Cloning, Expression, and Functional Characterization of Human Cyclooxygenase-1 Splicing Variants: Evidence for Intron 1 Retention

The commonly used nonsteroidal anti-inflammatory drugs (NSAIDs), such as aspirin, ibuprofen, and naproxen, are known for their effects to relieve pain and inflammation and to reduce fever. This therapeutic benefit derives primarily from the ability to inhibit the activities of certain cyclooxygenase (COX) enzymes that are critically involved in de novo prostaglandin biosynthesis (Vane, 1971). To date, two isoforms of the COX enzyme, COX-1 and COX-2, have been identified (DeWitt and Smith, 1988; Merlie et al., 1988; Hla and Neilson, 1992). These two isozymes have different expression patterns and distinct enzymatic properties. It has been suggested that COX-1, constitutively expressed in many tissues, is responsible for the physiological production of prostanoids, whereas COX-2, inducible under inflammatory conditions, is mainly responsible for the pathological production of prostanoids (O'Banion et al., 1992). Subsequent experiments determined that most of the clinically used COX inhibitors, such as ibuprofen and naproxen, were nonselective (i.e., they blocked both COX-1 and COX-2), albeit with varying potency ratios. Consequently, the major therapeutic limitation of the nonselective COX inhibitors has been their tendency to cause gastrointestinal lesions due to reductions in gut levels of homeostatic prostanoids via COX-1 inhibition. Accordingly, significant effort has been committed to the discovery and development of selective COX-2 inhibitors that would be expected to exhibit anti-inflammatory/analgesic efficacy with substantially improved gastrointestinal safety.

Acetaminophen is a widely used analgesic and antipyretic drug, particularly in the treatment of certain medically compromised or vulnerable populations, such as those with gastrointestinal diseases, children, and the elderly. Devoid of the antiedema effects of the nonselective cyclooxygenase inhibitors (i.e., NSAIDs) and selective COX-2 inhibitors, the mechanism of action of acetaminophen in the relief of pain and fever to a large extent remains unknown. Several lines of evidence have led to the hypothesis that one or more than one additional COX isoform may exist and constitute the therapeutic target of acetaminophen (for review, see Botting, 2000). Indeed, a protein comprising a third cyclooxygenase isozyme has been recently identified in canine cerebral cortex, and it is called COX-3 (Chandrasekharan et al., 2002). This canine COX-3 enzyme is actually a splice variant of the COX-1 gene, arising via the retention of intron 1, introducing an insertion of 90 bp at the 5′ terminus of the COX-1 mRNA. Moreover, several marketed NSAIDs and other analgesic drugs, including acetaminophen, were evaluated for their relative ability to inhibit COX-3 compared with COX-1 or COX-2 when functionally expressed in insect Sf9 cells. However, the existence of an orthologous splicing variant in human tissue has been questioned because, based on a search of the human genomic DNA database, the complete retention of intron 1 of human COX-1 would result in a frame shift that would be predicted to yield a prematurely terminated and COX-inactive protein (Dinchuk et al., 2003; Schwab et al., 2003). In contrast, molecular studies have suggested the existence of mRNA containing intron 1 of COX-1 in human, rat, and mouse tissues (Chandrasekharan et al., 2002; Dinchuk et al., 2003; Kis et al., 2003, 2004; Shaftel et al., 2003). Furthermore, a limited biochemical assessment also suggested the existence of the splicing variant at the protein level in human tissues (Chandrasekharan et al., 2002). In this study, we extensively investigated the existence of COX-1 splicing variants in human tissues at both mRNA and protein levels, precisely identifying the sequence and predicted products of three variant transcripts. Furthermore, we characterized the functional sensitivity of the COX-active variant to numerous marketed analgesic/anti-inflammatory drugs, with direct comparisons with COX-1.

