Identification of the Maturation Factor for Dual Oxidase

Dual oxidase 2 (DUOX2), an NADPH:O2 oxidoreductase flavoprotein, is a component of the thyroid H2O2 generator crucial for hormone synthesis at the apical membrane. Mutations in DUOX2 produce congenital hypothyroidism in humans. However, no functional DUOX-based NADPH oxidase has ever been reconstituted at the plasma membrane of transfected cells. It has been proposed that DUOX retention in the endoplasmatic reticulum (ER) of heterologous systems is due to the lack of an unidentified component required for functional maturation of the enzyme. By data mining of a massively parallel signature sequencing tissue expression data base, we identified an uncharacterized gene named DUOX maturation factor (DUOXA2) arranged head-to-head to and co-expressed with DUOX2. A paralog (DUOXA1) was similarly linked to DUOX1. The genomic rearrangement leading to linkage of ancient DUOX and DUOXA genes could be traced back before the divergence of echinoderms. We demonstrate that co-expression of DUOXA2, an ER-resident transmembrane protein, allows ER-to-Golgi transition, maturation, and translocation to the plasma membrane of functional DUOX2 in a heterologous system. The identification of DUOXA genes has important implications for studies of the molecular mechanisms controlling DUOX expression and the molecular genetics of congenital hypothyroidism.

Generation of H 2 O 2 at the apical membrane of thyroid follicular cells is essential for iodination of thyroglobulin by thyroid peroxidase and constitutes the rate-limiting step of thyroid hormone synthesis. Dual oxidases (DUOX1 and DUOX2) 2 appear to constitute the catalytic core of the H 2 O 2 generator (1,2). They are large homologs of the phagocyte gp91 phox /Nox2 NADPH-dependent oxidase with an N-terminal extension comprising a peroxidase-like domain. Although the crucial role of DUOX2 in thyroid hormonogenesis has been substantiated by reports of severe congenital hypothyroidism in patients with biallelic nonsense mutations (3), the understanding of structure, function, and regulation of DUOX has remained limited. The major obstacle for molecular studies of DUOX is the lack of a suitable heterologous cell system for DUOX-based functional NADPH oxidase expression. Transfected cells completely retain DUOX in the endoplasmatic reticulum (ER) (4 -8), suggesting that an unidentified component, essential for DUOX maturation, may be specifically expressed in tissues containing the functional enzyme.

EXPERIMENTAL PROCEDURES
Data Mining and Computational Analysis-Massively parallel signature sequencing (MPSS) data (9) were obtained from the NCBI Gene Expression Omnibus repository (www.ncbi.nlm.nih.gov/geo/; records GSE1747 and GPL1443). A thyroid specificity score, as defined by Jongeneel et al. (9), was calculated for signatures with frequency Ͼ100 tags per million (ϳ30 mRNA copies/cell) in the thyroid/parathyroid library. Tags with scores ϾϪ1 were mapped to the human genome assembly using BLAST. DUOXA homologs were identified by tBLASTn searches against the NCBI nr data base and trace archive and BLAT queries (at genome.ucsc.edu/) against assembled whole genome sequences. Orthologs were operationally defined as reciprocal best BLAST hits. Gene structures were deduced by spliced alignment maintaining maximum homolog similarity of the open reading frames (ORFs) and consensus splice junctions. Cladograms were constructed from ClustalX alignments (BLOSUM weight matrix, excluding gaps) using the Jones, Taylor, and Thornton (JTT) substitution model in PHYML 2.4.4 (10). SignalP 3.0 (11) and Phobius (12) were used to analyze signal peptides, transmembrane helices, and topology.
Heterologous Expression of DUOX2 and DUOXA2 Constructs-cDNA was synthesized with Superscript reverse transcriptase (Invitrogen) by oligo(dT) priming of total RNA from a normal human thyroid gland. The DUOX2 and DUOXA2 ORFs were amplified using native Pfu polymerase (Stratagene) and cloned into pcDNA3.1 (Invitrogen). Epitope-tagged constructs and fusions with enhanced green fluorescent protein (EGFP) were prepared by replacement or splicing-by-overlap extension using specifically designed primers. All constructs were verified by sequencing. HeLa cells were cultured and transfected as described (13).
