[CANCER RESEARCH 61, 2154–2161, March 1, 2001]

HuR, a RNA Stability Factor, Is Expressed in Malignant Brain Tumors and Bindsto Adenine- and Uridine-rich Elements within the 3* Untranslated Regions ofCytokine and Angiogenic Factor mRNAs1

L. Burt Nabors, G. Yancey Gillespie, Lualhati Harkins, and Peter H. King2

Departments of Neurology [L. B. N., P. H. K.], Neurosurgery [G. Y. G.], and Physiology and Biophysics [P. H. K.], University of Alabama at Birmingham, Birmingham, Alabama35233-7340, and Birmingham Veterans Affairs Medical Center [L. B. N., L. H., P. H. K.], Birmingham, Alabama

ABSTRACT

Tumors of the central nervous system (CNS) often have sustainedexpression of labile genes, including angiogenic growth factors and im-munosuppressive cytokines, which promote tumor progression. Stabiliza-tion of the RNA transcripts for these genes, such as vascular endothelialgrowth factor (VEGF), is an important molecular pathway for this up-regulation. HuR, a member of the Elav family of RNA-binding proteins,has been implicated in this pathway through its binding to adenine anduridine (AU)-rich stability elements (ARE) located in the 3* untranslatedregions (3*-UTRs) of the mRNA. Whereas three of the Elav family mem-bers (Hel-N1, HuC, and HuD) are restricted to young and mature neu-rons, HuR is more broadly expressed, including proliferating cells of thedeveloping CNS. Because RNA stabilization of labile genes may promotetumor growth, we analyzed and compared the expression pattern of HuRin 35 freshly resected and cultured CNS tumors to determine whetherthere was any correlation with tumor grade or histological type. We foundthat HuR mRNA was consistently expressed in all of the tumors, regard-less of cell origin or degree of malignancy. Using a novel HuR-specificpolyclonal antibody, we found that strong HuR protein expression waslimited to high-grade malignancies (glioblastoma multiforme and medul-loblastoma). Within the glioblastoma multiforme, prominent HuR expres-sion was also detected in perinecrotic areas in which angiogenic growthfactors are up-regulated. To further define its role as a potential RNAstabilizer, we analyzed whether HuR could bind to the stability motifswithin the 3*-UTRs of cytokines and growth factors linked to brain tumorprogression. We used a novel ELISA-based RNA binding assay andfocused on the 3*-UTRs of angiogenic factorsVEGF, COX-2, and (inter-leukin) IL-8 as well as the immunomodulating factorsIL-6, transforminggrowth factor (TGF)-b and tumor necrosis factor (TNF)-a as potentialRNA ligands. Our results indicated overall a very high binding affinity tothese RNA targets. A comparison of these ligands revealed a hierarchy ofbinding affinities with the angiogenic factors, and TGF-b showing thehighest (Kd of 1.8–3.4 nM), and TNF-a the lowest (Kd of 18.3 nM). Theexpression pattern of HuR, coupled with the RNA binding data, stronglysuggests a role for this protein in the posttranscriptional regulation ofthese genes in CNS tumors.

INTRODUCTION

Posttranscriptional regulation is emerging as an important controlpoint for gene expression in tumors. Many growth factors and cyto-kines integral to tumor proliferation and angiogenesis have ARE3

within the 39-UTR that govern transcript half-life (1, 2). Modulationof RNA stability can have a significant impact on mRNA abundanceand subsequent protein expression (3). In malignant gliomas, forexample, stabilization of VEGF mRNA is a major pathway for itsup-regulation in the hypoxic state (4, 5). Because of this adaptivemechanism, the tumor can then sustain growth by forming new bloodvessels. In another example, stabilization of autocrine-producedTNF-a mRNA in epithelial cancer cells may contribute to the ac-quired resistance of these cells to the cytolytic effects of exogenousTNF-a (6). Recently, HuR, a member of the Elav family of RNA-binding proteins, has been identified as a potentialtrans-acting factorin RNA stabilization. This observation stems from two lines of evi-dence. First, members of the Elav family have a strong bindingaffinity to AREs (7, 8). Second, the overexpression of HuR, in certaincell systems, has led to the stabilization of c-fosand VEGF transcriptsthat contain AREs in their 39-UTR (9–11). A recent study of cyclin Aand B1 regulation linked HuR with the RNA stabilization of thesegenes and enhanced tumor proliferation (12). The potential role ofHuR as a stabilizer of growth-related mRNAs is consistent with itsstrong expression pattern in proliferating cells of the developing CNS(13). As demonstrated in mutational studies inDrosophila, moreover,the Elav family is essential for the normal growth and development ofthe CNS (14, 15). Aberrant RNA stabilization of growth-relatedgenes, however, may promote the uncontrolled growth of a CNStumor. We were, therefore, interested in determining whether theexpression pattern of HuR in primary brain tumors correlated withhistological type or grade. To further delineate the potential role ofHuR in posttranscriptional gene regulation in brain tumors, we thenselected angiogenic and cytokine genes that are up-regulated in CNStumors and determined whether HuR could bind to their 39-UTRs.

MATERIALS AND METHODS

Tissues and Cell Lines.Paraffin-embedded surgical samples of humantissues were obtained from the University of Alabama Tissue ProcurementOffice. Cryopreserved fresh tissue samples were provided by the Division ofNeurosurgery Tissue Bank, University of Alabama. These latter tissues werecollected aseptically from surgically resected brain tumors that were deter-mined to be unnecessary for diagnosis. Tissues were debrided of blood andnecrotic material, subdivided into 100–200-mg portions and snap-frozen untilused. All of the samples were encoded to prevent patient identification.Tumors were graded according to the criteria established by WHO. Low-gradeastrocytomas, including pilocytic astrocytomas, are considered grades I and II,whereas GBM is grade IV. Normal brain tissue was obtained at autopsy.Glioma cell lines included D54 MG, U105 MG, U87 MG, U373 MG, D65MG, and U251 MG (16–18). Astrocytes cultured from human epilepsy pa-tients and the small-cell lung cancer cell line, NCI-N417, were used ascontrols. Glioma cells were cultured in DMEM mixed 1:1 with Ham’s nutrientmixture F-12 (DMEM/F12) supplemented to 2 mM with L-glutamine.

