Natural and Azido Fatty Acids Inhibit Phosphate Transport and Activate Fatty Acid Anion Uniport Mediated by the Mitochondrial Phosphate Carrier*

The electroneutral Pi uptake via the phosphate carrier (PIC) in rat liver and heart mitochondria is inhibited by fatty acids (FAs), by 12-(4-azido-2-nitrophenylamino)dodecanoic acid (AzDA) and heptylbenzoic acid (∼1 μm doses) and by lauric, palmitic, or 12-azidododecanoic acids (∼0.1 mmdoses). In turn, reconstituted E. coli-expressed yeast PIC mediated anionic FA uniport with a similar pattern leading to FA cycling and H+ uniport. The kinetics of Pi/Pi exchange on recombinant PIC in the presence of AzDA better corresponded to a competitive inhibition mechanism. Methanephosphonate was identified as a new PIC substrate. Decanephosphonate, butanephosphonate, 4-nitrophenylphosphate, and other Pi analogs were not translocated and did not inhibit Pi transport. However, methylenediphosphonate and iminodi(methylenephosphonate) inhibited both electroneutral Pi uptake and FA cycling via PIC. AzDA analog 16-(4-azido-2-nitrophenylamino)-[3H4]-hexadecanoic acid (3H-AzHA) bound upon photoactivation to several mitochondrial proteins, including the 30- and 34-kDa bands. The latter was ascribed to PIC due to its specific elution pattern on Blue Sepharose and Affi-Gel. 3H-AzHA photolabeling of recombinant PIC was prevented by methanephosphonate and diphosphonates and after premodification with 4-azido-2-nitrophenylphosphate. Hence, the demonstrated PIC interaction with monovalent long-chain FA anions, but with divalent phosphonates of short chain only, indicates a pattern distinct from that valid for the mitochondrial uncoupling protein-1.

The mitochondrial phosphate carrier (PIC) 1 (1)(2)(3)(4)(5)(6)(7)(8)(9)(10) belongs to the well established gene family of homologous mitochondrial anion carrier proteins (10 -12). It was thought to mediate a stoichiometric P i *H ϩ symport (7,13), but alternatively, a P i / OH Ϫ antiport has been proposed to be the most plausible mode (13,14). In addition to monovalent phosphate (15), arsenate and divalent monofluorophosphate (16) are also translocated. With the human genome available, a major effort will be made in the coming years to ascribe particular phenotypes to the revealed genes. Some proteins possess several functions; consequently, a complete spectrum of functions for a given protein should be known. It is not uncommon that a carrier fulfills several functions. For example, the ADP/ATP carrier is thought to participate in the so called mitochondrial permeability transition (17), which is also activated by fatty acids (18). Although they share a similar trans-membrane folding (10 -12), MACPs exert diverse functions and conduct anions of different charge and by different modes (symport, antiport, uniport). They all contain positively charged and membraneembedded arginines or lysines located on trans-membrane ␣-helices (11,12), which may contribute to the putative anion binding sites for fatty acid anions and other hydrophobic anions (12,19). Interaction of some MACPs with fatty acids seems to represent another common feature, probably their second phenotype. Consequently, probing carriers with artificial hydrophobic substrate analogs might reveal some new structure/ function relationships.
PIC has also been found to be partially inhibited by fatty acids; however, it has been interpreted in terms of a surface charge effect (31). Besides FAs, the only reported hydrophobic substrate analog was 4-azido-2-nitrophenylphosphate, which inhibited P i transport by PIC only upon photoreaction when it was covalently attached to PIC after UV irradiation (32). Hence, the specificity of FA interaction with PIC remains to be elucidated, mainly to determine whether FAs interfere with a putative H ϩ (OH Ϫ ) translocation pathway or with a P i binding site or translocation pathway in the PIC structure. It is not known whether an H ϩ (OH Ϫ ) translocation pathway is identical or overlapping with the P i pathway. A possibility that H ϩ flux concomitant to P i flux during the physiological electroneutral P i transport on PIC would be ensured by the FA cycling mechanism, as in UCP1 (23), is unlikely, since in this hypothetical case, when PIC would act as a phosphate/fatty acid antiporter, the P i transport should be activated by FAs.
In this paper, we studied both inhibition of PIC by fatty acids and other hydrophobic anions, namely hydrophobic phosphate analogs, as well as fatty acid cycling mediated by PIC. Our results demonstrate that the pattern of PIC interaction with hydrophobic anions is distinct from that of UCP1 (33) and other mitochondrial uncoupling proteins.
