Detailed studies by Aldridge reported in 1953 categorized serum esterases into those inhibited by diethyl p-nitrophenyl phosphate (E600 or PO) as B-esterases and those not inhibited by PO as A-esterases since they preferred the substrate p-nitrophenyl acetate to the longer chain esters [1]. The B-type esterases were inhibited by 10⁻⁷ to 10⁻⁸ M E600 and hydrolyzed the p-nitrophenyl butyrate at the same or higher rate than p-nitrophenyl acetate. In a second publication, Aldridge noted that the rabbit had the highest serum A-esterase of the species examined and the ratio of the serum to liver activity was very high while the opposite was true in rat [6]. The high activity of the rabbit plasma A-esterase facilitated the development of purification protocols for the human serum A-esterase as well as the cloning of both the rabbit and human cDNAs discussed below. Aldridge’s studies also indicated that PO and p-nitrophenyl acetate were hydrolyzed by the same enzyme. Later, there had been some controversy as to whether phenyl acetate and PO were hydrolyzed by one or two enzymes resulting in a reclassification of PON1 [3]. It has since been clearly shown that both substrates are hydrolyzed by PON1 as originally proposed by Aldridge [7, 8].

Studies on human A-esterase in the 1960s and 1970s revealed that plasma paraoxonase (POase) activity showed a large inter-individual variability in activity and that the activity was polymorphically distributed in human populations with differences among populations in the frequency of the “low activity allele” vs. the “high activity allele” (reviewed in [9]). In Northern European populations, approximately 50% of individuals are homozygous for the low activity allele, while populations of African and Asian origin had a higher frequency of individuals homozygous for the high activity allele with other populations having even higher frequencies of the high activity allele (reviewed in [9, 10]). The early studies made use of a single substrate, PO and presented data as histograms of activity (reviewed in [9]). In the early years, many different assays were developed for characterizing the A-esterase activity in individuals. These assays used variable conditions of pH, salt and inhibitor (EDTA) to characterize the A-esterase activity (reviewed in [11]). Based on the high variability of serum POase activity among individuals, it was proposed that individuals with high activity levels would be resistant to exposures of PS/PO. The oxon form of the insecticide is also included in exposures as most, if not all, exposures contain a variable percentage of the highly toxic oxon [12–14]. As noted below, human plasma POase does not appear to provide protection against PS/PO exposures as the catalytic efficiency of PO hydrolysis is too low.

Efforts aimed at purifying PON1 first revealed that gel filtration chromatography resolved two peaks of activity, a higher molecular weight peak and a lower molecular weight peak. The low molecular weight activity peak co-fractionated with albumin [11]. The availability of plasma from an analbuminemic individual provided solid evidence that the lower molecular weight activity was indeed associated with albumin. The activity associated with the albumin peak was completely absent in the plasma from the analbuminemic individual. Resolution of the two activity peaks allowed for the characterization of the pH optimum of each peak. The albumin-associated activity was active at only high pH values whereas the POase activity fractionating at higher molecular weight retained significant activity at pH 8.5. This study provided three key observations that facilitated purification and ultimately cloning of rabbit and human cDNAs; 1) the POase activity of the albumin peak had little activity below pH 8.5; 2) the POase activity associated with albumin was resistant to inhibition by EDTA; and 3) the albumin peak did not hydrolyze phenyl acetate. Unfortunately, some studies are still carried out that measure PON1 activity at high pH values. In an individual with low PON1 levels, at the high pH value, albumin hydrolyzes more PO than their PON1 [11].