Preparation of Recombinant Proteins. The cDNA for HuR was sub-cloned into the pGEX-5X-1 vector (Amersham Pharmacia Biotech, Piscat-

Received 6/14/00; accepted 1/4/01.The costs of publication of this article were defrayed in part by the payment of page

charges. This article must therefore be hereby markedadvertisementin accordance with18 U.S.C. Section 1734 solely to indicate this fact.

1 Supported by a Department of Veterans Affairs VISN 7 Career Development Award(to L. B. N.), by American Cancer Society RPG-97-111-01-CCE (to P. H. K.) and IR6-60-001-41 (to L. B. N.), by NIH CA71933 (to G. Y. G.), and grants from the PediatricBrain Tumor Foundation of the United States and from JCR Biopharmaceutical, Inc.Generous support provided by the Judge Richard Holmes Brain Tumor Research Fund.

2 To whom requests for reprints should be addressed, at Department of Neurology,University of Alabama at Birmingham, 1235 Jefferson Tower, Birmingham, AL 35233-7340. Phone: (205) 975-8116; Fax: (205) 934-0928.

3 The abbreviations used are: AU, adenine and uridine; ARE, AU-rich element; CNS,central nervous system; UTR, untranslated region; GBM, glioblastoma multiforme;VEGF, vascular endothelial growth factor; TNF, tumor necrosis factor; GAPDH, glycer-aldehyde-3 phosphate dehydrogenase; TGF, transforming growth factor; IL, interleukin;

GST, glutathioneS-transferase; RPA, RNase protection assay; Fc, frontal cortex; Oc,occipital cortex;Kd, dissociation constant.

2154

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

away, NJ) in frame with the GST tag and verified by sequence and restrictionanalysis. The plasmid was transformed into the BL21 strain ofEscherichia coliand induced as described previously (19). The protein was purified using aglutathione-Sepharose column based on the manufacturer’s specifications(Amersham Pharmacia Biotech). The protein was dialyzed extensively in PBSand quantified by the DC protein assay (Bio-Rad, Hercules, CA) using BSA asa standard. The biosynthesis of recombinant proteins HuC, HuD, Hel-N1, andHuR using the histidine tag have been reported previously (19).

RPA. TheHuRriboprobe template was obtained by subcloning aPstI/SacIdigestion fragment of theHuR cDNA (U38175; nucleotide 159–416; Ref. 8)into Bluescript SK (Stratagene, La Jolla, CA). Hel-N1, HuC, and HuD tem-plates were used as described previously (20). TheGAPDHgene was used asan internal standard for comparison purposes. Riboprobes were synthesizedwith [a-32P]UTP as described previously (20). RNA hybridizations and RNasedigestions were carried out as described previously (20, 21). To quantify RNAlevels, test and control bands were evaluated by phosphorimaging (MolecularDynamics, Sunneyvale, CA).

Preparation of Polyclonal Antibodies to HuR. Rabbit polyclonal anti-bodies to HuR were prepared commercially (Genosys, The Woodlands, TX).Animals were immunized with a synthetic peptide to the HuR NH2 terminuscontaining the initial 15 amino acid residues (MSNGYEDHMAEDCRG). Thisregion of HuR diverges significantly from the other human homologues ofElav (HuC, HuD, and Hel-N1; Ref. 22). Serum was affinity-purified on aHuRNH2 terminus peptide column, concentrated, and then stored in 50% glycerol.

Immunocytochemistry. Tissue sections from tumor samples (obtained atthe time of tumor biopsy or resection) were fixed in 10% buffered formalin,processed, and embedded in paraffin. Sections were cut at 8mm and driedovernight in a 50°C oven. Sections were deparaffinized in xylene and dehy-drated through a series of alcohols and hydrated in water. Sections werepretreated with pepsin (2.5 mg/ml; Biogenex Laboratories, San Ramon, CA) at37°C for 4 min and were rinsed with distilled water. Sections underwentepitope recovery using a modified protocol for Antigen Retrieval Citra Buffer(Biogenex Laboratories, San Ramon, CA). Briefly, citra buffer was prewarmedto 80°C in microwave oven. Sections were incubated in prewarmed bufferimmersed in a 48°C water bath for 150 min. Sections were rinsed and blockedwith 3% hydrogen peroxide (Sigma, St. Louis, MO). Sections were rinsed withTris buffer (pH 7.6) and blocked with FC receptor block (Innovex Bioscience,Richmond, CA) for 15 min at room temperature. Purified rabbit anti-HuRantibody was applied at 1:80 dilution and incubated in the refrigerator (at2–8°C) overnight. Sections were incubated with horseradish peroxidase-enhancing wash buffer (Innovex Bioscience). Detection and amplification ofsignal was achieved using Non-biotin Amplification Kit (Zymed Laboratories,South San Francisco, CA). Signal was developed using liquid DAB (InnovexBioscience), counterstained with hematoxylin I (Richard-Allan Scientific,Kalamazoo, MI), and rinsed with tap water. Sections were dehydrated througha series of alcohols into xylene and mounted with Secure Mount (Biosciences,Inc, Swedesboro, NJ).

