One Enzyme, Two Functions

The human enzyme paraoxonase-2 (PON2) has two functions, an enzymatic lactonase activity and the reduction of intracellular oxidative stress. As a lactonase, it dominantly hydrolyzes bacterial signaling molecule 3OC12 and may contribute to the defense against pathogenic Pseudomonas aeruginosa. By its anti-oxidative effect, PON2 reduces cellular oxidative damage and influences redox signaling, which promotes cell survival. This may be appreciated but also deleterious given that high PON2 levels reduce atherosclerosis but may stabilize tumor cells. Here we addressed the unknown mechanisms and linkage of PON2 enzymatic and anti-oxidative function. We demonstrate that PON2 indirectly but specifically reduced superoxide release from the inner mitochondrial membrane, irrespective whether resulting from complex I or complex III of the electron transport chain. PON2 left O2̇̄ dismutase activities and cytochrome c expression unaltered, and it did not oxidize O2̇̄ but rather prevented its formation, which implies that PON2 acts by modulating quinones. To analyze linkage to hydrolytic activity, we introduced several point mutations and show that residues His114 and His133 are essential for PON2 activity. Further, we mapped its glycosylation sites and provide evidence that glycosylation, but not a native polymorphism Ser/Cys311, was critical to its activity. Importantly, none of these mutations altered the anti-oxidative/anti-apoptotic function of PON2, demonstrating unrelated activities of the same protein. Collectively, our study provides detailed mechanistic insight into the functions of PON2, which is important for its role in innate immunity, atherosclerosis, and cancer.

The human enzyme paraoxonase-2 (PON2) has two functions, an enzymatic lactonase activity and the reduction of intracellular oxidative stress. As a lactonase, it dominantly hydrolyzes bacterial signaling molecule 3OC12 and may contribute to the defense against pathogenic Pseudomonas aeruginosa. By its anti-oxidative effect, PON2 reduces cellular oxidative damage and influences redox signaling, which promotes cell survival. This may be appreciated but also deleterious given that high PON2 levels reduce atherosclerosis but may stabilize tumor cells. Here we addressed the unknown mechanisms and linkage of PON2 enzymatic and anti-oxidative function. We demonstrate that PON2 indirectly but specifically reduced superoxide release from the inner mitochondrial membrane, irrespective whether resulting from complex I or complex III of the electron transport chain. PON2 left O 2 . dismutase activities and cytochrome c expression unaltered, and it did not oxidize O 2 . but rather prevented its formation, which implies that PON2 acts by modulating quinones. To analyze linkage to hydrolytic activity, we introduced several point mutations and show that residues His 114 and His 133 are essential for PON2 activity. Further, we mapped its glycosylation sites and provide evidence that glycosylation, but not a native polymorphism Ser/Cys 311 , was critical to its activity. Importantly, none of these mutations altered the anti-oxidative/anti-apoptotic function of PON2, demonstrating unrelated activities of the same protein. Collectively, our study provides detailed mechanistic insight into the functions of PON2, which is important for its role in innate immunity, atherosclerosis, and cancer.
Reactive oxygen species (ROS) 3 is a general term for several products that result from partial reduction of oxygen, with superoxide (O 2 . ) being the first radical in numerous reaction chains. Either spontaneously or by enzymatic dismutase reaction, O 2 . reacts to form hydrogen peroxide (H 2 O 2 ), which itself may be detoxified to water or further react to the hydroxyl radical. ROS, and also reactive nitrogen species (e.g. peroxynitrite; ONOO Ϫ ), differ in characteristics such as stability, membrane permeability, or reactivity. Enhanced ROS levels, a result of augmented production or insufficient removal, modify diverse signaling pathways and are associated with various disorders such as cardiovascular diseases (1) or cancer (2). Consequently, many strategies aim at reducing ROS back to physiological levels, and mitochondria are centrally involved because of their high potential in ROS production (3  (4 -6). Thus, alterations of these processes are two-edged and can lead to either prolonged survival or death, for which reason they are modified in several diseases (e.g. tumors) and targeted by pharmacological interventions. Interestingly, the human enzyme paraoxonase-2 (PON2) may serve a putative target gene, because it protects against atherosclerosis (7,8), but may also support apoptosis evasion by its anti-oxidative and anti-apoptotic function (9). 4 The paraoxonase family consists of three highly conserved proteins PON1, PON2, and PON3, which were structurally and experimentally identified as calcium-dependent, lactone-hydrolyzing enzymes with variations in substrate specificity (10 -14). The PONs differ in localization as PON1 mainly associates with circulating serum high density lipoprotein particles, whereas the intracellular PON2 appears in two spliced isoforms that predominantly localize to the endoplasmic reticulum, mitochondria, and (peri)-nuclear region (7,9,15). Endogenous substrates of PON2 are largely unknown, but it has been shown that its lactonase activity dominantly hydrolyzes the Pseudomonas aeruginosa signaling lactone 3OC12, for which reason PON2 may add to the anti-bacterial defense (16,17), a mechanism apparently sensitive to major calcium disturbances (18,19). Further, studies with PON2-deficient mice indicated that PON2 suppresses atherosclerosis, and several studies described a significant ROS-diminishing function of PON2 in various cells (8,9,18,20,21). Although fundamental to understanding PON2 function, it is yet unanswered whether the lactonase activity and anti-oxidative function are interrelated. Also, the mechanism and specificity of its anti-oxidative function are unresolved. These issues were addressed in the current study. We found that PON2 specifically diminished O 2 . released by both mitochondrial complex I and complex III at the inner mitochondrial membrane, presumably by acting on coenzyme Q10 (CoQ). In support, PON2 differs from superoxide dismutases (SODs) or vitamins in that it efficiently prevented O 2 .
