Developmental expression of functional cyclooxygenases in zebrafish

Abstract

Study of the cyclooxygenases (COXs) has been limited by the role of COX-2 in murine reproduction and renal organogenesis. We sought to characterize COX expression and function in zebrafish (z). Full-length cDNAs of zCOX-1 and zCOX-2 were cloned and assigned to conserved regions of chromosomes 5 and 2, respectively. The deduced proteins are 67% homologous with their human orthologs. Prostaglandin (PG) E₂ is the predominant zCOX product detected by mass spectrometry. Pharmacological inhibitors demonstrate selectivity when directed against heterologously expressed zCOX isoforms. Zebrafish thrombocyte aggregation ex vivo and hemostasis in vivo are sensitive to inhibition of zCOX-1, but not zCOX-2. Both zCOXs were widely expressed during development, and knockdown of zCOX-1 causes growth arrest during early embryogenesis. zCOX-1 is widely evident in the embryonic vasculature, whereas zCOX-2 exhibits a more restricted pattern of expression. Both zCOX isoforms are genetically and functionally homologous to their mammalian orthologs. The zebrafish affords a tractable model system for the study of COX biology and development.

Cyclooxygenases (COXs), also known as prostaglandin (PG) endoperoxide G/H synthases (EC 1.14.99.1), catalyze the conversion of arachidonate to PGs. Two different COX isozymes, COX-1 and COX-2, encoded by separate genes, have been identified (1, 2). COX-1 is expressed constitutively and is primarily responsible for PGs that maintain homeostatic function. Conversely, COX-2 is highly regulated by growth factors, tumor promoters, and cytokines. The recognition of such segregated actions of the COX isozymes has rationalized the development of selective COX-2 inhibitors.

Detailed elucidation of the role of COX isozymes in rodent model systems has been constrained by the importance of COX-2 during development (3). The zebrafish (Danio rerio) has emerged as an informative model organism for studies of vertebrate biology and genetics. Multiple phenotypes have been identified in chemical mutagenesis screens (4), of which some are strikingly reminiscent of human disease. For example, mutation of the novel basic helix–loop–helix transcription factor gridlock results in a stenosis of the aorta, which can be visualized by confocal microangiography and may be analogous to coarctation of the aorta in humans (5). The recent development of morpholino-based gene targeting technology allows for specific gene inactivation in zebrafish. Given the existence of phospholipase A₂, an enzyme that liberates arachidonic acid for subsequent COX metabolism in zebrafish (6), we hypothesized that the prostanoid biosynthetic pathway might be expressed and functional in this model system.

Materials and Methods

Isolation of zCOX-1 and zCOX-2.

Zebrafish embryos were obtained from wild-type AB strain fish and raised at 28.5°C as described (7). Zebrafish expressed sequence tag (EST) clones with high homology to human COX-1 (clone fc62e06.x1) and human COX-2 (clone fa92e05.y1) were identified in the Washington University (St. Louis) Genome Resources EST database and sequenced. EST clone fc62e06.x1 spanned the entire length of the putative zCOX-1 cDNA. Clone fa92e05.y1 contained 1,811 bp of the putative zCOX-2 cDNA including the 3′ untranslated region (UTR). Full-length zCOX-2 cDNA was obtained by performing 5′ anchored PCR using total RNA from adult zebrafish brains. Two nested PCRs were performed subsequently, and the final PCR product (≈450 bp) was subcloned (pCR2.1-TOPO, Invitrogen) and sequenced. Phylogenetic reconstruction was performed by the neighbor-joining method and confirmed by parsimony analysis with 1,000 bootstrap replicates by using bionavigator (Entigen, Sunnyvale, CA). zCOX-1 and zCOX-2 were physically mapped by using a zebrafish radiation hybrid panel (Research Genetics, Huntsville, AL) (8).

Eicosanoids.

Either the isoform nonselective inhibitor indomethacin or the selective COX-2 inhibitor NS-398 (Cayman Chemical, Ann Arbor, MI) were added to the tank water for 12 h. The effects on total body eicosanoid formation were measured by GC/MS (9) using homogenized tissue of adult zebrafish.

Heterologous Expression of COXs.