Serum rabbit PON1 is significantly more stable than human PON1 and was purified and sequenced to allow the design of probes for isolating and sequencing rabbit PON1 cDNA from a rabbit liver cDNA library [15]. The property of being capable of activity staining rabbit PON1 following SDS gel electrophoresis provided assurance that the protein sequenced was indeed PON1 [15]. Probe design based on the sequence of rabbit PON1 allowed for the isolation of the rabbit PON1 cDNA which in turn allowed for the isolation and characterization of human PON1 cDNA [16]. The human cDNA clones revealed the two common coding polymorphisms occurring in human PON1 (L55M and Q192R). The latter was subsequently shown to determine the catalytic efficiency of hydrolysis of PO [17, 18]. The Q192R polymorphism did not significantly affect the catalytic efficiency for hydrolysis of phenyl acetate (arylesterase activity – AREase) or diazoxon (DZO) but did affect the catalytic efficiency of PO, chlorpyrifos oxon (CPO) [19] and the nerve agents sarin and soman [20]. In addition to revealing the two common coding region polymorphisms, sequencing revealed single nucleotide variants (SNVs) in the 5′ and 3′ UTRs. Examination of the effects of the 5′ SNVs on expression of PON1 revealed that the C-108T 5′ UTR promoter region polymorphism in an Sp1 transcription factor binding site had a significant effect on PON1 expression with the C-108 allele expressing on average twice the level of PON1 compared with the T-108 allele [21–23]. An interesting report noted that a 3′ UTR SNV (CT at rs3735590) affected binding of a miRNA (miR-616) to the 3′ UTR of PON1 mRNA with the T allele having lower binding affinity, causing higher levels of expression of PON1 and reducing the risk of ischemic stroke and carotid atherosclerosis in individuals with the CT or TT genotype [24]. The effects of SNVs in the 3′UTR regions on gene expression plus the competing/stabilizing effects of RNA binding proteins (e.g. [25]) warrant further studies.

Sequencing studies in which the PON1 genes of 47 individuals (24 African-Americans and 23 Europeans) revealed 108 new PON1 polymorphisms, including 8 new promoter region SNVs, 9 additional 3′ SNVs and a new coding region SNV (W194X) [26]. Discovery of the latter SNV prompted the sequencing of PON1 genes where we had observed a discrepancy between the characterization of DNA SNVs and the determination of the functional PON1 status (Q192R polymorphism and PON1 activity levels – see below). Sequencing of the PON1 genes from individuals discrepant for the functional analyses and SNV analysis of the Q192R polymorphism revealed a P90L SNV, an Asp124missplice mutation and a partial deletion of the PON1 allele bearing the Q192 genotype [27]. Recently, a PON1 V109I SNV has been associated with ischemic stroke in African-Americans [28].

While the early measurements of variability in inter-individual activity levels of human PON1 suggested that high levels would protect against exposure to PS, it took the development of an animal model system to understand the ability of PON1 to protect against specific OP exposures. The earliest direct test of PON1 protecting against OP exposure was carried out by Main in 1956 [29] where he injected partially purified rabbit PON1 intravenously into rats raised the plasma activity of the A-esterase and decreased the toxicity of PO. Experiments with rats [30] and mice [31, 32] extended Main’s initial observations to include both PO and CPO. CPO was a much better human PON1 substrate than PO. Taken together, these experiments clearly demonstrated that high levels of injected rabbit PON1 provided some protection against PO, but significantly better protection against CPO. The question of the consequences of low PON1 levels was addressed with PON1 knockout mice (PON1^−/−) mice generated by Shih, Lusis and colleagues at UCLA [33]. The PON1^−/− mice were dramatically more sensitive to CPO and DZO than wild type mice, but surprisingly did not exhibit increased sensitivity to PO compared with PON1^+/+ mice [19]. The PON1^−/− mice also provided a model system for testing under physiological conditions the efficacy of purified [19] and recombinant engineered human PON1 [8] to protect against OP exposure when injected into the PON1^−/− mice. Experiments with the two purified human PON1₁₉₂ alloforms demonstrated that protection against specific OPs was determined by the catalytic efficiency with which PON1 hydrolyzed the specific OP. Again, these experiments demonstrated that human PON1 was not effective in protecting against PO exposures. The PON1_R192 alloform protected better than the PON1_Q192 alloform against CPO exposures while either alloform protected equally well against DZO exposure [19].