Chemical Syntheses-For 12-azidododecanoic acid, 100 mg of 12bromododecanoic acid methylester (0.36 mmol) and 116 mg of NaN 3 (1.8 mmol) were stirred in 2 ml of dry N,N-dimethylformamide at 20°C for 12 h. The reaction mixture was diluted with 6 ml of water and extracted with ether. The organic layer was dried and evaporated to an oil, which was hydrolyzed by stirring with 0.3 ml of 1 N methanolic KOH (20°C for 12 h) and subsequently diluted with water and acidified with 5 N HCl, and 12-azidododecanoic acid was extracted several times with ether. The organic fraction was washed with water, dried, and evaporated to a colorless oil (81 mg) that crystallized in a refrigerator. The product exhibited typical IR (KBr) spectral peaks at 2097 cm Ϫ1 (N 3 ) and 1710 cm Ϫ1 (COOH). When esterified with diazomethane and analyzed on a gas chromatograph with mass detector, the chemical purity found was 96%, MS: 228 (M ϩ 1 Ϫ 28).
Biological Material-Rat liver mitochondria were isolated from Wistar rats in 250 mM sucrose, 10 mM Tris-MOPS, 0.1 mM Tris-EGTA, pH 7.4, containing 0.5% BSA. BSA was omitted in the final washing. Mitochondria from trypsinized rat heart in 180 mM KCl, 5 mM Tris-Cl, 10 mM Tris-EDTA, pH 7.4, were isolated by differential centrifugation in 180 mM KCl, 5 mM Tris-Cl, pH 7.4, containing 0.5% BSA (36). Prior to swelling measurement, the KCl content was minimized by two washings in a sucrose medium.
The yeast phosphate carrier was expressed in Escherichia coli as described elsewhere (8). The aliquots of inclusion bodies containing about 3 mg of protein were suspended and washed two times in 10 mM Tris-Cl, 0.1 mM Tris-EDTA, pH 7.0. The washed pellet was presolubilized by 1.5 ml of 5 mM TEA-TES, 30 mM TEA 2 SO 4 , 0.1 mM Tris-EDTA, pH 7.2, containing 0.3% sodium lauroylsarcosinate. After centrifugation at 14,000 ϫ g for 2 min, the resulting pellet was solubilized in 0.75 ml of 5 mM TEA-TES, 30 mM TEA 2 SO 4 , 0.1 mM Tris-EDTA, pH 7.2 containing 1.67% sodium lauroylsarcosinate and 1% octylpentaoxyethylene.
Phosphate Transport in Mitochondria-Anion transport in mitochondria was indicated by osmotic swelling while detecting light scattering (LS) at 530 -550 nm as an apparent absorbance on a diode-array spectrophotometer (Spectronics 3000). LS intensity, reflecting the inverse volume, allows measurement at a protein concentration as low as 0.2 mg/ml in an optimum 40% isotonic medium osmolarity (270 mosmol as 100%). 44 mM KP i , 54 mM salts of monovalent anions, 36 mM salts of divalent anions, and 27 mM salts of trivalent anions were employed, all roughly corresponding to 108 mosmol at full ionization. The changes in the normalized reciprocal absorbance ␤ are directly proportional to the transport rates in min Ϫ1 (37) as follows, where P is protein concentration, (P 1 ϭ 1 mg/ml), and ␣ is a machine constant, 0.1163 for the Spectronics 3000.
Phosphate Transport in Proteoliposomes-The P i /P i antiport mediated by the reconstituted recombinant yeast PIC was determined by measuring [ 33 P]phosphate flux using a forward exchange procedure (8,9). Kinetics was evaluated by variations of the external [P i ], while 30 mM [P i ] was kept constant in the vesicle interior. Proteoliposomes prepared in a medium containing 30 mM [P i ] were washed on a Sephadex G25-300 soaked with the desired external [P i ], while PIC was blocked by mersalyl (0.3 mM). Transport was then initiated by the addition of 50 mM dithiothreitol. An inhibitor-stop assay has been employed, so that 64 mM pyridoxal phosphate was added after a given time. The sample was then passed through the Dowex column (1-X10, Cl Ϫ form) to remove the external label. Forward exchange rates were determined by fitting the time course of isotope equilibration to a single exponential, Y ϭ A(1 Ϫ e Ϫkt ) ϩ B, leading to the first order rate constant k (min Ϫ1 ). A specific activity in mol⅐min Ϫ1 (mg of protein) Ϫ1 was calculated using k, the known external [P i ] (the total P i amount in the sample volume) and the protein amount in the sample in mg (usually 4 g of PIC protein, which equals a lipid/protein ratio of 400 or a molar lipid/PIC monomer ratio of 18,100. Reconstitution was performed by 14 cycles of detergent removal on a single column as described elsewhere (8,9). Briefly, 94 l of PIC (e.g. 0.596 mg/ml) solubilized from inclusion bodies, 224 l of preformed liposomes (22.4 mg of lipid) from E. coli lecithin (Avanti Polar Inc.), and 140 l of Triton X114 were vortexed and supplied by stock solutions to obtain 1.4 ml of suspension containing PIC (e.g. 56 g), 30 mM KP i , 50 mM K-HEPES, pH 6.8. It was passed 14 times over a column filled with 0.6 g of Bio-Beads-SM2 (Bio-Rad).