A 2.1-kb BamHI/SalI fragment from EST clone fc62e06.x1 containing full-length zCOX-1 cDNA was subcloned into BamHI and XhoI sites in the mammalian expression vector pcDNA3.1 (Invitrogen). The zCOX-2 expression vector was constructed by cloning a 1,836-bp PCR fragment amplified from zebrafish brain cDNA onto pcDNA3.1 TOPO (Invitrogen). zCOX-2 cDNA-specific primers were designed that introduced a consensus Kozak sequence before the original start codon and removed the zCOX-2 3′ UTR (10).

COS-7 cells were grown in 6-well plates (10⁶ cells/well) and transiently transfected (Fugene6, Roche Molecular Biochemicals) with the expression constructs (1 μg/well). Forty-eight hours posttransfection, the cells were pretreated (1 h) with either indomethacin (10 nM to 1 μM) or NS-398 (30 nM to 3 μM) in Hanks' balanced salt solution and stimulated with arachidonic acid (20 μM, 30 min). PGE₂ was measured in the media by GC/MS. We compared the relative potencies of the two inhibitors separately in either zCOX-1- or zCOX-2-transfected cells because expression levels of the zCOX proteins could not be assessed in the absence of cross-reacting antibodies.

Thrombocyte Aggregation and Bleeding Time.

Thrombocyte function was assessed by visualizing platelet aggregation on microscopic slides. Blood was harvested from adult zebrafish after a 12-h exposure to the COX inhibitors indomethacin (10 nM) or NS-398 (100 nM). Whole blood aggregation was induced by adding ADP (16 μM) to 2 μl 1:4 citrated (3.8%) blood at room temperature and staining for glycoprotein IIb/IIIa as described (11). Adult zebrafish were exposed to indomethacin or NS-398 (10 nM-1 μM) for 12 h and then anaesthetized (tricaine 0.01%) for the bleeding time assay. Their caudal fins were cut off in the vascularized region under a dissecting scope, and time to the cessation of bleeding was measured.

RNA Localization.

zCOX-1 and zCOX-2 transcripts were detected by using reverse transcriptase–PCR and total RNA from adult organs and fertilized embryos at 2.5 hours postfertilization (hpf). Whole-mount in situ hybridizations were performed on zebrafish larvae (12). The zCOX-2 riboprobe (682 bp) was generated from BglII-cut EST clone fa92e05.y1 and contained the 3′ UTR. The zCOX-1 probe (full length) was made from the PstI-linearized EST clone fc62e06.1. We used a flk-1 riboprobe (full length, SmaI) as a marker for endothelium (13).

Morpholino Microinjection.

We designed two morpholino-modified oligonucleotides (morpholinos) for zCOX-1 and two for zCOX-2 that targeted separate sequences in the 5′ regions of the cDNAs. zCOX-1-A (5′-TCAGCAAAAAGTTACACTCTCTCAT-3′) and zCOX-2-A (5′-AACCAGTTTATTCATTCCAGAAGTG-3′) included the start codon, whereas zCOX-1-B (5′-AACAACAACGTCTTCAAAAGTGGTG-3′) and zCOX-2-B (5′-GCTGTTGAAGCAGAGATGCGTTACT-3′) sequences were upstream in the 5′ UTRs (Gene-Tools, Coravallis, OR). Four-base mismatch morpholinos were used as controls (zCOX-1-A4mis, 5′-TCAcCAAtAAGTTACACTgTCTgAT-3′ and zCOX-2-A4mis, 5′-AACgAGTaTATTCATTCCtGAAcTG-3). Solutions were prepared and microinjected into the yolk of one-cell stage embryos as described (14). Capped zCOX-1 RNA for rescue experiments was synthesized from the SalI-linearized pcDNA3.1-zCOX-1 plasmid by using SP6 polymerase (mMessage mMachine Kit, Ambion, Austin, TX).

Results

Zebrafish COX-1 and COX-2.

Sequence analysis of the putative zCOX-1 cDNA (2,094 bp) revealed a 1,794-bp ORF, preceded by 126 untranslated bases, and a 174-bp 3′ UTR. Although the zCOX-1 cDNA was considerably shorter than the human sequence (GenBank accession no. BC013734), which has an extended 3′ UTR of about 1.5 kb, the putative zCOX-2 cDNA composition (2,150 bp) was similar to the human ortholog (2,539 bp, GenBank accession no. XM_044882). It contained a 68-bp 5′ UTR, an 1,806-bp ORF, and a 276-bp 3′ UTR with AU-rich elements as is characteristic for COX-2s (15).