Since the PON1^−/− mice have no measurable activity against DZO, it was straightforward to follow the half-life of injected engineered recombinant human PON1. Experiments with PON1_K192 showed that this engineered PON1 variant provided good protection against DZO exposure when injected pre- or post-exposure and was able to provide protection for at least 48 h post-injection [8]. The R/Q192K substitution was based on the high rate of CPO hydrolysis by rabbit PON1_K192 [34].

Examination of the developmental time course of appearance of PON1 in newborn humans showed that at birth, babies had only one fourth to one third adult levels of PON1 and that they required 6 mo to 2 years to reach adult levels of PON1, indicating increased sensitivity of the very young to exposures of chlorpyrifos (CPS)/CPO or diazinon (DZS)/DZO. The oxon metabolites are listed in the exposures since levels of oxon are detected in most if not all exposures as noted above [12–14].

In addition to the PON1^−/− mice, Shih, Lusis and Tward also generated transgenic mice that expressed human PON1_Q192 (tgHuPON1_Q192) and HuPON1_R192 (tgHuPON1_R192) on the mouse PON1^−/− background which provided an animal model in which to characterize the two human PON1₁₉₂ alloforms under physiological conditions [35]. The experiments with the transgenic mice showed that the efficacy of protection by each human PON1₁₉₂ alloform was consistent with the experiments where purified human PON1 was injected into the PON1^−/− mice. Two other important observations came from experiments with the genetically modified mice. When the entire human PON1 genes (PON1_Q192 and PON1_R192) were introduced into the PON1^−/− mice complete with the 5′ promoter regulatory sequences, rather than following the human time course for developmental expression (6 mo to 2 years), the expression followed the developmental time course of PON1 appearance in mice, peaking at 3 weeks of age indicating a high degree of conservation of the regulatory components of the PON1 genes between man and mouse. Importantly, a comparison between the OP sensitivity of wild type mice (PON1^+/+) and PON1^−/− mice showed that by 4 days of age, the PON1^+/+ mice were already 2.5 times more resistant to CPO exposure than the PON1^−/− mice [36].

Taken together, all of the experiments on PON1 and resistance to OP exposure indicated that PON1 is important in modulating exposures to CPS/CPO and DZS/DZO, but not to PS/PO or the nerve agents sarin and soman. Engineering more catalytically efficient variants of PON1 will be required for treating exposures to OPs that are hydrolyzed at low catalytic efficiency by PON1. Work by Harel and colleagues provided a variant of PON1 that could be crystallized to provide a tentative 3D structure of PON1, a six-bladed beta-sheet propeller structure [37]. This structure has provided the basis for further engineering of PON1 for higher catalytic efficiency of hydrolysis of nerve agents (e.g. [38, 39]).

Phylogenetic analysis suggests that PON2 is the oldest PON family member, from which PON1 and PON3 have evolved [72]. PON2 does not have the OP-hydrolyzing activities of PON1, but as the other two PONs, it is a lactonase, displaying overlapping but distinct substrate specificities for lactone hydrolysis [73]. In particular, PON2 has the highest hydrolytic activity of the PONs toward a number of acyl-homoserine lactones (acyl-HCL), molecules which mediate bacterial quorum-sensing signals, important in regulating expression of virulence factors and in inducing a host inflammatory response [73–76]. Two common SNVs have been found in human PON2, an Ala/Gly substitution at position 147, and a Ser/Cys substitution at position 311 [2, 64]. The PON2 S311C SNV has been shown to affect lactonase activity [77]. Carriers of the Cys311 allele have been found to be at risk for myocardial infarction and other cardiovascular diseases (CVD), as well as for Alzheimer’s disease (AD) in several studies [78–82].