ELISA-based RNA-binding Assay. The 39-UTR riboprobe templates ofthe following growth factors and cytokines (accession number; nucleotidesequence) were obtained by reverse transcription-PCR from the U251 malig-nant glioma cell line: VEGF (Y08736; 1282–1913), COX-2 (U04636; 6904–7483), IL-6 (M54894; 691-1102), IL-8 (M28130; 3389–4198), TGF-b

(M60315; 1697–2859), and TNF-a (M10988; 795-1565). PCR products werecloned into pCR2.1-TOPO (Invitrogen, Carlsbad, CA). All of the clones wereverified by restriction mapping and sequence analysis. Biotinylated riboprobeswere synthesized as described previously (23). All of the probes were verifiedby visualization on an ethidium-stained polyacrylamide gel. Riboprobe con-centrations were determined using Ribogreen RNA Quantitation reagent (Mo-lecular Probes, Eugene, OR) in a fluorometer. The ELISA-based RNA bindingassay has been described previously (23). Briefly, 500 ng of recombinantGST-HuR fusion protein was plated in a volume of 50ml into individualELISA wells and allowed to adsorb overnight. On the following day, the plateswere washed, and biotinylated riboprobes were added at various amounts(range, 0.1–3.0 pmol) to 50ml of RNA-binding buffer (25 mM HEPES, 0.5 mMEGTA, 100 mM NaCl, 0.5 mM DTT, 4 mM MgCl2, 20 mM KCl, 0.05% NP40,0.5 mg/ml yeast tRNA, 0.05 mg/ml poly(A) RNA, 0.4 mM VRC, and 5%glycerol) and placed in the ELISA well for 30 min at room temperature. Afterextensive plate washing, a streptavidin-alkaline phosphatase conjugate wasadded and allowed to incubate for 30 min at room temperature. The plates were

then developed with ap-nitrophenyl phosphate (Sigma, St. Louis, MO) solu-tion, and the absorbance was determined at 405 nm. These values (in arbitraryunits) were plotted against RNA concentrations, and binding curves wereestimated with DeltaGraph software (SSPS, Chicago, IL). At first approxima-tion, the probes were in excess of the protein, and affinity constants wereestimated from the curves.

RESULTS AND DISCUSSION

The HuR Gene, in Contrast to the Neuronal-associated ElavHomologues, Is Ubiquitously Expressed in Primary CNS Tumors.Twenty-nine surgical samples of primary brain tumors and six cul-tured glioma cell lines were evaluated for expression ofHuRusing aRPA as described previously (20). We analyzed a diversity of tumortypes reflecting different cell lineages and degrees of invasiveness.Tumors were selected based on either being aggressively infiltrativeinto brain (GBM and medulloblastoma), low infiltrative capacity(pilocytic and low-grade astrocytoma) or generally confined extra-axial masses (meningioma). We also evaluated tumors capable ofdifferentiating down glial (GBM, pilocytic astrocytoma, and ependy-moma), neuroglial (medulloblastoma), or arachnoidal lineage (menin-gioma). As shown in Fig. 1, we detected nearly universal expressionof HuR in primary brain tumors and cultured tumor cells, regardlessof cell origin or degree of invasiveness. The only sample in whichHuRwas absent also had a weakGAPDHsignal (Fig. 1,Gb,Lane 9).Densitometric analysis of the bands, compared with the internalcontrol GAPDH, indicated that the highestHuR expression was ob-served in meningiomas, suggesting that elevated RNA levels did notcorrelate with higher-grade malignancies (Fig. 2). This observation,however, is based on the assumption thatGAPDH is equally ex-pressed among tumors. The other samples, including GBMs andmedulloblastomas, had lower but similarHuRRNA levels. The ubiq-uitous expression pattern is consistent with previous observations forHuR in lung cancer (24). Because the neuronal-associated homo-logues of the Elav family, includingHel-N1, HuC, andHuD, arerestricted to neurons and neuroendocrine tumors (20, 21, 25, 26), wewere interested in comparing the relative expression patterns of thesehomologues to that ofHuR in brain tumors of different cell origins.We quantified the RNA level of each family member within a tumorsample and expressed the value as a percentage of total Elav RNAexpression for that sample (Fig. 3). Because tissue samples werelimited, we analyzed only those with sufficient RNA to test all four ofthe Elav members. Thus, the percentage values shown in Fig. 3Bindicate relative rather than absolute expression levels. RepresentativeRPAs for each family member with different tumor samples areshown (Fig. 3A). Three protected bands were consistently observedwith HuC and are most likely attributable to alternative splicing (22).As shown in Fig. 3B, cells of glial or arachnoidal lineage, includingprimary astrocytes, malignant glioma cell lines, and meningiomas,had nearly exclusive expression ofHuR. One of the cell lines (D54MG) and two meningioma samples expressed low levels ofHel-N1but none of the other neuronal-associated homologues. The data,therefore, indicate that these glial and arachnoidal tumor types havethe capability of expressing neuroendocrine markers. The GBM sam-ples, on the other hand, had expression of all of the three neuronal-associated members, withHuR predominating. Because GBMs arehighly invasive, some of the biopsy specimens may have containednormal cortex, thus accounting for detection of the neuronal-associ-ated homologues. To help discern this possibility, we compared therelativeElav expression patterns between normal cortex (Fc and Oc)and the tumor samples. The expression patterns of theElav membersin Fc and Oc, withHuC showing the highest andHuR the lowestrelative RNA levels, corresponded well with previousin situ hybrid-

2155

HuR IS EXPRESSED IN MALIGNANT BRAIN TUMORS

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

ization studies in mouse brain (27). The predominant expression ofHuR in all of the GBMs is, therefore, reflective of tumor-specificexpression and not normal brain tissue [compareHuR levels in GBM(Fig. 1, Gb, Lanes 1–7) to Fc and Oc controls]. In four of the GBMsamples (Lanes 3,5, 6, and7), all of the three neuronal-associatedhomologues were detected, withHuC andHuD levels being compar-atively higher than that forHel-N1. This suggested the possibility thatthese specimens contained variable numbers of neuroglial cells. Insamples 2 and 4, however, neitherHel-N1 nor HuD was detected,indicating selective expression of some neuronal-associated markerswith these GBM tumor specimens. The two pilocytic astrocytomas,considered lower-grade glial tumors, contained reducedHuR levelsrelative to the GBMs but higher levels than did cortex (Fig. 1,PA,

samples 1–2). The presence of all of the three neuronal-associatedhomologues in pilocytic astrocytoma, with a pattern similar to Fc andOc, suggests that these specimens may also have included normalneuroglial elements. The predominance ofHuD expression, coupledwith low HuC expression, was also observed in two of the threeependymomas (Fig. 1,EP, Lanes 2and3), indicating that this tumortype also has the capacity to express neuronal-associated markers.

