Purification of a Novel Flavoprotein Involved in the Thyroid NADPH Oxidase

Hydrogen peroxide is the final electron acceptor for the biosynthesis of thyroid hormone catalyzed by thyroperoxidase at the apical surface of thyrocytes. Pig and human thyroid plasma membrane contain a Ca2+-dependent NAD(P)H oxidase that generates H2O2 by transferring electrons from NAD(P)H to molecular oxygen. We purified from pig thyroid plasma membrane a flavoprotein which constitutes the main, if not the sole, component of the thyroid NAD(P)H oxidase. Microsequences permitted the cloning of porcine and human full-length cDNAs encoding, respectively, 1207- and 1210-amino acid proteins with a predicted molecular mass of 138 kDa (p138Tox). Human and porcine p138Tox have 86.7% identity. The strongest similarity was to a predicted polypeptide encoded by a CaenorhabditiscDNA and with rbohA, a protein involved in theArabidopsis NADPH oxidase. p138Tox shows also similarity to the p65Mox and to the gp91Phox in their C-terminal region and have consensus sequences for FAD- and NADPH-binding sites. Compared with gp91Phox, p138Tox shows an extended N-terminal containing two EF-hand motifs that may account for its calcium-dependent activity, whereas three of four sequences implicated in the interaction of gp91Phox with the p47Phox cytosolic factor are absent in p138Tox. The expression of porcine p138Tox mRNA analyzed by Northern blot is specific of thyroid tissue and induced by cyclic AMP showing that p138Tox is a differentiation marker of thyrocytes. The gene of human p138Tox has been localized on chromosome 15q15.

Hydrogen peroxide is the final electron acceptor for the biosynthesis of thyroid hormone catalyzed by thyroperoxidase at the apical surface of thyrocytes. Pig and human thyroid plasma membrane contain a Ca 2؉ -dependent NAD(P)H oxidase that generates H 2 O 2 by transferring electrons from NAD(P)H to molecular oxygen. We purified from pig thyroid plasma membrane a flavoprotein which constitutes the main, if not the sole, component of the thyroid NAD(P)H oxidase. Microsequences permitted the cloning of porcine and human full-length cDNAs encoding, respectively, 1207-and 1210-amino acid proteins with a predicted molecular mass of 138 kDa (p138 Tox ). Human and porcine p138 Tox have 86.7% identity. The strongest similarity was to a predicted polypeptide encoded by a Caenorhabditis cDNA and with rbohA, a protein involved in the Arabidopsis NADPH oxidase. p138 Tox shows also similarity to the p65 Mox and to the gp91 Phox in their C-terminal region and have consensus sequences for FAD-and NADPHbinding sites. Compared with gp91 Phox , p138 Tox shows an extended N-terminal containing two EF-hand motifs that may account for its calcium-dependent activity, whereas three of four sequences implicated in the interaction of gp91 Phox with the p47 Phox cytosolic factor are absent in p138 Tox . The expression of porcine p138 Tox mRNA analyzed by Northern blot is specific of thyroid tissue and induced by cyclic AMP showing that p138 Tox is a differentiation marker of thyrocytes. The gene of human p138 Tox has been localized on chromosome 15q15.
