The COXes of Danio: From mechanistic model to experimental therapeutics

In a recent issue of PNAS, Grosser et al. (1) report studies that characterize cyclooxygenase-1 and –2 (zCOX-1 and zCOX-2) in zebrafish (Danio rerio). The importance of their work is that experimental systems that can be manipulated genetically have informed many areas of signal transduction, but, until now, studies of the prostaglandin system have largely been performed in mammals. The creation of genetically engineered mice that lack COX-1, COX–2 (2–4), or receptors for prostaglandins (5) has resulted in multiple advances to the field. However, in this case, the zebrafish may prove to be the central model system used to discover genes and molecular mechanisms that regulate developmental pathways. Hundreds of genes that control embryonic patterning and organogenesis have been uncovered in zebrafish by using forward genetic screens of ethynitrosourea-induced mutations that cause specific phenotypes (6, 7). Now, the power of this model is being used to elucidate pathologies relevant to human health. As illustrated by the recent study, zebrafish expressed sequence tags (ESTs) homologous to genes implicated in human disease are available through the Washington University Genome Resources EST-database. Further, the Sanger Center is rapidly sequencing the zebrafish genome. Already, it is clear that there are large syntenic regions between the human and zebrafish genomes (8), and both zCOX-1 and zCOX-2 were found to be within such regions, which helped confirm their identities (1).

What will this description of prostaglandin pathways in zebrafish yield? If history repeats itself, we will learn as many details about these pathways as we have when other signaling systems were dissected in model organisms. Further, analysis of zebrafish may be a powerful way to screen new pharmacological compounds for early clues of untoward effects, which is a potentially important application because nonsteroidal anti-inflammatory drugs (NSAIDs), which can block both COX-1 and –2, and the selective COX-2 inhibitors are among the most widely used products in the pharmaceutical industry (9). Another pharmacological application is found within the new comparative genomics analysis reported by Grosser et al. (1). The sequence of zCOX-2 shows two amino acid changes that previously had been thought to contribute to the structure which determined the specificity of the COX-2 inhibitors. But, they found that the pharmacological profile of the zebrafish enzymes was virtually identical to the mammalian enzymes (1). There is an obvious interest in having a detailed analysis of the enzymatic and pharmacological properties of the recombinant zCOX-2.

Defining the Unique Roles of COX-1 and COX–2

In 1988, the prostaglandin field had the elements of a mature field of investigation. The committed step in conversion of arachidonic acid to prostaglandins, a reaction catalyzed by COX (Fig. 1), was well characterized, and clinically useful inhibitors had been developed. The downstream steps were understood, and the remarkable achievements in the field had been recognized with a Nobel Prize a few years earlier. Further, the cDNA encoding the sheep COX had just been isolated (10, 11), and a wave of studies were underway to elucidate the mechanisms by which it was regulated. This comprehensible world didn't last long, as the literature quickly filled with experimental findings that didn't compute—i.e., they showed discrepancies between enzymatic activity and the amount of protein and lack of correlation with the quantity of mRNA. Virtually every laboratory in the field had evidence that something else was afoot. This era of confusion ended when a second gene encoding a closely related isoform was discovered, first in chicken (12) and mouse (13), and then in human (14, 15). There was a brief period of skepticism about whether COX-2 might be a pseudogene, because the first publications on the chicken and mouse forms didn't demonstrate activity, and the cDNA reported by Hla et al. (14) had very little activity, probably as a result of mutations at residues 165 and 438 as compared with the sequence reported by Jones et al. (15). But, work from many groups quickly showed that COX-2 accounts for the induced increase in synthetic capacity for prostaglandins in response to cytokines, growth factors, and tumor promoters (15–17).

Figure 1.

Pathway for prostaglandin synthesis. Arachidonic acid is found in membrane phospholipids in resting cells and is released as a free fatty acid by the action of phospholipase(s). It then can be a substrate for a variety of enzymes that catalyze the addition of oxygen, resulting in multiple products with diverse physiological actions. The free arachidonic acid itself has a variety of effects on intracellular signaling and can be recycled back to phospholipids. Cyclooxygenases –1 and –2 are the targets for NSAIDs and the specific COX-2 inhibitors.