Materials and Methods

Northern Blot Analysis. Northern blot analysis was used to assess the tissue distribution of human COX-1-related mRNAs. An antisense oligonucleotide (oligo), hcx1–12, corresponding to bp 26 to 63 within intron 1 of human COX-1, was used as a probe. The sequence of the antisense oligo was 5′-TGGCATTCAAGGCTCCACCAGGAGGCCAAGAAAATTCC-3′. The oligo probe was labeled at the 5′ end with [γ-³²P]ATP and then purified using a Sepharose spin column. Human multiple tissue Northern (MTN) blots, including human MTN blot, human II MTN blot, human III MTN blot, human brain MTN blot II, human brain MTN blot IV, and human cancer cell line MTN blot were purchased from BD Biosciences Clontech (Palo Alto, CA), and Human Different Types of Tumor mRNA blots were purchased from BioChain Institute, Inc. (Hayward, CA). Each blot was prehybridized with 5 ml of ExpressHyb solution (BD Biosciences Clontech) at 42°C for 60 min and then hybridized in the presence of 2 × 10⁶ cpm/μl of the human COX-1 intron 1 oligonucleotide probe at 42°C for 2 h. The blots were rinsed twice with 2× SSC/0.05% SDS and then washed twice with 0.1× SSC/0.1% SDS solution at 42°C for 2 h. Finally, the blots were exposed to X-ray film at –80°C from 24 to 72 h. Human β-actin was used as a control probe. The same blots were stripped with 0.5% SDS at 90°C for 10 min and then used for hybridization with the control probe at 65°C for 1 h. The blots were then washed twice with 0.1× SSC/0.1% SDS solution at 65°C for 2 h and exposed to X-ray film for about 1 h.

Generation of Polyclonal Antibody. A peptide (Ac-MSRECDPGARWGC-amide) derived from intron 1 of human COX-1 (called anti-human COX-1 intron 1 antibody) was synthesized and used for raising polyclonal antibodies in rabbits. Polyclonal antibody was raised and affinity purified using antigen peptide by BioSource International (Camarillo, CA) and was used for Western blot analysis.

Western Blot Analysis. Premade human multiple tissue total protein blots were purchased from BioChain Institute, Inc. Blots were blocked with 5% dry milk in 0.5% Tween 20, 100 mM NaCl, and 10 mM Tris-HCl, pH 7.4, for 2 h at room temperature and then incubated with a 1:1000 dilution of affinity-purified anti-human COX-1 intron 1 antibody in 5% dry milk/0.5% Tween 20, 100 mM NaCl, and 10 mM Tris-HCl, pH 7.4, at 4°C overnight. A 1:10,000 dilution of secondary goat anti-rabbit IgG conjugated with horseradish peroxidase (Pierce Chemical, Rockford, IL) was applied to the blot for 1 h at room temperature. Finally, the signals were visualized on X-ray film using an ECL Plus kit from GE Healthcare (Little Chalfont, Buckinghamshire, UK). The same blot was stripped and reprobed with a 1:2000 dilution of anti-human COX-1 monoclonal antibody (Sigma-Aldrich, St. Louis, MO) under the same conditions.

Cloning Human Genomic DNA Fragment Including Exon 1-Intron 1-Exon 2 of COX-1. To clone the genomic sequence of exon 1-intron 1-exon 2 of human COX-1, two oligonucleotides were designed for PCR amplification. The forward primer (5′-ATGAGCCGTGAGTGCGACCCCGGT-3′) was designed based on exon 1 and part of the intron 1 sequence of human COX-1, and the reverse primer (5′-CTACCTGGCGTGGGCGCCCCTGGGT-3′) was based on exon 2. PCR was performed using an Advantage-GC Genomic PCR kit and human genomic DNA purchased from BD Biosciences Clontech. The PCR product (∼190-bp fragment) was subcloned into pPCRScript cloning vector (Stratagene, La Jolla, CA) and then sequenced.

Cloning of cDNA Encoding the N Terminus of Human COX-1 Variants. To identify human COX-1 variants, a cDNA fragment was amplified from cDNA libraries of human brain and stomach (BD Biosciences Clontech) with the forward primer being the same as that used for amplifying the genomic DNA sequence, and the reverse primer (5′-TATGAACTTCCTCCTGAGCAGGAA-3′) corresponding to a position located at approximately bp 600 of human COX-1. The PCR was performed with Marathon-Ready human brain cDNA (BD Biosciences Clontech). After PCR, the approximately 0.6-kb PCR fragment was purified, polished, and subcloned into pPCRScript. Twenty to 25 independent clones were picked and subjected to double-stranded DNA sequencing analysis.

Expression of Human COX-1 and COX-1b₂ in Insect Sf9 Cells. For structural and functional studies, full-length human COX-1 and COX-1b₂ cDNA were assembled and subcloned into pFastBac1 (Invitrogen, Carlsbad, CA), a baculovirus expression donor vector. The recombinant bacmids were made in DH10Bac Escherichia coli cells after transposition of the construct. A high titer (1 × 10⁹ plaque-forming units/ml) of recombinant baculovirus was obtained after transfection of recombinant bacmid into insect Sf9 cells followed by two subsequent rounds of amplification. The expressed COX-1 and COX-1b₂ were observed in SDS-PAGE gels stained by Coomassie Blue R250, and their identities were further confirmed by Western blot using the anti-COX-1 and anti-human COX-1 intron 1 antibodies.

