Molecular Cloning of Mouse Glycolate Oxidase HIGH EVOLUTIONARY CONSERVATION AND PRESENCE OF AN IRON-RESPONSIVE ELEMENT-LIKE SEQUENCE IN THE mRNA*

Iron regulatory proteins (IRPs) control the synthesis of several proteins in iron metabolism by binding to iron-responsive elements (IREs), a hairpin structure in the untranslated region (UTR) of corresponding mRNAs. Binding of IRPs to IREs in the 5 * UTR inhibits translation of ferritin heavy and light chain, erythroid aminolevulinic acid synthase, mitochondrial aconitase, and Drosophila succinate dehydrogenase b, whereas IRP binding to IREs in the 3 * UTR of transferrin receptor mRNA prolongs mRNA half-life. To identify new targets of IRPs, we devised a method to enrich IRE-contain-ing mRNAs by using recombinant IRP-1 as an affinity matrix. A cDNA library established from enriched mRNA was screened by an RNA-protein band shift assay. This revealed a novel IRE-like sequence in the 3 * UTR of a liver-specific mouse mRNA. The newly identified cDNA codes for a protein with high homology to plant glycolate oxidase (GOX). Recombinant protein expressed in bacteria displayed enzymatic GOX activity. Therefore, this cDNA represents the first vertebrate GOX homologue. The IRE-like sequence in mouse GOX exhibited strong binding to IRPs at room temperature. However, it differs from functional IREs by a mismatch in the middle of its upper stem and did not confer iron-dependent regulation in cells. RNA-binding proteins nucleo-cytoplasmic identify the targets of a given RNA-binding protein. b - D - galactopyranoside (Boehringer Mannheim). White colonies were man-ually transferred onto 18 16 3 16 array master plates. Library Screening— Ferritin H and L chain cDNAs were identified by hybridization (on GeneScreen Plus membranes; DuPont NEN) with [ a - 32 P]dATP-labeled (Amersham Pharmacia Biotech), random-primed cDNA probes. Hybridization was carried out at 65 °C overnight in hybridization buffer (10% (w/v) dextran sulfate and 1% SDS) with 150 m g/ml salmon sperm DNA. To screen for new IRE-containing mRNAs, groups of 16 colonies (excluding the ferritin clones) were grown in 5 ml of L broth, and plasmid DNA was prepared using the QIAprep spin miniprep kit (Qia-gen). The plasmid mixture was linearized with Not I and purified on QIAquick columns (Qiagen). The eluted DNA was transcribed in vitro with T7-RNA polymerase in the presence of [ a - 32 P]CTP (Amersham Pharmacia Biotech). After ethanol precipitation, the labeled RNA was used to perform band shift assays. Positive groups were reanalyzed as subgroups of four colonies and finally as single clones. Plasmid DNA of final positives was analyzed by automated DNA sequencing. RNA Band Shift Assays— RNA-protein gel retardation assays were carried out as described previously (3). For library screening, RNA transcripts were incubated with 2 ng of purified recombinant human IRP-1. Unprotected RNA was degraded with 0.5 units of RNase T1 (Calbiochem) for 10 min, and 5 mg/ml heparin (Serva) was added with loading buffer before migration on a 6% nondenaturing gel. Heparin was omitted in competition band shift reactions. Competitor RNAs were tracer-labeled with [ a - 32 P]CTP (Amersham Pharmacia Biotech), puri- fied on 3% NuSieve agarose gels by migration onto NA45 DEAE paper (Schleicher and Schu¨ll) EDTA, and the decrease in absorption was measured at 605 nm (20). Construction of Plasmids— pBSII-GOXIRE for transcription of band shift probes was constructed by cloning a Dra I- Not I fragment of the GOX cDNA (nucleotides 1853–2028 plus the poly(A) tail) between sites Eco RV and Not I of pBluescript II SK( 1 ). For protein expression, the coding region of the GOX cDNA was amplified by polymerase chain reaction and cloned into the Sal I and Not I sites of pGEX-5X3 (Amer-sham Pharmacia Biotech) in frame with the GST sequence. The GOX IRE was inserted into the vector pL5-GH (6) as described (17). Plasmid pFer-GH was kindly provided by Dr. Matthias Hentze (EMBL, Heidel- berg, Germany). Translation Assay in Cells— Ltk 2 were stably transfected by the calcium phosphate method (21). Incorporation of [ 35 S]methionine and immunoprecipitation of secreted human growth hormone (hGH) were carried out as described previously (17, 22). were analyzed for RNA activity of with a 32 chain in but to activates fully all This serves as a