In contrast to glial and arachnoidal tumors, the medulloblastomatumor, which is capable of neuronal and/or glial differentiation, hadpredominant expression of the three neuronal-associated homologues(Fig. 3B,MD, samples 1–5). The expression pattern of these homo-logues, however, substantially differed from normal cortical tissues.In sample 1, for example,Hel-N1 was the only neuronal homologuedetected, similar to the glioma cell lines and meningiomas. On thebasis of developmental studies inXenopusand mouse, sole expressionof this neuronal homologue reflects a very early stage of neuronaldifferentiation (27, 28). This expression pattern was also observed inthe poorly neuroendocrine-differentiated variant of small-cell lungcancer (20). In the other four medulloblastoma samples,HuD expres-sion predominated (in contrast toHuC in normal cortex), suggestingmore advanced neuronal maturation (13, 20, 27, 28). Higher levels ofHuD have been correlated with favorable prognostic features in astudy of neuroblastoma tumors (21). Although the number of speci-mens investigated here were too few to establish definitive conclu-sions, the variable expression pattern of neuronal-associatedElavfamily members within medulloblastomas may be an important ge-netic footprint of tumor grade and prognosis.

In summary, all of the primary brain tumors expressedHuRRNA,with the highest levels detected in tumors of glial and arachnoidorigin. HeL-N1,HuC, andHuD, on the other hand, were detected atthe highest levels in medulloblastoma. The variable detection of thesehomologues in GBMs and ependymomas underscores the capacity ofthese tumors for expressing markers of neuronal differentiation.

HuR Is Expressed at the Protein Level in Highly ProliferativeBrain Tumors. The conserved RNA-binding domains within theElav family have made it difficult to develop specific antibodies to

Fig. 1. Results of a RPA analyzing the expres-sion of HuR in CNS tumors.Arrows, protectedbands forHuRandGAPDH.Gb, GBM;Gc, gliomacell line; Ep, ependymoma;Md, medulloblastoma;Pa, pilocytic astrocytoma;Mn, meningioma;P, un-protected probes (HuR and GAPDH);C, control(NCI-N417). For the Gc samples:Lane 1, D54 MG;Lane 2, U105 MG;Lane 3, U87 MG;Lane 4, U373MG; Lane 5, D65 MG; andLane 6, U251 MG.

Fig. 2. Analysis ofHuRexpression levels in various CNS tumors using densitometry.All of the values are expressed as a percentage of the internal GAPDH control.Gc, gliomacell line; Gb, GBM; Pa, pilocytic astrocytoma;Md, medulloblastoma;Mn, meningioma;Ep, ependymoma; Fc, frontal cortex; Oc, occipital cortex.

2156

HuR IS EXPRESSED IN MALIGNANT BRAIN TUMORS

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

individual members. The monoclonal antibody Mab16A11,e.g., im-munoreacts with all of the three neuronal-associated homologues (29).Polyclonal anti-Hu (or ANNA-1) serum from patients with paraneo-plastic neurological disease, on the other hand, reacts with all of thefour family members (19, 24, 27). Thus, to analyze HuR proteinexpression in primary brain tumors, we developed an anti-HuR poly-clonal rabbit antibody using a unique peptide sequence in the NH2

terminus of the protein. This portion of the Elav protein is the leastconserved among the different family members (22). We tested themonospecificity of this antibody by Western blot analysis using all ofthe four recombinant Elav proteins (Fig. 4). A reactive band wasobserved for HuR but not for the neuronal-associated members Hel-N1, HuC, or HuD (Fig. 4A,upper panel). Preabsorption of theanti-HuR IgG with recombinant HuR abrogated this immunoreactivity(not shown). The blot was stripped and probed with biotinylatedanti-Hu paraneoplastic IgG. Fig. 4A, lower panel, demonstrates thetypical immunoreactive pattern with all of the four family members(19). We next tested the antibody by immunocytochemistry usingU251 glioma cells in culture because this cell line robustly expressedHuRRNA (see Fig. 1B:Gc, Lane 6). As shown in Fig. 4B,left panel,there was intense staining, predominantly in the nucleus and to alesser extent in the cytoplasm. This localization pattern has beenobserved with the neuronal-associated homologues, using either an-ti-Hu paraneoplastic serum or Mab16A11 (29–31). Preabsorption ofthe antibody with recombinant HuR protein abolished the reactivity,underscoring the specificity of this immunostaining (Fig. 4B, rightpanel). With this novel antibody, we then analyzed different primarybrain tumors for HuR expression (summarized in Table 1). We choseimmunocytochemistry rather than Western blot analysis to allowaccurate localization of immunoreactivity within the tissue sample.

Fig. 3. Comparison of RNA expression levels for all of the four Elav family members (Hel-N1, HuC, HuD, and HuR) in various CNS tumors and tissues.A, representative RPAsfor each family member (and GAPDH control) as indicated.B, the relative RNA expression levels for each Elav family member in the tissue samples. Densitometric values for eachmember (and the GAPDH internal control) were determined by phosphorimaging and then summed to obtain a total Elav RNA expression value for the tissue sample. The relativeElav RNA expression for an individual family member represents the percentage of the total Elav RNA expression for that tissue sample.As, primary astrocyte in culture;Gc, gliomacell line; Mn, meningioma;Gb, GBM; Pa, pilocytic astrocytoma;Md, medulloblastoma;Ep, ependymoma;Fc, frontal cortex;Oc, occipital cortex.

