Activating ω -6 Polyunsaturated Fatty Acids and Inhibitory Purine Nucleotides are High Affinity Ligands for Novel Mitochondrial Uncoupling Proteins UCP2 and UCP3.*

UCP2 (the lowest Km values: 20 and 29 microm, respectively) for omega-6 polyunsaturated FAs (PUFAs), all-cis-8,11,14-eicosatrienoic and all-cis-6,9,12-octadecatrienoic acids, which are also the most potent agonists of the nuclear PPARbeta receptor in the activation of UCP2 transcription. omega-3 PUFA, cis-5,8,11,14,17-eicosapentaenoic acid had lower affinity (Km, 50 microm), although as an omega-6 PUFA, arachidonic acid exhibited the same low affinity as lauric acid (Km, approximately 200 microm). These findings suggest a possible dual role of some PUFAs in activating both UCPn expression and uncoupling activity. UCP2 (UCP3)-dependent H+ translocation activated by all tested FAs was inhibited by purine nucleotides with apparent affinity to UCP2 (reciprocal Ki) decreasing in order: ADP > ATP approximately GTP > GDP >> AMP. Also [3H]GTP ([3H]ATP) binding to isolated Escherichia coli (Kd, approximately 5 microm) or yeast-expressed UCP2 (Kd, approximately 1.5 microm) or UCP3 exhibited high affinity, similar to UCP1. The estimated number of [3H]GTP high affinity (Kd, <0.4 microm) binding sites was (in pmol/mg of protein) 182 in lung mitochondria, 74 in kidney, 28 in skeletal muscle, and approximately 20 in liver mitochondria. We conclude that purine nucleotides must be the physiological inhibitors of UCPn-mediated uncoupling in vivo.


EXPERIMENTAL PROCEDURES
Most of the chemicals were purchased from Sigma. 3 H-labeled nucleotides were from Amersham Pharmacia Biotech. Hydroxylapatite, Bio-Gel HTP and Bio-Beads SM2 were from Bio-Rad.
Materials for reconstitution were from the same sources as described elsewhere (39), materials for yeast fermentation were from Difco. Zymolyase 100T was from ICN. All other chemicals were of a reagent grade.

Yeast-expression of UCP2, UCP3
W303 yeast containing pCGS110 (or pYES) vectors with inserted cDNAs coding for human UCP2 and human UCP3 under control of an inducible Gal promotor and uraselection were donated by Ruth E. Gimeno  (1.5 vol%) lactate, 0.05% glucose, 0.17% yeast nitrogen base (Difco) 0.5% ammonium sulfate, 0.005% each of L-aminoacids (tryptophan, methionine, arginine, leucine, histidine) and adeninesulfate. After 24 hours, 0.2% galactose was added and cells were shaken for another 24-28 hours until the optical density of 1 was achieved. Mitochondria were prepared immediately after terminating fermentation using Zymolyase 100T (ICN) to cleave the cell wall (40). Protein extraction followed immediately. Alternations in different fermentation yields were compensated by taking the same amount of yeast mitochondrial protein, usually 30 mg. containing 1.67 % sodium lauroylsarcosinate (SLS) and 1% octylpentaoxyethylene. The resulted micellar solution was concentrated (thus partly depleted of SLS) and subsequently diluted (usually 1:1) in 20 mM Na-MES, pH 6. Protein content was estimated using the Amidoblack method (42).

Isolation and reconstitution of yeast-expressed UCP2, UCP3
Reconstitution with lipid protection was adopted from Klingenberg (43) to comply with 6-methoxy -N-(3-sulfopropyl)quinolinium (SPQ) fluorescent monitoring of ion fluxes (20,22,23,37,39). It included OctylPOE extraction of yeast mitochondria (30 mg protein) under lipid protection, the isolation step on HTP, detergent removal on Bio-Beads SM2 overnight, and external probe washing on Sephadex G25-300. The total amount of added lipid was 40.7 mg (egg yolk lecithin, type XI-E, Sigma, 4% cardiolipin and 1.6% L-α-phosphatidic acid). Additional lipids in 1 ml of final suspension, up to 20 mg, could have originated from the mitochondria. The internal medium for liposomes contained 84.4 mM TEA 2 SO 4 , 28.8 mM TEA-TES with 9.2 mM TEA, pH 7.2, 0.6 mM TEA-EGTA. The protein content of liposomes was also estimated by the Amidoblack method (42).