RNA-binding proteins play a central role in RNA processing, nucleo-cytoplasmic transport, localization, translation, or stability. This makes it desirable to identify all the targets of a given RNA-binding protein. Because many RNA recognition elements cannot be identified by hybridization or predicted with certainty by search programs, we developed an experimental approach based on the interaction between an RNAbinding protein and its target RNAs. As a model system we are investigating the iron regulatory proteins 1 and 2 (IRP-1 and IRP-2), 1 which become active in iron-deprived cells and bind with high affinity to structural RNA motifs, the iron-responsive elements (IREs) (1)(2)(3)(4)(5). IRPs repress translation of mRNAs with a cap-proximal IRE, as found in the mRNA of ferritin heavy (H) and light (L) chain (1,6), mitochondrial aconitase (7,8), erythroid 5-aminolevulinic acid synthase (9,10), and Drosophila succinate dehydrogenase b (SDHb) (8,11,12). As a consequence, IRP binding lowers iron storage and utilization. Moreover, binding of IRPs to five IREs in the 3Ј untranslated region (UTR) of transferrin receptor mRNA protects this otherwise unstable mRNA from degradation (3,13). This leads to more receptor synthesis and enhanced iron uptake.
Functional IREs form stem loop structures with a conserved CAGUGN loop sequence (where N is any nucleotide except G). An upper stem of five perfectly paired bases is separated from a lower stem by a single cytosine on the 5Ј side or by a cytosine preceded by two nucleotides with one unpaired nucleotide on the 3Ј side. Adopting an in vitro selection procedure, we and others have identified IREs with alternative loop sequences that bind to IRPs and of which some show a preferential interaction with IRP-1 or IRP-2 (14 -16). Such IRE mutants confer translational control in vivo when inserted into the 5Ј UTR of a reporter construct (17). However, none of these alternative IREs was detected in naturally occurring mRNAs to date.
Here we devised a method to enrich IRE-containing mRNAs by using recombinant human IRP-1 as an affinity matrix (Fig. 1). We then constructed an enriched cDNA library from mouse liver mRNA and screened it for IREs by RNA-protein band shift assays. This revealed a clone with strong homology to plant glycolate oxidase (GOX, or short-chain ␣-hydroxy acid oxidase, EC 1. 1.3.15). In plants, this enzyme participates in the glyoxylate cycle and catalyzes the oxidation of glycolate to glyoxylate. An IRE-like sequence was found in the 3Ј UTR of the mouse mRNA and analyzed with respect to its possible involvement in iron-dependent, post-transcriptional regulation.

EXPERIMENTAL PROCEDURES
Cell Culture-HL-60 and Ltk Ϫ cells were cultured in ␣-minimal essential medium; FTO2B rat hepatoma cells were cultured in 45% Dulbecco's modified Eagle's medium and 45% F-12 (Life Technologies, Inc.); and B16.F1 mouse melanoma cells were cultured in Dulbecco's modified Eagle's medium. All media were supplemented with 10% fetal calf serum. Cells were deprived of iron in medium with 100 M desferrioxamine (Desferal, a gift from Novartis, Basel, Switzerland) for 16 -20 h or iron-loaded in medium with ferric ammonium citrate (60 g/ml) for 4 -5 h.