Proton Uniport Induced by Fatty Acids in Proteoliposomes-PIC reconstitution was performed by the detergent removal method designed for fluorescent H ϩ transport monitoring by SPQ quenching on a Shimadzu fluorometer, RF5301 PC, as described elsewhere (20). To assay H ϩ fluxes, monitored as ␦H ϩ , i.e. changes in internal [H ϩ ] relative to the initial state, FAs were added to proteoliposomes containing recombinant yeast PIC, and after their redistribution to both sides of the membrane apparent as interior acidification, 1 M valinomycin was added to initiate H ϩ efflux. Measured fluorescence traces were converted into "H ϩ traces" (20), from which the derived rates in mM⅐s Ϫ1 (mg of lipid) Ϫ1 were multiplied by the internal vesicle volume (estimated from SPQ volume distribution) and divided by a surface of 1 mg of liposomes (in m 2 ) so that an H ϩ flux density per m 2 was obtained in pmol of H ϩ ⅐s Ϫ1 ⅐m Ϫ2 .
Photoaffinity Labeling-Mitochondrial photomodifications with AzDA and photoaffinity labeling with 3 H-AzHA were performed using protocols described for brown adipose tissue mitochondria (25). Rat liver or rat heart mitochondria (3 mg of protein) were resuspended in 10 ml of BSA-free isolation medium and shaken for 10 min in an ice bath. 3 H-AzHA (or nonradioactive AzDA) was added to reach a final concentration of 0.46 M (1.5 nmol/mg of protein), and the mixture was shaken in an ice bath, first in darkness for 10 min and then for 10 min under UV illumination with a 400-watt xenon arc lamp equipped with a WG 8 filter (Schott Glass, Germany) transmitting light above 270 nm. Labeled mitochondria were washed by three centrifugations with BSA and three without BSA at 8500 ϫ g. The last pellet was resuspended in 100 l of the sucrose medium.
Isolation of Rat Heart Phosphate Carrier-Chromatography on hydroxylapatite was conducted using stepwise fractionation on spin columns. Lower loads of mitochondrial octylpentaoxyethylene or Triton X-100 extract (45 mg of protein per 3 g of dry HTP) gave higher yields of PIC and AAC content. The labeled mitochondria were either applied to the HTP column, or proteins contained in the HTP pass-through (38) were photolabeled in the detergent micelles with 3 H-AzHA, using a protocol developed for UCP1 (24,25). Labeled proteins were further fractionated on blue Sepharose (2 mg/ml of column) using a modified procedure of Rojo and Wallimann (39). BS was prewashed, and the first four elutions (1 ml each) were also performed with 150 mM Na 2 SO 4 , 20 mM Na-HEPES, 0.2 mM Tris-EDTA, pH 7.0, containing 0.5% detergent. Thus, a "flow-through" and four "wash" fractions were obtained. A stepwise NaCl gradient (0.9, 1.5, 2, and 2.7 M) in 20 mM Na-HEPES, 0.2 mM Tris-EDTA, pH 7.0, with 0.5% detergent was subsequently applied, while the final elutions were performed four times with 1 ml of 150 mM NaCl containing 0.5% SDS. Since BS retained mostly AAC, other proteins, including PIC, were contained predominantly in the flow-through fraction. Proteins of the HTP or BS flow-through fractions were further separated using organomercurial affinity chromatography on Affi-Gel 501 (Bio-Rad) soaked in the same medium as BS. After loading samples, five elutions (0.5 ml each) were conducted with the BS medium and two subsequent series of five elutions by the BS medium with 1.5 and 30 mM mercaptoethanol, respectively. The Affi-Gel 501 was reported to retain the PIC (40), and hence elutions with 1.5 mM mercaptoethanol should yield PIC, since the combination of BS and Affi-Gel excludes most proteins of the HTP eluate.
