GLIA 53:733–743 (2006)
`
`AMP-Dependent Protein Kinase Alpha 2 Isoform
`Promotes Hypoxia-Induced VEGF Expression
`in Human Glioblastoma
`KATHRYN M. NEURATH,1 
`MARTIN P. KEOUGH,1 
`AND KEVIN P. CLAFFEY1*
`TOM MIKKELSEN,2
`1Department of Cell Biology, Center for Vascular Biology, University of Connecticut Health Center, Farmington, CT 06030-3501
`2Departments of Neurology and Neurosurgery, Henry Ford Hospital, Detroit, MI 48202
`
`KEY WORDS
`hypoxia; AMPK; VEGF; HIF-1; glioma
`
`ABSTRACT
`Tumor cells respond to hypoxic stress by upregulating a va-
`riety of genes involved in glucose uptake, glycolysis, and
`angiogenesis, all essential to maintaining nutrient avail-
`ability and intracellular ATP levels. Adenosine monopho-
`sphate-dependent kinase (AMPK) is a key sensor for cellu-
`lar homeostasis and is highly sensitive to changes in AMP:
`ATP ratios. The two catalytic AMPK alpha isoforms (AMPKa1,
`AMPKa2) were investigated with respect to their expres-
`sion,
`cellular distribution, and contribution to VEGF
`expression under hypoxic stress in human U373 glioblas-
`toma cells. Quantitative real-time PCR analysis showed
`AMPKa1 mRNA to be constitutively expressed in normoxia
`and hypoxia, whereas AMPKa2 mRNA levels were low in
`normoxia and significantly induced in hypoxia. Fluorescent
`immunohistochemistry showed that AMPKa2 protein redis-
`tributed to the nucleus under hypoxia, whereas AMPKa1
`remained distributed throughout the cell. The AMPK chem-
`ical
`inhibitor, 5-iodotubericidin, effectively repressed the
`hypoxic induction of VEGF mRNA levels and hypoxia in-
`ducible factor-1 dependent transcription. AMPKa2 repres-
`sion with RNA interference reduced hypoxia-induced VEGF
`mRNA and HIF-1 transcription, whereas AMPKa1 re-
`pression did not. Human glioblastoma cell lines U118 and
`U138 also showed hypoxia-induction of AMPKa2 as well as
`VEGF. Immunohistochemistry analysis of human astrocy-
`toma/glioma samples revealed AMPKa2 present in high
`grade gliomas within hypoxic pseudopalisading microenviron-
`ments. These data suggest that prolonged hypoxia promotes
`the expression and functional activation of AMPKa2 and
`VEGF production in glioma cell lines and glioblastoma multi-
`form tumors, thus contributing to tumor survival and angio-
`genesis in high grade human gliomas.
`VVC 2006 Wiley-Liss, Inc.
`
`INTRODUCTION
`
`is regulated by a number of angiogenic factors (Claffey
`et al., 1996; Dvorak, 2000). An important angiogenic fac-
`tor is vascular endothelial growth factor-A (VEGF), which
`is over expressed in response to environmental hypoxia
`(Denko et al., 2003; Michiels, 2004). Human glioblastomas
`are among the most highly vascularized tumors and ex-
`press high levels of VEGF (Kaur et al., 2004). A primary
`mechanism of increased VEGF expression is through the
`activation of hypoxia inducible factor-1 (HIF-1), a hetero-
`dimeric transcription factor (Semenza, 2000). Under nor-
`moxia, HIF-1a protein levels are regulated by ubiquitina-
`tion and rapid proteosomal degradation (Maxwell et al.,
`2001; Maxwell and Ratcliffe, 2002). In hypoxic conditions,
`HIF-1a degradation is inhibited; HIF-1a is stabilized,
`translocates from the cytoplasm to the nucleus, and di-
`merizes with HIF-1b/aryl hydrocarbon receptor nuclear
`translocator (ARNT). This complex associates with co-
`activators, such as p300/CREB, to induce transcription of
`genes containing hypoxia responsive elements (HREs),
`such as VEGF (Fedele et al., 2002; Lee et al., 2004).
`AMP-dependent kinase or AMP-activated kinase (AMPK)
`is a heterotrimeric protein composed of a catalytic alpha
`subunit, and two regulatory subunits, beta and gamma.
`The alpha subunit contains a serine/threonine protein
`kinase catalytic domain (Carling, 2004). AMPK has been
`identified as a primary sensor of cellular energy change
`by responding to increases in AMP:ATP ratios, concur-
`rent with hypoxia or nutrient depletion (Hardie, 1999;
`Hardie and Carling, 1997). AMPK is activated by the in-
`crease in AMP and in turn activates catabolic pathways
`and provides general repression of anabolic pathways by
`direct phosphorylation of downstream substrates, result-
`ing in cellular protection from metabolic or nutritional
`stress (Hardie et al., 2003). Previous reports have shown
`that AMPK is activated in hypoxia (Frederich et al., 2005;
`Lee et al., 2003; Marsin et al., 2002; Nagata et al., 2003).