COX Enzymatic Activity Assay. Microsomal protein fractions were prepared from Sf9 cells infected with recombinant virus for human COX-1 or COX-1b₂. Microsomal membrane protein (25 μg) was incubated with the indicated concentration of arachidonic acid in a volume of 150 μl of buffer containing 0.1 M Tris-HCl, pH 8.0, 5 mM EDTA, 2 mM phenol, and 1.5 μM hematin. After incubation at room temperature, the reaction was terminated by the addition of 40 μl of 1 N HCl. Samples were then neutralized with 40 μl of 1 N NaOH. The product, PGF_2α, was quantitatively measured using a PGF_2α immunoassay kit (Assay Designs, Inc., Ann Arbor, MI), as directed by the manufacturer.

Results

Northern Blot Analysis of the Expression of COX-1 Splicing Variants in Human Tissues. To demonstrate the existence of intron 1 retained in the human COX-1 variant transcripts, we first performed Northern blot analysis with an antisense oligonucleotide probe derived from intron 1 of COX-1. A Blast search of GenBank human genomic databases confirmed that, throughout the entire human genome, no significant homologous sequence existed, except intron 1 of COX-1. Therefore, under the high-stringency hybridization conditions used here, the probe would specifically hybridize with any transcripts containing intron 1 of the human COX-1 gene. It is well documented that there are at least two types of COX-1 transcript, a major 2.8 kb and a minor 5.2 kb, which is due to alternative polyadenylation of the COX-1 gene (Hla, 1996). Northern blot analysis with intron 1-specific probe, as shown in Fig. 1, also reveals that there are at least two types of COX-1 transcripts containing intron 1. The major transcript is slightly larger than the 4.5-kb molecular standard, which is very similar in size to the 5.2-kb minor transcript of COX-1. The 2.8-kb-like transcript was also observed, but it is not a major transcript. The large transcript containing COX-1 intron 1 is most abundant in human stomach, followed by skeletal muscle, heart, placenta, liver, pancreas, spleen, testis, adrenal gland, and kidney. It is also expressed at a relatively low level in brain, lung, prostate, small intestine, leukocyte, thyroid, spinal cord, lymph node, and trachea. The transcript was not detectable in thymus, ovary, colon, or bone marrow. The splicing variant was also detected in most of the brain subregions that we examined. Expression levels in the central nervous system were highest in the cerebral cortex and amygdala, followed in descending order by caudate nucleus, hippocampus, putamen, occipital pole, frontal lobe, temporal lobe, thalamus, corpus callosum, medulla, and spinal cord. These results are consistent with the finding of Chandrasekharan et al. (2002) and suggest that the intron 1-retained COX-1 gene is indeed expressed in a variety of human tissues. Here, we name the COX-1 splicing variants containing intron 1 using the COX-1b rather than COX-3 nomenclature (reflecting their derivation from the COX-1 rather than a separate gene).