Purification of Recombinant Human IRP-1-Bacteria expressing glutathione S-transferase (GST)-tagged human IRP-1 (18) were grown to A 600 0.7 and induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside (Boehringer Mannheim) for 5 h. Cells were harvested and lysed by sonication in phosphate-buffered saline and 1% Triton X-100. The lysate was cleared at 10,000 ϫ g for 10 min and adsorbed on glutathione-Sepharose beads (Amersham Pharmacia Biotech) for 1 h at 4°C. Beads were washed in a Bio-Rad Polyprep column, and bound IRP-1 was eluted in 50 mM Tris-Cl, pH 8.0, and 5 mM glutathione. The yield of IRP-1 was ϳ150 g/liter of bacterial culture. The eluate was diluted 2-fold in buffer A (20 mM Tris-Cl, pH 8.0, 5% glycerol, 8 mM 2-mercaptoethanol) and applied to a MonoQ column (Amersham Pharmacia Biotech). Bound protein was eluted with 0 -300 mM KCl in buffer A, and * This work was supported by the Swiss National Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Isolation of RNA-Total RNA was extracted as described by Chomczynski and Sacchi (19). Poly(A) ϩ RNA was prepared using the Poly(A)Tract mRNA isolation kit (Promega).
Affinity Purification of RNA-GST-tagged human IRP-1 (100 g) was immobilized on 50 l of glutathione-Sepharose (Amersham Pharmacia Biotech). Beads were washed extensively with binding buffer (10 mM HEPES, pH 7.6, 3 mM MgCl 2 , 40 mM KCl, 5% glycerol) containing 5 mg/ml heparin (Serva) and binding buffer alone. Total RNA (5 mg), renatured for 5 min at 70°C and 5 min at 42°C, was adsorbed to the IRP-coated beads in 4 ml of binding buffer with 0.2% 2-mercaptoethanol, 1 mM dithiothreitol, and 1 unit/l RNasin. After gentle agitation for 30 min at 25°C, unbound RNA was removed, and the beads were washed three times with binding buffer containing 5 mg/ml heparin and twice with binding buffer alone. Bound RNA was recovered by phenolchloroform extraction and precipitated with 2.5 volumes of ethanol.
To screen for new IRE-containing mRNAs, groups of 16 colonies (excluding the ferritin clones) were grown in 5 ml of L broth, and plasmid DNA was prepared using the QIAprep spin miniprep kit (Qiagen). The plasmid mixture was linearized with NotI and purified on QIAquick columns (Qiagen). The eluted DNA was transcribed in vitro with T7-RNA polymerase in the presence of [␣-32 P]CTP (Amersham Pharmacia Biotech). After ethanol precipitation, the labeled RNA was used to perform band shift assays. Positive groups were reanalyzed as subgroups of four colonies and finally as single clones. Plasmid DNA of final positives was analyzed by automated DNA sequencing.
RNA Band Shift Assays-RNA-protein gel retardation assays were carried out as described previously (3). For library screening, RNA transcripts were incubated with 2 ng of purified recombinant human IRP-1. Unprotected RNA was degraded with 0.5 units of RNase T1 (Calbiochem) for 10 min, and 5 mg/ml heparin (Serva) was added with loading buffer before migration on a 6% nondenaturing gel. Heparin was omitted in competition band shift reactions. Competitor RNAs were tracer-labeled with [␣-32 P]CTP (Amersham Pharmacia Biotech), purified on 3% NuSieve agarose gels by migration onto NA45 DEAE paper (Schleicher and Schü ll) and elution in 500 mM Tris-Cl, pH 7.2, and 2.5 M ammonium acetate at 65°C for 10 min. After phenol extraction and ethanol precipitation, RNAs were dissolved in water and diluted to equal molar concentration. Cytoplasmic extracts were prepared from iron-deprived FTO2B cells by lysis in binding buffer supplemented with 0.3% Nonidet P-40 and 1 mM phenylmethylsulfonyl fluoride. Nuclei were spun down at 13,000 ϫ g for 5 min at 4°C, and the supernatant was stored in aliquots at Ϫ70°C. Protein concentration was determined by the Bio-Rad protein assay.
In Vitro Transcription and Translation-Coupled in vitro transcription and translation of the mouse GOX cDNA was carried out with the TNT coupled reticulocyte lysate system (Promega) according to the manufacturer's instructions using T7-RNA polymerase. Proteins were labeled by the incorporation of [ 35 S]methionine (Amersham Pharmacia Biotech).
Expression of Recombinant GOX and Enzyme Assay-GOX was expressed as a GST fusion protein in E. coli DH5␣ and purified on glutathione-Sepharose as described for IRP-1. The yield was 0.5 mg/ liter of culture. Enzyme activity was determined at 25°C by incubation with 165 M 2,6-dichloroindophenol (sodium salt hydrate; Fluka), 660 M sodium glycolate (ICN) in 66 mM potassium phosphate, pH 8.3, and 1 mM EDTA, and the decrease in absorption was measured at 605 nm (20).