The synthesis of thyroid hormone is catalyzed by thyroperoxidase (TPO) 1 in the presence of H 2 O 2 (1) on the apical membrane of the thyroid follicular cells (2). The thyroid H 2 O 2 gen-erator is found in the apical plasma membrane of rat and pig open follicles (3,4), and is an NAD(P)H oxidase in the pig (5)(6)(7) and human (8) (7,9). Thyrotropin (TSH) induces the expression of thyroid NAD(P)H oxidase through a cAMP-dependent pathway in the dog (10) and pig (11) thyrocytes, and transforming growth factor ␤ counteracts the effect of TSH on pig cells (8). Electron transfer from NAD(P)H to molecular oxygen involves a membrane-bound flavoprotein (12) of unknown structure. Two flavoproteins involved in a mammalian NAD(P)H oxidase system have been cloned to date. One is the glycosylated flavoprotein gp91 Phox involved in the respiratory burst oxidase, also called "phagocyte oxidase" (for a review, see Ref. 13). The second is the p65 Mox , which participates in the activity of the "mitogenic oxidase" found in vascular smooth muscle cells (14). Both require the small membrane subunit p22 Phox for their activity (13,15). A functional NAD(P)H oxidase, generating H 2 O 2 in a Ca 2ϩ -dependent manner, has been solubilized by treating pig thyroid plasma membranes with detergents at a high salt concentration (16). The flavoprotein became unable to reduce molecular oxygen after the first step of its purification by octyl-Sepharose chromatography, suggesting that it requires one or more additional component(s) to generate H 2 O 2 (16). However, the partially purified flavoprotein is still able to catalyze the NADPH-dependent reduction of the electron scavengers nitroblue tetrazolium and potassium ferricyanide (K 3 Fe(CN) 6 ). We used this property to detect and quantify the flavoprotein activity during its purification (16). We have now purified the thyroid oxidase flavoprotein as a first step in defining the molecular structure of the enzyme. Peptide microsequences were used to design a gene-specific primer for rapid amplification of the 3Ј cDNA ends (3Ј-RACE). RACE by PCR provided full-length porcine and human cDNAs encoding a protein called p138 Tox , which comprises 1207 (porcine) and 1210 (human) amino acids. p138 Tox is similar to a predicted 1506amino acids protein encoded by a Caenorhabditis elegans gene and with other flavoproteins involved in plant and mammalian NADPH oxidases. The p138 Tox contains two Ca 2ϩ -binding motifs (EF-hand) which could be involved in the control of the thyroid NADPH oxidase by Ca 2ϩ . We also find that p138 Tox is a new specific marker of thyrocyte differentiation. Finally, the gene of human p138 Tox was located by FISH on chromosome 15q15. * This work was supported in part by grant 7298 from "Association pour la Recherche sur le Cancer", France. 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 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM
FAD-Agarose Chromatography-NADPH oxidase was solubilized from pig thyroid plasma membrane, and the flavoprotein partially purified by chromatography on octyl-Sepharose CL 4B (16) except that FAD was omitted from the washing buffer containing CHAPS and from the elution buffer containing 1% Triton X-100. Batches of eluted protein were stirred overnight at 4°C with 350 l of FAD-agarose equilibrated with 50 mM sodium phosphate, pH 7.2, containing 2 mM MgCl 2 , 0.25 M sucrose, 1 mM NaN 3 , 0.5 mM CaCl 2 , 0.4 mM EGTA, 0.5 mM dithiothreitol, 0.1 M KCl, and 0.2% Triton X-100 (buffer A). 10 mM CHAPS was then added to the suspension, which was again gently stirred overnight at 4°C. The mix was then centrifuged (2000 ϫ g, 30 min at 4°C). The FAD-agarose pellet was washed twice with 3 ml of buffer B (buffer A plus 10 mM CHAPS) and once with 1.5 ml of the same buffer and suspended in SDS-gel sample buffer (1.7% final) (17). The suspension was mixed gently overnight at room temperature and centrifuged. Proteins eluted by SDS-gel sample buffer were recovered in the supernatant.
Proteolytic Cleavage and Amino Acid Sequencing-Analytical SDS-PAGE (8% acrylamide) was performed essentially as described by Laemmli (17). Gels were stained with 0.003% Amido Black in 45% methanol, 10% acetic acid and then extensively washed in distilled water. The band of interest (3.5 g of protein) was removed from the gel, cut into small pieces, dried under vacuum, and incubated in 50 mM Tris (pH 8.6), 0.03% SDS, 0.2 g/ml endoproteinase Lys-C (overnight at 35°C) for amino acid sequencing. The supernatant volume was reduced to 200 l under vacuum and injected onto an AX300 (2.1 ϫ 30 mm) precolumn (Perkin-Elmer) followed by a 218TP52 Vydac reversedphase high performance liquid chromatography column (2.1 ϫ 250 mm). Peptides were eluted with a 2-35% (v/v) linear gradient of 0.1% trifluoroacetic acid in acetonitrile over 90 min at a flow rate of 200 l/min. The major peaks were sequenced.