`
`Hypoxia plays a critical role in regulating tumor
`growth and angiogenesis. The cellular response to hy-
`poxia is adaptive and necessary to maintain the minimal
`energy levels required for cell survival. Once a tumor
`reaches a volume greater than a few mm3, regions of hy-
`poxia begin to occur and neovascularization is essential
`for tumor survival (Folkman et al., 1989). Angiogenesis,
`the growth of new blood vessels from pre-existing blood
`vessels to supply the tumor with oxygen and nutrients,
`
`Grant sponsor: NIH; Grant number: NCI CA-064436; Grant sponsor: The
`Patrick and Catherine Weldon Donaghue Foundation.
`
`*Correspondence to: Kevin P. Claffey, Ph.D., Center for Vascular Biology,
`EM028, Department of Cell Biology-MC3501, University of Connecticut Health
`Center, 263 Farmington Ave., Farmington, CT 06030-3501.
`E-mail: claffey@nso2.uchc.edu
` 
`
`These authors contributed equally to this work.
`
`Received 18 October 2005; Accepted 13 January 2006
`
`DOI 10.1002/glia.20326
`
`Published online 3 March 2006 in Wiley InterScience (www.interscience.
`wiley.com).
`
`VVC 2006 Wiley-Liss, Inc.
`
`L'OREAL USA, INC. EX. 1014
`
`733
`
`

`

`734
`
`Target
`
`huCyclophilinA
`huAMPKa1
`huAMPKa2
`huVEGF
`
`NEURATH ET AL.
`
`TABLE 1. qRT-PCR Primers
`0
`0
`Forward Primer (5
`
`– 3
`
`)
`
`CTGGACCCAACACAAATGGTT
`AGGAGAGCTATTTGATTATATCTGTAAGAATG
`CGGCTCTTTCAGCAGATTCTGT
`CGAGGGCCTGGAGTGTGT
`
`Reverse Primer (5
`
`0
`
`0
`– 3
`
`)
`
`CCACAATATTCATGCCTTCTTTCA
`ACACCAGAAAGGATCTGTTGGAA
`ATCGGCTATCTTGGCATTCATG
`GGCCTTGGTGAGGTTTGATC
`
`Repression of total AMPKa with chemical inhibitors or
`dominant negative isoform over expression represses HIF-
`1 dependent transcription (Hwang et al., 2004; Lee et al.,
`2003). However, there are two distinct AMPK alpha iso-
`forms—AMPKa1 and AMPKa2. Although highly homolo-
`gous, it is becoming clear that AMPKa1 and AMPKa2
`have exclusive functions. Mouse genetic deletion studies
`have shown that AMPKa2-null mice demonstrate glucose
`intolerance and reduced insulin sensitivity, while the
`AMPKa1-null mice do not show any appreciable altera-
`tions (Viollet et al., 2003a; Viollet et al., 2003b). However,
`the potential roles for the selective AMPK alpha isoform
`regulation and activity in response to hypoxia have not
`been clearly evaluated. The studies performed here inves-
`tigated the role of AMPK in the response to hypoxia in
`U373 glioma cells and whether the catalytic alpha iso-
`forms of AMPK show differential gene expression or func-
`tional roles in this response. RNAi technology targeting
`AMPK alpha isoforms selectively repressed protein levels
`and isoform specific activities. These studies indicate that
`AMPKa2, but not AMPKa1,
`is selectively induced in
`hypoxic conditions and significantly contributes to VEGF
`expression in human glioma cells as well as in high grade
`glioblastoma tumors.
`
`MATERIALS AND METHODS
`Cell Line and Culture Conditions, Antibodies,
`and Chemical Effectors
`
`Human glioma cell lines, U373, U118, U138, and U87,
`were maintained at 37°C in Dulbecco’s Modified Eagle
`Medium (DMEM) (Invitrogen, Carlsbad, CA) supplemen-
`ted with 10% FBS, and penstrep (100 ug/ml). Cells were
`cultured either under normoxic conditions (5% CO2,
`21% O2, 74% N2) or hypoxic conditions (5% CO2, 2%
`O2, 93% N2) as determined previously to be effective for
`maximal activation of VEGF expression without cytotox-
`icity through 48 h treatment (Claffey and Robinson,
`1996; Shih et al., 1999). Antibodies used for immuno-
`blots: AMPKa1 and AMPKa2 were from US Biological
`(Swampscott, MA), purchased p-AMPK(Thr172) and
`phospho-p70S6K(Thr389) from Cell Signaling Technology
`(Beverly, MA), and b-actin from Abcam (Cambridge, MA).