Laemmli SDS-PAGE was conducted, either on a Mighty Small II apparatus (Hoefer, minigels), or on a Protean IIxi (Bio-Rad, 15-cm gels). In both cases, high resolution 17% acrylamide gels were cast with an acrylamide/bisacrylamide ratio of 150:1, allowing for an expansion of the 30 -40-kDa region (38). Parallel gels to those for autoradiography were silver-stained using the bichromate method. Coomassie Bluestained gels were treated with an autoradiography enhancer ENTEN-SIFY and dried between plastic follies and a filter paper under vacuum. Dried gels with peeled-out follies were exposed in the steel cassettes on Kodak Scientific Imaging Film X-Omat, AR-5, for 5-10 days (mitochondria) or 10 -21 days at Ϫ70°C. Films were developed by Dektol (Kodak) for 25 min and fixed in Kodak fixer for 5 min, both at 21°C. Percentages of the remaining 3 H-AzHA label were quantified from the band density compared with controls using one-dimensional image analysis software (EDAS 40, Digital Science, Kodak).

Inhibition of Phosphate Transport by Natural and Azido
Fatty Acids-Rapid electroneutral phosphate uptake was induced in rat liver (Fig. 1a) or rat heart (Fig. 1b) mitochondria as a passive swelling initiated by nigericin in 44 mM potassium phosphate, pH 7.4. Participation of PIC is indicated by electroneutrality and by the specific inhibitory pattern (1-6) when both NEM and mersalyl are inhibiting (Fig. 1a). FAs such as lauric (Fig. 2a) and palmitic acid only inhibited P i uptake in high concentrations. Lauric acid exhibited a K i of 250 M (Fig.  2a). Contrary to a rather weak inhibition by natural FAs, 10 M AzDA inhibited P i uptake by more than 90% (Fig. 1a), and the estimated K i was 3.8 M (Fig. 2a). A substantial inhibitory strength was exhibited by heptylbenzoic acid (K i of 89 M; Fig.   2b). 12-Azidododecanoic acid was inhibiting with a lower potency (K i of 310 M; Fig. 2b). The protein-independent swelling (induced by nigericin in potassium acetate) was only affected by 10% with 10 M AzDA, suggesting that AzDA produces no major nonselective side effect on the inner mitochondrial membrane. AzDA, even at 50 M doses, did not inhibit swelling in sodium acetate, which reflects the electroneutral Na ϩ /H ϩ antiport (Fig. 2a). No inhibition of the latter by any FA tested was found, nor inhibition of pyruvate carrier (41), thus demonstrating specificity for PIC.
Interestingly, the FA derivatives, which were previously found (42,43) to be unable to flip-flop across the lipid bilayer, such as 12-hydroxylauric (Fig. 2b), phenylvaleric, and dodecanedioic acid, did not affect P i transport. Since the other FAs tested, including 12-azidododecanoic acid, 2 were confirmed to possess the ability of fast flip-flop, we suggest that it is a specific (U-shape) conformation in the membrane (42) that prevents these inactive FAs from interacting with PIC. Photo-2 P. Jezek, unpublished data. activated AzDA also inhibited P i uptake in mitochondria with an apparent IC 50 of 4 M. Since the mitochondria were first irradiated in the presence of AzDA, washed with BSA, and then reisolated, the apparent IC 50 is not directly comparable. Coincidentally, however, it is the same as the above reported K i value; IC 50 of 4 M was obtained when mitochondria preincubated first with AzDA in the dark were washed with BSA and reisolated. Consequently, UV illumination does not seem to strengthen the AzDA inhibition. AzDA binding is apparently so tight that no effective washing can eliminate it.
Kinetics of Fatty Acid Inhibition of Phosphate Transport-To establish the type of inhibition by fatty acids, kinetic measurements with the reconstituted E. coli-expressed yeast PIC have been performed. P i /P i exchange was measured as a forward 33 P uptake into proteoliposomes containing recombinant PIC, while varying external P i between 1 and 30 mM. Kinetics of such P i /P i exchange determined in the absence or presence of 100 M AzDA was found to agree better with a competitive mechanism (Fig. 3). Although data cannot definitively distinguish between the competitive and noncompetitive type of kinetics, fits of the competitive model gave better agreement. The derived V max in control was 0.55 mol of P i ⅐min Ϫ1 ⅐(mg of protein) Ϫ1 , K m was 6.5 mM, and derived K i from the data measured with 100 M AzDA was 99.5 M. When the direct plots V versus [P i ] were fitted by nonlinear regression to the Michaelis-Menten equation, the derived V max values in control and with 100 M AzDA were 0.74 and 0.77 mol of P i ⅐min Ϫ1 ⅐(mg of protein) Ϫ1 , respectively, and K m was 9 mM in control.
