Interactions of cyclophilin with the mitochondrial inner membrane and regulation of the permeability transition pore, and cyclosporin A-sensitive channel.

Mammalian mitochondria possess an inner membrane channel, the permeability transition pore (MTP), which can be inhibited by nanomolar concentrations of cyclosporin (CS) A. The molecular basis for MTP inhibition by CSA remains unclear. Mitochondria also possess a matrix cyclophilin (CyP) with a unique N-terminal sequence (CyP-M). To test the hypothesis that it interacts with the MTP, we have studied the interactions of CyP-M with rat liver mitochondria by Western blotting with a specific antibody against its unique N terminus. Although sonication in isotonic sucrose at pH 7.4 refraction sediments with submitochondrial particles at 150,000 x g. We show that the interactions of this CyP-M pool with submitochondrial particles are disrupted (i) by the addition of CSA, which inhibits the pore, but not of CSH, which does not, and (ii) by acidic pH condition, which also leads to selective inhibition of the MTP; furthermore, we show that the effect of acidic pH on CyP-M fully prevents the inhibitory effect of H+ on the MTP (Nicolli, A., Petronilli, V., and Bernardi, P. (1993) Biochemistry 32, 4461-4465). These data suggest that CyP-M inhibition by CSA and protons may be due to unbinding of CyP-M from its putative binding site on the MTP. A role for CyP-M in MTP regulation is also supported by a study with a series of CSA derivatives with graded affinity for CyP. We show that with each derivative the isomerase activity of CyP-M purified to homogeneity is similar to that displayed at inhibition of MTP opening, CyP-M (but not CyP-A) and decreased efficiency at MTP inhibition is obtained by substitution in position 8 while a 4-substituted, nonimmunosuppressive derivative is a as effective as the native CSA molecule, indicating that calcineurin is not involved in MTP inhibition by CSA.

Mitochondria from a variety of sources possess a regulated inner membrane channel, the permeability transition pore (MTP). 1 Pore opening is dependent on both the transmembrane potential difference (1) and on matrix pH (2) and is modulated by a variety of effectors acting at multiple sites (for recent reviews, see Refs. 3 and 4). Among pore inhibitors, CSA stands out for its potency (apparent I 50 is in the submicromolar range) and for its selectivity (no other mitochondrial functions appear to be affected by this drug) (Refs. 3 and 4 and references therein). In fact, it is because of inhibition by CSA (5-7) that the nature of the "permeability transition" (a Ca 2ϩ -dependent increase of inner membrane permeability to solutes with molecular mass Յ 1500 Da) was recently recognized as being due to opening of a channel (8 -10), as first proposed by Hunter and Haworth in 1979 (11, 12), rather than to a permeability change of the membrane lipid phase (for review, see Ref. 13).
Inappropriate MTP opening is becoming increasingly recognized as a causative event in cell injury by a variety of conditions, including ischemia (14 -16). On the other hand, and although a role of the pore in cellular Ca 2ϩ homeostasis appears likely (17)(18)(19), most questions regarding its physiological function, regulation, and molecular structure await an answer.
A soluble mitochondrial extract displays PPiase activity, which can be inhibited by CSA and CSG but not by CSH or FK506 (20). Since this pattern is shared by the permeability transition, it has been suggested that a mitochondrial CyP mediates CSA inhibition of the pore (20). Although a 20-kDa CyP isoform (CyP-M) with a unique N-terminal sequence has later been isolated from mitochondria (21), its role in the permeability transition is unclear.
A study of radiolabeled CSA binding to mitochondria defined two classes of high affinity CSA-binding sites, suggesting the possible existence of further CS-binding proteins besides CyP-M (22). In keeping with this hypothesis, a 10-kDa protein could be labeled with a photoactive CSA derivative in a membrane-associated fraction obtained after sonication of mitochondria, and it was suggested that the labeled protein was part of the membrane CSA receptor on the pore (23).
The question of whether CSA interacts directly with the pore, or rather whether its inhibitory effects on the pore are mediated by CyP-M is a fundamental one. Besides the intrinsic importance for the mechanistic aspects of MTP function, assessing this point has a specific relevance for any strategy aimed at pore isolation. Indeed, if CyP-M is required for pore opening, any protocol designed for its reconstitution must include this protein.