formation but did not diminish O 2 . resulting from CoQ-independent sources. We further demonstrate independent functions of PON2 because mutants devoid of lactonase activity still exhibited full protection against ROS and apoptosis. Together with results indicating that PON2 shares the same hydrolytic mechanism as PON1, we reveal fundamental details of the function and mechanism of PON2.
Site-directed Mutagenesis of PON2-We used the QuikChange or Finnzymes site-directed mutagenesis kits (Stratagene/New England Biolabs, respectively) according to the supplier's instructions with primers given in the supplemental material. Sequenced plasmids were transfected as before (9). Protein expression was verified by Western blotting; PON2 mutant-GFP cDNAs were resequenced.
Cell Fractionation-Mitochondria were enriched according to standard procedures (see supplemental material). Membranes of fresh mitochondria were prepared by brief sonication and recentrifugation.
ROS Detection-Unless stated otherwise, ROS was detected as previously published (9,18). Where indicated, cells were pre- PON2 Lactonase Activity-Deglycosylation and determination of PON2 lactonase activity were previously published (9,18). Importantly, in every experiment, PON2 mutant expression levels were analyzed by Western blotting using a ChemiDoc XRS imaging system (Bio-Rad) equipped with QuantityOne 4.6.7 software, normalized to ␣-tubulin, and subsumed comparing levels with PON2-WT.
Software, Statistics, and Image Acquisition-GraphPad Prism 5 was used for calculation or statistical evaluation using one-or two-way analysis of variance and/or non-linear regression curve fitting. p Ͻ 0.05 was considered significant. Adobe Photoshop was used for image acquisition. If necessary, only brightness and/or contrast were changed simultaneously for all areas of any blot.
-We previously showed that PON2 reduced DMNQ-generated ROS detected with carboxyl-H 2 DCFDA or the luminol derivative L-012 (9), but these analyses were unable to deduce the particular radical diminished by PON2. Because pretreatment with cell-permeable PEG-SOD but not PEG-catalase (to diminish O 2 . or H 2 O 2 , respectively) significantly lowered ROS levels in such analyses (Fig. 1A), we concluded that PON2 diminished superoxide. To test this more directly, we used dihydroethidium, which under oxidative stress is oxidized to O 2 . -specific 2-hydroxyethidium and unspecific ethidium. After cells were loaded with dihydroethidium and treated with DMNQ, significantly increased 2-hydroxyethidium levels were found in naive, but not PON2-overexpressing, cells (Fig. 1B), which verifies the above conclusion. PON2 appears specific for O 2 . because it did not alter levels of other radicals, namely H 2 O 2 and ONOO Ϫ (Fig. 1, C and D).
To test for direct protein actions, O 2 . was generated in vitro by xanthine/xanthine oxidase (X/XO) in the presence of recombinant or mammalian PON2, but these PON2 preparations were indistinguishable from controls (Fig. 1E). This implies an indirect mechanism of PON2 and the necessity of a co-factor, or given the lack of redox-active centers, PON2 itself could serve as a co-factor for other systems. However, the chief O 2 . -diminishing activities of SOD1 and SOD2 were not changed by PON2 overexpression (Fig. 1F and  . but could not diminish this radical once it is produced. This was tested with CoQ-independent ROS sources. We stimulated mitochondrial membranes from naive or PON2-overexpressing cells with ␣-ketoglutarate, which produces ROS from membrane-associated but CoQ-independent ␣-ketoglutarate-dehydrogenase (25). Under such conditions, similar O 2 . levels were observed (Fig. 2C). Likewise, when O 2 .
was generated by X/XO, mitochondrial membranes from PON2-overexpressing cells were indistinguishable from controls (Fig. 2D), in contrast to stimulation with succinate, rotenone, or antimycin (see above . to oxygen (26), but its expression was unchanged by PON2 (Fig. 2E), and cytochrome c siRNA did not . was generated by X/XO and detected by L-012 in the presence of recombinant human PON2 (rhPON2) or bovine serum albumin (100 ng; upper panel) or PON2-His 6 (2.5 g; lower panel). F, named cells were DMNQ-treated (2 h) and analyzed for SOD1 ϩSOD2 activities by zymography. Band intensity reflects enzyme activity.

