma growth and progression [1,2,3,18,19,20,21,22]. We therefore evaluated whether the expression of any member of the Egln prolyl-hydroxylase family correlated with the Hif-expression pattern during the hypoxic response of glioma cells. Since it previously has been reported that glioma cells differentially expressed Hif-2a relative to neural stem cells (NSCs) under hypoxia [2], we utilized NSCs as a control in these experiments.
We first compared the expression profiles of several hypoxiaresponse pathway components ( Figure 1A) in Rt-glioma (F98; Rtglioma) cells to Rat NSCs (Rt-NSCs) by RT-PCR ( Figure 1B) and Western ( Figure 1C). Egln1 transcript levels increased significantly under hypoxia (1% O 2 ), consistent with reports indicating that Egln1 is hypoxia-responsive ( Figure 1B) [30]. Egln2 mRNA was neither highly nor differentially expressed in Rt-glioma or Rt-NSCs, despite the use of several primer sets to probe for this transcript ( Figure 1B; data not shown). On the other hand, Egln3 protein was undetectable in Rt-glioma cells (normoxia and hypoxia) whereas Egln3 in Rt-NSCs exhibited a dramatic increase upon hypoxia induction ( Figure 1B, C).
Rt-glioma cells and Rt-NSCs each expressed Hif-1a (mRNA and protein, Figure 1B-C) and up-regulated the Hif gene target, Vegf, confirming that a hypoxic response had successfully been induced. In contrast, Rt-glioma cells, but not Rt-NSCs exhibited Hif-2atranscript and protein (Figure 1B-C) [2]. Interestingly, Klf5 expression correlated with the relative absence of Egln3 and presence of Hif-2a, as Klf5 was not detected in Rt-NSCs and was up-regulated in Rt-glioma cells ( Figure 1B, C). Klf5, a member of the Kruppel-like factor family of transcriptional regulators, encodes a key mediator of angiogenesis and arterial development [35,36].
We next examined whether this hypoxic gene response profile was conserved within human glioma (U87; Hu-glioma) cells, using mouse NSCs (Ms-NSCs) and Rt-glioma cells as controls. RT-QPCR ( Figure 2A) and Western (Fure 2B) confirmed the absence of Egln3 expression in Hu-glioma cells and a hypoxia-dependent increase in Ms-NSCs. Similar to Rt-glioma cells, Hu-glioma cells upregulated both Hif-2aand Klf5 upon hypoxic insult (Figure 2A, B). These data suggest that Klf5 may be downstream of Hif-2ain glioma cells, as observed in liver hemangioma cells [9].
To provide additional evidence that Klf5 can respond downstream of Hif-2a in glioma cells, we expressed a shRNA designed to specifically silence Hif-2a [38] in Hu-glioma cells. shHif-2a severely curtailed the expression of Hif-2atranscript and protein ( Figure 4) to nearly undetectable levels in Hu-glioma cells under hypoxia, while not significantly affecting Hif-1aConsistent with the hypothesis that Klf5 acts downstream of Hif-2a, knockdown of Hif-2ain Hu-glioma cells resulted in <70% reduced expression of Klf5 ( Figure 4). Collectively, these data suggest that Klf5 expression can serve as a downstream readout for Hif-2aactivity ( Figure 1A). Our results also raise the possibility that Klf5 mediates certain Hif-dependent downstream functions within glioma.
Egln3 suppresses Hif-2a, Oct4 and Klf5 expression within cultured glioma cells Since the relative absence of Egln3 correlated with Hif-2a expression in glioma cells, we predicted that the expression of Egln3 might be sufficient to influence the hypoxic response of these cells. To assess this possibility, we inducibly-expressed Egln3 and assayed for effects in the Hif expression pattern of glioma cells.
Hu-glioma cells exhibited a dosage-dependent induction of Egln3 with doxycycline (Dox) levels of 0-1mg/mL. 2-4 fold elevation of Egln3 protein expression was achieved compared to Ms-NSCs over this Dox concentration range. Relative to Hif-1a, Hif-2a protein levels were sensitive to Egln3 after a 6hr treatment with 0.25mg/mL Dox under hypoxia whereas higher induction levels (1mg/mL Dox) destabilized both Hif-1a and Hif-2a ( Figure 5A). This analysis was necessary to determine a working concentration of Dox to evaluate the functional significance of Egln3 within glioma cells. Low levels of Egln3 induction (0.25mg/ mL Dox) reduced the transcription of Klf5 and the Hif-2a transcriptional target, Oct4 [18] by <50% and <70%, respectively within Hu-glioma cells relative to hypoxia-treated controls ( Figure 5A). Similarly, for Rt-glioma cells infected with a Doxinducible Egln3 construct, the presence of 1mg/mL Dox (6hrs) upregulated the expression of Egln3 protein and caused <50% decreases in Hif-2a protein levels after a 6hr exposure to hypoxic conditions without significantly altering Hif-1a expression ( Figure 5B). The increased expression of Egln3 coincided with <75% decreases in Klf5 and Oct4 ( Figure 5B). Therefore, the introduction of Egln3 into Hu-and Rt-glioma cells was biased to uncouple the hypoxia-initiated adaptive responses mediated by Hif-2a relative to Hif-1a.