Fig. 4. Analysis of a novel, NH2 terminus-specific polyclonal anti-HuR antibody.A,Western blot of recombinant Elav proteins indicating that the anti-HuR IgG immunoreactsonly with recombinant HuR and not with the neural-specific members, HeL-N1, HuC, orHuD (top panel). The same blot was stripped and probed with anti-Hu IgG showingimmunoreactivity with all of the four family members (bottom panel).B, immuncyto-chemical analysis of U251 glioma cells in culture with the anti-HuR IgG indicating theimmunoreactivity to be primarily nuclear in distribution [(1)panel]. The staining isabolished by preabsorbing the antibody with recombinant HuR [(2)panel].

2157

HuR IS EXPRESSED IN MALIGNANT BRAIN TUMORS

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

The most striking finding was the intense staining of the high-gradetumors (GBMs and medulloblastomas) and the minimal reactivity inthe low-grade tumors (pilocytic astrocytomas and meningiomas; Fig.5). The predominant nuclear pattern of immunoreactivity in the GBMsand medulloblastomas was similar to that of U251 cells (Fig. 4B), andpreabsorption of the antibody with recombinant HuR abrogated thisreactivity as shown in Fig. 5,the (2) panels. This pattern wasconsistently observed in all of the eight GBM tumors analyzed (seeTable 1). Histologically, the GBM and medulloblastoma tumors dem-onstrated high cellular density with many mitotic figures. The asso-

ciation between HuR protein expression and cell proliferation wasrecently demonstrated in colorectal carcinoma cells in which reducedHuR expression levels (by transfection of antisense HuR) correlatedwith reduced tumor growth (12). Reduced stabilization of cyclin Aand B1 mRNAs was implicated in the altered growth phenotype.Another recent report of HuR expression in malignant lung tumorsalso indicated a correlation of HuR protein expression with degree ofmalignancy (32).

Extensive vascular proliferation and areas of necrosis typical for aGBM were also observed (Fig. 6). Interestingly, tumor cells adjacentto regions of necrosis showed intense staining. These regions, whichare relatively hypoxic, have increased expression of angiogenic genesincluding VEGF, COX-2, andIL-8 (33–35). The colocalization andup-regulation of HuR in these regions meshes with the previousobservation in other tumor cells in which HuR participates in thehypoxia-induced stabilization and up-regulation ofVEGF (9). Incontrast to the high-grade tumors, very little immunostaining was

Table 1 Summary of anti-HuR immunostaining in primary brain tumors

Tumor HuR (1)/total

GBM (WHO IV/IV) 8/8Astrocytoma (WHO I–II/IV) 0/3Meningioma 1/9Medulloblastoma 1/1

Fig. 5. Immunohistochemical anlaysis of pri-mary brain tumors using the anti-HuR antibody.Prominent immunoreactivity [(1)panels] is seenin tumor cells of GBM (Gb) and medulloblastoma(Md). Insets, the predominant nuclear localizationof HuR with lesser cytoplasmic staining. Reactivitywithin pilocytic astrocytoma (Pa) is seen only incells that resemble reactive astrocytes rather thantumor cells. In meningioma (Mn), the staining isprimarily limited to endothelial-like cells ratherthan tumor cells. The immunostaining is abolishedby presabsorbing the serum with recombinant HuR[(2) panels].Bar, 50 mm.

2158

HuR IS EXPRESSED IN MALIGNANT BRAIN TUMORS

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

detected in low-grade astrocytoma and meningioma tumors. Thosecells that were positive, moreover, were not the primary tumor and didnot have a predominant nuclear pattern. In pilocytic astrocytoma, forexample, the subset of cells that stained positive was morphologicallydifferent from the tumor cells. They demonstrated a plump cytoplasmwhere most of the immunostaining was localized. They resembledreactive astrocytes rather than neoplastic cells. In the meningiomatumor, the small subset of positively staining cells morphologicallyresembled endothelial cells rather than tumor cells. This pattern of lowHuR immunoreactivity was observed in three of three low-gradeastrocytomas and eight of nine meningiomas. The dissociation be-tweenHuR RNA and protein expression in low-grade brain tumorsand normal brain has also been observed in normal, nonproliferatingmouse tissues, suggesting that HuR is regulated at the translationallevel (27). In summary, the correlation of HuR expression with highlyproliferative CNS tumors meshes with other findings in other tumorsand supports a role for HuR in the posttranscriptional regulation ofgene expression in malignant brain tumors.

HuR Displays a Hierarchy of Binding Affinity for 3 *-UTRs ofGrowth Factor and Cytokines Linked to Brain Tumor Progres-sion and Angiogenesis.In non-CNS tumor cells, HuR has beenlinked to the stabilization and up-regulation ofVEGF and c-fosmRNAs through its interaction with AREs in the 39-UTR (9–11). Ourobservation that HuR is expressed in proliferating CNS tumorsprompted us to determine whether HuR could bind to the 39-UTRs ofcytokine and growth factor mRNAs that have similar expressionpatterns.COX-2 andIL-8, for example, along withVEGF, are angio-genic genes with ARE-bearing 39-UTRs that are up-regulated inperinecrotic regions within a GBM (33–36). To analyze this RNAbinding, we used a novel, ELISA-based assay as described recently(23). This assay permits rapid assessment and comparison of therelative binding affinity of HuR for different RNA ligands. The39-UTR probes are schematically diagrammed in Fig. 7A and show thenumber and distribution of AREs (as defined by an AUUUA orAUUUUA motif). We used the VEGF regulatory sequence (VRS)within the VEGF 39-UTR, which has previously been shown to bindHuR by gel-shift assay (9). HuR had a high binding affinity to all ofthe three angiogenic factor 39-UTRs (Fig. 7B). No binding wasdetected to the control transcript,pBSK, underscoring the specificity