PON2 Prevents Mitochondrial Superoxide Formation
abrogate the effect of PON2, albeit cytochrome c levels were reduced to ϳ60% (not shown). Second, PON2 did not increase O 2 . dismutation because lysates from naive or PON2-overexpressing cells had similar H 2 O 2 levels secondary to X/XO-generated O 2 . (Fig. 2F).
PON2 Has Independent Activities-To test whether ROS reduction is linked to enzymatic activity, we identified residues crucial to hydrolytic activity. PON2 shares ϳ60% identity with PON1. For PON1, key catalytic residues are known, and its hydrolytic activity, but not glycosylation, is essential to its anti-atherogenic properties (11,(27)(28)(29). We created point mutations in PON2 that correspond to the PON1 residues required for hydrolytic activity. Confocal microscopy of the GFP-tagged proteins revealed subcellular targeting identical to PON2-WT (not shown). To map the glycosylation site of PON2, glycosidase treatment with subsequent Western blotting was applied to lysates from cells overexpressing PON2-WT, -N254A, -N269A, -N323A, or -N254A/N323A (Fig. 3A). The mass of PON2-N269A was similar to the WT; thus, Asn 269 is not glycosylated. In contrast, PON2-N254A or -N323A migrated between glycosylated and deglycosylated PON2-WT, and double mutant N254A/N323A was identical to deglycosylated PON2-WT. Thus, PON2 is glycosylated on both Asn 254 and Asn 323 , which is consistent with Stoltz et al. (30).
To analyze PON2 activity, we used one of its putative natural substrates, 3OC12 (12,16,18), under conditions that estimate initial hydrolytic rates. For all experiments, activities were normalized to expression levels (see "Experimental Procedures") to exclude that differences result from dissimilar expression. Our previous studies showed that 3OC12 activity is specific for PON2 in EA.hy 926 cells (16,18). Although PON2-WT overexpression increased 3OC12 hydrolysis ϳ2.3-fold, this was blunted by single mutations H114Q or H133Q (Fig.  3B), conforming with crucial functions in PON1 where His 115 deprotonates water to create a lactone-attacking hydroxide and His 134 establishes H 115 basicity (11,28). A similar effect was observed with PON2-D268A, which likely disturbs binding of the catalytic calcium. This emphasizes the requirement of the two enzyme-bound calcium ions, which adds to previous studies (11,16) and suggests similar hydrolytic mechanisms of PON1 and PON2. However, although glycosylation was dispensable for secretion and activity of PON1 (27), Stoltz et al. (30) showed that glycosylation of PON2 influenced lactonase activity. In accord with the latter, we also found that PON2-N254A/N323A lacked activity (Fig. 3B). To exclude that lost activity resulted merely from . formation. A, indicated cells were loaded with Mito-HE, treated with solvent, antimycin, or rotenone for 2 h, and analyzed by fluorescence-activated cell sorting. n ϭ 3-6; ***, p Ͻ 0.001. B, mitochondrial membranes from indicated cells were loaded with MCLA and treated with succinate. C, similar to B but stimulating with ␣-ketoglutarate. D, intact mitochondria, mitochondrial membranes (mem), or matrix fractions of named cells were loaded with MCLA and exposed to X/XO-generated O 2 . . E, named cells were analyzed for cytochrome c and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) protein expression. F, H 2 O 2 secondary to X/XO-generated O 2 . was measured in the presence of native lysates from naive (ϮPEG-SOD) or PON2-GFP-overexpressing cells. HA, hemagglutinin. disturbed initial folding due to lacking glycosylation, we also analyzed lysates with prior deglycosylation, upon which active PON2-WT again lost activity (not shown). In contrast to Stoltz et al. (30), however, we did not observe any impact of natural polymorphism Ser/Cys 311 on 3OC12 hydrolysis (Fig. 3B).