Construction of Plasmids-pBSII-GOXIRE for transcription of band shift probes was constructed by cloning a DraI-NotI fragment of the GOX cDNA (nucleotides 1853-2028 plus the poly(A) tail) between sites EcoRV and NotI of pBluescript II SK(ϩ). For protein expression, the coding region of the GOX cDNA was amplified by polymerase chain reaction and cloned into the SalI and NotI sites of pGEX-5X3 (Amersham Pharmacia Biotech) in frame with the GST sequence. The GOX IRE was inserted into the vector pL5-GH (6) as described (17). Plasmid pFer-GH was kindly provided by Dr. Matthias Hentze (EMBL, Heidelberg, Germany).
Translation Assay in Cells-Ltk Ϫ were stably transfected by the calcium phosphate method (21). Incorporation of [ 35 S]methionine and immunoprecipitation of secreted human growth hormone (hGH) were carried out as described previously (17,22).

GST-tagged Human IRP-1 Can Be Used as an Affinity Matrix for IRE-containing mRNAs-
The approach for the identification of targets of IRP-1 is depicted in Fig. 1A. It comprises the selection IRE-containing mRNA on an IRP-1 affinity column, the construction of an enriched cDNA library, and its screening by RNA-protein band shift assays. As shown in Fig.  1B, an affinity column with recombinant IRP-1 can selectively bind IRE-containing mRNAs. For this, the human IRP-1 mutant C437S that binds IREs constitutively (23) was expressed with a GST tag in bacteria, purified to remove bacterial RNA, and adsorbed on glutathione-Sepharose. Total input RNA, unbound RNA, washes, and bound RNA were analyzed by Northern blotting (Fig. 1B). Quantification by the PhosphorImager revealed specific retention of 50% of the IRE-containing ferritin tracer RNA as well as human transferrin receptor mRNA. In contrast, Ͻ0.5% of the control ␤-actin and glyceraldehyde-3phosphate dehydrogenase mRNA without an IRE was retained on the column. Thus, immobilized IRP-1 can be used to enrich IRE-containing mRNAs at least 100-fold.
The column was also tested for the binding of human SDHb mRNA when we noticed an IRE in Drosophila SDHb mRNA (11). The Northern blot was rehybridized with a rapid amplification of cDNA ends polymerase chain reaction product of the human SDHb cDNA 5Ј end including part of the coding sequence (24). Human SDHb mRNA was not retained on the IRP-coated beads (Fig. 1B). This agrees with conclusions drawn from the human SDHb gene sequence that revealed no IRE-like motif upstream of the coding region (25). It suggests that human SDHb expression is not regulated by IRP-1.
Analysis of Enriched mRNAs: Putative Glycolate Oxidase mRNA Is Bound by IRP-1 in Vitro-To isolate new mRNAs with an IRE, we incubated the IRP-1 column at a preparative scale with total mouse liver RNA. Bound mRNAs were eluted, reverse-transcribed, and cloned directionally to obtain an oligo(dT)-primed cDNA library enriched in IRE-containing inserts. 4608 recombinant clones were identified by blue-white selection and transferred to 18 master plates with 256 clones each. Among 22 randomly picked recombinant cDNAs, we found clones coding for mouse ferritin H and L chains. Additional ferritin cDNAs were expected at high frequency in the library and identified by hybridization of colony lifts. 56 ferritin H chain and 214 ferritin L chain clones were found, representing a frequency of 1.2 and 4.6%, respectively.