Egln3 decreases glioma progression in vivo
Based upon a critical role for Hifs in promoting tumor progression [10,11,12,13,14,15,16], it was reasonable to speculate that the induction of Egln3 might attenuate glioma growth. To determine if Egln3 could influence the kinetics of glioma development in vivo, we stereotaxically-engrafted Hu-glioma cells expressing Egln3 under the control of a Dox-inducible promoter into the cerebral cortex of immuno-compromised (NSG) mice. After 2 weeks of engraftment, mice were fed Dox daily and the experiment terminated upon the initial manifestation of neurological symptoms, an indicator of glioma formation ( Figure 6A). While mock-treated Hu-gliomas developed within 9-12 weeks, Hu-gliomas induced to express Egln3 did so consistently between 14-16 weeks, as represented by a Kaplan-Meier plot in Figure 6B. In contrast with mock-treated Hu-glioma cells that always initiated tumor formation, a single animal injected with Hu-glioma Egln3 +Dox failed to develop neurological symptoms after a <18 week period ( Figure 6B). Microscopic analysis of the engraftment site of this Hu-glioma Egln3 +Dox animal confirmed that tumor formation had failed to occur.

Egln3 promotes normalization of glioma capillary morphology in vivo
Since Egln3 was sufficient to induce decreases in Hif signaling in vitro ( Figure 5), the expression of Egln3 might also influence blood vessel formation within glioma. To test this possibility, we assessed CD31 (PECAM-1)-labeled blood capillaries within Hu-gliomas that were allowed to progress until neurological symptoms developed, facilitating an analysis of tumors of comparable size. As is typical of the inefficient blood vessels of solid tumors such as glioblastoma [39,40,41], the vascular architecture of Hu-glioma tumors were distended and haphazardly organized, with a significant percentage of capillary branches exhibiting diameters between 26-50mm (<20%) or capillary diameters of .50mm (<20%). In contrast, non-tumor capillaries of adult mouse brain had diameters of 1-to-25mm ( Figure 6C, D). Strikingly, induction of Egln3 normalized tumor capillary morphology, reducing the percentage of disorganized and distended vascular branches (0% tumor capillaries .25mm in diameter; Figures 6C, D).