We finally tested whether mutations that abrogate lactonase activity also influence ROS reduction. Importantly, PON2-H114Q and -H133Q (and also -S311C) were similar to WT in their ability to reduce ROS (Fig. 4A), which implies independent hydrolytic and anti-oxidative functions. Similarly, additional experiments demonstrate direct mitochondrial functions of PON2-WT and -H114Q on steps subsequent to O 2 .
formation (i.e. cardiolipin peroxidation, cytochrome c release, etc.), suggesting that the anti-apoptotic effect of PON2 is based on ROS reduction (not shown); 4 in accordance, PON2-H114Q and/or PON2-H133Q equaled PON2-WT in preventing induction of apoptosis, which we here assessed by investigating activation of caspase-3, loss in intracellular ATP levels, and annexin V/7-aminoactinomycin D flow cytometry (Fig. 4, C-E, respectively). The fact that PON2-H114Q and PON2-H133Q prevented apoptosis somewhat better than PON2-WT agrees with their slightly better performance in reduction of ROS.

DISCUSSION
ROS and reactive nitrogen species are not just harmful but also vital to cell function, and some clinical trials demonstrated even augmented mortality after anti-oxidant treatment (31)(32)(33). Hence, the outcome of anti-oxidant treatment depends on the overall context, the cell type, efficient targeting, and the nature of the radical species and its detoxification product. Previous reports on PON2 demonstrated its anti-oxidative func-tion (8,9,18,20)  . production from both complex I and complex III, which together suggests a modulation of or interaction with CoQ. A PON2 mutant lacking solely the anti-oxidative effect would be very valuable in studying ROS-related functions of the protein. However, despite all effort, we and other laboratories did not find such mutants. For PON1, enzymatic activity may give rise to its anti-oxidative role, but our data suggest that this is different for PON2. We also inserted mutations other than the ones reported here (not shown). For instance, PON2-C283A and -C283V were constructed as the corresponding residue in PON1 may be involved in its anti-oxidative function. However, this PON2 mutation was hardly tolerated by cells and did not appear to impact on the reported activities. Further, we truncated the amino-terminal residues 1-26 as this region likely represents the membrane attachment site of PON2 such that its deletion may have altered localization and function, but cells did not express it to reasonable levels. We also used the 12-residue-shorter isoform-2 of PON2 (9), but this mutant appeared to induce cell cycle arrest for yet unknown reasons (not shown), which confounds analyses of cell death. Given the fact that all named mutants did not impact on ROS and considering the now characterized anti-oxidative mechanism of PON2, we argue that PON2, in the simplest model, may act as an insulator for CoQ to prevent coincidental superoxide production at the inner mitochondrial membrane, a principal that applies to several proteins (23). Apparently, this may culminate in cancer as some tumors could exploit PON2 and enhance its expression to evade ROS-mediated apoptotic signaling. 4 Together with the fact that PON2 most likely lacks redox-active groups or co-factors (e.g. those present in SODs), this collectively implies that there may not exist such a mutant that inhibits the anti-oxidative activity of PON2 unless its proposed interactions with CoQ, folding, or targeting are disturbed.
Most of our understanding on PONs come from PON1, especially due to structural data and functional characteristics (11,28,29). Despite differences in substrate specificity and localization (i.e. secretion) of the PONs, their high conservation suggested similar enzymatic mechanisms. Indeed, the histidine dyad His 114 /His 133 in PON2 and His 115 /His 134 in PON1 are similarly fundamental. Also, it is known that PONs stripped of their calcium lose activity (16), which we confirmed by mutating residues likely involved in calcium binding (D268A). An exception, however, seems to be glycosylation because unglycosylated PON1 still had ϳ50% activity (35), whereas deglycosylated PON2 was inactive. Unlike PON1, PON2 is not present on high density lipoprotein particles (20); thus, glycosylation may dictate structural arrangements critical to the specific environment or substrate binding in particular. Further, the role of the polymorphism Ser/Cys 311 of PON2 remains controversial. Recently, Stoltz et al. (30) claimed that Cys 311 impaired PON2 activity, but we did not observe this effect. One putative confounder may come from technical differences (we measured 3OC12 hydrolysis directly, whereas Stoltz et al. (30) employed a Bio-assay). Also, we measured initial rates at a 3OC12 concentration of 50 M, whereas Stoltz et al. (30) used 10 M 3OC12 and did not measure initial rates. However, even if PON2-Cys 311 hydrolyzed 3OC12 slower than WT (30), we found no importance to the reduction of ROS or apoptosis.
Together, we demonstrate that PON2 prevents mitochondria-derived superoxide formation, identify catalytically essential residues, and provide evidence that the anti-oxidative and hydrolytic activities are separate. This allows discrimination of putative roles of PON2 in host-pathogen interactions, atherosclerosis, or cancer.