Vanderbilt Institute of Chemical Biology



Discovery at the VICB







The Complex Interplay of Substrates with Cyclooxygenase-2



By: Carol A. Rouzer, VICB Communications
Published:  September 22, 2015



Kinetics studies and mathematical modeling provide critical insights into the mechanism of an enzyme that modulates multiple processes in health and disease.


The cyclooxygenase enzymes (COX-1 and COX-2) catalyze the first two biosynthetic reactions leading to the prostaglandins (PGs), a class of potent lipid signaling molecules. These enzymes play an important role in both health and disease as PGs modulate a wide range of processes, including fever, pain, inflammation, uterine contractions, renal function, and hemostasis. The cyclooxygenases are the primary site of action of the widely used nonsteroidal anti-inflammatory drugs, such as aspirin, ibuprofen, and naproxen. The precursor for PG biosynthesis is the polyunsaturated fatty acid, arachidonic acid (AA, Figure 1A), and both cyclooxygenase isoforms utilize this substrate with similar efficiencies. However, the larger active site of the COX-2 isoform allows it to utilize bulkier ester and amide derivatives of AA that are poor substrates for COX-1. Among the COX-2-selective substrates is 2-arachidonoylglycerol (2-AG, Figure 1B), a lipid signaling molecule notable for its agonist activity at the cannabinoid receptors. The action of COX-2 on 2-AG produces glyceryl esters of PGs (PG-Gs). In vitro kinetics studies suggest that COX-2 metabolizes AA and 2-AG with similar efficiencies when present alone; however, in intact cells, the ratio of PG-G to PG formation is much lower than would be expected based on the relative availability of 2-AG and AA. This led Vanderbilt Institute of Chemical Biology members Larry Marnett and Carlos Lopez to investigate the interaction of AA and 2-AG with COX-2 in detail [M. M. Mitchener, et al., (2015) Proc. Natl. Acad. Sci. U.S.A., published online September 21, DOI: 10.1073/pnas.1570307112].



Figure 1.
The cyclooxygenase reaction. (A) Both COX-1 and COX-2 can utilize arachidonic acid (AA) to form prostaglandin G2 (PGG2) in a reaction that occurs at the cyclooxygenase active site. The enzymes then reduce PGG2 to prostaglandin H2 (PGH2) at their peroxidase active site. (B) 2-Arachidonoylglycerol (2-AG) is an efficient substrate only for COX-2. The products formed from 2-AG by COX-2’s cyclooxygenase and peroxidase active sites are the glyceryl esters of PGG2 (PGG2-G) and PGH2 (PGH2-G), respectively. Both PGH2 and PGH2-G are subject to further metabolism, leading to the biologically active prostaglandins (PGs) and prostaglandin-glyceryl esters (PG-Gs).


The Marnett lab investigators first showed that enrichment of membrane phospholipids with AA led to higher levels of both AA and 2-AG in mouse macrophages exposed to an inflammatory stimulus that triggers PG and PG-G formation. Despite these higher levels of substrate, however, the cells produced no additional PGs, and their biosynthesis of PG-Gs actually decreased. Furthermore, treatment of the macrophages with an inhibitor of cytosolic phospholipase A2 to prevent release of AA for PG biosynthesis resulted in markedly reduced AA levels while 2-AG levels were unchanged. These cells produced very low amounts of PGs, as expected, but biosynthesis of PG-Gs increased. These results demonstrated an inverse correlation between levels of AA and PG-G biosynthesis, leading the investigators to hypothesize that AA suppresses 2-AG utilization by COX-2.


Structurally, COX-2 is a homodimer, but prior studies had shown that only one of the two subunits of the enzyme contains the required heme cofactor. Hence only one subunit is catalytically active. Binding of nonsubstrate fatty acids and enzyme inhibitors at the second subunit modulates COX-2 activity, suggesting that it serves a regulatory, allosteric role. The Marnett investigators began their research with careful kinetics studies using each substrate individually. These experiments yielded key kinetic parameters for each substrate and demonstrated that at high concentrations, 2-AG binds to COX-2’s allosteric subunit, forming an inactive complex. Thus, 2-AG exhibited substrate inhibition that was not observed with AA. The researchers then went on to study the kinetics of PG and PG-G formation when both substrates were present in the reaction mixture simultaneously. They discovered that each substrate suppressed utilization of the other; however, the suppression of 2-AG utilization by AA was much greater than the suppression of AA utilization by 2-AG.


The simplest mechanism by which AA can suppress 2-AG utilization (or vice versa) is by simple competition of the two substrates for COX-2’s active site. The Marnett investigators tested the hypothesis that simple competition could explain their observations by simulating the behavior of an enzyme that exhibits competition between the two substrates for the enzyme’s active site and substrate inhibition in the case of 2-AG (Figure 2). For these studies, they used the kinetic parameters measured for the individual substrates in their initial experiments. As expected, this model predicted that the presence of each substrate inhibits utilization of the other; however, it overestimated the suppression of AA utilization by 2-AG and underestimated the suppression of 2-AG utilization by AA when its predictions were compared to the actual experimental data.