In this study, we have isolated CSA-binding proteins from a subcellular fraction enriched in mitochondria by affinity chromatography on immobilized CSA. We show that CyP-M is the main and probably the only mitochondrial receptor for CSA and address the question of its role in MTP function with two approaches. In the first, we have studied the interactions of CyP-M with intact rat liver mitochondria by Western blotting with a specific antibody against its unique N terminus. We show that CyP-M is a matrix protein that is engaged in interactions with the mitochondrial inner membrane, and that these interactions are partially retained by SMP. Interactions of CyP-M with SMP are disrupted by the addition of CSA and by mildly acidic pH values, conditions that also lead to selective inhibition of the MTP. On the other hand, CyP-M interactions with the inner membrane are unaffected by CSH, which does not inhibit the MTP, while the effects of acidic pH can be prevented by DPC, which also prevents MTP block by H ϩ (2). These observations have been exploited to devise a novel protocol for CyP-M purification to homogeneity. In the second approach, we have then studied a series of CSA derivatives with graded affinity for CyP-A for their inhibition of CyP-M and of MTP. We find that with each derivative the potency at inhibition of the PPiase activity of purified CyP-M, relative to that of CSA, is similar to that displayed at inhibition of MTP opening. Structure-function analysis of CS derivatives reveals that decreased binding to CyP-M (but not CyP-A) and decreased efficiency at MTP inhibition is obtained by substitutions in position 8, while a 4-substituted, nonimmunosuppressive derivative, which does not bind calcineurin, is as effective as the native molecule at inhibition of the MTP and of CyP-M. Taken together, these data (i) suggest that CyP-M binding is involved in MTP function and that pore inhibition by CSA and H ϩ may be due to unbinding of CyP-M from its putative binding site on the MTP and (ii) show that inhibition of calcineurin activity is not required for the inhibitory effects of CSA on MTP.

MATERIALS AND METHODS
Preparation of Subcellular Fractions-Rat liver mitochondria were prepared by standard differential centrifugation (24). SMP were prepared as follows. The mitochondrial stock solution was diluted to about 30 mg ϫ ml Ϫ1 in 0.25 M sucrose, 10 mM Tris-HCl, pH 7.4, 0.1 mM EGTA-Tris. 2-ml aliquots were transferred to 6-ml glass vials (Packard Instruments Co., Meriden, CT), and subjected to three cycles of sonication of 3 min each at room temperature, with 15-min intervals on ice in a G112SP1T water bath Sonifier (Laboratory Supplies Co., New York, NY). The samples were centrifuged at 8,000 ϫ g for 10 min to remove unbroken mitochondria, and the resulting supernatants were centrifuged for 1 h at 150,000 ϫ g in the rotor 70 Ti of a Beckman Ultracentrifuge. The supernatants were decanted, the pellets and the tube walls were rinsed twice with 0.25 M sucrose, 10 mM Tris-HCl, pH 7.4, 0.1 mM EGTA-Tris, and the SMP pellets were resuspended in the desired buffer (1 ml of buffer/30 mg of starting mitochondrial protein), as detailed in the figure legends.
Affinity Chromatography-To identify mitochondrial CSA-binding proteins, isolated mitochondria (250 mg of protein) were extracted with 25 ml of 0.15 M NaCl, 10 mM NaP i , pH 7.4, 2% Triton X-100, 0.23 mM phenylmethylsulfonyl fluoride, 0.83 mM benzamidine for 30 min on ice. After removal of insoluble material by centrifugation at 120,000 ϫ g in the rotor 50.2 Ti of the Beckman ultracentrifuge, the extract was diluted with an equal volume of 0.15 M NaCl, 10 mM NaP i , pH 7.4, 0.23 mM phenylmethylsulfonyl fluoride, 0.83 mM benzamidine, and incubated overnight at 4°C with 0.5 ml of a D-Ala-(3-amino)-8-CSA Affi-Gel 10 matrix preequilibrated with 0.15 M NaCl, 10 mM NaP i , pH 7.4, 0.5% Triton X-100, 0.23 mM phenylmethylsulfonyl fluoride, 0.83 mM benzamidine (column buffer). The column was sequentially washed with 10 bed volumes of column buffer, 5 bed volumes of column buffer contain-ing 0.5 M NaCl, and 10 bed volumes of column buffer, followed by elution with 0.42 mM CSA in column buffer and 6 M urea in NaCl-free column buffer. 1-bed volume fractions were collected, and 15 l of relevant fractions (as indicated in the legend to Fig. 1) were analyzed by SDS-PAGE and silver staining.