Translation predicted a 597-aa zCOX-1 protein and a 601-aa zCOX-2 protein. zCOX-1 showed a 12-aa N-terminal insertion that was absent in zCOX-2, whereas zCOX-2 contained an 18-aa C-terminal motif that was not present in zCOX-1, characteristic distinctions in virtually all species (Fig. 1). Multiple alignment with the human COX sequences revealed a 67% identity between the species for each isoform. The highest homologies were found in the epidermal growth factor and catalytic domains (≥72%). Analysis of class-specific amino acids showed that expected glycosylation sites, potential heme coordinating histidines, and virtually all residues that are considered critical for catalysis were conserved in both putative zCOX isozymes (ref. 16; Fig. 1). The volume of the arachidonate-binding channel of mammalian COX-1 enzymes is determined by Ile-523, His-513, and Ile-434 (17–19). These residues are substituted by Val-523, Arg-513, and Val-434 in human COX-2, which opens access to a side pocket in the arachidonate-binding channel for selective COX-2 inhibitors. Surprisingly, both zCOX-1 and zCOX-2 had Val and Arg in positions 523 and 513. The Ile-434–Val substitution, however, was conserved (Fig. 1).

Figure 1.

(A) Alignment of the deduced zebrafish COX-1 and COX-2 aa sequences with their human orthologs. Dots indicate identity, and gaps are represented by dashes. Critical residues for catalysis (Arg-120, Tyr-355, Tyr-385), aspirin acetylation (Ser-530), heme coordination (His-207, His 388) are highlighted by boxes. The putative N-glycosylation sites (Asn-69, Asn-144, Asn-410, Asn-577) are marked by *. Ile-413–Val, His-513–Arg, and Ile-523–Val are highlighted by boxes.

The genetic distances between the zebrafish COXs and other known COXs calculated from the amino acid sequences are shown in a dendrogram (Fig. 2). The zebrafish sequences locate to the respective arms of the other vertebrate COX-1 and COX-2 enzymes, which are separated from the invertebrate COX of the coral species Gersemia fruticosa (20) and Plexaura homomalla (21). Parsimony analysis confirmed the branch topology of the distance phylogeny, indicating that mammalian and teleost COXs were indeed sister groups (data not shown).

Figure 2.

Neighbor-joining phylogenetic distance dendrogram of vertebrate COX-1 and COX-2 proteins and ancestral invertebrate coral COX enzymes (Gersemia fruticosa and Plexaura homomalla).

Chromosomal Assignment of Zebrafish COX-1 and COX-2.

The zCOX-1 gene was unambiguously mapped on zebrafish linkage group (LG) 5 in between markers Z6880 and Z14143 by using a radiation hybrid panel. This mapping places zCOX-1 between 65.1 cM (Z6880) and 74.4 cM (Z14143) from the telomeric end of LG 5. Interestingly, the synteny between zCOX-1 and two other genes, RXRG and NOTCH1B, was identical to that described for hCOX-1 on human chromosome 9 (22). zCOX-2 was assigned to LG 2, between markers Z6569 and Z9944, placing it between 35.8 and 38.3 cM on the map. Again the syntenic relations were conserved in this region, as COX-2 maps into close vicinity of CPLA2 both on zebrafish LG 2 and human chromosome 1 (22).

Eicosanoid Formation in Zebrafish.

In mammals, arachidonic acid is the major COX substrate in vivo, although other polyunsaturated fatty acids, such as ω-3 fatty acids, may function as alternate substrates. PG biosynthesis was profiled in homogenized adult zebrafish. PGE₂ was the predominant product (9.1 ± 1.3 pg/mg tissue, n = 20). The stable hydrolysis product of prostacyclin (PGI₂), 6-keto PGF_1α, was much less abundant (0.33 ± 0.06 pg/mg tissue, n = 10). Minimal concentrations of TxB₂ were detectable (0.08 ± 0.008 pg/mg tissue, n = 5). PGE₃, 6-keto-PGF_2α, and TXB₃ were all at or below the detection limit (≈0.01 pg/mg tissue) in zebrafish on a brine shrimp diet (data not shown).