Clones that did not code for ferritins were screened for the presence of an IRE by an RNA-protein band shift assay. Plasmids were linearized behind the cDNA insert, transcribed in vitro with T7-RNA polymerase, and incubated with recombinant purified IRP-1, first in groups of 16, then in subgroups of 4, and finally as single clones. Because some transcripts were RNase T1-resistant and resembled IRE⅐IRP band shift complexes in their migration, we had to exclude false positives by omitting IRP-1 in the assay. 14 cDNA clones were finally identified as IRE-positive and sequenced. Of these, 11 had escaped detection by hybridization and coded for ferritin L or H chain. The remaining three clones originated of a single mRNA species with homology to plant glycolate oxydase and contained in the 3Ј UTR an IRE-like sequence close to the polyadenylation site ( Fig. 2A, bold). However, a mismatch in the upper stem distinguished this putative IRE from other known IREs (Fig.  2B). We also noticed different 3Ј end sequences among the isolated clones, one with the poly(A) tail beginning at 20 bases from the polyadenylation site and another one extending 40 bases further ( Fig. 2A, underlined) to a second poly(A) tail start site. Both classes of cDNAs were also found in an independent cDNA library (data not shown). This difference was not further investigated and probably represents alternative polyadenylation sites.
Rehybridization of our cDNA library with the longest cDNA insert identified three more clones, one of them with a fulllength cDNA of 2.0 kb. This clone contained an open reading frame of 1110 nucleotides coding for a protein of 370 amino acids with a predicted molecular mass of 41.0 kDa and a pI of 7.6 (GenBank accession number AI104312). In vitro transcription and translation of the plasmid yielded a protein with the predicted size (Fig. 2C). A search in the data bases revealed that this clone had Ͼ55% amino acid identity (74% similarity) to GOX from plants (Fig. 3), suggesting strongly that it might encode a mouse GOX. A kidney-specific isozyme from rat longchain hydroxyacid oxidase (26) was likewise related but less homologous than the plant enzyme (48% identical and 59% similar). We also identified several expressed sequence tag clones of human origin, which could be aligned to the mouse sequence and served to reconstruct part of the human GOX sequence (Fig. 3). It exhibited 87.9% amino acid identity (92.2% similarity) with the mouse protein. Moreover, a genomic sequence from Caenorhabditis elegans (GenBank accession number AF016448) previously predicted to encode GOX (27) was 48.9% identical (60.2% similar) to that of mouse GOX (Fig. 3).
cDNA Codes for a Liver-specific Mouse Glycolate Oxidase-Spinach GOX has been crystalized, and amino acid residues involved in the active site, the substrate binding, and tight interaction with FMN are known (28). Here we found that all of these residues are conserved in our cDNA (Fig. 3), supporting the idea that it encodes a mouse GOX. To test this hypothesis, we investigated the enzymatic activity of recombinant protein.
The cDNA was subcloned into the plasmid pGEX-5X3 for expression with an amino-terminal GST tag in bacteria. A spectrophotometric assay for GOX activity (20) was set up for crude bacterial extracts. Control extracts carrying plasmid without insert showed no reduction of 2,6-dichloroindophenol in the presence of excess sodium glycolate (660 M) as a substrate (Fig. 4A). In contrast, extracts from GOX-transformed bacteria oxidized glycolate at a rate of 69 pmol/s per g of crude protein. Moreover, GOX activity co-purified with the fusion protein on glutathione-Sepharose beads (Fig. 4B). This confirmed that the newly isolated cDNA encodes mouse GOX.
In animals, two isozymes of ␣-hydroxyacid oxidase were re- ported (29,30), one expressed in the kidney with preference for long chain ␣-hydroxyacids and the liver-specific short chain ␣-hydroxyacid oxidase, also known as GOX. This prompted us to investigate the tissue specificity of the mouse GOX mRNA. Northern blot analysis of RNA from various mouse tissues and fibroblast cells revealed GOX mRNA expression exclusively in liver but not in spleen, skeletal muscle, kidney, embryos, or fibroblasts (Fig. 5A).
Mouse GOX IRE Binds to IRPs with High Affinity in Vitro but Not in Cells-The presence of an IRE-like sequence in the 3Ј UTR suggested that GOX mRNA might be regulated by IRPs. Therefore, rat hepatoma cells (FTO2B) were either treated with iron or the iron chelator desferrioxamine or with iron after desferrioxamine. GOX mRNA levels were measured on Northern blots and normalized to ␤-actin mRNA (Fig. 5B). No effect was observed when iron-deprived cells were compared with controls. But when iron salt was added, GOX mRNA was up-regulated 2.3-and 2.8-fold, respectively. The RNA binding activity of IRPs in FTO2B cells was tested by band shift experiments and responded as expected (31) to different iron-loading conditions (not shown).