`Chemical effectors AICAR (5-Aminoimidazole-4-carboxa-
`mide 1-b-D-ribofuranoside) and 5-iodotubericidin were pur-
`chased from Sigma-Aldrich (St. Louis, MO).
`
`Northern Blot Analysis
`
`Total RNA was extracted using RNeasy RNA extrac-
`tion kit (Qiagen, Chatsworth, CA). Northern blot was
`performed as described previously (Claffey et al., 1998;
`
`GLIA DOI 10.1002/glia
`
`Hong et al., 2003). Hybridization was carried out over-
`night at 65°C with [a32P]dCTP-labeled human VEGF 165
`Acc I/Nco I fragment (823 base pairs) and an [a32P]
`dCTP-labeled human Glut-1 probe BamH I/BamH I frag-
`ment (2473 base pairs) (ATCC, Rockville, MD). A ribo-
`some-associated protein cDNA probe, 36B4, was used as
`a loading control. Blots were washed at high stringency
`(1% SDS, 1X SSC at 60°C) and exposed to Kodak MR
`film. Quantification was determined by densitometry
`using ImageQuant Software.
`
`RNA Interference
`
`siRNA oligos were obtained from Dharmacon (Lafay-
`ette, CO) and were used to specifically target either
`AMPKa1 or AMPKa2 subunits of the AMPK heterotri-
`mer. RNAi transfections were performed with Oligofec-
`tamine (Invitrogen) in OptiMEM according to manufac-
`turer’s protocols at various concentrations. Post-transfec-
`tion, cells were allowed to recover overnight and refed
`24 h before treatment with normoxic or hypoxic conditions.
`RNAi experiments were performed at least three times to
`assure representative results.
`
`SDS-PAGE and Immunoblots
`
`Cytoplasmic extracts were obtained as described pre-
`viously (Claffey et al., 1998). Protein extracts were sepa-
`rated on 10% SDS-PAGE and transferred to nitrocellu-
`lose membranes. Immunoblots were performed using
`various primary antibodies and horseradish peroxidase-
`conjugated species appropriate IgG secondary antibodies
`diluted in blocking buffer according to manufacturer’s
`protocols. Blots were developed using ECL reagents.
`
`VEGF ELISA Assay
`
`VEGF capture ELISAs were performed as described
`previously (Shih et al., 1999). Culture supernatants in
`triplicate were collected from cell culture wells and cleared
`by centrifugation. The samples were buffered with Tris-HCl
`pH 7.5 to a final concentration of 1 mM prior to analysis.
`
`Quantitative Real-Time PCR
`
`Cellular total RNA was harvested using RNeasy kit
`(Qiagen) and reverse transcribed with random hexamer
`primers using PowerScript RT Strips (BD Biosciences,
`Palo Alto, CA). Quantitative PCR primers were designed
`using ABI Primer Express software for use with a SYBER
`Green detection kit (Qiagen). Sequences of qRT-PCR pri-
`mers are in Table 1.
`Samples in duplicate or triplicate were run on an ABI
`7900 thermocycler for 40 cycles, and results were ana-
`
`

`

`DISTINCT ROLES FOR AMPKa ISOFORMS IN HYPOXIA
`
`735
`
`lyzed using GraphPad software using a mathematical
`model described by Lui and Saint (Liu and Saint, 2002).
`Results were normalized to an internal control gene,
`huCyclophilin A.
`
`Luciferase Assay
`
`Cells were transfected with siRNA oligos, as described
`above, along with a 5X HIF-1 promoter element luci-
`ferase reporter (Claffey et al., 1998), luciferase control
`vector, or CMV driven beta-galactosidase vector, and
`treated with either normoxia or hypoxia for 24 h. Cells
`were harvested with a passive lysis buffer from assay
`kit, and luciferase activity was measured according to
`supplier’s instructions (Promega, Madison, WI). Luci-
`ferase activity was normalized to beta-galactosidase ex-
`pression as determined with a similar kit (Promega).
`
`Immunofluorescence
`
`Cells were grown on coverslips prior to exposure to
`normoxia or hypoxia. Cells were washed with PBS and
`fixed with 10% formalin in PBS. Cells were incubated in
`0.01% Triton X-100 detergent for 3 min, washed in PBS,
`and blocked in 3% BSA/PBS at RT for 1 h. Primary anti-
`bodies were added at a 1:50 dilution (5 ug/ml) in 3%
`BSA/PBS and incubated at RT for 1 h. Cells were washed
`and a secondary antibody at a 1:800 dilution was added
`and incubated for 30 min at RT. Finally, coverslips were
`washed thoroughly with PBS and mounted onto slides
`with PBS:Glycerol.