Egln3 suppresses the expression of Klf5 and the Hif transcriptional targets, Oct4 and Vegf in vivo
Consistent with our in vitro experiments, a quantitative analysis of total tumor transcript levels indicated that Dox-induced Egln3 reduced Vegf by <80%, Klf5 by <90% and Oct4 by <90% in Rt-gliomas ( Figure 6G). Intermediate reductions in the expression of these genes were achieved in Rt-glioma Egln3 -Dox tumors due to a basal level of Egln3 expression by our construct (data not shown). Therefore, Egln3 decreased the expression of Klf5 and several hypoxia-regulated targets in vivo. Catalytically-inactive Egln3 H196A partially suppresses glioma growth While many of the known effects of Egln family members are mediated through the hydroxylation of Hifs and possibly other targets, catalytically inactive Egln1 has been shown to maintain hydroxylase-independent functions through additional signaling interactions [26]. To address whether the tumor suppressive phenotype of Egln3 was intrinsically linked to hydroxylase activity, we generated an enzymatically defective Egln3 mutant through the alteration of a catalytically critical histidine residue (i.e. H196A) [26,34,43]. Consistent with previous studies [43], Egln3 H196A was expressed at levels comparable to wild-type Egln3 in Hu-glioma cells in vitro and did not promote the degradation of Hif-1a or Hif-2a ( Figure 7A). In addition, quantification of total tumor Vegf, Klf5 and Oct4 transcript levels indicated that Egln3 H196A induction did not inhibit Hif-mediated transcription whereas tumors expressing wild-type Egln3 displayed decreases in Vegf, Klf5 and Oct4 in vivo ( Figure 7B). In fact, Egln3 H196A evoked increases in both Klf5 and Oct4 that we speculate may be due to potential dominant-negative effects of Egln3 H196A expression within glioma upon other signaling pathways (e.g. Egln1).
We next ascertained whether Egln3 H196A affected the kinetics of tumor progression by injecting Hu-glioma cells expressing Egln3 or Egln3 H196A under the control of a Dox-inducible promoter into the flanks of NSG mice. One week following engraftment, mice were fed Dox daily and the experiment terminated after 30 days. Consistent with our intracranial engraftment assays (Figure 6), Egln3 expression dramatically reduced tumor size ( Figure 7C, D). Intriguingly, the hydroxylase-deficient Egln3 H196A , exhibited tumor growth intermediate to mock-treated and Egln3-expressing human gliomas. Collectively, these data confirm a key role for Egln3 in glioma progression through the hydroxylation of Hifs and through the participation of Egln3 in non-hydroxylase-dependent signaling pathways (see Discussion).

Egln3 is sufficient to attenuate glioma progression
In this study, we have assessed the effects of Egln3 upon the development of glioma in vivo. According to The REMBRANT Database, some human gliomas exhibit a highly aggressive, low Egln3-expression phenotype (i.e. .2-fold decreases in Egln3 relative to basal brain levels), analogous to Hu-and Rt-gliomas in which Egln3 protein was not detected. This raised the possibility that the down-regulation of Egln3 might contribute to the progression and aggressiveness of some gliomas by accentuating the activity of Hif-mediated signaling within glioma cells. Given the role of Hifs in the progression of a variety of tumor types, Eglns could be promising candidate suppressors of tumor growth. Further experimentation demonstrated that Eglns influence other cellular processes as well, since several non-Hif targets of Eglns have been identified including cyclin D1, NF-kappab and the b2adrenergic receptor [26,33,34]. More recently, the Egln1 hydroxylase has been shown to have a Hif-independent function on amphiregulin regulation in breast carcinoma [44]. Our data suggest that Egln3 follows this precedent within glioma as well. Therefore, the interfacing of Eglns with Hif and non-Hif factors appears to influence tumorigenesis, highlighting the need to mechanistically dissect Egln3 signaling pathways in glioma and  We hypothesize that induction of Egln3 during the initial stages of tumor formation impaired cellular factors that are proposed to support tumor initiation such as Hif-2a and Oct4 [2,19] and that later stage glioma progression was abrogated through reduced angiogenic signaling within the tumor. Within cultured glioma cells it has been suggested that Egln3 can support cell viability, by attenuating exaggerated Hif signaling [30]. The dosage and timing of Egln3 induction as well as the targeted expression of Egln3 to specific glioma cell types likely will prove to be important considerations in directing the tumor suppressing potential of Egln3. Therefore, while the overall effect of Egln3 was to suppress Hu-glioma growth, our findings also raise the prospect of additional Egln3 effectors within glioma.

Egln3 as a vascular normalization factor within glioma
Relative to healthy tissue, many solid tumor types including glioblastoma are believed to overproduce pro-angiogenic factors, resulting in the distension and disorganization of tumor vasculature [39,40,41]. We noted that Egln3 reduced tumor Vegf production and induced the normalization of capillary structure. Vascular normalization through the administration of angiogenic inhibitors has arisen as a promising means of enhancing the circulation and efficacy of chemo-and radio-therapeutics [45,46]. Interestingly, it has been hypothesized that angiogenesis inhibitors can be dosed to simultaneously decrease capillary number and normalize vessel structure, avoiding increased tumor growth kinetics [45]. According to this hypothesis, our observed decreases in tumor aggression upon Egln3 induction would be consistent with depressed levels of angiogenesis within Egln3-expressing Hugliomas. Identifying cellular signaling factors that can be modulated to elicit vascular normalization has therefore become of great interest [45]. Nitric oxide and Rgs5 have been revealed as key regulators [47,48] and the haplodeficiency of Egln1 within host mice (i.e. not within engrafted tumor cells) also influenced intra-tumor capillary structure [31]. We propose that Egln3 encodes an additional, molecular participant that governs tumor vascular normalization. Clearly Egln3 participates as a regulator of many fundamental cellular processes, highlighting the complex nature of Egln3 function in glioma formation.