Screening of Possible Hydrophobic Substrates of Mitochondrial Phosphate Carrier-Interaction of PIC with FAs suggested that PIC could also interact with some other amphiphilic anions. Therefore, we evaluated whether hydrophobic phosphate analogs are transported by PIC or act as competitive inhibitors. Contrary to the alkylsulfonate translocation by UCP1 (33), which is faster with the increasing chain length, decanephosphonate and butanephosphonate were not transported at a significant rate by PIC, and neither inhibited P i transport up to a 0.7 and 100 mM dose, respectively. We found that only phosphonate with the shortest chain, methylphosphonate, is the PIC substrate. Methylphosphonate exhibited the same transport characteristics as P i transport, including NEM sensitivity (Fig. 4) and inhibition by lauric acid, 12-azidododecanoic acid, and AzDA (Fig. 4).
We also attempted to evaluate whether some other amphiphilic compounds are transported by PIC or inhibit P i transport in mitochondria. While screening various mono-, di-, and trialkylsulfonates and benzene mono-, di-, and trisulfonates or anions derived from phosphate and phosphonate, such as phosphoformate, phosphopyruvate, phosphogluconate, and 4-nitrophenylphosphate, we found no such case. Particularly, we confirmed that phosphoformate does not inhibit the net electroneutral P i uptake, as reported previously (8). On the contrary, methylenediphosphonate and iminodi(methylenephosphonate) were found to be strong inhibitors ( Fig. 5a and b).
FIG. 2. Dose responses for FA inhibition of phosphate uptake in rat liver mitochondria. a, inhibition by lauric acid (OE) and AzDA (q); b, inhibition by heptylbenzoic (f) and 12-azidolauric (ƒ) acids versus no effect of 12-hydroxylauric acid (E). All FAs were added directly to the assay medium. a, the inhibitory dose responses are compared with the data illustrating no effect of AzDA on the swelling of rat liver mitochondria in sodium acetate, monitoring the function of the Na ϩ /H ϩ antiporter (Ⅺ). Theoretical fits (solid lines) to the Hill equation are also shown (for which the K i values of 3.8 M for AzDA, 250 M lauric acid, 89 M heptylbenzoic acid, and 310 M 12-azidolauric acid and the Hill coefficients (n H ) of 0.93, 1.07, 1.02, and 1.00, respectively, were derived on the assumption of 100% inhibition at infinite concentration). Measurements were performed as described in the legend to Fig. 1 with the exception of Na ϩ /H ϩ antiporter testing, for which 54 mM sodium acetate, 5 mM Tris-MOPS, pH 7.2, containing 2 M rotenone and 0.25 g/ml antimycin, was used. The K i values derived for their inhibition were 4.9 and 5.2 mM, respectively. However, they were not transported by PIC. Also, although 4-azido-2-nitrophenylphosphate (AzNPP i ) was previously found to inhibit swelling in NH 4 -P i only when photoactivated (32), in our study it inhibited the nigericin-induced P i transport via PIC in the dark with a K i of 1.5 mM (Fig. 5a). Its analog, lacking the azido group, 4-nitrophenylphosphate, did not exhibit an inhibitory effect.