Purification of CyP-M-CyP-M was purified from SMP isolated as above. Our purification strategy capitalized on the original finding that CyP-M binding to inner membranes can be disrupted by acidic pH values (see Fig. 4). SMP were resuspended in 0.25 M sucrose, 20 mM sodium acetate, pH 5.0, 0.1 mM EGTA-Tris at approximately 10 mg ϫ ml Ϫ1 and centrifuged for 1 h at 150,000 ϫ g in the rotor 70 Ti of a Beckman ultracentrifuge. The supernatants, containing essentially all of CyP-M, were adjusted to pH 8.2 with Tris base and further fractionated essentially as described by Connern and Halestrap (21). Briefly, the supernatants (about 20 ml) were applied to a 1 cm 2 ϫ 8 cm Q-Sepharose column (Pharmacia Biotech Inc.). The flow-through, containing all of CyP-M, was loaded onto a 1 cm 2 ϫ 2 cm S-Sepharose column (Pharmacia). After washing with 65 bed volumes of 10 mM Tris-HCl, pH 8.2, 2 mM EDTA-Tris, 0.5 mM DTT and then with the same buffer containing 20 mM NaCl, the S-Sepharose column (which under these conditions retains CyP-M) was finally eluted with 0.1 M NaCl, 10 mM Tris-HCl, pH 9.5, 2 mM EDTA-Tris, 0.5 mM DTT. The fractions containing CyP-M (assessed by Western blotting with a monospecific antibody, see Fig. 4) were pooled, dialyzed against 10 mM Tris-HCl, pH 8.2, 2 mM EDTA-Tris, 0.5 mM DTT, and then against 50 mM MES, pH 6.0, 2 mM EDTA, 0.5 mM DTT, and finally loaded onto an HR 5/5 mono-S HPLC column (Pharmacia) and eluted with a 0 -250 mM NaCl gradient. CyP-M eluted at approximately 0.1 M NaCl, and was analyzed for purity by both SDS-PAGE followed by Coomassie Blue or silver staining and by Western blotting (see Fig. 4).
Antibody Preparation-For preparation of a monospecific antibody, CyP-M was purified by affinity chromatography exactly as described above and concentrated by precipitation with trichloroacetic acid as described previously (25). The 20-kDa CSA-eluted protein of interest was transferred to an Applied Biosystems (Foster City, CA) ProBlott nylon membrane as described under "Analytical Techniques" below, stained with Coomassie Blue according to the manufacturer's instructions, dried, and subjected to amino acid sequence analysis with an Applied Biosystems 477A Sequenator equipped with on-line phenylthiohydantoin aminoacid analyzer (model 120A). The N-terminal sequence, determined by Dr. Patrizia Polverino De Laureto of the Centro di Ricerca Interdepartimentale per le Biotechnologie Innovative Biotechnology Center of the University of Padova, was ASDGGARGAN-SSSQNPLV. A synthetic peptide with sequence ASDGGARGANS-SSQC, and an irrelevant peptide with sequence KVEKIG/EGTYGV-VYK were prepared with an Applied Biosystems 431-A automatic peptide synthesizer and validated by sequencing by Dr. Oriano Marin, CRIBI Biotechnology Center and Department of Biochemistry, University of Padova. After conjugation of the CyP-M peptide to maleimideactivated keyhole limpet hemocyanin (Pierce), the peptide conjugate was injected subcutaneously in New Zealand rabbits (1 mg/rabbit at multiple sites in complete Freund's adjuvant). After 3 weeks, and then at three 4-week intervals, the rabbits were boosted with 1 mg of peptide conjugate in incomplete Freund's adjuvant, and the appearance of specific antibodies monitored by Western blotting of either affinity-purified CyP-M or of an ASDGGARGANSSSQC peptide conjugated to bovine serum albumin.
Analytical Techniques-SDS-PAGE was performed according to Laemmli (26) in acrylamide-bisacrylamide slab gels, which were stained with either Coomassie Blue or silver as specified in the figure legends, or transferred to nitrocellulose (0.22-m pore size, Hoefer, San Francisco, CA) in CAPS-NaOH, pH 11.0, 10% MetOH (overnight at 4°C, 2 mA/cm 2 ). In the experiment of Fig. 2, the nitrocellulose sheet was stained with Ponceau Red, photographed, destained with dH 2 0, and subjected to Western blotting with the monospecific antibody against the N-terminal sequence of CyP-M. In the experiments of Figs. 3, A and B, and 4, two parallel gels were run, one being fixed and stained and the other transferred to nitrocellulose and subjected to Western blotting as above. The reaction was detected with a secondary alkaline phosphatase-conjugated goat anti-rabbit antibody (Sigma).