Western Blot Analysis of the Human COX-1 Splicing Variant with Intron 1 Retention. To further determine whether the intron 1-retained COX-1 transcript is indeed translated in human tissues, a polyclonal antibody was raised and affinity purified with a peptide derived from the intron 1 sequence. The anti-intron1 antibody was specific, because it was able to differentiate human COX-1 and its intron 1-retained splicing variant; moreover, its immunoreactive signal was blocked by preincubation of the antibody with antigen peptide (Fig. 2). The same set of antibodies was used for tissue distribution analysis via Western blot. Figure 3, A and B, illustrate the analysis of the premade human multiple tissue total protein blots. With anti-human COX-1 intron 1 antibody, two major and specific immunoreactive proteins, ∼75 and ∼55 kDa, were identified in human total protein lysates (Fig. 3A). The 75-kDa protein, presumably representing human COX-1b, is expressed at a relatively high level in heart, brain, kidney, liver, skeletal muscle, stomach, and small intestine as well as in various subregions of human brain. No 75-kDa protein was detected in lung, pancreas, spleen, colon, rectum, uterus, prostate, testes, placenta, or spinal cord. In contrast, the 55-kDa protein, the size of which is similar to another COX-1 splicing variant (PCOX-1a) described by Chandrasekharan et al. (2002), was expressed more broadly than 75-kDa protein in the human tissues that we tested. A couple of smaller uncharacterized proteins were also identified (about 15 kDa). However, no prematurely terminated 8.7-kDa peptide (82 residues; Fig. 5A) was detected when selected samples were subjected to 18% SDS-PAGE. The specificity of COX immunoreactivity was further confirmed by probing the same blot under the same conditions with the antibody preabsorbed by 20-fold (w/w) antigen peptide (data not shown). As an internal control, the same blot was stripped and reprobed with anti-human COX-1 antibody (Fig. 3B). The expression pattern of the 75-kDa protein is consistent with that of the major intron 1-retained COX-1 mRNA (Fig. 1), with the exception of placenta, in which there was a high mRNA level but no 75-kDa protein was detected. The size of the large protein (i.e., 75 kDa on 4–20% SDS-PAGE) detected here by anti-human COX-1 intron 1 antibody is different from that of the previous report, in which it was approximately 65 kDa and smaller than COX-1 (Chandrasekharan et al., 2002). The calculated molecular mass of COX-1b is about 6 kDa larger than COX-1, since it bears about 55 extra residues at its amino terminus (31 amino acids encoding by intron 1 and 24 amino acids of uncleaved signal peptide). Our Western blot result clearly demonstrates the intron 1-retained splicing variant to be larger than that of COX-1 (i.e., ∼75 versus ∼70 kDa) on a 4 to 20% SDS-PAGE. In addition, our results suggest that human COX-1 splicing variant is expressed broadly in human tissues.

Retention of COX-1 Intron 1 in Human Tumor Tissues and Cancer Cell Lines. It is well documented that both COX-1 and COX-2 are associated with carcinogenesis, and their overexpression is believed to be involved in the development of tumors in several tissues (Oshima and Taketo, 2002; Zha et al., 2004). To determine whether intron 1-retained COX-1 splicing variant is up-regulated under certain pathological conditions, such as in a tumor, three premade human tumor tissue mRNA blots were chosen for use in Northern blot analysis with intron 1-specific probes. As shown in Fig. 4A, when normalized to β-actin expression, the expression levels of intron 1-retained COX-1 mRNA in most of human tumor tissues, including rectum, kidney, lung, ovary, and uterus, were not significantly changed. In contrast, the mRNA level of intron 1-retained COX-1 in colon tumor tissue was elevated compared with that of nondiseased tissues. Further experiments are required to confirm this finding. The expression level of intron 1-retained COX-1 mRNA in human carcinoma cell lines was also tested under the same conditions. The expression levels (normalized to β-actin mRNA level) in human carcinoma cell lines were highest in leukemia chronic myelogenous K-562, followed by HeLa cell S3, whereas the expression level in malignant melanoma G-361, colorectal adenocarcinoma SW480, lymphoma Burkitt's Raji, lung carcinoma A549, leukemia promyelocytic HL-60, and leukemia lymphoblastic MOLT-4 were relatively low.

To further determine whether COX-1 variant proteins were also overexpressed in selected human tumor tissues, Western blot analysis of total protein from both normal and tumor brain, breast, colon, and stomach tissues was also performed with intron 1-specific antibody. However, no significant changes in protein level (for both 75- and 55-kDa proteins) were observed when comparing several human tumor tissues with normal tissues (data not shown).

Human COX-1 Gene and Cloning of cDNA Encoding the N Terminus of Human COX-1b. The human COX-1 gene, localized on chromosome 9, comprises 11 exons and 10 introns that span approximately 22 kb. COX-1 is encoded by exon 1 through exon 11, whereas the novel splicing variant COX-1b, called COX-3 by Chandrasekharan et al. (2002), is encoded by the same number of exons plus a retained intron 1 between exon 1 and exon 2. However, if the entire intron 1 is retained in COX-1b, then the sequence obtained from the human genomic database would be out of reading frame. Therefore, we first sought to confirm the published genomic sequence in this region. To accomplish this, two oligonucleotides were designed for amplifying the genomic sequence of exon 1-intron 1-exon 2 of COX-1 from human genomic DNA. The forward primer was designed based on exon 1 and part of the intron 1 sequence, and the reverse primer was based on exon 2. The PCR product (∼190-bp fragment), including exon 1-intron 1-exon 2, was subcloned into pPCRScript cloning vector and sequenced. Sequence analysis indicated that the sequence of intron 1 was consistent with that published in the human genome database. Accordingly, if the entire intron 1 is retained in COX-1b, then its length (i.e., 94 bp) would lead to a shift in the reading frame and hence a prematurely terminated and substantially shortened COX-1b protein. Therefore, the expression of intron 1-retained COX-1 and the size of COX-1b demonstrated here by Western blot analysis must arise via either retention of the entire intron 1 followed by the removal of some nucleotides at the mRNA level (i.e., RNA editing) or by partial retention of intron 1. In both cases, an open reading frame leading to the production of an active COX enzyme would be preserved.