Possible Fatty Acid Cycling Mediated by Phosphate Carrier-Although inhibitory to P i transport, FAs were previously found to induce H ϩ uniport in proteoliposomes containing PIC, sensitive to diphosphonates (20). As for AAC and UCPs, it has been interpreted in terms of FA cycling (23,28), in which uniport of anionic FA is mediated by PIC and the neutral FA diffuses back across the lipid bilayer, thus carrying H ϩ . Fig. 6a illustrates the dose responses for lauric acid-induced H ϩ uniport (lauric acid cycling) in proteoliposomes with the reconstituted recombinant PIC in the absence or presence of 10 mM methylenediphosphonate (MDPh). The difference between them gives a net H ϩ flux density sensitive to MDPh, hence a portion of transport that can be ascribed to PIC. It follows a Michaelis-Menten kinetics (Fig. 6b). The FA cycling via UCP1 was found 2 to be insensitive to MDPh. The residual H ϩ uniport is very similar to the one observed in protein-free liposomes (20). Other FAs tested gave similar kinetics ( Fig. 6c and d), with V max (derived from MDPh-sensitive fluxes, in nmol of H ϩ ⅐s Ϫ1 ⅐ (mg of lipid) Ϫ1 ) highest for oleic (1.55) and decreasing for heptylbenzoic (1.49), myristic (1.35), lauric (0.96; Fig. 6b) and 12-azidolauric (0.55) acids. Note that FAs that are unable to flip-flop, such as 12-hydroxylauric acid, gave a background H ϩ flux of 0.17 nmol of H ϩ ⅐s Ϫ1 ⅐(mg of lipid) Ϫ1 that was identical to the background H ϩ flux in the absence of FAs (Fig. 6c). The apparent affinity to PIC (inverse K m ) decreased in nearly the same order as

H-AzHA Labeling of Phosphate Carrier in Mitochondria-
Using a rather small amount (1.5 nmol/mg of protein) of highly tritiated azido-FA ( 3 H-AzHA) incubated with rat heart or rat liver mitochondria and subsequently illuminated by UV light, only a small portion of numerous mitochondrial proteins were labeled with 3 H-AzHA (Fig. 7). This is very similar to the photolabeling of brown adipose tissue mitochondria, which yielded UCP1 as the major labeled band (25). Among the labeled proteins, the most apparent were 30-and 34-kDa bands as illustrated by a typical autoradiogram (Fig. 7). These two most likely correspond to AAC (22) and PIC, respectively. Our further steps led to the verification of this assumption.
3 H-AzHA Labeling of Partially Purified Phosphate Carrier-To confirm that PIC is the protein interacting with 3 H-AzHA, we chose to separate proteins of the hydroxylapatite eluate by two subsequent affinity chromatography steps, selective enough to determine the final product as PIC. We solubilized rat heart and liver mitochondria by octylpentaoxyethylene or Triton X-100, passed the extracts through HTP, and attempted to photolabel containing proteins. As a result, the 30-and 34-kDa bands were again photolabeled with 3 H-AzHA (Fig. 8a). In a further experiment, just the 34-kDa band but not the 30-kDa band was found in the flow-through fraction of the Cibacron blue affinity agarose column (BS column) and still retained the 3 H-AzHA label attached (Fig. 8a). In turn, AAC (30-kDa monomer) was tightly bound to the BS column and could be eluted as reported before (39), either at higher NaCl concentrations or with SDS (Fig. 8a). Also, the 30-kDa band in these fractions retained the 3 H-AzHA label. It was identified as AAC by Western blots (22). The "upper" bands were found only in minute amounts in the intermediate NaCl fractions of the BS column.
When we further fractionated the BS flow-through fraction on Affi-Gel 501 (an organo-mercurial affinity matrix) and eluted first with a medium containing low and then high mercaptoethanol concentration, the intermediate fractions yielded only a 34-kDa band, which still retained the 3 H-AzHA label (Fig. 9, sample 1). Also, when the HTP pass-through containing the labeled proteins was applied on the Affi-Gel, the resulting flow-through contained most of the typical HTP bands, among which only AAC was labeled with 3 H-AzHA (not shown). Hence, one may expect that the protein retained in the Affi-Gel was predominantly the PIC protein (40). Identical results were obtained when rat liver or heart mitochondria was first labeled with 3 H-AzHA and then fractionated on HTP, BS column (Fig.  8b), and Affi-Gel (Fig. 9).
Prevention of 3 H-AzHA Photolabeling by 4-Azido-2-nitrophenylphosphate, Methanephosphonate, and Diphosphonates-To evaluate whether the P i binding domain of PIC is close or overlaps with the putative FA binding site, we performed competition studies and tested which phosphate analogs would prevent the 3 H-AzHA photoaffinity labeling of recombinant yeast PIC. Percentages of the remaining 3 H-AzHA label were quantified from the band density compared with controls. There was no change in the 3 H-AzHA photolabeling when preincubations with phosphonoformate or undecanesulfonate were performed (Fig. 10). Pyrophosphate and 4-nitrophenyl-phosphate partly prevented the 3 H-AzHA label. Photolabeling was prevented by preincubations with the newly identified substrate, methanephosphonate (10 mM), and was completely prevented by 10 mM methylenediphosphonate and iminodi-(methylenephosphonate) (Fig. 10). The amount of the 3 H-AzHA label on yeast PIC also decreased after the preceding photolabeling with 4-azido-2-nitrophenylphosphate (Fig. 10).  Fig. 8a) or as isolated rat heart mitochondria (2) (sample of Fig. 8b). Both samples were fractionated on three columns: first on hydroxylapatite, its pass-through fraction (1.5 mg of protein) was loaded onto a blue Sepharose column, and the resulting flow-through was loaded onto the Affi-Gel 501 column, which was washed and then eluted with 1.5 mM mercaptoethanol. Shown are autoradiograms (A) and the corresponding PAGE gels (P) run in parallel. Positions of molecular mass (M) standards are indicated.