RACE by PCR-RACE was performed using the SMART RACE cDNA amplification kit (CLONTECH, Palo Alto, CA) according to the protocol of the supplier with 1 g total mRNA from porcine thyrocytes that had been cultivated for 4 days with 10 M forskolin or with human thyroid Marathon-Ready cDNAs. PCR products were cloned in pCR-XL-TOPO (Invitrogen, Inc.) for sequencing. Sequencing was performed three times in the 5Ј to 3Ј direction and three times in the reverse direction.
RNA Isolation and Northern Blot Analysis-The method of Chom-czynski and Sacchi (19) was used to extract RNA from all cells and porcine tissues. For Northern blot analysis, total RNA (20 g) was denaturated and electrophoresed in a 1% agarose gel containing formaldehyde. Denaturated RNAs were transferred by diffusion blotting onto a nylon membrane (Stratagene, La Jolla, CA) using 20 ϫ SSC (1 ϫ SSC ϭ 0.15 M NaCl, 15 mM sodium citrate). The membrane was first incubated in 0.25 M sodium phosphate buffer (pH 6.8) containing 1 mM EDTA, 7% SDS, 10 mg/ml bovine serum albumin for 4 h at 65°C and then was hybridized overnight at 65°C with the heat-denatured probe cDNA. The cDNA probe was ␣-32 P-labeled by random priming extension. The membrane was washed three times for 20 min in 2 ϫ SSC, 0.1% SDS at 65°C and then was autoradiographed at Ϫ80°C using Hyperfilm MP (Amersham Pharmacia Biotech). The cDNA probe used was the porcine 5Ј-RACE cDNA EcoR1 insert (4 kb) of pCR-XL-TOPO. The sense and antisense 5Ј-RACE PCR primers (Eurogentec Bel SA) were 5Ј-CCCACACACACTTCATCTGGCACAGCCTGC-3Ј and 5Ј-TGG-CACAGCCTCTGAGCAGTTCTTCTC-3Ј. Fluorescent in Situ Hybridization-The probe used to locate the p138 Tox gene was the pCR-XL-TOPO vector containing the human 4-kb product of the 5Ј-RACE. The probe preparation and FISH conditions were essentially as described previously (20). Labeled probe (40 ng) with competitor (120 ng of human placental DNA) were denatured and placed directly on the slides without incubation at 37°C. The signal was detected as described by Valent et al. (21).

RESULTS AND DISCUSSION
Purification and Molecular Cloning of the Flavoprotein-Porcine flavoproteins bound to the FAD-agarose and eluted with SDS migrated in SDS-PAGE as a single major band of apparent molecular mass of 180 kDa, and two diffuse bands with apparent molecular masses of 150 and 130 kDa (Fig. 1). Digestion of the major 180-kDa protein (see "Experimental Procedures") with endoproteinase Lys-C generated two peptides, peptide A: AVVPPPRLYTEALQEK, and peptide B: X(V)QLIN(R/P)QDQ(T)HFVHHYEN(P), whose sequences were not found in protein data bases. Peptide B was used to design a gene-specific primer for the 3Ј-RACE. We carried out 5Ј-RACE and then amplified porcine and human full-length cDNAs by reverse transcriptase-PCR. These were cloned and sequenced. The porcine cDNA (GenBank TM accession number AF181973) is 4996-bp long, and the human cDNA (GenBank TM accession number AF181972) has 5160 bp; they are 78.1% identical (93.5% in coding region). The open reading frame (porcine: 3621 bp; human: 3630 bp) encodes a predicted protein with a theoretical molecular mass of 138 kDa. Kosak sequences (A/G)CCATGG were present at the translation start codon (porcine: CCACCATGG; human: CTACCATGG). The predicted protein, named p138 Tox for "p138 thyroid-oxidase", contains peptides A and B, confirming the cloning of the purified 180-kDa flavoprotein.