Usually, lipid-to-protein ratio of about 1000 or higher was obtained. The identity of UCP2 (UCP3) in HTP-pass-through was verified using peptide mapping assisted by MALDI-TOF mass spectroscopy after in-gel trypsine cleavage of the PAGE-separated acetone-precipitated samples.

Fluorescent monitoring of H + fluxes in proteoliposomes.
Valinomycin-induced H + fluxes in the presence of various FAs were monitored by the SPQ quenching method (20,22,23,37,39) (46). Note that the extent of this acidification is unchanged in the presence of ATP, reflecting the undistorted SPQ response. Consequently, we could detect that the observed H + fluxes were up to 50% inhibited by various purine nucleotide (PN) di-and triphosphates (vide infra). High PN concentrations decreased also the extent of SPQ response to H + efflux, which may reflect a possibility that a single protein per vesicle exists in our preparation (or if several proteins exist they have the same orientation). This is feasible due to our high lipid-to-protein ratio 3 and such observations were made for years with UCP1 or plant UCP not only for SPQ monitoring of H + fluxes, but also for Cland alkylsulfonate fluxes (22,23,37,40,47). Fig.1a,b illustrate the inhibition of the lauric acid-induced H + efflux in proteoliposomes containing UCP2 or UCP3. However, ATP did not affect the slow H + efflux induced at a 10-fold higher valinomycin concentration by lauric acid in proteoliposomes containing the extract from the mitochondria of W303 yeast not expressing UCP2 or UCP3 (Fig.1c). The W303-extract was prepared and passed via the HTP column in the same way as for yeast expressing UCP2 or UCP3. The resulting FA-induced H + efflux increased with the increasing initial amount of W303 yeast mitochondria. It is likely mediated by yeast carriers of the HTP pass-through, such as the ADP/ATP carrier (48) or the phosphate carrier (39,49). In turn, the ATP insensitivity of H + efflux with the reconstituted W303-extract and only slight (<10%) inhibition of FA-induced H + efflux in UCP2-or UCP3-proteoliposomes by CAT or methylenediphosphonate suggest that the ATPsensitive H + fluxes in UCP2-(UCP3)-proteoliposomes are indeed due to the UCP2 (UCP3) function, respectively. Various natural FAs were able to induce H + efflux in proteoliposomes containing UCP2 (Tab.I) and UCP3. Among all tested FAs, the fastest H + efflux was found for oleic acid and myristic acids followed by polyunsaturated FAs. Also the H + efflux induced by PUFAs was inhibited up to 50% by ATP present in the external medium ( Fig.1d and vide infra). Similar results were found with UCP3. Moreover, additions of ω-6 PUFAs exhibited the highest extent of flip-flop acidification.
Thus, cis-8,11,14-eicosatrienoic acid (C20:3 ω-6) had on average a double extent of flip-flop acidification when compared to lauric acid (Fig.1a,d). This finding would suggest that the physicochemical properties of FA movement in the membrane and their partition coefficient K p could contribute to their maximum cycling rates. Also flip-flop rate of fatty acids was previously found to increase with increasing unsaturation (50). In order to elucidate the involved structure/ kinetic relationships we further studied the kinetics of the FA activation of UCP2-mediated H + efflux.