In several tissues, PON2 has been shown to exhibit antioxidant properties [65]. PON2 antagonizes oxidative stress generated by various sources in the intestine of humans and rats [68], in human vascular endothelial cells [71], in lung epithelial carcinoma cells [76], in Caco-2/15 intestinal epithelial cells [83], and in mouse macrophages [67]. These antioxidant effects of PON2 are believed to play a major role in preventing the atherosclerotic process, as shown by studies indicating that PON2 over-expression decreases atherosclerotic lesions, while the opposite is true in PON2-null (PON2^−/−) mice [84, 85]. In macrophages, PON2 has been suggested to protect against accumulation of triglycerides and oxidative stress, thereby attenuating the development of vascular complications in diabetes [86, 87]. Mitochondria are a major source of free radical-related oxidative stress [88], and the preponderant localization of PON2 in mitochondria would support a role for this enzyme in protecting cells from oxidative damage. In HeLa cells, PON2 has been shown to bind to coenzyme Q10 that associates with mitochondrial complex III, and PON2 deficiency causes mitochondrial dysfunction [70]. In human endothelial cells PON2 has been shown to reduce, indirectly but specifically, the release of superoxide from the inner mitochondrial membrane, without affecting levels of other radicals such as hydrogen peroxide (H₂O₂) and peroxynitrite [89]. Also of interest is that the C311S SNV, which influences lactonase activity [77], does not appear to affect PON2’s antioxidant properties, suggesting independent hydrolytic and antioxidant functions [89]. PON2 also appears to exert anti-inflammatory effects. In the gastrointestinal tract, PON2 antagonizes oxidative and inflammatory processes that may affect mucosal integrity [68]. The absence of PON2 in PON2^−/− mice exacerbates the macrophage inflammatory response [90]. Furthermore, PON2 acts as a potent anti-inflammatory agent against the inflammatory response caused by administration of pyocyanin (a quorum sensing signal factor) [91].

PON2 mRNA has also been found in mouse and human brain [2, 64, 65, 92], and PON2 protein has been detected in mouse [5, 85] and monkey brain [93]. In a series of recent studies, the protein expression of PON2 has been characterized in mouse brain [66, 94, 95]. The highest levels of PON2 protein were found in three dopaminergic regions, the substantia nigra, the striatum, and the nucleus accumbens, with lower levels in cerebral cortex, cerebellum, hippocampus and brainstem. In every brain region, PON2 levels were higher (by ~2–3-fold) in female mice than in male mice. The higher levels of PON2 in dopaminergic areas are of interest, as they may be related to the higher levels of oxidative stress, due to dopamine metabolism, present in these regions. The regional distribution and gender difference of PON2 was confirmed by measurements of its lactonase activity [measured by dihydrocoumarin (DHC) hydrolysis] and of PON2 mRNA levels [66]. In brain, and to a lesser extent in kidney and testis, but not in all tissues, the PON2 antibody recognized two bands, the lower at MW ~43 kDa, which corresponds to the reported MW of PON2, and an upper band at MW ~55 kDa. This upper band had been found at times by some investigators [71, 84, 96–98], but not by others [65, 68, 83, 99], and may represent a PON2 alloform, in accordance with the two mRNA splice variants [64, 71, 76]. However, its exact nature and function have not yet been defined. Though PON2 is known to have four putative N-linked glycosylation sites at asparagine residues [77], deglycosylation experiments indicated that both putative alloforms are glycosylated [66]. Thus, purification of the upper band, and its analysis (e.g. by mass spectrometry) are needed to identify its structural features and other potential post-translational modifications. Nevertheless, neither band was detected in brain from PON2-deficient mice [66]. Also of interest, PON1 was detected at very low levels in all brain areas and did not show any regional brain or gender differences [66]. Such low levels in tissue homogenates may be due to residual blood, as no PON1 could be detected in striatal astrocytes or neurons [95]. PON3 was not detected in any brain region (either homogenate or cells).