FIG. 8. Phosphate carrier and ADP/ ATP carrier as two proteins photolabeled by 3 H-AzHA.
Shown are autoradiograms (A) and the corresponding PAGE gels (P) run in parallel. Either partially purified proteins of the HTP passthrough (a) or the rat heart mitochondria (b) were photolabeled. a, 3 H-AzHA was preincubated with the HTP pass-through fraction, prepared from Triton X-100-extracted rat heart mitochondria, and the sample was irradiated by UV light and passed over Sephadex G25-300. The labeled sample (1.5 mg of protein) was further fractionated on blue Sepharose. b, labeling of rat heart mitochondria was performed first (cf. Fig. 7), and subsequently separations on HTP and blue Sepharose columns were conducted.

DISCUSSION
We have demonstrated the ability of fatty acids to inhibit (strongly for azidonitrophenyl-FAs) the mitochondrial phosphate carrier (PIC) and to function as its monovalent anionic substrates in a FA cycling process. Furthermore, photolabeling of PIC with azidonitrophenyl-FAs has also been performed. In this way, we have tried to characterize the corresponding FA binding site and its relationship to the P i binding site. We have found that these two sites reside close to each other (or might overlap), since the inhibition of P i /P i exchange by FAs is most likely of a competitive type, and covalently bound AzNPP i , the newly identified substrate methanephosphonate, and the inhibitory diphosphonates did prevent 3 H-AzHA photolabeling.
A question now arises concerning the nature of FA interaction with PIC. An identity of the FA binding site and the P i binding site could be excluded by the lack of PIC interaction with long-chain alkylphosphonates. On the contrary, a proximity of these two sites may be the origin for the observed competitive inhibition of P i /P i exchange by FAs, for prevention of 3 H-AzHA binding by some P i derivatives, and for the correlation between K i values of FAs inhibiting P i uptake and the apparent affinity of cycling FAs. We have intentionally selected FAs covering a range of affinities or K i values. FA cycling occurred at higher FA concentrations as suggested by higher K m values (ϳ100 M) when compared with UCP1 (ϳ10 M; Ref. 23). The presence of a 4-azido(2-nitrophenyl) group enhanced the inhibitory ability of FAs, but it is not exclusively this portion of AzDA that mediates the inhibitory effect, since the closest analog, -azido-dodecanoic acid, and natural FAs inhibit as well. The enhancement of the inhibitory strength could originate from the interaction within the P i binding site (anion pathway) of PIC, since AzNPP i , attached by photoreaction to the carrier, does interfere with 3 H-AzHA photolabeling. This fact again confirms the proximity of the P i and FA binding sites.
Our findings extend the previous report of Wojtczak and Załuska (31), who interpreted the observed FA inhibition of P i transport in rat liver mitochondria as a surface charge effect.
Our data indicate a rather specific effect of FAs on P i transport via PIC. A surface charge should also probably inhibit the H ϩ efflux coupled to the Na ϩ uptake by the Na ϩ /H ϩ antiporter or the pyruvate*H ϩ symport via the pyruvate carrier. None of these effects was observed. Also, several inactive FA derivatives (42,43), such as 12-hydroxydodecanoic acid, if causing the surface charge effect, should inhibit the PIC as well. Again this was not observed. Finally, butyl-and decylphosphonate should cause even higher inhibition due to a surface charge, but they did not, even at millimolar concentrations.
Concerning the specificity of FA interaction, it is interesting to ask whether the demonstrated photolabeling of PIC with 3 H-AzHA or FA cycling resulted from the existence of a specific, preformed, FA binding site or whether it instead reflects the overall carrier hydrophobicity and positive charges inside the membrane. This calls for further studies. Nevertheless, prior to being combusted, natural FAs will be present in the membrane due to their high partition coefficient and may potentially inhibit PIC and may cycle via PIC in either of the two cases described above. Consequently, FAs should be considered as important regulators of oxidative phosphorylation (29), since they affect PIC (Ref. 20 and this work) and AAC (18,19,21,22), besides the other mitochondrial carriers (30) and uncoupling proteins (19,(23)(24)(25)(26)(27).