We further explored these explanations by investigating more precisely the sequence(s) of COX-1b at the cDNA level. Complimentary DNA fragments were amplified from human brain and stomach cDNA libraries, with the forward primer being the same as that used for amplifying the genomic DNA sequence and the reverse primer corresponding to bp position 600 of human COX-1. After PCR, the 0.6-kb PCR fragment was subcloned and sequenced. Double-stranded DNA sequencing analysis revealed the existence of three distinct COX-1b splicing variants as shown in Fig. 5. The first and major type (86 of 93 total independent clones), named COX-1b₁, is one variant in which the entire 94 bp of intron 1 is retained, leading to a shift in the reading frame and the introduction of a stop codon approximately 249 bp down-stream. Therefore, this variant type encodes a very short peptide that is most likely a COX-inactive protein. The second type (Fig. 5B), named COX-1b₂ (5 of 93 total clones), retains almost the entire intron 1, but it is missing a guanidine at position 72 (from start codon), leading to a short, self-rectifying shift in the reading frame, which encodes a full-length and likely COX-active protein. Like the second type of splicing variant, the third type (Fig. 5C), named COX-1b₃ (2 of 93 total clones), also retains almost the entire intron 1, but it is missing a cytosine at bp position 50, leading to another short, self-rectifying shift in the reading frame, which encodes a slightly different but also full-length and likely COX-active protein. The peptides encoded by intron 1-retained COX-1 variants are significantly different between human and canine versions, exhibiting less than 25% identity (Fig. 5D).

COX-1b₂ Enzymatic Activity. Based on its differential sensitivity to inhibition by a selection of NSAIDs or acetaminophen, the novel intron 1-retained splicing variant of canine COX-1 originally described by Chandrasekharan et al. (2002) was called COX-3. To determine whether a human splicing variant of COX-1 also exhibits such differential sensitivity, we expressed the putatively active human COX-1 variant, COX-1b₂, in insect Sf9 cells and characterized its enzymatic properties compared with that of human COX-1. As shown in Fig. 6A, arachidonic acid was converted to PGF_2α by both human COX-1 and COX-1b₂ in a concentration-dependent manner, with K_m values of 1.3 and 0.54 μM, respectively. The V_max value of COX-1 (6.23 pmol/mg/min) was significantly higher than that of COX-1b₂ (3.07 pmol/mg/min), indicating that COX-1 may be more efficient than COX-1b₂ in converting arachidonic acid to PGF_2α under the experimental conditions described.

The sensitivities of both recombinant COX-1 and COX-1b₂ to the inhibition by NSAIDs were also compared under identical conditions. Twelve NSAIDs and some common analgesic/antipyretic drugs, including acetaminophen, phenacetin, and dipyrone, were tested for their inhibitory activities at human COX-1 and COX-1b₂. These compounds inhibited the activities of COX-1 and COX-1b₂ at IC₅₀ values ranging from less than 20 nM to more than 10 mM. In contrast to canine COX-3 (Chandrasekharan et al., 2002), neither acetaminophen and phenacetin nor dipyrone exhibited significantly different potency for inhibiting human COX-1 versus COX-1b₂ (Fig. 6B; Table 1).

Discussion

Acetaminophen serves an important role in the analgesic and antipyretic pharmacopeias, especially in the treatment of patients who cannot tolerate NSAIDs. However, thus far, the lack of a well defined mechanism of action has substantially impeded efforts to generate improved analogs, whereas the existence of one or more than one additional COX isoform has been suspected for many years. The discovery of a canine COX-1 splice variant via the retention of intron 1, which was shown to be expressed in brain and sensitive to inhibition by acetaminophen has been called COX-3 (Chandrasekharan et al., 2002). However, the existence of an orthologous splicing variant in human tissue has been questioned due to a frame shift and premature termination that occurs through the retention of the entire intron 1. Here, we confirm the existence of multiple molecular forms of a COX-1 splicing variant containing intron 1 and demonstrate the expression of one or more of these variants in several human tissues at both mRNA and protein levels. We also show that one of these COX-active variants, COX-1b₂, shows no differential sensitivity compared with COX-1 to inhibition by acetaminophen or several marketed NSAIDs.