The immediate question was why are there two isoforms. The answer seemed to be that COX-1 is a housekeeping enzyme that produces a low basal amount of prostaglandins for virtuous uses (e.g., protect gastric epithelium), whereas the inducible COX-2 serves to make larger amounts of prostaglandins for nefarious purposes (e.g., inflammation, carcinogenesis). We now know that the situation is more complicated, because mice lacking COX-2 show abnormal renal development (3, 4). Moreover, both traditional NSAIDs and the selective COX-2 antagonists have an unfavorable effect on renal function, demonstrating that COX-2-derived products support normal renal function (9). Likewise, COX-2-derived products are produced by normal vasculature (18), and their inhibition might result in an unfavorable alteration in the hemostatic-thrombosis balance. Both isoforms contribute to reproductive success—COX-2 being required for fertilization and implantation, and COX-1 for parturition—and there is some ability for one isoform to compensate for some of the actions normally supplied by the other (19). But let us return to the question: why is there a need for COX-2, because increasing the level of COX-1 would be a simpler way to regulate the amount of prostaglandin? What were the possible explanations? The two isoforms might use different substrates; COX prefers arachidonic acid but can use other polyunsaturated fatty acids. Indeed, Laneuville et al. (20) showed that COX-2 is more promiscuous in its substrate choices than is COX-1, but both of them clearly prefer arachidonic acid, and the impact of COX-2 using other fatty acids seems to be modest. Another possible reason for two isoforms was that they might yield different products. This is only true to the extent that different substrates are used. When arachidonic acid is acted upon, the product profile from the two isoenzymes is virtually the same and depends on the downstream enzymes within the relevant cell or tissue. Maybe they are found in different cells? No, they often coexist. Are they found in different locations within the cell? No, both are in the endoplasmic reticulum and the nuclear envelope (21).

Where Will the Zebrafish Help with These Questions?

Cardiovascular Actions of COX-2.

One surprise over the past couple of years has been evidence of an unanticipated role for COX-2 in the cardiovascular system. The first observation was that a COX-2-selective inhibitor decreased the synthesis of prostacyclin (18), presumably from the endothelial cells, which was worrisome because this prostaglandin has protective actions in cardiovascular disorders by virtue of its ability to inhibit the activation of platelets and to relax blood vessels. Conversely, the COX-2 inhibitor did not block the platelet synthesis of thromboxane, which has the opposite actions—activation of platelets and constriction of vessels. Thus, the net effect of COX-2 inhibition was to alter the balance toward a prothrombotic state (22). The importance of this shift is not yet clear, but the issue was highlighted when a clinical trial of another COX-2 inhibitor of the same class showed increased cardiovascular mortality in arthritis patients who received the COX-2 inhibitor compared with patients who received a traditional NSAID (reviewed in ref. 9). The zebrafish offers a new tool to define the precise effect of COX-2 (and COX-1) inhibitors on prostaglandin pathways in the vascular system. Grosser et al. (1) defined the patterns of expression in the blood vessels and showed that zebrafish thrombocytes behave much like human platelets with respect to the products generated by COX-1.

Developmental Events Influenced by Prostaglandins.

Given the studies with knockout mice, it could be argued that we don't need another system to study the influence of COX-1 and COX-2 on development. This is where the intrinsic properties of the zebrafish add particular value (see below). Analyses of the mouse have been informative but have shortcomings, at both practical and theoretical levels. The system reported here will allow more detailed mechanistic questions to be addressed. This process has begun, even in this first report. as the authors found that a “knock-down” of COX-1, but not of COX-2, resulted in marked developmental defects that could be rescued by injecting a cDNA encoding the zebrafish COX-1. This system seems to be an excellent experimental one to address whether COX-1 and –2 have inherently different functions in development. For example, will injection of the zCOX-2 cDNA also reverse the developmental defect in the COX-1 knock-down? One would postulate that it would, if the authors are correct on a related point. The developmental block was a surprise, because the knockout of COX-1 in the mouse did not result in such a phenotype. The authors propose that the mouse embryos were spared this fate because they received prostaglandins from the placenta. If this conclusion is correct, then the proper prostaglandin should complement the phenotype irrespective of the cellular or enzymatic origin. This idea suggests an even more straightforward experiment—test whether prostaglandin E₂ (PGE₂) added to the tank water will complement the phenotype in the COX-1 knock-down embryos.