of this binding. The affinity toIL-8 (Kd of 1.8 nM) was slightly higherthan forVEGF andCOX-2 (Kd of 3.4 and 3.1 nM, respectively). Aswith VEGF, however, we analyzed only a segment of the entire39-UTR for COX-2, which is;2.5 kb. TheIL-8 39-UTR was fulllength based on published sequence (37). BothCOX-2 andIL-8 havenot previously been shown to bind HuR, although the 39-UTR probescontained a similar number of AREs as that ofVEGF (Fig. 7A). Thecolocalization of HuR with the angiogenic factor genes in GBMtumors, coupled with these high binding affinities, supports, but doesnot prove, a role for HuR in their posttranscriptional regulation. Therelative hypoxia present in the perinecrotic tumor zones of the GBMraises the intriguing possibility that HuR may be involved in thehypoxia-induced stabilization ofVEGF in GBM (as it is in other

Fig. 6. Immunohistochemical analysis of aGBM demonstrates intense HuR immunoreactivityin areas of tumor and endothelial proliferation (p)and adjacent to regions of tumor necrosis (N).Bar,100 mm.

Fig. 7. HuR binds with high affinity to the stability regions within the 39-UTRs ofcytokines and angiogenic growth factor mRNAs linked to glioma progression.A, aschematic drawing of the 39-UTR probes analyzed using an ELISA-based RNA bindingassay (23). The number and relative distribution of AREs (as defined by an AUUUA orAUUUUA motif) within each probe is illustrated.B, the binding curves for angiogenicfactor 39-UTRs including VEGF, COX-2, and IL-8. A control riboprobe, pBSK, which hasno AREs is shown as well.C, a representative binding curve for the immunomodulatorycytokines, TGF-b, IL-6, and TNF-a. All of the binding experiments were repeated intriplicate.

2159

HuR IS EXPRESSED IN MALIGNANT BRAIN TUMORS

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

tumor cells; Refs. 4, 5, 9, 35, 38). Although posttranscriptional reg-ulation of COX-2 is an important mechanism for its up-regulation inother tumors, the role for this pathway inCOX-2 (andIL-8) expressionin GBMs has not been elucidated (39–41).

Whereas the binding affinities for the angiogenic factor transcriptswere high, those of the immunomodulating cytokines,TGF-b, TNF-a,andIL-6, were more variable. These autocrine factors play an impor-tant role in the interplay between the immune system and the braintumor, and may confer a growth advantage to the tumor (42, 43).TGF-b, e.g.,had one of the highest affinities for HuR (Kd of 2.3 nM)whereasTNF-a andIL-6 had the lowest (18.3 and 10.6, respectively).TGF-b had an;8-fold higher-binding affinity for HuR than didTNF-a. The TGF-b 39-UTR also had the fewest AREs (six in total)which were dispersed over the entire probe, whereas theTNF-a hadeight clustered AREs (Fig. 7A). This discrepancy in binding affinitiessuggests that optimal RNA binding may be related to otherciselements within the UTR or its secondary structure, rather than to theabsolute number of AREs. The original RNA-binding studies withHel-N1, using an immunoprecipitation technique, indicated that thisfamily may have binding preferences for other U-rich motifs (7). TheTGF-b 39-UTR had many of these additional motifs (not shown).Additionally, the Elav proteins have three RNA recognition motifs,each of which may bind to a different RNA ligand (44). The maximalsaturation levels for each of the curves (including the angiogenicfactor mRNAs), however, were similar, indicating that the RNAligands were likely binding to the same site, but with differentaffinities. Qualitatively, HuR and other members of the Elav familyhave been shown to bind a number of 39 UTRs using different assays(7, 8, 25, 26, 45–48). The ELISA-based RNA binding assay, althoughstill in vitro, permits a rapid comparison of binding affinities withdifferent RNA ligands (23). The hierarchy of affinities observed heresuggests that ARE-bearing 39-UTRs are not alike with respect to HuRbinding. Specificity of HuR binding, therefore, may partly be dictatedby these affinity differences.

Our results indicate that: (a)HuR, unlike the neuronal-associatedElav homologues, is ubiquitously expressed in tumors of the CNS atthe RNA level; (b) HuR protein is strongly expressed in highlyproliferative tumors including GBMs and medulloblastomas in con-trast to weak expression in low-grade brain tumors; (c) intense HuRimmunoreactivity was observed in perinecrotic regions within theGBMs in which angiogenic growth factors are up-regulated; and (d)HuR can bind with high affinity to the 39-UTRs of critical cytokineand growth factor mRNAs involved in CNS tumor proliferation andangiogenesis, suggesting a posttranscriptional regulatory role for thisprotein in malignant brain tumors.

REFERENCES

1. Shaw, G., and Kamen, R. A conserved AU sequence from the 39 untranslated regionof GM-CSF mRNA mediates selective mRNA degradation. Cell,46: 659–667, 1986.

2. Chen, C. A., and Shyu, A. AU-rich elements: characterization and importance inmRNA degradation. Trends Biochem. Sci.,20: 465–470, 1995.

3. Ross, J. mRNA stability in mammalian cells. Microbiol. Rev.,59: 423–450, 1995.4. Stein, I., Neeman, M., Shweiki, D., Itin, A., and Keshet, E. Stabilization of vascular

endothelial growth factor mRNA by hypoxia and hypoglycemia and coregulation withother ischemia-induced genes. Mol. Cell. Biol.,15: 5363–5368, 1995.

5. Damert, A., Machein, M., Breier, G., Fujita, M. Q., Hanahan, D., Risau, W., andPlate, K. H. Up-regulation of vascular endothelial growth factor expression in a ratglioma is conferred by two distinct mechanisms. Cancer Res.,57: 3860–3864, 1997.