One week's exposure of zebrafish to 3 μM indomethacin or 10 μM NS-398 (each n = 5) did not appear to influence their activity. The effects of the agents on PG biosynthesis were analyzed in homogenized fish. Treatment with indomethacin for 12 h resulted in concentration-dependent suppression of total body PGE₂ formation (n = 7, P < 0.05 vs. control) with an IC₅₀ of 13 nM (Fig. 3A). Similarly, the COX-2 selective inhibitor NS-398 also inhibited total PGE₂ in a concentration-dependent fashion with an IC₅₀ of 93 nM (n = 7; P < 0.05 vs. control) (Fig. 3B). PGI₂ was similarly inhibited by both compounds; however, the dose dependency was less pronounced (data not shown).

Figure 3.

(A and B) Inhibition of PGE₂ formation in adult zebrafish by (A) indomethacin (IC₅₀ 13 nM) and (B) NS-398 (IC₅₀ 93 nM). The data are means (±SEM) of seven independent experiments. (C and D) Analysis of inhibitor selectivity in COS-7 cells overexpressing either (C) zCOX-1 or (D) zCOX-2. Data are means (±SEM) of four experiments. (E–G) Zebrafish thrombocyte aggregation induced ex vivo by ADP (16 μM) after exposure of adult zebrafish to either indomethacin or NS-398. (E) A representative experiment showing single thrombocytes after ADP stimulation in blood from an indomethacin-treated fish. Immunostaining for GPIIb/IIIa was used as a thrombocyte marker. (F) Thrombocyte aggregation induced by ADP in blood from an NS-398-exposed fish. (G) In indomethacin (Indo)-treated fish, the frequency of observed aggregation was reduced to four of 10 experiments, whereas vehicle or NS-398-treated zebrafish exhibited platelet aggregation in all experiments either spontaneously (black bars) or ADP-induced (hatched bars) in 10 experiments per treatment group. (H) Bleeding time (seconds) in adult zebrafish after exposure to various concentrations of indomethacin and NS-398. Only the nonselective inhibitor indomethacin resulted in prolongation of bleeding time. Data are means (± SEM) of five experiments per group.

Heterologous Expression of zCOXs.

We confirmed activity of the proteins by detection of PGE₂ in COS-7 cells transfected with either pCDNA3.1-zCOX-1 or pCDNA3.1-zCOX-2 upon addition of the substrate archidonic acid (20 μM). Vector control-transfected cells generated only minor amounts of PGE₂ under these conditions (data not shown). The relative selectivity of COX inhibitors was assessed by preincubating the transfected cells with indomethacin (10 nM-1 μM) and NS-398 (30 nM-3 μM). Although both suppressed PGE₂ to a similar degree in cells expressing zCOX-2, indomethacin was distinctly more potent than NS-398 in zCOX-1-transfected cells (Fig. 3 C and D). The effective concentrations for both compounds were compatible with the IC₅₀s of indomethacin and NS-398 for COX-1 and COX-2 in mammalian systems (23, 24).

Thrombocyte Aggregation and Bleeding Time in COX Inhibitor-Treated Zebrafish.

Aggregation of thrombocytes was detected by GPIIb/IIIa immunostaining in whole blood (Fig. 3 E and F). Adult zebrafish were treated with vehicle, indomethacin (10 nM), or NS-398 (100 nM) at concentrations previously determined as their IC₅₀ for suppression of PGE₂ formation in live fish (see Fig. 3 A and B). In five of 10 vehicle-treated fish, harvesting of blood led either to spontaneous formation of visible aggregates (n = 2) or the complete clearance of thrombocytes from the samples at baseline (n = 3). Additional stimulation with ADP (16 μM) induced aggregation in all five remaining vehicle-treated samples. Basal thrombocyte activation upon blood withdrawal was observed in three of 10 indomethacin-pretreated zebrafish. Stimulation with ADP did not induce aggregate formation in six of the remaining seven samples. By contrast, basal thrombocyte activation was seen in six of 10 NS-398-exposed fish. All four remaining samples were activated upon addition of ADP and showed aggregate formation (Fig. 3 E–G). Bleeding time was prolonged by indomethacin (≥100 nM), but not by NS-398 (30 nM-3 μM, Fig. 3H), consistent with the effects of COX inhibitors in humans.

zCOX-1 and zCOX-2 Expression.

We assessed constitutive zCOX isoform expression in adult zebrafish by reverse transcriptase–PCR (Fig. 4a). zCOX-1 transcripts were detected in all analyzed tissues except brain. The most robust zCOX-2 signals were detected in gills, intestine, and testes, followed by heart, skeletal muscle, and the brain. zCOX-2 message was absent in adult liver. Interestingly, transcripts of both isoforms were detected in ovaries, consistent with maternally expressed genes. Indeed, both transcripts were still detectable by reverse transcriptase–PCR (data not shown) in fertilized oocytes before the onset of zygotic transcription 2.5 hpf.