`
`Human Glioma Sample Immunohistochemistry
`
`Paraffin sections were received from Tom Mikkelsen
`at the Hermelin Brain Tumor Center at Henry Ford
`Hospital in Detroit, Michigan. Slides were processed af-
`ter antigen retrieval in sodium citrate buffer pH 6.0 for
`20 min at 95°C. Slides were blocked with PowerBlock
`(Biogenex, San Ramon, CA) for 10 min; and a 1:200 dilu-
`tion primary antibody, anti-AMPKa1 or AMPKa2 (US
`Biological), was incubated overnight at 4°C. Biotinolyated
`secondary was incubated for 30 min at RT and slides were
`developed using ABC (Vector Laboratories, Burlingame,
`CA) and DAB (Electron Microscopy Sciences, Hatfield, PA)
`reagents. Methyl green was added as a counter stain.
`Slides were imaged using white light microscopy on a
`Zeiss Axioplan microscope and quantified using ImagePro
`Plus software (MediaCybernetics, Silver Spring, MD).
`
`Statistical Analysis
`
`Data from individual experiments were represented as
`mean 6 standard error unless otherwise stated. Statisti-
`cal comparison of groups was performed using a 2-tailed
`Student’s t-test with appropriate tests for equal var-
`iances. Statistical significance was defined and indicated
`as P < 0.05 (*) or P < 0.01 (**).
`
`Fig. 1. Chemical modulators of AMPK affect hypoxia induced VEGF
`expression. U373 cells incubated in Normoxia or Hypoxia treated with
`0.5 mM AICAR (A), 10 lM 5-iodotubercidin (5-I), or no treatment (NT).
`A: VEGF, Glut-1, and 36B4 mRNA expression determined by northern
`blot. B: Hypoxia time course of VEGF mRNA normalized to 36B4
`mRNA expression. C: HIF-1 dependent luciferase expression normal-
`ized to beta-galactosidase transfection control from cellular lysates.
`(**P  0.01.)
`
`RESULTS
`Chemical Modulators of AMPK Affect
`Hypoxia Induced VEGF Expression
`
`To determine if AMPK has a role in the regulation of
`the hypoxia-inducible gene, VEGF, the AMPK activating
`5-aminoimidazole-4-carboxamide-1-b-4-ribofura-
`agent,
`noside (AICAR), and an AMPK inhibitor, 5-iodotuberi-
`cidin, were used to treat human U373 glioblastoma cells
`under normoxic and hypoxic conditions. Cells were treated
`with AICAR, 5-iodotubericidin, or no treatment control
`and exposed to hypoxia for 2, 4, 8, and 12 h. The treated
`cells were harvested for total RNA and northern blots per-
`formed to analyze VEGF and Glut-1 mRNA expression
`levels, Fig. 1A. No treatment controls showed a large
`increase in VEGF mRNA in hypoxia as compared to nor-
`moxia at 12 h. The addition of AICAR (0.5 mM) showed
`increases over the hypoxic induction of VEGF and Glut-1
`
`GLIA DOI 10.1002/glia
`
`

`

`736
`
`NEURATH ET AL.
`
`cells exposed to normoxic or hypoxic conditions. Under
`normoxic conditions the AMPKa1 isoform was the predo-
`minant isoform, showing 2- to 4-fold more mRNA than
`the AMPKa2 isoform. Exposure of U373 cells to 12 h of
`hypoxia showed little change in the baseline ratio of
`AMPKa1 to AMPKa2 mRNA levels. However, over a
`24 h hypoxic treatment, a significant 3-fold increase of
`AMPKa2 mRNA was observed over normoxic controls,
`Fig. 2A.
`A direct analysis of AMPKa isoform protein expression
`over a time course of hypoxia treatment showed a similar
`pattern of increasing AMPKa2 protein levels. The levels of
`phosphorylated AMPK (p-AMPK) also increased with ex-
`posure to hypoxia. AMPKa1 protein levels did not signifi-
`cantly change with the hypoxic time course, Fig. 2B.
`In order to determine whether hypoxia affects AMPKa
`isoform intracellular distribution, isoform-specific fluores-
`cent immunohistochemistry was employed on cells exposed
`to normoxic or hypoxic conditions for 24 h, Fig. 3. No
`significant change was observed in the localization of
`AMPKa1 between normoxia and hypoxia, showing a wide-
`spread nuclear and cytoplasmic distribution. AMPKa2 was
`found to redistribute within the cell from a uniform cyto-
`plasmic/nuclear distribution to a predominantly nuclear
`localization under hypoxia. These data demonstrate that
`AMPKa2, but not AMPKa1, is regulated at the mRNA
`and protein levels by hypoxia in U373 cells.