PPiase activity was determined with the spectrophotometric method of Fischer et al. (27) and analyzed according to Harrison and Stein (28). The concentration of CyP-M or of recombinant human CyP-A (a generous gift of Dr. Mauro Zurini, Sandoz Pharma AG) was adjusted to give an observed first-order rate constant for the catalyzed reaction about 5-fold larger than that of the uncatalyzed reaction.
MTP opening was followed as the rate of absorbance or 90°C light scattering change at 540 nm in 0.20 M sucrose, 10 mM Tris-MOPS, pH 7.4, 5 mM succinate-Tris, 1 mM P i , 20 M EGTA-Tris, and 2 M rotenone. Mitochondria (0.5 mg ϫ ml Ϫ1 ) were incubated in thermostatted, magnetically stirred cuvettes (final volume, 2 ml, 25°C). After accumulation of 40 M Ca 2ϩ , 0.5 mM EGTA-Tris was added to prevent opening of the Ca 2ϩ channel, and MTP opening was then triggered by the addition of 0.2 M carbonyl cyanide p-trifluoromethoxyphenyl hydrazone (see Ref. 29 for further details and representative traces).
Liposomes reconstituted with mitochondrial proteins were prepared, and their permeability to solutes was measured as described in Ref. 30, and measurements of SMP membrane potential with oxonol VI (Molecular Probes, Eugene, OR) were performed as described previously (31).
CSA Derivatives and Other Reagents-D-MeSer-3-CS was synthesized from CSA as described previously (Ref. 32, and for more details see Ref. 33). MeVal-4-CS was obtained by total synthesis using the Merrifield technique (to be published elsewhere). Val-2-MeBmt(6.7-DH)-1-CS was obtained by palladium-catalyzed reduction (H 2 in ethanol) of natural CSD. Norvaline-2-CS is natural CSG. D-Lys(dansyl)-8-CS was synthesized as described in Ref. 34. The structure of all CS derivatives is given in Table I. All chemicals were of the highest purity commercially available.

RESULTS
In order to identify mitochondrial CSA-binding proteins, a Triton X-100 extract of a subcellular fraction enriched in mitochondria was subjected to affinity chromatography on a D-Ala-(3-amino)-8-CSA Affi-Gel 10 affinity matrix. CSA specifically and reproducibly eluted three major proteins with apparent molecular weights of 18, 20, and 22 kDa (lanes 2-7, arrows).
The minor component of about 50 kDa could not always be observed in repeats of this experiment. Elution with 6 M urea reproducibly released two proteins of 30 and 32 kDa, and occasionally higher molecular weight proteins (lanes 9 -13).
The major CSA-eluted proteins were identified by aminoterminal sequencing. The 18-kDa species had sequence VNPTVFFDI, which is identical to the amino-terminal sequence of cytosolic rat brain CyP-A. The 20-kDa species had the sequence ASDGGARGANSSSQNPLV, which matches that obtained by Connern and Halestrap (21) on a matrix PPiase isolated by ion exchange chromatography. The 22-kDa species had sequence NDKKKGPKVTVKVYFDF, which is identical to residues 26 -42 of rat kidney CyP-B (for a review on CyPs, see Ref. 35). Thus, as suggested earlier (25), only the 20-kDa species can be considered a selective mitochondrial CyP (termed CyP-M in this paper), while the 18-and 22-kDa species, which can be purified from endoplasmic reticular fractions (25), are most likely localized in mitochondria-associated endoplasmic reticulum.
The urea-eluted proteins of 30 and 32 kDa revealed a blocked N terminus. The 32-kDa species was identified as porin (experiments not shown) by Western blotting with a specific antibody against the rat liver protein (a generous gift of Professor Vito De Pinto, University of Catania, Italy), while the identity of the 30-kDa species remains unknown. Since under no circumstances could these proteins be eluted with CSA, the nature of their binding to the affinity matrix remains unclear, and was not investigated further in this paper.