Figure 4.

(a) Analysis of zCOX-1 and zCOX-2 transcript expression in adult zebrafish. We assessed constitutive zCOX isoform expression in adult zebrafish by reverse transcriptase–PCR using total RNA from freshly dissected adult organs. PCR products (500 bp and 162 bp) and the corresponding negative controls were visualized by agarose gel-electrophoresis. Expression of zCOX-1 (b, e, h, k, n, and q), zCOX-2 (c, f, i, l, o, and r), and flk-1 (d, g, j, m, p, and s) by whole-mount in situ hybridization in zebrafish larvae at 96 hpf. zCOX-1 (b), zCOX-2 (c), and the endothelial marker flk-1 (d) are expressed in the carotid artery (ca) and the pharyngeal arches (white arrow) (lateral view). zCOX-1 (b) and flk-1 (d) are also present in cranial arteries (arrowheads). (e–g) Ventral views showing expression in the pharyngeal arches (arrowheads). While flk-1 staining (g) highlights the vasculature in the center of the arches [arch arteries 1 and 2 (opercular artery); 3–6, ha, hypobranchial artery (35)], zCOX-1 (e) and zCOX-2 (f) appear to be expressed in more peripheral structures of the arches. (h–j) Oblique ventral views. zCOX-1 (h), zCOX-2 (i), and flk-1 (j) colocalize and are expressed within sprouting gill arteries (*) that are apparent as protrusions at the caudal surface of arch arteries 3–6. (k–m) Histological cross sections through these vessels (*, section rotated by 90°) show flk-1 (m) expression in endothelial cells. zCOX-2 is highly expressed in the vessel wall of the sprouting gill vasculature (l). zCOX-1 expression is less intense and appears to be present in both endothelial and wall structures (k). (n–p) Endothelial expression of zCOX-1 was seen the ciliary arteries of the eye (n, arrowheads), which were also visible in flk-1 (p), but not in zCOX-2-stained embryos (o). zCOX-1 and zCOX-2 were both differentially expressed in other structures of the developing eye (n and o). (q–s) All three genes were expressed in the vasculature of the intestine. Endothelial cells are highlighted by arrowheads. (Magnifications: (b–j) ×200, (k–m and q–s) ×630, n–p) ×400.)

Both genes were expressed in multiple locations during embryonic development. Expression in the developing vasculature was particularly striking. zCOX-1 and zCOX-2 both were expressed in the carotid arteries and the vasculature of the pharyngeal arches at 96 hpf. Both COX isoforms colocalized with flk-1 on whole-mount analyses of carotid arteries (Fig. 4 b–d), while their expression within the pharyngeal arches diverged. In ventral views of whole mounts, zCOX-1 and zCOX-2 appeared to be expressed on the periphery of the arches, whereas flk-1 was restricted to their centers (Fig. 4 e and f). Whole mounts viewed obliquely showed flk-1, zCOX-1, and zCOX-2 expression in sprouting gill vasculature, derived from the third through sixth pharyngeal arches (Fig. 4 h–j). Expression of flk-1 localized to the endothelium of the sprouting gill arteries in histological sections, whereas zCOX-2 expression was restricted to the vessel wall. zCOX-1 staining, although less intense, appeared to be expressed in both endothelium and vessel wall (Fig. 4 k–m). zCOX isoform expression was also seen in other larval vascular beds. Endothelial expression of zCOX-1 was evident in cranial vessels such as the nasal ciliary artery of the eye (Fig. 4 n–p), as well as the aorta and the segmental (not shown) and intestinal arteries (Fig. 4 q–s). By contrast, additional vascular expression of zCOX-2 was seen only in the intestine, where it largely overlapped with zCOX-1 and flk-1.

Morpholino Microinjection.