`
`Inhibition of AMPK Alpha Isoform Expression
`by RNA Interference
`
`In order to evaluate possible AMPK alpha isoform-spe-
`cific functions, RNA interference (RNAi) protocols were
`performed to repress AMPK alpha isoform gene expres-
`sion and function. Figure 4A shows AMPKa1 mRNA
`levels determined by quantitative RT-PCR when cells
`were transfected with either AMPKa1- or AMPKa2-tar-
`geted siRNA oligos. Post-transfected samples show that
`AMPKa1 mRNA levels were inhibited at greater than
`90% in both normoxic and hypoxic conditions when
`treated with AMPKa1 RNAi. AMPKa2 siRNA transfection
`did not significantly affect AMPKa1 mRNA levels. Conver-
`sely, Fig. 4B shows a similar pattern for the AMPKa2
`RNAi. AMPKa2 mRNA was repressed greater than 90% in
`normoxia and hypoxia by AMPKa2 RNAi, and AMPKa1
`siRNAs did not significantly affect AMPKa2 mRNA levels.
`To determine the effectiveness of the RNAi treatments
`at the protein level, AMPKa isoforms were evaluated by
`direct immunoblot of total cell extracts taken 60 h after
`transfection, Fig. 4C. The AMPKa1 RNAi repressed
`AMPKa1 protein in normoxia and hypoxia when com-
`pared to mock transfected controls. Hypoxia alone greatly
`increased AMPKa2 protein, as was observed in Fig. 2B.
`The AMPKa2 siRNA was effective at repressing AMPKa2
`protein levels in hypoxia, although the protein levels in
`normoxia were not significantly affected. These data
`demonstrate that RNAi against the individual AMPKa iso-
`forms was effective at repressing the targeted AMPKa
`mRNA and protein in U373 glioblastoma cells.
`
`Fig. 2. AMPKa isoforms are differentially expressed under hypoxic
`conditions. A: AMPKa1 (white bars) or AMPKa2 (dark bars) isoform
`mRNA expression as determined by qRT-PCR from U373 cells incu-
`bated in normoxia (N) or hypoxia (H) over 12 and 24 h periods. (*P 
`0.05.) B: Immunoblots on cellular lysates exposed to a hypoxia time
`course detected for AMPKa1, AMPKa2, and p-AMPK.
`
`mRNAs as compared to vehicle control. The addition of
`5-iodotubercidin (10 lM) completely repressed the hy-
`poxic expression of VEGF and Glut-1 mRNAs. In time
`course analysis, 5-iodotubercidin effectively repressed VEGF
`mRNA levels from 2 to 12 h in hypoxia, Fig. 1B.
`To evaluate the potential influence of AMPK on HIF-1
`mediated transcription, required for VEGF mRNA in-
`duction under hypoxic conditions, an HIF-1 dependent
`transcriptional assay was performed with the AMPK in-
`hibitor, 5-iodotubericidin. Transcriptional activity was
`measured using a transiently transfected HIF-1 respon-
`sive element driven luciferase reporter normalized to
`beta-galactosidase transfection baseline control, Fig. 1C.
`Under hypoxia, there was a significant, 12-fold increase
`in HIF-1 dependent transcription. This induction was
`completely inhibited to normoxic levels by AMPK inhibi-
`tor (5-iodotubericidin) treatment prior to hypoxic induc-
`tion. There was no regulation of HIF-1 dependent tran-
`scription using AICAR alone in a similar assay (data not
`shown).
`
`Differential Expression and Subcellular
`Localization of AMPK Isoforms Under
`Hypoxic Conditions
`
`In order to determine whether AMPKa isoforms are
`differentially regulated by hypoxia in human glioblas-
`toma cells, the mRNA expression of the two AMPKa iso-
`forms was determined by quantitative RT-PCR in U373
`
`GLIA DOI 10.1002/glia
`
`

`

`DISTINCT ROLES FOR AMPKa ISOFORMS IN HYPOXIA
`
`737
`
`Fig. 3. AMPKa isoforms are dif-
`ferentially localized under hypoxic
`conditions. AMPKa1 or AMPKa2 pro-
`tein localization as determined by im-
`munofluorescent histochemistry under
`normoxic (N) and hypoxic (H) condi-
`tions.
`
`The Effect of AMPKa Isoform Repression
`by RNAi on Hypoxia-Induced AMPK Activation
`and Downstream Signal Transduction
`
`In order to determine the effect of AMPK alpha isoform
`repression on functional AMPK protein, we examined the
`level of total phosphorylated AMPK, which detects both
`p-AMPKa1 and p-AMPKa2 isoforms at Thr172, an activat-
`ing phosphorylation site (Woods et al., 2003). Figure 5A
`shows that hypoxia increased p-AMPK and that applica-
`tion of siRNA targeted to AMPKa1 repressed the total
`p-AMPK considerably under normoxic and hypoxic condi-
`tions. AMPKa2 RNAi also repressed total p-AMPK under
`normoxia and hypoxia, although to a lesser extent than
`AMPKa1. Since AMPKa1 mRNA and protein are more
`abundant in the U373 cells, AMPKa2 repression only par-
`tially affects the total amount of p-AMPK protein.