Confocal Laser Scanning Microscopy-Indirect immunofluorescence of permeabilized cells has been described previously (13). For surface staining, cells were incubated with rat anti-HA clone 3F10 and/or mouse anti-c-myc clone 9E10 (both from Roche Applied Science) at 1 g/ml in Hank's buffered saline solution/10 mM Hepes, pH 7.4, 1% bovine serum albumin at 4°C. Rabbit anti-calnexin was obtained from StressGen. Images were captured on a Nikon Eclipse E800 equipped with PCM2000.
Analysis of N-Glycosylation-Postnuclear supernatants (in 50 mM Tris/ HCl, pH 8.0, 150 mM NaCl, and proteinase inhibitors) were adjusted to 0.5% SDS, 0.4 mM dithiotreitol and denatured, at room temperature, for 30 min. Samples were deglycosylated with N-glycosidase F (PNGase F) and endoglycosidase H (Endo H) (both from New England Biolabs) according to manufacturer's recommendations, followed by SDS-PAGE under reducing conditions and Western blotting as described (13).
Measurement of H 2 O 2 Generation-Release of H 2 O 2 was determined by reaction with cell-impermeable 10-acetyl-3,7-dihydroxyphenoxazine (14) (Amplex Red reagent, Invitrogen) in the presence of excess peroxidase, producing fluorescent resorufin. Briefly, cell monolayers were incubated, with or without 10 M diphenyleneiodonium (DPI), in Dulbecco's phosphate-buffered saline supplemented with 50 M Amplex Red reagent and 0.1 unit/ml horseradish peroxidase for 1 h at 37°C. Relative fluorescence units (excitation/ emission: 535/595) were corrected for Amplex Red oxidation in wells containing non-transfected cells and converted into H 2 O 2 concentrations using a calibration curve. Renilla luciferase activity from co-transfected pRL-Tk plasmid (Promega) was used as internal control as described (13).

Identification of Novel Genes in the DUOX1/DUOX2 Intergenic Region-
We used MPSS data for 32 normal human tissues (9) to identify novel transcripts with predominant expression in thyroid gland. One of the extracted tags mapped to an uncharacterized locus (LOC405753) oriented head-to-head to DUOX2 in the ϳ16-kbp DUOX1/DUOX2 intergenic region. For reasons outlined below, we called the corresponding gene DUOX maturation factor 2 (DUOXA2). 3 Based on human-mouse homology (Riken clone 9030623N16Rik), and supported by contig assembly of expressed sequence tags (ESTs), it comprises six exons, confirmed by reverse transcription-PCR amplification from human thyroid tissue (GenBank TM accession number DQ489734). The putative transcription start site defined by clone DKFZp686C04213 maps to a GpC rich region (Fig. 1A). This site is 135 bp from the 5Ј terminus of a spliced DUOX2 EST (BI045475) on the opposite strand. A single polyadenylation signal (Fig.  1A) is supported by all mapped 3Ј ESTs. We confirmed a specific transcript of the expected size (1.3 kbp) by Northern blot analysis (Fig. 1B), which also validated the MPSS-based expression profiling: DUOXA2 mRNA was by far most abundant in thyroid, with lower levels in salivary glands reflecting the known expression profile of DUOX2 (1,2,15).
The DUOXA2 ORF is initiated within a Kozak consensus (gcagcATGa) and spans all six exons. The encoded 320-amino acid protein was strongly pre-dicted to comprise five membrane-integral regions, including a reverse signal anchor with external N terminus (type III) (Fig. 1C). The three NX(S/T) consensus sites for N-glycosylation are clustered within an extended external loop connecting the second and third transmembrane helices.