Kinetics of UCP2-mediated H + efflux induced by lauric acid -Kinetics of lauric acid activation
of H + efflux (FA cycling) in UCP2 proteoliposomes is illustrated in Fig. 2a. H + efflux rates were evaluated with and without 2.5 mM ADP. Assuming Michaelis-Menten kinetics, we constructed Eadie-Hofstee plots for total rates (right panel in Fig. 2a, Table I) and for differential rates (not shown), when the rates measured with ADP were subtracted from the control rates. The kinetic parameters obtained for differential rates were used to construct the theoretical fit by the Michaelis-Menten equation (dotted line in Fig. 2a). This fit reflects the kinetics of uniformly oriented UCP2 molecules with the PN-binding site exposed outside. Mostly, this kinetics was similar to that measured with external ADP or ATP, which reflects the opposite UCP orientation.
The Eadie-Hofstee plot for the experiment with externally added ADP is nearly parallel to the control plot (Fig. 2a). Thus, kinetics in the absence and presence of ADP exhibits almost equal K m , while V max with ADP was about half of the control. As with UCP1, this fact reflects a noncompetitive and perhaps an allosteric type of nucleotide inhibition. Although subjected to errors, the differential rates also yielded an Eadie-Hofstee plot parallel with those for control or ADP. The apparent K m s for lauric acid and UCP2 or UCP3 (Tab. I) were higher than those found for UCP1 (22). Note, that this has been also observed for CoQ 10 -activated E.coli-expressed UCP2 and UCP3 (21). ing total FAs was observed for all FAs tested. When comparing H + flux densities per square µm (39) (Tab. I), the densities in UCP2-or UCP3-proteoliposomes were much higher than the densities for background H + fluxes in the protein-free liposomes. They were usually also higher than the H + flux densities estimated in the vesicles containing extracts from W303 yeast mitochondria (Tab. I). The Eadie-Hofstee plots constructed for total H + fluxes in µmol H + .min -1 .(mg protein) -1 were again parallel with those for H + fluxes in the presence of external ATP or ADP and mostly also with those for differential rates (dotted lines in Fig. 2 a-d).
The apparent affinity for various FAs (taken as the inverse K m ,Tab. I) was quite similar but decreasing in order for heptylbenzoic > palmitic > lauric > linoleic > arachidonic > oleic > myristic acid. Note that PUFAs (vide infra) and nonphysiological heptylbenzoic acid exhibited higher affinity than naturally abundant FAs. The V max values and turnover numbers per dimer are also listed in the Table I. The turnovers represent the minimum estimates, since not all measured protein is likely to be active. Among abundant FAs, the highest turnover was found for oleic acid, while it decreased for myristic > linoleic > arachidonic ≥ palmitic ≥ lauric acid. Heptylbenzoic acid exhibited the lowest evaluated turnover. Only 10% H + flux was observed for caprylic acid and less than 5% for 12-hydroxylauric acid, an inactive FA which is unable to flip-flop across the lipid bilayer (46).
Note that H + fluxes in the absence of FAs amounted to 7 to10% of V max for lauric acid (Tab. I).

No effect of Coenzyme Q 10 on UCP2 and UCP3-mediated H + efflux induced by lauric acid -
We have also tested the effect of the presumed activating cofactor, Coenzyme Q 10 . No effect of oxidized CoQ 10 (1 to 5 µM) was observed when added directly to the assay (Fig.1a, right panel), nor when oxidized CoQ 10 was added to lipids during extraction of yeast mitochondria and formation of vesicles (not shown). This indicates that in our reconstituted system the recombinant yeastexpressed UCP2 and UCP3 are intact and do not need further activation other than by FAs. These results, however, do not entirely exclude the existence of CoQ 10 activation of UCPs, since the effective CoQ 10 dose could be extracted from yeast mitochondria. Nevertheless, with E.coliexpressed proteins CoQ 10 activation did not exceed 30% -1 µM CoQ 10 at 100 µM lauric acid activated UCP2 to 132%; 0.1 µM CoQ 10 activated UCP2 to 103%, UCP3 to 135% 4 .