General Organization of p138 Tox -Human and porcine p138 Tox are 86.7% identical and contain 1210 and 1207 amino acids, respectively. p138 Tox contained three main domains ( Fig.  2A). The N-terminal region had similarities to peroxidases. Residues 49 -206 of the human p138 Tox are 27% identical to a region of pig TPO that excludes the heme-linked proximal and distal histidines. There are four putative sites of N-glycosylation on the human p138 Tox and three on the pig protein ( Figs.  2A and 3A), all in a predicted extracellular N-terminal domain. The difference between the 138-kDa theoretical molecular mass and the 180-kDa apparent mass of the flavoprotein could reflect the presence of complex sugars at these N-glycosylation sites. A median domain is similar to the Ca 2ϩ -binding proteins containing Ca 2ϩ EF-hand motifs. The C-terminal domain is similar to gp91 Phox and p65 Mox , the large subunits of phagocyte (13) and vascular smooth muscle cell NADPH oxidases (14).
Similarity to Non-mammalian Proteins-The overall sequence of p138 Tox is very similar to that of a predicted protein from C. elegans (GenBank TM accession number AF043697). In addition to its glycosylated N terminus being similar to peroxidases, the Caenorhabditis protein has two EF-hand motifs in the middle of the primary structure ( Fig. 2A) which are conserved in p138 Tox ( 494 DKDGNGYLSFREF 506 and 530 DLDENG-FLSKDEF 542 in human sequence). These Ca 2ϩ -binding sites could be involved in the direct activation of the thyroid H 2 O 2 generator by Ca 2ϩ . The flavoprotein p138 Tox is also similar to rbohA, a homolog of gp91 Phox involved in the respiratory burst oxidase of Arabidopsis thaliana, which also contains two Ca 2ϩbinding EF-hand motifs in its hydrophilic N-terminal domain (22).
Hydrophilicity plots obtained by the method of Kyte and Doolittle (23) showed similar distributions of seven hydrophobic stretches in Caenorhabditis protein and in p138 Tox , indicating that the proteins share a well conserved tertiary structure (Fig. 2B). The first five hydrophobic stretches are probably membrane-spanning domains, according to the TMpred program (24), with the glycosylated N-terminal outside the cell and the EF-hand motifs logically located inside. However, p138 Tox is also a truncated mammalian version of the Caenorhabditis protein, which has an additional 331 amino acids at its N terminus. Arabidopsis rbohA has a similar, but even shorter, N terminus ( Fig. 2A), so that the first membrane-spanning domain is absent as well as the peroxidase-like glycosylated domain.
Similarities to Mammalian Oxidases-The C-terminal regions of p138 Tox , rbohA, and the Caenorhabditis protein are very similar to the gp91 Phox and p65 Mox , with a similar distribution of six hydrophobic stretches. An isoalloxazine FADbinding motif of gp91 Phox ( 338 HPFTLTS 344 ) is conserved in human (978 -984) and porcine (975-981) p138 Tox as shown (Fig. 3B) by the multiple alignment of human and porcine p138 Tox sequences with human gp91 Phox sequence made with the CLUSTALW program (25). The ribose ( 410 GIGVTPF 416 ) and adenine ( 534 VFXCGP 539 ) NADPH-binding motifs of gp91 Phox are also found in the C-terminal region of human (1048 -1054) and porcine (1045-1051) p138 Tox and should lie in a cytosolic domain. The four histidine residues, which are probably involved in heme binding in gp91 Phox are conserved in p138 Tox .