`One essential function for AMPK is to repress protein
`translation by blocking mTOR/S6 kinase pathways
`through phosphorylation of the TSC1/2 complex (Inoki
`et al., 2003). To determine whether the mTOR pathway
`inhibition was selective to either AMPK alpha isoform,
`levels of phospho-p70S6K were evaluated in AMPK RNAi
`repressed cells. Hypoxia alone significantly reduced the
`amount of phospho-p70S6K signal in U373 glioblastoma
`cells, Fig. 5B. This data supports the observation of
`Krause et al., 2002, which states that AMPK activation
`leads to the repression of phospho-p70S6K through mod-
`ulation of the mTOR pathway. Under normoxic condi-
`tions, AMPKa1 and AMPKa2 RNAi resulted in increased
`levels of phospho-p70S6K. It was found that the RNAi
`treatment for both AMPKa isoforms increased the phos-
`pho-p70S6K in hypoxia, with the repression of AMPKa2
`demonstrating a slightly greater increase in phospho-
`
`p70S6K. These data indicate that both AMPK alpha iso-
`forms contribute to the repression of the protein transla-
`tional pathways by repressing active p70S6K under
`hypoxia. This also confirms that RNAi to AMPK alpha
`isoforms is effective at altering downstream pathways of
`AMPK and is sufficient to study the AMPK alpha iso-
`forms independently.
`
`AMPKa2 Selectively Contributes
`to Hypoxia-Induced VEGF Expression
`at the mRNA and Protein Levels
`
`In an effort to examine the potential selectivity for the
`AMPKa isoforms in regulating important hypoxia respon-
`sive genes, the hypoxic induction of VEGF under condi-
`tions where AMPKa1 or AMPKa2 is repressed with RNAi
`treatments was examined. Hypoxia responsive U373 cells
`were transfected with mock or AMPKa isoform specific
`siRNA oligo pools. To assess RNAi effectiveness for each
`experiment, the level of AMPKa1 or AMPKa2 specific
`mRNA repression was determined to be greater than 90%
`of mock transfected control by qRT-PCR (data not shown).
`Total cellular RNA was analyzed for VEGF mRNA expres-
`sion by qRT-PCR, and the level of secreted VEGF in condi-
`tioned media from the same cells was determined by
`ELISA. Figure 6A shows the quantitative levels of VEGF
`mRNA levels determined for normoxic and hypoxic cells
`corresponding to mock transfected, AMPKa1 RNAi, or
`AMPKa2 RNAi treated U373 cells. Only the AMPKa2
`RNAi
`treated cells showed repressed hypoxic-induced
`VEGF mRNA at a significant level (50% of non-transfected
`cells), whereas AMPKa1 RNAi had no effect. Similarly, the
`conditioned media of the AMPKa2 RNAi treated cells
`showed significant repression of secreted VEGF protein, as
`
`GLIA DOI 10.1002/glia
`
`

`

`738
`
`NEURATH ET AL.
`
`Fig. 5. RNAi targeted to AMPKa isoforms affects p-AMPK and a
`downstream AMPK target. Cellular lysates isolated from U373 cells
`treated with no siRNA oligos (NT) or with siRNA oligos against
`AMPKa1 (a1) or AMPKa2 (a2) under normoxia (N) or hypoxia (H). A:
`Immunoblot for p-AMPK (Thr172) and b-actin. B: Immunoblot for phos-
`pho-p70S6K (Thr389) (p-p70S6K) and b-actin.
`
`Inhibition of AMPKa isoforms by RNAi is effective at the
`Fig. 4.
`mRNA and protein levels. RNA and cellular lysates isolated from U373
`cells treated with no siRNA oligos (NT) or with siRNA oligos against
`AMPKa1 or AMPKa2 under normoxia (white bars or N) or hypoxia
`(dark bars or H). A: AMPKa1 isoform mRNA expression. B: AMPKa2
`isoform mRNA expression. C: Immunoblots on cellular lysates for
`AMPKa1, AMPKa2, or b-actin. (**P  0.01.)
`
`measured by ELISA assay, Fig. 6B. Protein lysates from
`the experiment were also evaluated for VEGF protein
`levels by ELISA assay, and AMPKa2 repression resulted
`in similar decrease of VEGF protein production at the
`time of harvest (data not shown). The novel finding that
`AMPKa2 repression selectively affects hypoxia induced
`VEGF expression supports the hypothesis that the
`AMPKa isoforms have distinct functions.