PON2 protein levels (by Western blot), mRNA (by qRT-PCR) and activity (by DHC hydrolysis) were also examined in astrocytes and in neurons isolated from several brain regions. PON2 was significantly higher in astrocytes than in neurons in all brain regions, with the highest levels in cells isolated from the striatum. Striatal neurons and astrocytes isolated from female mice expressed higher levels of PON2 than the same cells from male animals. PON2 was also present in cortical microglia, at levels similar to those found in neurons [66]. The sub-cellular distribution of the PON2 protein was assessed in cerebellar granule neurons and cerebellar astrocytes, and found to be similar in astrocytes and neurons. Differences were found in the localization of the 43 kDa lower band (the putative PON2) and the upper 55 kDa band (the putative PON2 alternate alloform). In both cell types, the highest levels of 43 kDa PON2 were found in mitochondria, followed by membranes (microsomes), in agreement with previous observations in HeLa cells [70]. PON2 was not detected in the cytosolic, nuclear, or cytoskeletal fractions. In contrast, the upper band PON2 alloform was expressed at highest levels in the nucleus and the cytoskeleton, neither of which contained significant levels of the 43 kDa band [66].

As in other peripheral tissues, PON2 exerts a protective effect toward oxidative stress and neuroinflammation in brain cells. The cytotoxicity of two known oxidants, H₂O₂ and 2,3-dimethoxy-1,4-naphtoquinone (DMNQ), was investigated in cerebellar and striatal astrocytes and neurons isolated from wild-type (PON2^+/+) and PON2^−/− mice. In all instances, cells from mice lacking PON2 were more susceptible to the toxicity of both compounds, by a factor of 5 to 11-fold [66]. The protection afforded by PON2 to neurons and astrocytes was related to its ability to scavenge reactive oxygen species (ROS) upon exposure to oxidants. For example DMNQ (10 μM) increased ROS to ~400% of basal in neurons from PON2^−/− mice, and only 170% in the same cells from PON2^+/+ mice [66]. Levels of glutathione (GSH), which represents the main cellular defense factor against oxidative stress, did not differ between cells isolated from PON2^−/− and PON2^+/+ mice, suggesting that the differential susceptibility to oxidants was primarily due to the presence or absence of PON2 [66].

The higher levels of PON2 in tissues, including the brain, from female mice may be related to a positive modulatory effect by estrogens. In striatal astrocytes from male mice, 17β-estradiol caused a time- and concentration-dependent increase in the levels of PON2 protein; a 12–24 h exposure with 200 nM estradiol increased PON2 expression to the levels found in female striatal astrocytes [94]. Interestingly, in female astrocytes, estradiol could further increase PON2 expression, by a factor of about 2.5-fold. The estradiol effect was due to transcriptional activation of the PON2 gene, and was mediated by activation of estrogen receptor-alpha [94]. In ovariectomized mice, PON2 levels (protein and mRNA) were significantly reduced in striatum, cerebral cortex and liver, approaching the levels found in male mice. Striatal astrocytes and neurons from male mice were more sensitive to H₂O₂ and DMNQ-induced oxidative stress and ensuing cytotoxicity [94]. Though gender-dependent differences in other cell defense mechanisms cannot be excluded, it is noteworthy that levels of GSH did not differ between genders. Another important aspect is the lack of gender difference in susceptibility in cells from PON2^−/− mice. Striatal astrocytes from PON2^−/− mice of either gender were highly susceptible to oxidant-induced toxicity, as expected, but there were no significant female/male differences. In CNS cells from PON2^+/+ male mice, exposure to estradiol (200 nM, 24 h) provided protection toward toxicity induced by the two oxidants. This is not surprising, as neuroprotective actions of estrogens are well known [100–103]. However, the protective effect of estradiol was absent in cells from PON2^−/− mice, suggesting that a major mechanism of estrogen neuroprotection may be represented by induction of PON2 [94]. The functional consequences of a higher expression of PON2 in females may have several ramifications. First, similar gender differences were also found in rats, humans [94], and non-human primates[93]. With regard to neurodegenerative diseases, the role of oxidative stress in the etiopathology of Parkinson’s disease (PD) is well established [104]; of note is that the incidence of PD is 90% higher in males [105, 106]. Even though dopaminergic areas (striatum, substantia nigra, and nucleus accumbens) have the highest levels of PON2 in both genders, levels in females are still 2- to 3-fold higher than in males [66, 94, 95]. Lower PON2 levels in dopaminergic neurons in males may thus provide fewer defenses against oxidative stress. In this regard, of much interest are the recent findings that activation of dopamine D2 receptors in the kidney positively modulates PON2 expression, leading to a decrease in ROS production [107]. In the CNS, the highest levels of dopamine D2 receptors are found in the striatum, nucleus accumbens, substantia nigra and olfactory tubercle [108], areas that also have the highest level of PON2 expression [66] (Giordano et al., unpublished results). If a similar mechanism as observed in kidneys also occurs in the CNS, the loss of dopamine associated with PD would lead to decreased PON2 levels, thus fostering a spiral of events further aggravating neurodegeneration. Furthermore, as PON2 is expressed in most tissues, and levels appear to be higher in females in each tissue examined [66], the reported higher sensitivity of males to oxidative stress in heart, the higher susceptibility of males to atherosclerosis and to infections, may all be related to a differential expression of PON2 [109–112].