We identified a single DUOXA2 paralog in the human genome. We will refer to this locus, annotated as "homolog of Drosophila Numb-interacting protein," as DUOXA1. It is immediately adjacent, in tail-to-tail orientation to DUOXA2 and extends, via untranslated exons, into the DUOX1 promoter region. DUOXA1 mRNA was predominantly expressed in thyroid gland and, at lower level, in esophagus (Fig. 1B). Two transcripts of ϳ2.9 and ϳ3.5 kbp were detected, compatible with alternative splicing of 5Ј-untranslated exons and the use of alternative 3Ј-polyadenylation signals (data not shown). The DUOXA1 respectively. The shaded box indicates the MPSS signature extracted by data base mining, which also contains the 3Ј-polyadenylation signal (underlined). Arrows in the upstream genomic region indicate the transcriptional start sites of DUOX2 on the opposite strand, as previously determined by 5Ј-rapid amplification of cDNA ends (23) and as evidenced by EST BI045475. B, multiple tissue Northern blot analysis for DUOXA2 and DUOXA1. C, strongly favored topology model for DUOXA2, depicting five membrane-integral regions including a reverse signal anchor with external N terminus (type III). Identical topology was predicted for the DUOXA1 paralogs. D, plot of the residue-wise posterior probability for transmembrane location in DUOXA2. Data were calculated with the hidden Markov model-based predictor Phobius (12). Note the potential for two additional membrane-spanning regions in the Drosophila homolog (mol-PA). E, maximum likelihood protein cladogram illustrating the relationship of DUOXA homologs (multiple protein alignment shown in Fig. 3S in the on-line supplement). Bootstrapping values for 100 replicates are shown at the nodes. The schematic to the right summarizes the results of the microsynteny analysis. In Caenorhabditis elegans, DUOXA homolog and duox are on distinct chromosomes; in D. melanogaster, they are on the same chromosome, but ϳ14 Mbp apart. Note that for clarity evolutionary recent tandem duplications of the protostomal duox loci are not shown.
ORF was confirmed by sequencing from human thyroid cDNA (GenBank TM accession number DQ489735).
By spliced alignment, we deduced the gene structures of all DUOXA homologs in 10 other vertebrate whole genome assemblies. The splicing sites of all structures were conserved at the single codon level (exon alignment shown in Fig. 2S in the on-line supplement). Remarkably, the bidirectional DUOX/DUOXA arrangement was conserved throughout the vertebrate lineage (Fig. 1E, accession numbers of genomic contigs available in Table 4S in the on-line supplement). Teleosts have a single DUOX/DUOXA arrangement, which has undergone tandem duplication to an inverted repeat (DUOX2/ DUOXA2/DUOXA1/DUOX1) before the amphibian divergence. Analyzing unassembled genomic contigs, we mapped the evolutionary event leading to the bidirectional association of DUOX and DUOXA before the divergence of echinoderms, since linkage of the loci was present in Strongylocentrotus purpuratus. Thus, conserved microsynteny in deuterostomes was a strong predictor for cooperation between DUOX and DUOXA.
The protostomes C. elegans and D. melanogaster lack a DUOXA homolog in the vicinity of their respective duox loci. They do, however, each harbor a single ancient DUOXA homolog. For instance, Drosophila moladietz (mol) encodes a 474-amino acid protein that exhibits 39% amino acid identity over 256 amino acids with human DUOXA1.
Functional Rescue of DUOX2 by DUOXA2-To test whether DUOXA2 can reconstitute DUOX2 activity in a heterologous system, we expressed either DUOX2, DUOXA2, or both in HeLa cells and measured H 2 O 2 released into the culture medium. Transfection of either DUOX2 or DUOXA2 alone did not result in increased H 2 O 2 generation compared with nontransfected cells, confirming previous results for DUOX2 (2,4). Remarkably, co-transfection of DUOX2 with DUOXA2 rescued DUOX2 activity as indicated by the significant amounts of H 2 O 2 released from the cells (Fig. 2A). The H 2 O 2 release triggered by DUOX2/DUOXA2 co-transfection was completely blocked by the flavoprotein inhibitor DPI (Fig. 2A).