`
`Contribution of AMPKa Isoforms
`to Hypoxia-Induced Transcription Through
`HIF-1 Dependent Mechanisms
`
`the individual
`In order to assess the control of
`AMPKa isoforms on VEGF mRNA transcription, isoform
`
`GLIA DOI 10.1002/glia
`
`Fig. 6. RNAi targeted to AMPKa isoforms affects VEGF mRNA and
`protein expression. Total RNA and cellular lysates isolated from U373
`cells treated with no siRNA oligos (NT) or with siRNA oligos against
`AMPKa1 or AMPKa2 under normoxia (white bars) or hypoxia (dark
`bars). A: Fold change in VEGF mRNA expression as determined by
`qRT-PCR. B: Fold change in total VEGF secretion into conditioned
`media as determined by VEGF ELISA assay. (*P  0.05.)
`
`specific involvement in HIF-1 dependent transcription
`was tested. Figure 7 shows a transient transfection with
`an HIF-1 dependent luciferase reporter normalized to
`the baseline transcription of a CMV-driven beta-galacto-
`sidase reporter. Hypoxia resulted in a 4.5-fold induction
`of HIF-1 dependent luciferase activity at 24 h. AMPKa2
`inhibition by RNAi repressed HIF-1 dependent expres-
`sion by 30% as compared to mock transfected control.
`Interestingly, AMPKa1 RNAi treatment showed a dou-
`bling of HIF-1 dependent luciferase activity compared to
`
`

`

`DISTINCT ROLES FOR AMPKa ISOFORMS IN HYPOXIA
`
`739
`
`and hypoxic conditions, which correlates with the high
`AMPKa2 expression observed by qRT-PCR analysis.
`These data indicate that the hypoxic regulation of
`AMPKa2 occurs in at least three glioblastoma cell lines.
`The U87 cells show no hypoxia regulation, but have a
`constitutive high AMPKa2:AMPKa1 ratio consistent
`with the continuous VEGF production observed. These
`data support the hypothesis that AMPKa2 not only is a
`significant contributor to the response of glioblastoma
`cells to hypoxic conditions, particularly with respect to
`VEGF production, but that the balance between AMPKa1
`and AMPKa2 levels might be one mechanism that contri-
`butes to the hypoxic response in different cells.
`
`AMPKa2 Protein Expression May Be
`Associated with High Tumor Grade
`Clinical Glioma Samples
`
`In a retrospective analysis of low, mid, and high grade
`clinical human glioma samples, archived paraffin slides
`were evaluated for AMPKa1 and AMPKa2 protein expres-
`sion by immunohistochemistry. Representative examples
`are shown in Fig. 9A. AMPKa1 expression did not signifi-
`cantly change with glioma grade. AMPKa2 expression was
`nearly identical in low and mid grade glioma samples.
`However, in high grade glioma samples, AMPKa2 expres-
`sion was increased in individual cells. Maximal AMPKa2
`expression was typically observed in and around areas of
`vascular occlusion and pseudopalisades, Fig. 9B. High
`grade human glioma tumors with areas of pseudopalisades
`have been reported to be hypoxic, and correlate with HIF-
`1a and VEGF expression (Brat et al., 2004; Zagzag et al.,
`2000). Thus, our data suggest that high levels of AMPKa2
`correspond to high grade gliomas and hypoxic microenvir-
`onments defined by pseudopalisading histopathologies.
`
`DISCUSSION
`
`Glioma responses to hypoxic microenvironments are
`likely to affect tumor angiogenesis, progression, and in-
`vasion (Brat et al., 2003). AMPK is a central regulator
`of cellular homeostasis under stress and is activated by
`hypoxia (Hardie et al., 2003). We hypothesized that
`AMPK plays a crucial role in the response to hypoxia in
`glioblastoma. AMPK alpha isoforms are known to be dif-
`ferentially expressed in various tissues (Jorgensen et al.,
`2004; Stapleton et al., 1996), which suggests that the
`AMPK alpha isoforms could have distinct functions. To
`investigate this possibility, we evaluated the regulation
`of AMPK alpha isoforms in the human glioblastoma cell
`line, U373. AMPKa1 showed a constitutive pattern of
`expression in U373 cells cultured under normoxic or
`little AMPKa2 was
`hypoxic
`conditions. Conversely,
`detected in U373 cells in normoxia, but AMPKa2 expres-
`sion at the mRNA and protein level was induced by hy-
`poxia. AMPKa2 was also observed to concentrate in the
`nucleus under hypoxic conditions. The novel observation
`that AMPKa2 is induced in hypoxia, along with its nu-
`indicated that AMPKa2
`clear subcellular localization,
`may have distinct functions from AMPKa1.
`
`GLIA DOI 10.1002/glia
`
`Fig. 7. RNAi targeted to AMPKa isoforms affects HIF-1 dependent
`luciferase transcription. Cellular lysates isolated from U373 cells treated
`with no siRNA oligos (NT) or with siRNA oligo pools against AMPKa1
`(a1) or AMPKa2 (a2) and an HIF-1 dependent luciferase reporter vector
`expression normalized to beta-galactosidase transfection control from
`cellular lysates. (*P  0.05 and **P  0.01.)