While there is a substantial amount of work on the modulation of PON1, which has been summarized in several reviews [113–115], more limited research has been carried out on PON2. In macrophages, PON2 expression is increased by oxidative stress [67], and in vascular cells by endoplasmic reticulum stress modulated via an endoplasmic reticulum stress element-like sequence found to be present in the promoter region of PON2 [71]. Arachidonic acid [116], unesterified cholesterol [117], the licorice phytoestrogen glabidrin [118], and the hypocholesterolemic drug atorvastatin [119] also upregulate PON2 expression in various cell types. Urokinase plasminogen activator upregulates PON2 in macrophages via NADPH oxidase and the transcription factor SREBP-2 [120, 121], while in mouse fibroblasts dexamethasone increases PON2 mRNA levels [122]. In one study, pomegranate juice was found to increase PON2 in macrophages [123], while quercetin was reported to increase PON2 mRNA and protein in macrophages in vitro [124]. The latter was confirmed in mouse astrocytes [95]. Extracts of Yerba mate (Ilex paraguariensis) have been reported to increase PON2 mRNA and lactonase activity in macrophages in vitro and after in vivo administration to healthy women [125].

PON3 was the last member of the PON family of proteins to be described [2] and is the least characterized. Like PON2, PON3 cannot hydrolyze OPs [72], but it retains lipo-lactonase and N-acyl homoserine lactone activities. The activity of PON3 has been reported to be calcium-dependent, like PON1 [131]. Interestingly, PON3 has a higher catalytic activity for statin lactones than PON1 [73, 127]. It is for this reason that statin lactones (such as lovastatin, spironolactone and canrenone) are commonly used to monitor PON3 activity. There are very few studies on polymorphisms in the PON3 gene. In a study with healthy subjects from southern Italy, the authors identified 3 silent (G51G, G73G, G99G) and 2 missense (S311T, G324D) variants in the exons III, IV and IX of PON3 [132]. These SNVs exhibited very low frequency in comparison with the frequency of PON1 and PON2 coding region SNVs. The effects of these missense variants on PON3 activity have yet to be evaluated. The same polymorphisms were studied in children with diagnosed inflammatory bowel disease, but no relationship between the PON3 genetic variants and disease was observed [133]. Haplotype associations between several SNVs in the PON gene cluster and AD were reported in a cohort of Caucasian and African Americans [134], suggesting a possible important role of PONs in AD that has not been ascertained to date. More recently, six PON3 promoter SNVs in linkage disequilibrium were significantly associated with changes in serum PON3 concentration in a healthy Mediterranean cohort [135]. The same authors studied these six SNVs in human immunodeficiency virus (HIV)-infected patients [136] and in coronary artery disease and peripheral artery disease patients [137]. However, no differences were found in these PON3 gene promoter SNVs and their haplotypes between patients and controls, suggesting that PON3 genotype neither influences serum PON3 concentration nor the course of these human diseases.