Structural Differences between gp91 Phox and p138 Tox -Whereas the FAD and NADPH binding sites of gp91 Phox are clearly present in p138 Tox , alignment of the two sequences indicates some differences. For example, the N-glycosylation sites of p65 Mox and gp91 Phox are absent from the p138 Tox , rbohA, and Caenorhabditis protein C-terminal regions. p138 Tox also differs from gp91 Phox in some other sequence features, which could be relevant to the different specific mechanisms involved in their activation. Previous biochemical studies (7,9) showed that micromolar concentrations of Ca 2ϩ were necessary and sufficient to cause complete, reversible activation of thyroid oxidase, suggesting that interaction of the flavoprotein with cytosolic factors such as p47 Phox and p67 Phox is not required, unlike the respiratory burst oxidase. Indeed, three of the four motifs involved in p47 Phox binding to gp91 Phox subunit are not present in p138 Tox (Fig. 3B). Furthermore, the tervalent arsenoxide phenylarsine oxide (PAO), a hydrophobic molecule that reacts with two vicinal thiol groups in proteins, alters the structure of the porcine thyroid oxidase so that electron transfer occurs more slowly but without Ca 2ϩ (26). The porcine and human p138 Tox contain only two adjacent cysteinyl residues (Fig. 3A), which are located in the first predicted membrane-spanning domain ( 268 C and 269 C), just before the intracellular segment containing the two EF-hand motifs. They are probably the binding site of PAO, reinforcing the idea that this region, absent from gp91 Phox , is crucial for the control of the thyroid NAD(P)H oxidase activity by Ca 2ϩ . PAO also affects the respiratory burst oxidase, but in a different manner, because it essentially prevents the human (27) and bovine (28) oxidase activation by cytosolic factors through its binding to gp91 Phox . This particular effect could result from the binding of PAO to the only two adjacent cysteinyl residues of gp91 Phox , located just before the first peptide involved in activation by p47 Phox ( 85 CCSTRVRRQL 94 ), a sequence which is absent from p138 Tox (Fig. 3B).
Tissue Distribution of the p138 Tox mRNA- Fig. 4A shows that the 5-kb p138 Tox mRNA was detected by Northern blot only in thyroid tissue. This strongly supports the idea that the cloned flavoprotein actually participates in the specific function of thyrocytes, the synthesis of thyroid hormones, as an essential component the H 2 O 2 generator associated with TPO. The thyroid-specific expression of the p138 mRNA also justifies the designation of the NADPH oxidase as "thyroid oxidase" or "tox." Flavoprotein p138 Tox as a Marker of Thyrocyte Differentiation- Fig. 4B shows that porcine thyrocytes cultured without forskolin do not contain the 5-kb p138 Tox mRNA, whereas its concentration is maintained, or even enhanced, by cyclic AMP. These data are in good agreement with those of Raspé and Dumont (10), who reported that TSH induces H 2 O 2 generator activity in dog thyrocytes by activating the adenylate cyclase cascade. They also corroborate our results showing that the TSH and/or forskolin stimulate NADPH oxidase activity in pig thyrocytes (11), which seems to be correlated with a higher expression of the flavoprotein (16).
Localization of the p138 tox Gene on 15q15-FISH was used to locate the p138 Tox gene in the human genome (Fig. 5). We analyzed 50 normal metaphases and detected no fluorescent signal in 10 metaphases because of FISH efficiency. But a fluorescent signal was observed at chromosomal band 15q15 in 40 of them. The fluorescent signal was present on only one of the two chromosomes 15 in 10% of positive metaphases, due again to the efficiency of FISH or to the detection procedure used. 15q15 is not presently considered to be a susceptibility locus for any thyroid disease.
We have demonstrated that the thyroid NAD(P)H oxidase is different from other known mammalian NADPH oxidases. Its flavoprotein is not only a new homolog of gp91 Phox and p65 Mox , but it also has specific structural features that can account for its biochemical properties. Among them is the presence of two calcium binding EF-hand motifs, which have only been found in non-mammalian NADPH oxidases to date. Its tissue specificity and cAMP-regulated expression makes p138 Tox a new marker of thyrocytes, which could be relevant in thyroid disorders.