Co-expression of DUOXA2 Permits ER-Exit of DUOX2 and Plasma Membrane Targeting via the Secretory Pathway-Lack of DUOX2 activity in heterologous systems has been associated with absence of DUOX2 at the plasma membrane (4). To directly test whether reconstitution of active DUOX2 by DUOXA2 is indeed due to translocation of DUOX2 to the plasma membrane, we HA-tagged DUOX2 at its extracellular domain (HA-DUOX2; tag inserted between Asp 27 and Ala 28 ). Non-permeabilized cells showed strong anti-HA plasma membrane signals in cells co-transfected with HA-DUOX2 and DUOXA2 (Fig. 2B). Untransfected cells, or cells transfected with either DUOXA2 or HA-DUOX2 alone, were devoid of surface fluorescence (Fig. 2B and data not shown).
To determine whether DUOXA2-induced surface expression of DUOX2 involved ER-to-Golgi transition of DUOX2, we analyzed the maturation of DUOX2 N-glycan moieties using specific glycosidases. Whereas all N-glycans are cleavable by PNGase F, the ER-derived high-mannose type N-glycans become resistant to Endo H once they have been modified by Golgi-localized enzymes. HA-DUOX2 expressed in HeLa cells migrated as a single band on SDS-PAGE and was sensitive to full deglycosylation by Endo H, consistent with published data (4). In contrast, co-transfection with DUOXA2 resulted in the appearance of a second DUOX2 species with slightly decreased mobility and complete resistance to deglycosylation by Endo H (Fig. 2C). These findings resembled those previously obtained with endogenous DUOX2 protein (4,5,16), indicating that expression of DUOX2 in our reconstituted system involved normal maturation of DUOX2 within the secretory pathway.
Characterization of DUOXA2 as ER-resident Protein-DUOXA2 could be an integral part of a DUOX2 complex, endowing a holocomplex with the ability to exit the ER and reach the plasma membrane. We, therefore, determined whether myc-tagged DUOXA2 alone or in combination with DUOX2 would be detectable at the plasma membrane. Of several constructs tested, only DUOXA2 with N-(myc-DUOXA2) or C-terminal (DUOXA2-myc/His) attached myc tags were fully functional in rescuing DUOX2 activity as assessed by H 2 O 2 generation and HA-DUOX2 plasma membrane targeting (data not shown). However, neither myc-DUOXA2 nor DUOXA2-myc/His was detectable at the plasma membrane (data not shown), although they had the expected size on Western blot analysis (Fig. 3A) and intracellularly co-localized with HA-DUOX2 (Fig. 3B).
To exclude that this was due to a discrepancy between the modeled and actual DUOXA2 membrane topology or due to masking of the N-terminal epitope tag, we fused EGFP/myc to the C terminus of DUOXA2 (an N-terminal fusion was not functional). As shown in Fig. 3C, DUOXA2-EGFP/myc did not co-localize with HA-DUOX2 at the plasma membrane, the latter delineated by anti-HA surface staining. The intracellular distribution of DUOXA2-EGFP/myc (and of DUOXA2-myc/His) showed a similar distribution pattern as the ER-marker calnexin (Fig. 3D).
To further corroborate that DUOXA2 is indeed an ER-resident protein, we analyzed the maturation of DUOXA2-myc/His N-glycosylation in cells coexpressing HA-DUOX2. We found that the N-glycans of DUOXA2-myc/His were exclusively of the high-mannose type (Fig. 3E). In contrast, detection of HA-DUOX2 in the same samples demonstrated, again, that about half of DUOX2 protein had been subject to Golgi modification of its glycosylation (data not shown, compare Fig. 2C). Collectively, these results indicate that DUOXA2 is not an integral part of a DUOX2 enzyme complex at the plasma membrane but an ER-resident protein promoting ER exit and maturation of DUOX2. It should be noted that N-glycosylation of DUOXA2 supports our topology model (Fig. 1C), since the apparent molecular weight of the N-glycan moieties (ϳ10 kDa) indicates N-glycosylation of all three consensus sites.