The physiological function of PON3 is less clear as this PON member has been poorly investigated. PON3 seems to be more potent than PON1 in protecting LDL from oxidative modification in vitro, although PON3 concentration in serum is about 2 orders of magnitude lower than PON1 [127]. In addition, unlike PON1, liver PON3 expression is not affected by oxidized phospholipids (HepG2 cells) or a high-fat diet (mouse liver) [126].

Circulating PON3 has been studied in a variety of human oxidative stress-related diseases with the objective to explore if disease states are associated with changes in the levels of PON3 concentration, as seen with PON1. Indeed, a significant increase of PON3 concentration has been reported in chronic liver disease [138], HIV-infection [136], and coronary and peripheral artery disease [137]. However, a more recent study in patients with autoimmune disease (systemic lupus erythematosus and type 1 diabetes) has shown significant depletion of PON3 protein in HDLs of patients with autoimmune disease and subclinical atherosclerosis [139]. Of note is that the technique used to measure PON3 in these studies is different (in-house serum ELISA [135] in the first studies vs. HDL LC-MS/MS in the latter study). Interestingly, in the HIV study, the authors also studied possible changes in the distribution of PON3 in lipoproteins with disease. Lipoproteins were fractionated by FPLC. They found that in non-infected participants, PON3 was exclusively detected in HDLs, while in HIV-infected subjects a substantial amount of PON3 was measured in the smallest HDL and LDL particles [136]. The HDLs measured in patients with autoimmune disease with and without subclinical atherosclerosis were separated by ultracentrifugation, and presence of PON3 in LDLs was not studied [139]. All in all, PON3 may be a useful analytical biomarker of human oxidative-stress related diseases.

PON3 has been shown to protect murine macrophages against oxidative damage, although cellular PON3 activity is decreased under oxidative stress [67]. These early in vitro results suggested an atheroprotective role for PON3 in vivo that was demonstrated in mice overexpressing human PON3 [140]. In the first study, Shih and colleagues found that elevated PON3 expression in human PON3 transgenic mice [either wild-type or LDL receptor knockout (LDLR^−/−) mice on the C57BL/6J background] significantly reduced diet-induced atherosclerotic lesions [140]. Surprisingly, this protective effect was male-specific and not driven by HDL, as the authors could not detect PON3 activity in mouse serum. Another striking finding from Shih and colleagues was a role for PON3 in attenuating the development of obesity in male mice. In the study led by Ng et al., adenoviral expression of human PON3 protected apolipoprotein E knockout (apoE^−/−) mice against progression of atherosclerosis [141]. In particular, elevated levels of PON3 enhanced cholesterol efflux, decreased LDL oxidation, and increased antioxidant properties of HDL, factors that promote slowing down the atherosclerotic process. Furthermore, the authors demonstrated that although human and rabbit PON3 associate with HDL [126, 127], endogenous mouse PON3 is undetectable in serum, in accordance with similar previous observations by the same group [140, 142]. Neither mice with adenoviral human PON3 expression nor mice that did not receive adenoviral PON3 had detectable PON3 protein in serum or HDL [141], suggesting that PON3 remains associated with cells in mice. In fact, adenovirus-mediated transgene expression was detected in a number of tissues, with liver showing a 2-fold increase in PON3 lactonase activity. This observation was supported by immunohistochemical analysis of PON3 expression in normal mouse tissues [5] and by Pon3 mRNA expression via in situ hybridization [143]. Both studies described PON3 expression in a wide range of mouse tissues, with much higher Pon3 mRNA expression in newborn mice compared to adult mice [143]. It should be noted that PON3 mRNA and protein expression in murine macrophages [67] and in specific segments of human and mouse gastrointestinal tract [96] had already been reported. Altogether, these results suggest that mouse PON3 may exert its anti-oxidative effect locally in a variety of epithelia and cells.