These results suggested that IRP might be involved in the control of GOX expression. In support, we mapped the IRP-1 binding region of the mRNA between nucleotides 1884 and 1974 in the 3Ј UTR encompassing the IRE-like sequence ( Fig.  2A, bold). An A:A mismatch in the upper stem of this putative IRE (Fig. 2B) suggested, however, that it might affect protein binding. Therefore, the affinity of the GOX IRE for IRPs was compared with that of ferritin H chain IRE in competition band shift assays (Fig. 6). Cytoplasmic extract prepared from irondeprived FTO2B cells was incubated with labeled GOX or ferritin IRE in the presence of increasing molar excess of nonlabeled competitor RNA (GOX IRE, ferritin IRE, or tRNA). Surprisingly, at 25°C, both probes bound equally well to IRP-1 and IRP-2 (Fig. 6). Encouraged by this result, we inserted the putative IRE into the 5Ј UTR of vector pL5-GH containing a constitutive ferritin promoter and the human growth hormone gene (hGH) (6). A wild-type ferritin IRE in this vector confers IRP-mediated repression of hGH translation in iron-deprived cells (Ref. 10 and Fig. 7A). However, we were unable to detect translational regulation in response to iron deprivation in stably transfected mouse L cells with the GOX IRE construct (Fig.  7A). Yet, control band shift experiments with extracts from the same cells revealed strong induction of IRP-1 in response to FIG. 3. Evolutionary conservation of glycolate oxidase. Using the program Pileup of the Genetics Computer Group package (52), the amino acid sequence of rat kidney long chain 2-hydroxyacid oxidase (L-HAO; GenBank accession number X67156) was aligned to the deduced amino acid sequence of the mouse GOX cDNA as well as to other GOX sequences from human (partial sequence reconstructed from several expressed sequence tags), C. elegans (GenBank accession number AF016448), ice plant (GenBank accession number U80071), pumpkin (GenBank accession number D14044), spinach (GenBank accession number J03492), Arabidopsis thaliana (GenBank accession number AL021710), and rice (GenBank accession number AF022740). Active site residues (A) and amino acids involved in FMN (F) or substrate (S) binding of the spinach enzyme (28) are indicated.
iron deprivation (Fig. 7B). Thus, it seemed plausible that the GOX IRE was not binding to IRPs in cells. Therefore, we measured carefully the IRE-IRP interaction at higher temperatures in vitro (Fig. 8). We found a strong temperature dependence, the interaction being entirely lost between 34 and 37°C. The stem loop of the GOX IRE is probably thermodynamically unstable because of the A:A mismatch in the upper stem.

DISCUSSION
Screening for Targets of mRNA-binding Proteins-We explored whether affinity chromatography with immobilized RNA-binding IRP-1 can contribute to the identification of new targets for post-transcriptional gene regulation by this protein. Such a method would be generally useful in view of the rapidly growing list of RNA-binding proteins. Only few methods to search for targets of RNA-binding proteins have been described. They comprise either systematic evolution of ligands by exponential enrichment with naturally occurring mRNAs (32) or immunoprecipitation of mRNA-protein complexes followed by reverse transcription-polymerase chain reaction (33). The present study indicates that mRNA affinity purification on recombinant protein and construction of an enriched cDNA library combined with screening by RNA-protein band shift assays is a viable approach. Our cDNA library was clearly enriched for IRE-containing clones. Although a normal human liver cDNA library comprises ϳ0.5% ferritin H and L chain clones (34), the frequency was increased to 5.8% in the enriched library (1.2% ferritin H chain clones and 4.6% L chain clones). This was less than expected from the mRNA enrichment step but can be explained by a high frequency of void plasmids.
The data with human SDHb mRNA (Fig. 1B) illustrate that by affinity purification it is also possible to determine whether a candidate mRNA binds to a regulatory protein in vitro. FIG. 4. Enzymatic activity of recombinant mouse glycolate oxidase. The activity of GOX was measured at 25°C by the reduction of 2,6-dichloroindophenol (DCIP) coupled to the oxidation of glycolate. Enzyme activity was detected in crude extract of bacteria expressing GST-tagged GOX protein (A; q) but not in extract from mock-transfected bacteria with the parental plasmid (A; E). The enzyme was adsorbed on glutathione-Sepharose (B; f), and only low residual activity was found in the soluble supernatant (B; Ⅺ). No activity could be adsorbed from extracts of mock-transfected bacteria (B; E).