Nucleotide inhibition of UCP2 and UCP3 -Also the above kinetic data demonstrated that the observed H + fluxes were up to 50% inhibited by various PN di-and triphosphates 3 . The inability of complete inhibition by the external PN can be explained by two equally distributed orientations of UCP molecules in the membrane, with PN-binding sites exposed outside and inside (22). Hence, we could evaluate the inhibitory dose-responses for lauric acid-induced H + uniport while assuming a decrease to a 50% rate as 100% inhibition. Stock nucleotide solutions that are usually acidic had to be buffered by Tris-base and assay pH was carefully checked to be constant at 7.2, otherwise, artificial "inhibition" was observed due to a decrease in rates by acidic pH. For reconstituted UCP2, the lowest K i was exhibited by ADP (350 µM, Fig. 4), while ATP exhibited higher K i (445 µM) reflecting a slightly lower affinity of ATP to UCP2 (Fig. 4). These results were confirmed by three ω-6 Polyunsaturated fatty acids and purine nucleotides as UCP2 and UCP3 ligands 12 independent experiments. The apparent affinity for GTP was slightly lower than for ATP (Tab. II) and for GDP it was even lower (Tab. II). The lowest affinity was found for AMP (Fig. 4, Tab. II); indeed it is similar to low affinity of AMP and GMP vs. ATP or GTP, previously documented for UCP1 (51,52). Similar results were obtained with UCP3 (Tab. II). The magnitudes of K i s for PN inhibition are higher than those found for UCP1, as also tested with recombinant yeast expressed UCP1 (Tab. II), but are lower than those reported for E. coli-expressed protein (20). Similar data but with lower affinities (higher K d s) and lower numbers of binding sites were obtained for UCP2 and UCP3 expressed in E. coli ( Fig. 6; Tab. III). CAT had no effect (Fig. 6), thus reflecting that no ADP/ATP carrier contamination is possible with this expression (as well as no Coenzyme Q carryover). In spite of this, the obtained K d s were three times higher, possibly also due to remaining lauroylsarcosinate present. We can conclude that the observed saturated binding reflects the nucleotide binding sites of recombinant UCP2 and UCP3 in both cases of expression. The higher affinity and higher proportion of binding sites obtained for yeast-expressed proteins reflects the advantage of this system and the lipid protection used.

Binding of 3 H-GTP and 3 H-ATP to isolated
3 H-GTP binding to isolated mitochondria of several tissues -In order to demonstrate the existence of native UCPs in intact tissues and to show the relevancy of the binding method not only for recombinant, but also for native proteins, we have evaluated 3 H-GTP binding in the presence of CAT in mitochondria isolated from rat liver, skeletal muscle, kidney and lung (Fig. 7).
In agreement with findings of Peqeuer et al. (19), we found the highest number of 3 H-GTP binding sites in lung mitochondria (182 ± 18 pmol/mg protein). This is about four times lower than the usual amount of 3 H-GTP binding sites reflecting mostly the UCP1 molecules in BAT mitochondria (53). However, the evaluated K d of 0.43 ± 0.03 µM reflects an even higher affinity than found with yeast-expressed human UCP2. The estimated total number of 3 H-GTP binding sites was lower (74 ± 22 and 28 ± 6 pmol/mg protein, K d s 0.3 ± 0.06 µM and 0.14 ± 0.02 µM) in kidney and skeletal muscle mitochondria, respectively, accounting for~10 and ~30 times less than for UCP1 in BAT mitochondria. The lowest amount was found in liver mitochondria (21 ± 4 pmol/mg protein; K d 0.23 ± 0.03 µM). The measured proportions between the numbers of 3 H-GTP binding sites in the studied tissues correlate well to the proportions of UCP2 mRNA typically found in these tissues (5-7; 19). Hence, even if not all 3 H-GTP binding sites could be ascribed to the native UCP2 (UCP2 and UCP3 in skeletal muscle), the interfering part should represent a minor portion. Consequently, these data represent the first demonstration of the existence of high affinity nucleotide binding to native UCP2 (UCP3).