Figure 2. 
Simple competitive model to explain the interaction of AA and 2-AG with COX-2. The COX-2 enzyme comprises a catalytic site (green, left) and an allosteric site (red, right). In this model, AA and 2-AG compete for the catalytic site. Binding of either substrate to this site leads to formation of its respective product and inhibits the formation of products from the other substrate. In addition, kinetics studies showed that 2-AG binds to the allosteric site of COX-2, forming a complex that does not form product. Thus, at high 2-AG concentrations, PG-G formation is inhibited.


When the simple competitive model did not explain the simultaneous interaction of COX-2 with its two substrates, the Marnett lab turned to a more complex classical model of enzyme inhibition. This model took into account the presence of COX-2’s allosteric subunit by hypothesizing that binding of AA to the allosteric site prevents 2-AG utilization by the enzyme. Depending on the effect of the presence of 2-AG in the catalytic site on the affinity of AA for the allosteric site, this model is called uncompetitive, noncompetitive, or mixed inhibition (Figure 3). All three types of inhibition were modeled using the experimental parameters measured for each substrate in the initial kinetics experiments. Of the three options, the uncompetitive model most closely explained the experimental data. However, all three models assume that the complex containing AA in the allosteric site is completely inactive with regard to 2-AG utilization. This assumption predicts that, at very high concentrations of AA, no PG-Gs can be formed from 2-AG. This was not the case, however, in that experimental data showed residual PG-G formation in the presence of high AA concentrations. In addition, this model did not fully address the effects of 2-AG on AA utilization.


Figure 3. Model for uncompetitive/noncompetitive/mixed inhibition of 2-AG oxygenation by AA. In addition to competing for the active site and substrate inhibition in the case of 2-AG, this model proposes that AA binds to the allosteric site of COX-2, forming a complex that cannot utilize 2-AG. If AA binding to the allosteric site only occurs when substrate is in the active site, the inhibition is uncompetitive. If binding of AA to the allosteric site occurs equally whether or not 2-AG is in the active site, the inhibition is noncompetitive. If binding of AA to the allosteric site occurs with different affinities depending on whether or not 2-AG is in the active site, the inhibition is mixed.



These initial attempts to explain COX-2’s behavior in the presence of both substrates failed to explain all aspects of the experimental data, leading the investigators to conclude that a much more complex model was required. Thus, they hypothesized that both substrates are capable of binding at either the catalytic or the allosteric site of COX-2 and that the presence of a substrate in the allosteric site will modulate the activity of the catalytic site. That modulation could be to increase or decrease enzymatic activity. To determine if this COX-2 reaction model (CORM, Figure 4) could explain the experimental data, the Lopez laboratory applied a combination of rule-based mathematical modeling and Bayesian statistics to assign the most probable kinetic parameters to each of the interactions between AA and 2-AG with either of the COX-2 subunits. Wherever possible, experimental kinetic parameters were used to minimize the number of values to be determined mathematically. The results demonstrated that, indeed, CORM was more consistent with the experimental data than any of the simpler models evaluated.



Figure 4.
  COX-2 reaction model (CORM) for the interaction of AA and 2-AG with COX-2. The model proposes that both AA and 2-AG can bind to either COX-2 subunit, and that binding of a substrate in the allosteric site alters the activity of the catalytic site. A Bayesian statistical approach was used to determine the most likely kinetic parameters for each step in the model (shown in blue text) using experimental parameters (shown in red text) when available. Figure reproduced with permission from M. M. Mitchener, et al., (2015) Proc. Natl. Acad. Sci. U.S.A., published online September 21, DOI: 10.1073/pnas.1570307112 . Copyright 2015, M. M. Mitchener, et al.



An evaluation of the kinetic parameters of CORM provides some interesting insights regarding the interaction of COX-2 with its substrates. For example, the dissociation constants suggest that binding of a substrate to the allosteric subunit is only favored after binding to the catalytic subunit has occurred. In addition, at most substrate combinations tested, complexes containing two different substrates are favored over those containing only one or two identical substrate molecules. Binding of AA to the allosteric subunit has little effect on AA utilization in the catalytic subunit, but it substantially decreases 2-AG utilization. In contrast, binding of 2-AG in the allosteric subunit stimulates the utilization of AA while inhibiting utilization of 2-AG. These results help to explain how AA markedly suppresses PG-G formation from 2-AG while 2-AG has only modest effects on PG formation from AA.


CORM provides some important insights into the mechanism of COX-2 activity that could not have been discovered from kinetics studies using individual substrates. The model demonstrates that COX-2 is subject to allosteric modulation by physiological substrates, a finding that may help to explain the poor cellular production of PG-Gs by cells responding to stimuli that cause substantial AA release. In addition, the discovery that substrates bind to the allosteric site of the enzyme is important to fully understand how inhibitors modulate COX-2 activity as at least some of these molecules are believed to act by binding at the allosteric site of the enzyme. Finally, CORM opens the possibility that there are other, naturally occurring modulators of COX-2 activity that act through allosteric site binding. Identification of these modulators may provide key insights into the function of this important enzyme.


View PNAS article: "Competition and allostery govern substrate selectivity of cyclooxygenase-2"










The Vanderbilt Institute of Chemical Biology, 896 Preston Building, Nashville, TN 37232-6304, phone 866.303 VICB (8422), fax 615 936 3884
Vanderbilt University is committed to principles of equal opportunity and affirmative action. Copyright © 2014 by Vanderbilt University Medical Center