FIG. 5.
Liver-specific expression of glycolate oxidase mRNA and effect of iron levels. A, total RNA (20 g) extracted from mouse Ltk Ϫ cells or mouse tissues was analyzed on Northern blots with an antisense riboprobe specific for mouse GOX mRNA. Cross-hybridization to 28 S rRNA provides a loading control. B, FTO2B rat hepatoma cells were cultured for 20 h in medium with either 60 g/ml ferric ammonium citrate (FAC), 100 M desferrioxamine (Des), or 100 M desferrioxamine for 20 h followed by 60 g/ml ferric ammonium citrate (Des/ FAC) in fresh medium for an additional 5 h. Poly(A) ϩ RNA (2.5 g) from each batch of cells, as well as untreated control cells, was analyzed on a Northern blot next to 15 g of mouse liver total RNA. The blot was first hybridized to a GOX-specific antisense riboprobe and then stripped and rehybridized to a mouse ␤-actin probe. The intensity of radioactive signals was quantified on a PhosphorImager. Knowledge of the entire mRNA or precise target sequence is not required, because a Northern blot with column fractions can be hybridized with a probe corresponding to any fragment of the candidate RNA. From the results with this particular mRNA we conclude that IRPs are probably not involved in regulation of human SDHb expression.
Our data demonstrate that screening by in vitro RNA-protein band shift requires certain controls, notably for the specificity and temperature of the interaction. In the mouse GOX mRNA, the IRE-related sequence binds IRPs at 25°C but not at 37°C. It fits with the lack of translational repression of the GH construct with the 5Ј GOX IRE despite IRP activation (Fig.  7). These results indicate that GOX mRNA is not regulated by IRPs and that the stimulation of expression by high iron in a liver cell line must be attributable to a different mechanism.
At 25°C, quite unexpectedly, the GOX IRE bound as strongly to rat IRP-1 and IRP-2 as the ferritin H chain IRE (Fig. 6), despite an A:A mismatch in the center of the upper stem (Fig. 2B). We have previously made a similar observation with a mutant ferritin H chain IRE containing a G:G mismatch in the same position and that displayed ϳ25% of wild-type binding affinity (14). These results seem to reflect a certain tolerance for noncanonical base pairing in the upper stem. Its thermodynamic instability explains probably why this IRE does not bind at higher temperature and, when placed in the 5Ј UTR, does not render translation of a human growth hormone reporter mRNA iron-dependent in cells.
Considering the relatively limited number of clones analyzed in the present screen, we cannot exclude that further IREcontaining candidate mRNAs may exist in the liver. However, if further clones are to be found, they can be expected at a frequency at least as low as that of the GOX mRNA, which was 50 -100-fold less frequent than that of ferritin mRNA.
Possible Role of Mammalian Glycolate Oxidase-The second main conclusion from our study concerns the first isolation of a mammalian GOX cDNA. Its amino acid sequence was 55% identical (74% similar) to GOX (also termed short-chain 2-hydroxyacid oxidase) from plants. The plant enzyme contains an FMN cofactor and catalyzes oxidation of glycolate to glyoxylate. The three-dimensional structure of spinach GOX has been resolved by x-ray crystallography (28), which allowed determination of active site residues and side chains involved in FMN binding. All of these amino acids are perfectly conserved in the isolated clone (Fig. 3), suggesting that it codes for mouse GOX. This conclusion is supported by the enzymatic activity of the purified recombinant protein (Fig. 4).