`
`mock controls. There was no effect on the total HIF-1a
`protein levels when cells were treated with either AMPK
`alpha isoform siRNA oligos as determined by direct im-
`munoblot (data not shown). These data suggest that at
`least some of the effect on VEGF protein by AMPKa2 is
`through an HIF-1 dependent transcription activity in
`spite of not affecting the total level of HIF-1a protein.
`
`AMPKa2 mRNA Is Regulated by Hypoxia
`in Several Glioblastoma Cell Lines
`
`To determine if AMPKa2 regulation by hypoxia and
`contribution to VEGF production is selective to U373
`cells, three additional human glioblastoma cell lines were
`subjected to hypoxia and RNA isolated for qRT-PCR analy-
`sis, and conditioned medias evaluated for VEGF levels. In
`all cell lines tested, U373, U118, U138, and U87, there
`was no apparent regulation of AMPKa1 mRNA levels by
`hypoxia, Fig. 8A. Interestingly, in addition to U373 cells,
`U118 and U138 cells demonstrated increases in AMPKa2
`mRNA when treated with hypoxia, Fig. 8B. In contrast,
`the U87 cell line showed significant levels of AMPKa2
`mRNA levels in both normoxic and hypoxic conditions.
`To analyze the balance between AMPKa1 and AMPKa2
`levels in each cell line, we calculated the AMPKa2 to
`AMPKa1 ratios as defined by mRNA expression levels,
`Fig. 8C. This analysis was revealing in that U373 cells
`demonstrated the most significant
`increase in the
`AMPKa2:AMPKa1 ratio with hypoxia and that U87 cells
`had a high AMPKa2:AMPKa1 ratio in both normoxic
`and hypoxic conditions. The U118 and U138 appeared
`similar to the U373 line, showing higher AMPKa2:
`AMPKa1 ratios in hypoxia.
`To compare the observed levels of AMPK alpha iso-
`forms with VEGF expression and secretion, the condi-
`tioned medias from the same set of cells were tested for
`VEGF protein by ELISA. The expression and secretion
`of VEGF was upregulated in all three glioblastoma lines
`demonstrating hypoxic increases in AMPKa2, namely,
`U373, U118, and U138 cells, albeit at varying levels,
`Fig. 8D. Interestingly, U87 cells demonstrated a continu-
`ous and high level of VEGF production in both normoxic
`
`

`

`740
`
`NEURATH ET AL.
`
`Fig. 8. AMPKa2 is regulated by hypoxia in multiple glioblastoma cell
`lines. Glioblastoma cell lines treated normoxia (white bars) or hypoxia
`(dark bars). A: AMPKa1 isoform mRNA expression by qRT-PCR. B:
`
`AMPKa2 isoform mRNA expression by qRT-PCR. C: Ratio of AMPKa2
`mRNA expression to AMPKa1 mRNA expression. D: Total VEGF secre-
`tion into conditioned media as determined by VEGF ELISA assay.
`
`Hypoxia has been shown to regulate the expression of
`genes required for cellular adaptation to stress (Helfman
`and Falanga, 1993; Shih and Claffey, 1998). One of the
`classic hypoxia regulated genes is the angiogenic growth
`factor VEGF. We have previously shown that VEGF is
`significantly induced in U373 cells in response to hy-
`poxia (Claffey and Robinson, 1996). Chemical inhibition
`of AMPK with 5-iodotubericidin completely abolished
`the hypoxic induction of VEGF mRNA. Since VEGF is
`regulated transcriptionally in hypoxia through HIF de-
`pendent mechanisms (Fedele et al., 2002; Semenza,
`1999), we utilized an HIF-1 dependent luciferase reporter
`construct to evaluate whether AMPK modulates HIF-1
`dependent transcription. Repression of AMPK with 5-
`iodotubericidin abrogated the HIF-1 dependent luciferase
`activity. Investigative studies aimed at determining the
`role for AMPK in hypoxic responses have incorporated
`the overexpression of dominant negative AMPKa1 or
`AMPKa2 isoforms, or chemical inhibitors (Hwang et al.,
`2004; Lee et al., 2003; Nagata et al., 2003; Yun et al.,
`2005). These studies are limited in their ability to assign
`specific activities to AMPK alpha isoforms due to the
`likely saturation of the beta and gamma subunits with
`dominant negative overexpression and lack of isoform
`specificity of chemical inhibitors (Stapleton et al., 1996;
`
`Woods et al., 2000). Our findings with AMPK chemical
`inhibition confirm these previous reports, but have the
`same limitations to distinguish AMPK alp