6. Wang, E., Ma, W. J., Aghajanian, C., and Spriggs, D. R. Post-transcriptional regu-lation of protein expression in human epithelia carcinoma cells by adenine-uridinerich elements in the 39-untranslated region of tumor necrosis factor-a messengerRNA. Cancer Res.,57: 5426–5433, 1997.

7. Levine, T. D., FenBiao, G., King, P. H., Andrews, L. G., and Keene, J. D. Hel-N1:an autoimmune RNA-binding protein with specificity for 39uridylate-rich untrans-lated regions of growth factor mRNAs. Mol. Cell. Biol.,13: 3494–3504, 1993.

8. Ma, W., Cheng, S., Campbell, C., Wright, A., and Furneaux, H. Cloning andcharacterization of HuR, a ubiquitously expressed Elav-like protein. J. Biol. Chem.,271: 8144–8151, 1996.

9. Levy, N., Chung, S., Furneaux, H., and Levy, A. Hypoxic stabilization of vascularendothelial growth factor mRNA by the RNA-binding protein HuR. J. Biol. Chem.,273: 6417–6423, 1998.

10. Fan, X. C., and Steitz, J. A. Overexpression of HuR, a nuclear-cytoplasmic shuttlingprotein, increases thein vivo stability of ARE-containing mRNAs. EMBO J.,17:3448–3460, 1998.

11. Peng, S. S-Y., Chen, C-Y. A., Xu, N., and Shyu, A-B. RNA stabilization by theAU-rich element binding protein. EMBO J.,17: 3461–3470, 1998.

12. Wang, W., Caldwell, M. C., Lin, S., Furneaux, H., and Gorospe, M. HuR regulatescyclin A and cyclin B1 mRNA stability during cell proliferation. EMBO J.,19:2340–2350, 2000.

13. Wakamatsu, Y., and Weston, J. A. Sequential expression and role of Hu RNA-bindingproteins during neurogenesis. Development,124: 3449–3460, 1997.

14. Campos, A. R., Grossman, D., and White, K. Mutant alleles at the locuselav inDrosophila melanogasterlead to nervous system defects. J. Neurogenet.,2: 197–218,1985.

15. Yao, K., Samson, M., Reeves, R., and White, K. Geneelav of Drosophila melano-gaster: a prototype for neuronal-specific RNA binding protein gene family that isconserved in flies and humans. J. Neurobiol.,24: 723–739, 1993.

16. Bigner, D. D., Bigner, S. H., Ponten, J., Westermark, B., Mahaley, M. S., Ruoslahti,E., Herschman, H., Eng, L. F., and Wikstrand, C. J. Heterogeneity of Genotypic andphenotypic characteristics of 15 permanent cell lines derived from human gliomas.J. Neuropathol. & Exp. Neurol.,40: 201–229, 1981.

17. Ponten, J., and Westermark, B. Properties of human malignant glioma cellsin vitro.Med. Biol., 56: 184–193, 1978.

18. Ponten, J., and Macintyre, E. H. Long-term culture of normal and neoplastic humanglia. Acta Pathol. Microbiol. Scand.,74: 465–486, 1968.

19. King, P. H., Redden, D., Palmgren, J. S., Nabors, L. B., and Lennon, V. A. Hu antigenspecificities of ANNA-I autoantibodies in paraneoplastic neurological syndromes.J. Autoimmun.,13: 435–443, 1999.

20. King, P. H. Differential expression of the neuroendocrine genes Hel-N1 and HuD insmall-cell lung carcinoma: evidence for down-regulation of HuD in the variantphenotype. Int. J. Cancer,74: 378–382, 1997.

21. Ball, N. S., and King, P. H. Neuron-specificHel-N1 andHuD as novel molecularmarkers of neuroblastoma: a correlation ofHuD messenger RNA levels with favor-able prognosis. Clin. Cancer Res.,3: 1859–1865, 1997.

22. King, P. H. Hel-N2: a novel isoform of Hel-N1 which is conserved in rat neural tissueand produced in early embryogenesis. Gene,151: 261–265, 1994.

23. King, P. H. RNA-binding analysis of HuC and HuD with theVEGF and c-myc39-untranslated regions using a novel ELISA-based assay. Nucleic Acids Res.,28:E20, 2000.

24. Nabors, L. B., Furneaux, H. M., and King, P. H. HuR, a novel target of anti-Huantibodies, is expressed in non-neural tissues. J. Neuroimmunol.,92: 152–155, 1998.

25. Szabo, A., Dalmau, J., Manley, G., Rosenfeld, M., Wong, E., Henson, J., Posner,J. B., and Furneaux, H. M. HuD, a paraneoplastic encephalomyelitis antigen, containsRNA-binding domains and is homologous toelavand sex-lethal. Cell,67: 325–333,1991.

26. King, P. H., Levine, T. D., Fremeau, R. T., and Keene, J. D. Mammalian homologuesof DrosophilaELAV localized to a neuronal subset can bindin vitro to the 39-UTRof mRNA encoding the Id transcriptional repressor. J. Neurosci.,14: 1943–1952,1994.

27. Okano, H. J., and Darnell, R. B. A hierarchy of Hu RNA-binding proteins indeveloping and adult neurons. J. Neurosci.,17: 3024–3037, 1997.

28. Good, P. J. A conserved family ofelav-like genes in vertebrates. Proc. Natl. Acad.Sci. USA,92: 4557–4561, 1995.

29. Marusich, M. F., Furneaux, H. M., Henion, P. D., and Weston, J. A. Hu neuronalproteins are expressed in proliferating neurogenic cells. J. Neurobiol.,25: 143–155,1994.

30. Dalmau, J., Furneaux, H. M., Cordon-Cardo, C., and Posner, J. B. The expression ofthe Hu (paraneoplastic encephalomyelitis/sensory neuronopathy) antigen in humannormal and tumor tissues. Am. J. Pathol.,141: 881–886, 1992.