Given the existence of both isoforms during embryogenesis, we screened for developmental perturbation in embryos that were microinjected with zCOX antisense morpholino oligonucleotides. Two separate morpholinos (A and B) and a 4-base mismatch (4mis) control were designed for each gene. Upon injection of a range of concentrations into one-cell stage embryos (0.55–4 ng/embryo), we noticed a considerable retardation of early (6–12 hpf) embryonal development by both morpholinos targeting zCOX-1, as compared with mismatch and anti-zCOX-2 morpholinos. Lethality increased dramatically because of complete growth arrest in concentrations ≥2 ng (2 ng: lethality measured at 12 hpf >30% for zCOX-1-A and zCOX-1-B, for all other morpholinos 10–15%). Lethality was less than 15% in both zCOX-1 morpholino groups at concentrations less than 2 ng. However, the majority of embryos were significantly delayed in epiboly, a process comprised of rapid cell division and migration that accompanies the formation and invagination of the germ layers (Fig. 5 A and B). Growth retardation and consequent lethality were dose dependent (data not shown). In some zCOX-1 mismatch morpholino-injected embryos, mild retardation was observed, but only at high concentrations (2 and 4 ng), probably reflective of residual antisense activity (25). The majority of the delayed embryos died between 12 and 18 hpf or showed severe malformations such as defective head, short tail, or abnormal somites (not shown). We coinjected capped zCOX-1 RNA with the zCOX-1 morpholinos to demonstrate their specificity by phenotypic rescue. While the fraction of delayed embryos in the uninjected control group, the zCOX-RNA (200 pg)-injected, and the mismatch morpholino groups (0.85 ng) was ≤10%, zCOX-1-A and zCOX-1-B morpholinos (0.85 ng) caused severe growth retardation in approximately 66% and 50%, respectively, of the embryos (Fig. 5C). Coinjection of the zCOX-1 morpholinos with zCOX-1 RNA reversed this ratio and resulted in over 60% of normally developing embryos (Fig. 5C), consistent with successful rescue.

Figure 5.

Morpholino knockdown of zCOX-1. (A and B) Micrographs of zebrafish embryos (Normarski optics) at 11 hpf. Embryo (A) was injected with a control mismatch morpholino (0.85 ng) in the one-cell stage and has progressed normally to the tailbud stage. Embryonal cells have migrated around the yolk and the embryonal axis has been established. Embryo (B) was injected with 0.85 ng zCOX-1 antisense morpholino and is severely delayed in its development. Embryonal cells are covering only 40% of the yolk (40% epiboly). (C) Phenotypic rescue by zCOX-1 RNA. Embryos were classified as normally developed (green bars) or delayed (red bars) at 12 hpf. Lethality at 12 hpf is shown by the black bars. Embryos were microinjected in the one-cell stage with either mismatch morpholino (mis), 0.85 ng morpholino zCOX-1A (A), or 0.85 ng morpholino zCOX-1B (B). The ratio of delayed/normal embryos was reversed when COX-1 RNA (200 pg) was injected simultaneously with the antisense morpholinos as compared with the morpholinos alone. Each treatment group consists of injections of four different egg lays and minimally 160 individual embryos in total. Data are means (±SEM) of percent embryos in each set of delivered eggs.

No perturbation of development by morphological criteria was observed when the zCOX-2 morpholinos were injected and these embryos could be raised to adulthood.

Discussion

A limitation to the use of mouse models for elucidating the biology of COXs has been the diverse reproductive defects reported in the COX-2 knockout mouse and the impact of gene deletion on cardiorenal development in that species. For these reasons, we sought to characterize a complementary model system, which might afford insight into the biology and pharmacology of the COX isozymes. The zCOXs exhibit many genetic, pharmacological, and functional characteristics of the human enzymes. Additionally, we use a gene knockdown approach to reveal a unique role for COX-1 in development.

Several lines of evidence support the contention that we cloned the zebrafish orthologs of the human COX isozymes. Sequence conservation ranges from 65% to 73%. All of the putative domain structures are conserved, as is the isoform-specific sequence divergence in their signal peptides and their carboxyl termini. Tyrosine and arginine residues, key to the catalytic activity of the enzymes, are also conserved, as are potential heme-coordination sites and the serine target for aspirin acetylation (26). Another distinction between the isozymes, which is conserved among species, is the presence of multiple AU-rich motifs in the extended 3′ UTR of COX-2. Both zCOXs elaborate PGE₂ when expressed in a heterologous expression system and PGs are formed in zebrafish tissue. Finally, both genes map to synteny groups that correspond to segments of conserved map orders on human chromosomes 9 and 1, the loci of human COX-1 and COX-2, respectively.