Transcript expression analysis
RT-PCR was performed as described [42] with primers specific for each gene. Glioma RNA was prepared by resecting the tumor, dounce homogonizing the tissue and purifying total RNA using RNeasy Plus Mini Kit (Qiagen, Valencia, CA). Relative mRNA levels of Egln1, Egln3, Hif-1a, Hif-2a, Vegf, Oct4 and Klf5 were determined by reverse transcription-quantitative polymerase chain reaction (RT-QPCR), as described [2] using GoScript Reverse Transcriptase (Promega Corp,). Fold changes were calculated as a percentage relative to the relevant corresponding control after normalized to actin.   Relative protein levels were calculated by using ImageJ (National Institute of Health, MD) and Adobe Photoshop (Adobe Systems, San Jose, CA) software using Actin as a normailization standard.

DNA manipulations
The cDNA encoding Hif-1a (NM_001430) was provided by Gregg Semenza, Johns Hopkins University School of Medicine, Baltimore, MD. Degradation-resistant variants of human Hif-1a and human Hif-2a (cloned from a human endothelial cell cDNA library; NM_001530) were generated by altering proline 564 or proline 531 to alanine by site-directed mutagenesis of Hif-1a and Hif-2a, respectively. The coding region for Egln3 (NM_028133) was isolated from a mouse NSC cDNA library by PCR. The hydroxylase-deficient variant of mouse Egln3 was generated by altering histadine 196 to alanine by site-directed mutagenesis. The sequences of all Hif and Egln3 constructs were verified. The mutated Hifs and Egln3 were cloned into a Dox-inducible retroviral vector, pTre-tight modified to include the hygromycin resistance gene (Clontech, Mountain View, CA) and introduced into glioma cells.

Intracranial and subcutaneous injections of glioma cells
NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ; The Jackson Laboratory, Bar Harbor, Maine) were positioned in a stereotaxic frame (David Kopf Instruments, Tujunga CA) and 100,000 Huglioma or 50,000 Rt-glioma cells in a volume of 1mL were intracranially engrafted into the cerebral cortex, as described [2]. 1mg doxycycline (Dox) was administered in food ad libitum per day for <1-3 weeks. For Hu-and Rt-gliomas this Dox regimen was started 2 weeks or 1 week following cell engraftment, respectively. Upon onset of neurological symptoms, mice were sacrificed for tumor analysis. For subcutaneous injections, 1610 6 Hu-glioma cells were mixed 1:1 with growth factor reduced matrigel (BD Biosciences, Bedford, MA), injected in the flanks of NSG mice and dox was administered as described above starting 1 week following engraftment. All mouse procedures were carried out in accordance with applicable IACUC and federal guidelines and protocols were approved by the Animal Care and Use Committee of the University of California, Berkeley (Animal Use Protocol # R318-1011B).

Microscopy
Images were obtained with an IX71 microscope system (Olympus America Inc, San Diego, CA)/Retiga 2000R cooled camera (QImaging, Surrey, BC Canada). Anti-CD31 antibody was used to label and assess Hu-glioma capillaries (BD Biosciences, San Jose, CA). Secondary antibodies coupled to the Alexa-555 dye (Invitrogen) was used as described [42]. Rt-glioma capillaries were labeled using and endothelial-specific isolectin GS-IB 4 from Griffonia simplicifolia conjugated to Alexa-594 (Lectin; Invitrogen) at 1mg/ml, according to the manufacturers instructions. Nuclei were stained with 49,6-Diamidino-2-phenylindole, dihydrochloride (DAPI; Anaspec, Fremont, CA). Tumor capillaries were delineated by labeling tumor sections with CD31 or EC Lectin. Capillary diameter was then quantified relative to scale bars to determine capillary diameter. For each experimental group, .500 capillaries of 10 randomly chosen tumor sections were chosen from 3 glioma engrafted mice.

Statistical Analysis
Statistical analysis was performed using a standard two-tailed T test.