We have also revealed for the first time that the "hydrophobic" P i analog, methanephosphonate, is a good PIC substrate. This is unrelated to the FA inhibition, since alkylphosphonates with a longer chain were found not to be transported. They did not inhibit P i uptake as well. These findings suggest that the putative hydrophobic part of P i binding site cannot accommodate larger alkylphosphonate analogs. However, we found that the P i binding site also interacts with P i analogs of medium size such as methylenediphosphonate, iminodi(methylenephosphonate), and AzNPP i , which inhibited P i uptake. This confirms the previous finding of inhibition by photoactivated AzNPP i (32). We found that AzNPP i together with diphosphonates belongs to nontransportable P i analogs. Thus, a specific gate of PIC does not allow sulfate and tungstate to pass (7), and, on the other hand, it is able to accommodate the methyl group and part of the phosphonomethyl and phosphonomethylimino groups, or a phenyl attached to the phosphate. Wohlrab et al. (7) hypothesized that some acid residues of yeast PIC, such as Glu 163 , Glu 164 , Glu 192 , and Glu 196 might ensure specificity for phosphate and exclude interaction with sulfate. Interestingly, glutamate Glu 190 on UCP1 conveys the pH dependence of nucleotide di-and triphosphates binding to this protein (44). Phosphonoformate, previously reported to inhibit P i /P i exchange (8), seems not to fit in the revealed pattern, since it is not transported and does not inhibit the net electroneutral P i transport. Perhaps the short formyl (carboxyl) group attached on phosphonate cannot cause the inhibition as methylenephosphonate does.
Diphosphonates that inhibit FA cycling enabled by PIC were also identified as the first known nontransported substrate analogs that inhibit the putative FA anion uniport. No such inhibition was found in the case of uncoupling proteins. UCP1or PUMP-mediated FA cycling is inhibited by undecanesulfonate, which is the translocated anionic substrate (19,(23)(24)(25)(26)(27)33). The existence of FA cycling via PIC indicates that this is a more general phenomenon not related exclusively to UCPs. However, it is a ligand-gated (purine nucleotide-gated) regulation of FA cycling, what is distinct for UCPs (23,27). It is known that nucleotides, as intermediate size ligands, interact with several membrane ␣-helices of UCP1 (45). Also, phosphate might interact with several trans-membrane ␣-helices of PIC, as shown by the site-directed mutagenesis (7), revealing five critical res- FIG. 10. Photoaffinity labeling of recombinant yeast phosphate carrier with 3 H-AzHA and its prevention by various phosphocompounds. Autoradiograms of PAGE-separated yeast PIC solubilized from inclusion bodies (equal amounts, three experiments, A, B, and C) as described under "Experimental Procedures" are shown. Samples were photolabeled with 3 H-AzHA (5.2 nmol/mg total protein, corresponding to a stoichiometry of 0.3 per dimer) in controls (C, 100%) or after preincubations with 10 mM methanephosphonate (C 1 Ph, remaining 37% of the label), 10 mM phosphonoformic acid (PFA, 99%), 50 mM sodium pyrophosphate (PP i , 44%), 1 mM sodium undecanesulfonate (C 11 SO 3 , 99%), 10 mM methylenediphosphonate (MDPh, 9%), 10 mM iminodi(methylenephosphonate) (IDPh, 12%), 10 mM 4-nitrophenylphosphate (NPP i , 40%), or after preceeding photolabeling with nonradioactive 10 mM 4-azido-2-nitrophenylphosphate (AzNPP i , 30%). For details of SDS-PAGE (17.5% acrylamide, its ratio to bisacrylamide 150:1) and autoradiography, see "Experimental Procedures." idues for yeast PIC spread along the whole sequence: His 32 and Asp 39 on the first; Glu 126 and Glu 137 on the second; and the Asp 236 on the third trans-membrane ␣-helix, respectively. Among them, His 32 , Glu 126 , and Glu 137 were proposed to form a putative proton cotransport pathway (7). However, due to functional reasons, we cannot conclude that the FA binding site is identical with the H ϩ /OH Ϫ binding site of PIC, as we did for UCP1, for which FAs were documented to enter into its anion binding site (24,25).
In conclusion, FAs interact with PIC in a hydrophobic binding site that lies in proximity to or overlaps the P i binding site, which might represent a slightly hydrophobic internal domain in PIC. When amphiphiles such as AzDA or native FAs interfere with the domain, the transport process is inhibited. Upon interaction with PIC, FAs might also reach the opposite side of the membrane, which leads to FA cycling and uncoupling. Both inhibitory and cycling effects could lead in vivo to a fine regulation of oxidative phosphorylation efficiency.