COX inhibitors suppressed formation of PGs by intact zebrafish. Similar IC₅₀s were determined for an isoform nonselective nonsteroidal antiinflammatory drug, indomethacin, and for a selective COX-2 inhibitor, NS-398. Two lines of evidence were obtained to support isoform selectivity for the latter compound in the zebrafish. First, indomethacin was a more effective inhibitor of zCOX-1-derived PGE₂ formation, whereas both drugs were similarly effective inhibitors of zCOX-2-derived PGE₂. Interestingly, previous studies of the mammalian enzymes have shown that mutation of three residues positioned in the hydrophobic channel that permits access of the arachidonic acid substrate to the active site of COX(Ile-523–Val, His-513–Arg, and Ile-434–Val) opens the side pocket relevant to binding of selective COX-2 inhibitors (16). Although two of these substitutions are evident in both COX isozymes in zebrafish, sufficient distinction is maintained to permit selective inhibition of zCOX-2 by NS-398. Second, only the isoform-nonspecific inhibitor prevented spontaneous and induced thrombocyte aggregation and increased the bleeding time, whereas the COX-2 inhibitor was ineffective. This finding suggests that zebrafish thrombocytes, like mature human platelets (27), express only COX-1.

The predominant COX product of the adult zebrafish is PGE₂. Although its tissue source is unknown, it is the most abundant COX product of microvascular endothelium in humans (28). PGI₂, the predominant product of macrovascular endothelium in humans, was a minor product of whole zebrafish. PGE₂ plays an important role in angiogenesis (29), regulation of blood pressure (30), and hemostasis (31) in mice. Marked inhibition of PGE₂ synthesis by both nonselective and COX-2-selective inhibitors suggests that COX-2 is the major source of zebrafish PGE₂. Indeed, constitutive expression of the zCOX-2 transcript was apparent in blood vessels and the intestine, likely sources of PGE₂, in adult fish.

Parsimony analysis of cloned COXs suggests that invertebrate coral COXs may be common ancestors of the vertebrate COX isozymes and that divergence of the two isoforms occurred later, during the separation of vertebrates from invertebrates. Arachidonic acid is released by phospholipase A₂ in zebrafish (32), and the EST database reveals the likely presence of zebrafish orthologs of human PG receptors. Thus, it is probable that the entire PG biosynthetic/response pathway is present in zebrafish. In vivo assays, using fluorescent reporter substrates that visualize molecular processes in the translucent larvae, may emerge to represent elegant screening tools in large-scale genetic screens (32).

A comparative advantage of the zebrafish as a model system is the ease of detection of gene expression during development. For example, during development, zCOX-2 was evident in vascular structures of the pharyngeal arches, in the carotid artery, and in the intestinal vasculature. Although zCOX-2 appears to colocalize with zCOX-1 and flk-1 in endothelial cells of the intestinal vasculature, this isoform appears to be exclusively expressed in nonendothelial cells of the vessel wall of sprouting pharyngeal arch arteries.

Although targeted gene disruption is not yet feasible in zebrafish, microinjected antisense morpholino oligonucleotides afford specific, but transient, gene inactivation (knockdown) that identifies gene-specific effects during development. The potential utility of this approach is illustrated by such a knockdown of zCOX-1. We used two independent morpholinos and caused early embryonal growth arrest and lethality during epiboly, a complex process involving morphogenetic cell movements of involution, conversion, and extension between 4 and 10 hpf. The use of mismatch controls, zCOX-2 morpholinos, and phenotypic rescue with concomitant zCOX-1 RNA injection supports the specificity of the maneuver, which reveals a unique role for zCOX-1 during development. Mutant zebrafish embryos that are characterized by developmental arrest during epiboly show severe defects in cell migration (33). In COX-1-deficient mice such an early developmental phenotype may have been obscured by maternal generation of PGs (34).

In summary, the zebrafish affords an additional model system in which the biology of COX isozymes and their products may be investigated. The detectable expression of both isozymes during development as well as in the adult, the apparent recapitulation of the human pharmacology of enzyme inhibitors, the applicability of zebrafish to high throughput analysis of drug effect and mutant analysis, and the potential for gene inactivation and overexpression all suggest that studies in zebrafish will contribute substantially to our understanding of this complex system.

Abbreviations

COX

cyclooxygenase

PG

prostaglandin

UTR

untranslated region

hpf

hours postfertilization

EST

expressed sequence tag