From these results and evidence detailed below we conclude that our clone is full length. The sequence encompassing the start codon (GCCACAAUGU) matches well the Kozak consensus (GCC(A/G)CCAUGG) for translation start sites (35), and the size of the predicted protein was in good agreement with the in vitro transcription-translation product analyzed on SDS-PAGE (Fig. 2C). Second, several homologous clones were isolated from a commercial mouse liver cDNA phage library. Only one of those clones contained a slightly longer 5Ј sequence (not shown) but had no additional upstream start codon. Third, a probe derived from the isolated clone detected a single band of 2.2 kb on a Northern blot, the size expected for the predicted, polyadenylated mRNA (Fig. 5A). Fourth, alignment to homologous sequences in the data base revealed high conservation close to the amino terminus (Fig. 3).
Mammalian GOX activity had been discovered as early as 1940 (36), but to our knowledge no vertebrate GOX cDNA or gene has been cloned to date. In animals, two 2-hydroxy acid oxidases have been reported with expression in either liver or kidney, one with specificity for short chain 2-hydroxyacids and the other for long chain 2-hydroxyacids, respectively (30,37). Our cDNA corresponds to the liver enzyme (Fig. 5B). The sequence of a rat kidney long chain 2-hydroxyacid oxidase (26) is similar and belongs to the same family of proteins (Fig. 3). Mammalian liver GOX has been located in peroxisomes (38 -40). This localization might be mediated by the carboxyl-terminal sequence SKI (Fig. 3), which resembles the consensus SKL sequence for peroxisomal import (41,42). However, studies with mutants of the localization tripeptide behind luciferase suggested that SKI is a poor import signal (43). Further work is required to analyze this issue for mammalian GOX.
Glycolate oxidase contains an FMN cofactor (20,44) that catalyzes the oxidation of glycolate to glyoxylate and possibly to oxalate (45)(46)(47). GOX could be implicated in the detoxification of glycolate via the synthesis of oxalate. Alternatively, it might provide building blocks for the synthesis of glycine and serine, because the carbon atoms of 14 C-labeled glycolate are detected in these amino acids (48). FIG. 7. Effect of the GOX IRE in a 5 position on translation of human growth hormone. The putative GOX IRE was subcloned into the 5Ј UTR of the vector pL5-GH (6), which codes for hGH transcribed from a constitutive promoter. Mouse Ltk Ϫ cells were stably transfected and labeled for 4 h with [ 35 S]methionine, and secreted hGH was immunoprecipitated (A). To detect IRP-mediated inhibition of translation, cells were preincubated overnight with 100 M iron chelator desferrioxamine (Des) and compared with cells treated with 60 g/ml ferric ammonium citrate (Fe). A 10-fold inhibition of hGH translation was observed with the positive control construct harboring the ferritin IRE (Fer-GH) (10), but no difference was visible with the GOX IRE construct (Gox-GH). B, cytoplasmic extracts of the same cells were analyzed for RNA binding activity of IRPs with a 32 P-labeled ferritin H chain IRE. The IRE-IRP band shift assay shows strong activity of IRP-1 and IRP-2 in iron-deprived cells (Des; Ϫ2-me) but very weak activity in iron-loaded cells (Fe; Ϫ2-me). The in vitro addition of 2% 2-mercaptoethanol (ϩ2me) to extracts activates fully all cytoplasmic IRP-1. This serves as a control for equal loading.
FIG. 8. Temperature dependence of IRP binding to the GOX IRE. Cytoplasmic extracts made from B16.F1 cells were mixed at different temperatures with a molar excess of [␣-32 P]CTP-labeled GOX IRE and analyzed by the band shift assay. RNase T1 was added after 10 min to digest excess IRE and to prevent its binding to IRP during gel migration at 4°C. IRP-1 and IRP-2 are not inactivated at 37°C, as shown in the last lane, where the sample was shifted back to 25°C before adding RNase T1.
In oil seed plants, GOX acts as part of the glyoxylate cycle. This pathway is required in early stages of seedling growth for the conversion of fatty acid stores into sugars (49). Because of its importance in plant growth, GOX cDNAs have been cloned in several plant species. The question arises of whether lipid mobilization and gluconeogenesis involving glyoxylate cycle enzymes are a speciality of plants only. Both isocitrate lyase and malate synthase activity have been detected in liver peroxisomal fractions and in adipose tissue of mammals (50,51), where they are induced by fasting. This suggests that under certain conditions, mammalian GOX might also feed glyoxylate into this pathway. The present cDNA clone should help clarify the physiological role of GOX.