Obviously, retention of the entire 94-bp intron 1 of human COX-1 will lead to a shift in the COX-1 translational open reading frame, thereby encoding a prematurely truncated and most likely COX-inactive peptide. Unlike COX-1b in rat tissue (Snipes et al., 2005), no prematurely truncated 8.7-kDa peptide was detected in human tissues. However, the present Western blot results clearly indicate the existence in human tissues of one or more proteins containing intron 1 of COX-1. In other words, some transcripts of the COX-1 splicing variant are apparently capable of undergoing a reading frame self-correction, thereby resulting in the translation of a full-length, COX-active protein. Furthermore, using gradient SDS-PAGE, the size of the immunoreactive protein was shown to be slightly larger than COX-1, which is consistent with containing 55 extra residues by retention intron 1 and an uncleaved signal peptide. This result also argues that the retention of intron 1 of COX-1 is neither simply a splicing error in human tissue nor a peculiarity of canine genetics.

There are at least two possible explanations that may account for the observed restoration of the reading frame shift. First, single-nucleotide polymorphisms (SNPs) are common. In fact, many human COX-1 SNPs (15 SNPs in coding regions and five in intronic regions within 60 bp of an exon) have been identified, and among them, at least seven cases result in amino acid changes (Ulrich et al., 2002). All of these SNPs are rare and account for less than 4% of the population. In addition, until now, no SNP has been identified in the intron 1 region. However, this finding was based on screening relatively small populations of human samples. Whether any of the human COX-1 variants described here are derived from SNPs remains to be determined by screening relatively large populations of human samples. The second possible explanation is RNA editing, involving the deletion of a single nucleotide. RNA editing is a powerful mechanism to expand genome capacity, and it is widespread in eukaryotes. RNA editing includes nucleotide substitution, insertion, and deletion after transcription (Gott, 2003). By using intron 1-specific primers, we were able to identify three types of cDNA fragments encoding the amino terminus of COX-1 with intron 1 retained. In addition to the truncated form, two minor types of cDNA fragments were identified, containing almost the entire intron 1 but missing either a guanidine at position 72 (COX-1b₂) or a cytosine at position 50 (COX-1b₃). Both of these variations comprise a short and self-rectifying shift in the reading frame, which leads to the translation of full-length COX-active proteins, at least in COX-1b₂. Regardless, we do not believe that either of these variant clones was detected due to PCR or cloning errors, because they were independently amplified and subcloned from four independent experiments, comprising approximately 100 clones (20–25 clones were picked and sequenced for each amplification). In fact, the same sequence of COX-1b₂ was previously identified by Wang et al. (1993).

COX-1 is an integral membrane glycoprotein (Otto et al., 1993; Spencer et al., 1999). Presumably, the signal peptide at its amino terminus plays a key role in proper insertion of COX-1 into the membrane and also in glycosylation. Insertion of a peptide encoded by intron 1 upstream of the signal peptide, theoretically disabling the function of the signal peptide, may change the topology of the protein. This could potentially result in a different glycosylation pattern as well as a change in the enzyme activity and selectivity. Unlike canine COX-3 (Chandrasekharan et al., 2002), human COX-1b₂ is apparently not heavily glycosylated (data not shown). This may partially explain why the COX-1b₂ has a lower V_max value than COX-1 (Fig. 5). Although both COX-1 and COX-1b₂ bear the same active sites or catalytic domains, differences in enzyme topology and glycosylation sites may alter enzyme activity and selectivity; however, side-by-side comparison of enzyme properties between COX-1 and COX-1b₂ from the same species (i.e., human) shows no significant differences between them (Fig. 4), suggesting that human COX-1b₂ is not likely to be the therapeutic target of acetaminophen.

In summary, the existence of intron 1-retained COX-1 variants in human tissues was confirmed at both mRNA and protein levels, and their molecular identity was revealed. However, it remains to be clarified in further investigations what the physiological roles of COX-1b₂ or COX-1b₃ are or for which, if any, drugs they serve as the therapeutic target. In addition, we have to keep in mind that the majority of splicing variant COX-1b₁ only encodes a short peptide, which is unlikely to possess any COX activity, as previously demonstrated with its rat counterpart (Snipes et al., 2005); therefore, its physiological significance remains to be revealed.