DISCUSSION
In this work we have evaluated in detail the possible phenotypes of the novel uncoupling proteins UCP2 and UCP3. We have identified their best up-to-date known activating and inhibitory ligands.
We have demonstrated that two (C18 and C20) ω-6 PUFAs are the most potent activators of UCP2 (UCP3), whereas among purine nucleotides, ADP is the most potent inhibitor. We have for the first time demonstrated the binding of natural 3 H-labeled nucleotides to recombinant UCP2 (UCP3) proteins and to the mitochondria of several tissues.
Concerning activating FAs, we have clearly demonstrated that all physiologically abundant long chain FAs, saturated or unsaturated, activate H + translocation in UCP2-and UCP3-proteoliposomes. This is demonstrated by the parameters of their activating (FA cycling) kinetics. We cannot explain why Rial et al. (36) could not find any response with all their tested FAs except for all-transretinoic acid in yeast mitochondria of yeast expressing UCP2. Among our tested FAs, we have found that oleic acid exhibited the highest rate, but ω-6 PUFAs, such as cis-8,11,14-eicosatrienoic (C20:3 ω-6) and cis-6,9,12 octadecatrienoic (C18:3 ω-6) exhibited both high V max and the highest apparent affinity (1/K m ). As such they were very efficient in inducing H + uniport mediated by UCP2.
Hence, we have shown that "more efficient" FAs do indeed exist and are activating UCP2mediated H + uniport in lower concentrations (amounts) than those required for other natural FAs.
Their low in vivo abundance is balanced by their high activating profile. This is valid for the two ω-6 PUFAs which were identified by us as the best UCP2 activators. For example, the typical content as 1.9 and 2.9 µg of C20:3 ω-6, or 0.8 and 0.4 µg of C18:3 ω-6, was identified per mg of phospholipids in rat liver and kidney, respectively (54). The amount of C20:3 ω-6 in human plasma phospholipids is equivalent to 100 µM (55). One can speculate that if 10% of these amounts would be cleaved off, a substantial activation of UCP2 will occur. A slightly lower efficiency with regard to UCP2 was found for ω-3 PUFAs, EPA (C20:5 ω-3) and docosahexaenoic acid (C22:6 ω-3), but due to the very high content of the latter in the brain or retina tissues (56), activation of UCP2 by C22:6 ω-3 is very plausible. The efficiency of C20:5 ω-3 and C22:6 ω-3 was still higher than that of arachidonic or oleic, or lauric and palmitic acid (equally effective as lauric). Our results have a great physiological relevance, since ω-6 PUFAs are among the most efficient activators of PPARβ ω-6 Polyunsaturated fatty acids and purine nucleotides as UCP2 and UCP3 ligands 15 (17) and together with ω-3 PUFAs also to potent activators of PPARγ (18) and PPARα (57). In conclusion, our findings suggest their possible dual role in activating both UCP2 (UCP3) expression and the uncoupling activity.
Our results have also shown that protonophoric phenotypes of UCP2 and UCP3 do not differ qualitatively from the UCP1 phenotype. Their per-dimer-turnovers (estimated from V max and total protein) fall into the range of hundreds per second (Table II). It is similar to the turnover of 94 s -1 that can be derived from the rate of 86.7 µmol H + min -1 (mg protein) -1 for lauric acid at 25 o C and CoQ 10 -activated E. coli-expressed UCP2 (21). Also, the maximum per-dimer-turnover reported for UCP1 (133 s -1 , 26) is comparable. The same is true for plant UCP (99 s -1 ,47). The situation is more complex when one compares experimental apparent affinities taken as 1/K m . Since, we did not subtract a contribution of non-protein related H + leak, the real K m values could be lower. Our K m values (except for a few or those for PUFAs) are higher than those reported for UCP1 and for the phosphate carrier (49). Nevertheless, similar high K m values can be derived from the results reported by Klingenberg (21).