Following up on these observations, a protective role of PON3 in obesity has recently been confirmed in vivo in PON3 knockout (PON3^−/−) mice [130]. In this study, the authors found that lack of PON3 led to alterations in bile acid metabolism, increased body weight and increased atherosclerotic lesions compared to wild-type mice on a high fat diet. In addition, PON3 deficiency seemed to result in impaired mitochondrial respiration and mitochondrial superoxide levels, and increased hepatic expression of inflammatory genes. A recent study in patients with systemic lupus erythematosus reported an inverse association between PON3 concentration in HDLs and body mass index, supporting a role of PON3 in adiposity [139].

PON3 has also been related to immune-mediated enteropathies such as inflammatory bowel disease and celiac disease. In the study by Rothem et al., the authors found that PON3 (and PON1) mRNA expression was repressed in certain parts of the large intestine of Crohn’s patients and non-treated celiac disease patients, and in patients with ulcerative colitis [128]. They also localized PON3 in the endoplasmic reticulum of cultured PON3-GFP transfected HT29 and CaCo-2 cells, suggesting local synthesis and secretion of PON3 by intestinal cells.

Despite of PON3’s beneficial role in protecting against a variety of oxidative-stress related diseases, an unexpected finding by Schweikert and colleagues demonstrated that PON3, like PON2, has an oncogenic role in human cancers [129]. PON3 was found to be upregulated in various human cancer tissues, protecting those cells against mitochondrial superoxide-mediated apoptosis. In this regard, PON2 and PON3 seem to play similar roles in cancer, with PON3 being much more overexpressed in cancer cells than PON2. Although knocking down PON3 from certain cancer cells did not enhance susceptibility to chemotherapeutic drugs in vitro, PON3 may have potential as an anti-cancer target.

The finding that PON2 and PON3 have similar roles in cancer prompted the same research group to study if, similar to PON2, PON3 is involved in protecting against P. aeruginosa infections [91]. It is known that pyocyanin, an essential virulence factor secreted by the bacterium, causes oxidative stress (ROS) in vivo, which results in cell damage [144], and leads to a pro-inflammatory response resulting in interleukin-8 release [145]. Schweikert and colleagues described that in vitro exposure of pyocyanin to different cell lines led to ROS production, and activation of NF-kB, the major pathway involved in inflammatory response (via interleukin-8 release) [91]. More importantly, PON3 (and PON2) overexpression significantly prevented pyocyanin-induced ROS production and had a dramatic effect diminishing NF-kB-mediated inflammatory response. Furthermore, the quorum sensing signal N-acyl homoserine lactone promoted calcium influx in cells, which resulted in a calcium-dependent inactivation of PON2 (as previously demonstrated [76]) and PON3 activity. Altogether, PON3 and PON2 have important roles in the defense against P. aeruginosa virulence, although the bacterium can inactivate PON2 and PON3 activity during the infection.

The case of clopidogrel (an anti-platelet drug used in treating coronary artery disease, peripheral vascular disease and cerebrovascular disease) requires special comment. In 2011, Bouman et al. reported that PON1 was a major determinant of clopidogrel efficacy [152]. In a letter to the editor, many members of the PON1 research community expressed their concerns about the methodology used in their study [153]. PON1 was subsequently shown to generate an endothiol inactive metabolite [154, 155] while the two steps required for the bioactivation of clopidogrel to the active metabolite involve cytochromes P450.

The lesson learned from the studies of PONs in considering their roles in metabolism of various endogenous metabolites as well as xenobiotics is that the catalytic efficiency of hydrolysis is key in determining the physiological relevance of a given hydrolytic activity. For example, it was thought for a half-century that PON1 would provide protection against parathion/paraoxon. The studies by Li et al. [19] showed that while PON1 hydrolyzed PO in vivo, the catalytic efficiency was too low to provide protection against exposure whereas, the catalytic efficiency of DZO and CPO hydrolysis was sufficient to protect against exposures to these two OPs. The clopidogrel and homocysteine thiolactone studies described above provide additional examples of the importance of assuming a physiological relevance of a given PON in metabolism of a specific compound without solid in vivo data to support the conclusion.