The zebrafish system also could be used to address a related question in prostaglandin signaling: is topography important? Prostaglandins are secreted and typically are thought to act over only a short distance (Fig. 2). That is an autocrine or paracrine form of signaling, rather than an endocrine form of signaling at remote sites. (Autocrine or paracrine signaling is true only of physiological events—pharmacological quantities of prostaglandins clearly result in remote responses). They bind to G protein-coupled receptors on the surface of the target cells and elicit responses. Some prostaglandins also can signal within the cell (intracrine) by virtue of binding to the PPAR family of transcription factors (23). There are experimental results suggesting that signaling that requires COX-2 might depend on the enzyme being in the same cell that generates the downstream response (ref. 24, with comment in ref. 25). The zebrafish embryo presents a ready mechanism to explore these issues, because the cells are accessible from the water, and one could test whether a cell or organ must express either COX-1 or –2 to have a certain signaling response, or whether providing the prostaglandin externally would suffice.

Figure 2.

Prostaglandins can signal by means of multiple mechanisms. Prostaglandins usually are secreted and then bind to G protein-coupled receptors on the target cells. In most cases, it is thought that the target is either the synthesizing cell itself (autocrine signaling) or a neighboring cell (paracrine). In some cases, particularly with pharmacological use of prostaglandins, they circulate in the blood to effect a signal in remote cells (endocrine). Lastly, prostaglandins can bind to nuclear transcription factors to induce gene expression, and this can occur within the same cell (intracrine).

Given the existence of two COX isoforms in humans, a key issue is the extent of functional overlap. Whereas the genomic complexity of zebrafish is significantly less that that of humans, there are supernumerary clusters of some important regulatory genes in zebrafish. For example, invertebrates have one HOX gene cluster; mice and humans each have four, and zebrafish have seven (26, 27). What at first glance seems to be a disadvantage for genetic analysis turns out to be an asset. The duplicated genes in zebrafish have divided both the functions and expression domains of the primordial vertebrate genes. Thus, gene duplication in zebrafish allows geneticists to eliminate a subset of functions and expression domains for a gene family, which yields a fine dissection of pathways while avoiding the pleiotropic effects of a gene knockout in mice. Grosser et al. (28) obtained an early embryonic phenotype by morpholino treatment against zCOX-1 but not zCOX-2. However, as described for the HOX clusters, there might be additional COX isoforms in zebrafish. zCOX-1 is in a syntenic region with NOTCH 1B, which is duplicated in the zebrafish genome (29). If there are additional COX isoforms in zebrafish, their divergent functions could be assessed by the targeting of each isoform and by combinations of multiple isoform targeting and replacements of isoform functions by RNA rescue.

Zebrafish, Cancer, and COXes.

COX-2 (and perhaps COX-1) plays a critical role in many cancers (reviewed in ref. 30). Based on the ability to knockout genes selectively, cancer studies have moved forward rapidly in the mouse model (31). However, many genes that have roles in carcinogenesis also have specific, if sometimes poorly understood, roles in embryogenesis, organogenesis, or physiology that can preclude a straightforward analysis of the effects on cancer. Studies in zebrafish often can circumvent these limitations. For example, the roles of COX-1 and -2 in reproduction in mice (19) may not limit similar studies in zebrafish.

As a specific application, zebrafish might be used to address the issue of how COX-2 promotes carcinogenesis. One mechanism seems to be by preventing apoptosis (32, 33), which allows the propagation of cells with damaged genomes. Whether the antiapoptotic actions derive from a signal generated by products of COX-2 or by removal of a proapoptotic signal is a current topic. There is abundant evidence that unesterified (free) arachidonic acid within cells can signal apoptosis (33, 34). A counter argument has been that mice with a germline APC mutation, which results in intestinal polyposis, are protected when they have been engineered to lack cPLA₂ (35, 36), which is present in zebrafish (37). However, the results leave many questions unresolved, because the effects were only in the small intestine. Futher, other phospholipases may contribute to the cellular level of arachidonic acid. The zebrafish embryo would seem to be an excellent place to test whether the ability of COX-2 to block apoptosis derives from a reduction in the arachidonic acid levels in the cell or the production of a prostaglandin. On a broader scale, zebrafish will allow the roles of COX-1 and –2 and other components of the prostaglandin pathway (e.g., the initiating phospholipases and the prostaglandin receptors) to be scrutinized in an efficient genetic model.