Concerning PN inhibition, the derived K i s for UCP2 and UCP3 were higher than those for UCP1 in our reconstituted system, but all K i s for ATP and ADP were within the range of hundreds of µM. We must admit that sulfate used in our assay media strongly decreases the PN affinity to UCP1 (25) and presumably also to UCP2, UCP3, which might explain the discrepancies between our K i s and those measured by Klingenberg (21,58) and partly also between the K i s (Tab. II) and binding constants K d s (Tab. III). We chose sulfate to ensure that the anion used would be membrane impermeant, would not quench the SPQ probe, and that UCPs would not transport it.
Thus, sacrifying the efficiency of nucleotide inhibition, we ensured high K + diffusion potential and the correct probe response. Our media may better simulate in vivo conditions. We have also demonstrated that ADP is a slightly stronger UCP2 inhibitor than ATP. The same finding has previously been reported for UCP3 (58). GTP and GDP exhibited a little bit lower inhibitory ability.
However, the differences between ATP and GTP were not found for binding constants K d s. It is also not certain whether the observed small difference in ATP and ADP affinities in vitro can This indicates that if binding to the ADP/ATP carrier contributes to the total binding in the samples obtained by yeast-expression, it represents a minor contribution. The data obtained with CAT then reflect net binding to UCP2. For E. coli-expressed UCP2 and UCP3, similar data were found, but with higher K d s and lower binding site numbers. This may reflect the continued presence of interfering lauroylsarcosinate or that only a portion of the population of UCPn molecules has an intact conformation. Nevertheless, the affinity derived as reciprocal K d is about three times higher for the native UCP2 in mitochondria than for recombinant yeast-expressed UCP2. Obviously, the natural insertion of UCP2 molecule in the internal membrane is superior to the micellar solution.
The existence of high affinity 3 H-GTP-binding sites in mitochondria suggests that these sites may indeed reflect the binding to native UCP2 (UCP2 plus UCP3 in skeletal muscle mitochondria). This conclusion is strongly supported by our finding of the highest number of 3 H-GTP binding sites in the mitochondria of tissues where UCP2 content was reported to be high, i.e.
in lung (19) and kidney (27). On the contrary, we found a ~10 times lower number of 3 H-GTPbinding sites in the mitochondria of liver, an organ which has almost no UCP2 expression (6,7), unless it is stimulated, e.g. by cytokines (2,9). This fact again suggests that the detected 3 H-GTPbinding sites could be predominantly formed by UCP2. Surprisingly, skeletal muscle mitochondria, presumably containing both UCP2 and UCP3, exhibited only a little bit higher content of 3 H-GTPbinding sites than liver mitochondria. However, this estimate fits well with the protein measurements using antibodies (19). In conclusion, the maximum UCP2 content found in lung mitochondria (~200 pmol/ mg protein, Fig. 7) is about four times less abundant than the content of UCP1 in BAT by guest on March 24, 2020 http://www.jbc.org/ mitochondria (53). In skeletal muscle, the UCP2 + UCP3 content derived from the total number of 3 H-GTP-binding sites was ~7 times smaller than the UCP2 in the lung, and in the liver, UCP2 is even less abundant. The finding of small amounts of UCP2 in isolated mitochondria supports the concept of UCP2 (UCP3) as a stress protein, which can be elevated as required under various physiological situations (2,5).
It would seem to be trivial to note that no superoxide activation was required in our experi-   Table I.  Table I.    The derived number of binding sites corresponded to 182 ± 18 pmol/mg protein, n = 4, in lung; to 74 ± 22 (n = 3), 28 ± 6 (n = 2), and 21 ± 4 pmol/mg protein (n = 5), in kidney, skeletal muscle, and liver mitochondria, respectively. ± Errors of x-intercepts were taken from linear regression fits.