31. Graus, F., Cordon-Cardo, C., and Posner, J. B. Neuronal antinuclear antibody insensory neuronopathy from lung cancer. Neurology,35: 538–543, 1985.

32. Blaxall, B. C., Dwyer-Nield, L. D., Bauer, A. K., Bohlmeyer, T. J., Malkinson, A. M.,and Port, J. D. Differential expression and localization of the mRNA binding proteins,AU-rich element mRNA binding protein (AUF1) and Hu antigen R (HuR), inneoplastic lung tissue. Mol. Carcinog.,28: 76–83, 2000.

33. Deininger, M. H., Weller, M., Streffer, J., Mittelbronn, M., and Meyermann, R.Patterns of cyclooxygenase-1 and -2 expression in human gliomasin vivo. ActaNeuropathol.,98: 240–244, 1999.

34. Desbaillets, I., Diserens, A. C., Tribolet, N., Hamou, M. F., and Van Meir, E. G.Upregulation of interleukin 8 by oxygen-deprived cells in glioblastoma suggests arole in leukocyte activation, chemotaxis, and angiogenesis. J. Exp. Med.,186:1201–1212, 1997.

35. Shweiki, D., Itin, A., Soffer, D., and Keshet, E. Vascular endothelial growth factorinduced by hypoxia may mediate hypoxia-initiated angiogenesis. Nature (Lond.),359: 843–845, 1992.

36. Desbaillets, I., Diserens, A. C., de Tribolet, N., Hamou, M. F., and Van Meir, E. G.Regulation of interleukin-8 expression by reduced oxygen pressure in human glio-blastoma. Oncogene,18: 1447–1456, 1999.

37. Mukaida, N., Shiroo, M., and Matsushima, K. Genomic structure of the humanmonocyte-derived neutrophil chemotactic factor IL-8. J. Immunol.,143: 1366–1371,1989.

2160

HuR IS EXPRESSED IN MALIGNANT BRAIN TUMORS

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

38. Shweiki, D., Neeman, M., Itin, A., and Keshet, E. Induction of vascular endo-thelial growth factor expression by hypoxia and by glucose deficiency in multicellspheroids: implications for tumor angiogenesis. Proc. Natl. Acad. Sci. USA,92:768 –772, 1995.

39. Sheng, H., Shao, J., Dixon, D. A., Williams, C. S., Prescott, S. M., DuBois, R. N., andBeauchamp, R. D. Transforming growth factor-b1 enhances Ha-ras-induced expres-sion of cyclooxygenase-2 in intestinal epithelial cells via stabilization of mRNA.J. Biol. Chem.,275: 6628–6635, 2000.

40. Ristimaki, A., Narko, K., and Hla, T. Down-regulation of cytokine-induced cyclo-oxygenase-2 transcript isoforms by dexamethasone: evidence for post-transcriptionalregulation. Biochem. J.,318: 325, 1996.

41. Vane, J. R., Bakhle, Y. S., and Botting, R. M. Cyclooxygnases 1 and 2. Ann. Rev.Pharmacol. Toxicol.,38: 97–120, 1998.

42. Van Meir, E. G. Cytokines and tumors of the central nervous system. Glia,15:264–288, 1995.

43. Jennings, M. T., and Pietenpol, J. A. The role of transforming growth factorb inglioma progression. J. Neuro-Oncol.,36: 123–140, 1998.

44. Chung, S., Jiang, L., Cheng, S., and Furneaux, H. Purification and properties of HuD,a neuronal RNA-binding protein. J. Biol. Chem.,271: 11518–11524, 1996.

45. Liu, J., Dalmau, J., Szabo, A., Rosenfeld, M., Huber, J., and Furneaux, H. Paraneo-plastic encephalomyelitis antigens bind to the AU-rich elements of mRNA. Neurol-ogy, 45: 544–550, 1995.

46. Chagnovich, D., Fayos, B. E., and Cohn, S. L. Differential activity of ELAV-like RNA-binding proteins in human neuroblastoma. J. Biol. Chem.,271: 33587–33591, 1996.

47. Chung, S., Eckrich, M., Perrone-Bizzozero, N., Kohn, D. T., and Furneaux, H. TheElav-like proteins bind to a conserved regulatory element in the 39-untranslated regionof GAP-43 mRNA. J. Biol. Chem.,272: 6593–6598, 1997.

48. Joseph, B., Orlian, M., and Furneaux, H. p21waf1 mRNA contains a conservedelement in its 39-untranslated region that is bound by the Elav-like mRNA-stabilizingproteins. J. Biol. Chem.,272: 20511–20516, 1998.

2161

HuR IS EXPRESSED IN MALIGNANT BRAIN TUMORS

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

2001;61:2154-2161. Cancer Res   L. Burt Nabors, G. Yancey Gillespie, Lualhati Harkins, et al.   Factor mRNAs

Untranslated Regions of Cytokine and Angiogenic′within the 3Tumors and Binds to Adenine- and Uridine-rich Elements HuR, a RNA Stability Factor, Is Expressed in Malignant Brain

  Updated version

  http://cancerres.aacrjournals.org/content/61/5/2154

Access the most recent version of this article at:

   

   

  Cited articles

  http://cancerres.aacrjournals.org/content/61/5/2154.full#ref-list-1

This article cites 45 articles, 24 of which you can access for free at:

  Citing articles

  http://cancerres.aacrjournals.org/content/61/5/2154.full#related-urls

This article has been cited by 44 HighWire-hosted articles. Access the articles at:

   

  E-mail alerts related to this article or journal.Sign up to receive free email-alerts

  Subscriptions

Reprints and

  .pubs@aacr.orgDepartment at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

  Permissions

  Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/61/5/2154To request permission to re-use all or part of this article, use this link

on May 16, 2018. © 2001 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from