Identical affinity purification protocols were carried out in the presence of Ca 2ϩ (which promotes opening of the MTP) or of ADP and Mg 2ϩ (which inhibit it synergistically with CSA, Refs. 36 and 37), and no major differences emerged relative to the results presented in Fig. 1 (not shown). We then focused on CyP-M, the only obvious candidate as a mitochondrial CSA receptor and therefore as the target for MTP inhibition by CSA.
In order to study the interactions of CyP-M with mitochondria, we prepared rabbit antisera against a synthetic peptide modelled on the unique rat CyP-M N-terminal sequence. The experiments depicted in Fig. 2 document the properties of one such antiserum. It can be seen that a single 20 kDa band was recognized in Western blots of total mitochondrial proteins (lanes 1) and in the CSA eluate of the D-Ala-(3-amino)-8-CSA affinity matrix (lanes 3). The reaction was specific, in that it was not detected with an identical dilution of preimmune serum (lanes 2), and it was selective for the N terminus of CyP-M in that (i) no reactivity was detected with either CyP-A or CyP-B (lanes 1 and 3, compare panels A and B) and (ii) antibody binding could be prevented by the synthetic N-terminal-like peptide ASDGGARGANSSSQC (lanes 4) but not by an irrelevant KVEKIGEGTYGVVYK peptide (lanes 5). Thus, this antibody is a useful tool to probe interactions of CyP-M with mitochondria.
In the experiments reported in Fig. 3, we have studied the distribution of CyP-M between a membrane and a soluble fraction obtained by ultracentrifugation of a suspension of sonicated mitochondria in a sucrose-based medium (panels A and B). CyP-M was clearly detectable both in the membrane-asso- Lateral chain amino acid 1 Lateral chain amino acid 11 mitochondrial inner membrane, and that this interaction is retained after sonication and despite the large dilution of the matrix following disruption of mitochondria. The nature of these interactions was investigated further by exposing this SMP preparation to a variety of conditions (Fig.  3, panel C). CyP-M could be released by isotonic NaCl (lanes 2, compare with lanes 1), suggesting that CyP-M membrane interactions are influenced by the ionic strength and demonstrating that all CyP-M is bound to the outer surface of our SMP preparation and accessible to the external milieu. Importantly, a large fraction of CyP-M could be released by CSA, which inhibits the pore, but not by CSH, which does not (lanes 3 and 4, respectively, compare with lanes 1), providing a first indication that CyP-M (un)binding to the inner membrane may be instrumental in MTP operation.
Release of CyP-M could also be detected after exposure of the SMP preparation to an acidic medium (lanes 5, compare with lanes 7), and this effect could be prevented by treatment of SMP with DPC prior to acidification (lanes 6, compare with lanes 7). We have previously shown that matrix acidification is accompanied by MTP closure, due to reversible protonation of histidyl residues, and that the inhibitory effects of matrix H ϩ on the MTP can be prevented by carbethoxylation of critical histidyl residues with DPC (2). Thus, these experiments further support our hypothesis on MTP⅐CyP-M interactions. It should be noted that a small proportion of CyP-M could be detected in the soluble fraction of the controls (SMP resedimented in sucrose medium at pH 7.4, panel C, lanes 1 and 7), most likely a dilution effect. With no exception, however, exposure to CSA, acidic pH, or isotonic NaCl resulted in an increase of the CyP-M released in the soluble fraction relative to the controls.
To investigate the role of CyP-M interactions in MTP regulation directly, we tried to detect MTP activity in SMP by a variety of methods. These included (i) a study of the permeability properties of liposomes reconstituted with SMP proteins (30) (these liposomes were permeable to sucrose and raffinose as are intact mitochondria after MTP opening; however, sucrose permeation was insensitive to CSA and was unaffected by Ca 2ϩ or EGTA, and therefore could not be unequivocally related to MTP (not shown)) and (ii) measurements of SMP membrane potential with the probe oxonol-VI (31) under a variety of conditions that should promote or inhibit MTP opening. We detected a decrease of membrane potential after treatment with Ca 2ϩ and phenylarsine oxide (two powerful MTP agonists), but again this depolarization was insensitive to CSA (experiments not shown). Thus, from these experiments we could not obtain clear evidence for MTP operation in SMP. As a result, we could not assess a role for CyP-M associations in MTP regulation directly. While our efforts in this direction are continuing, we decided to further probe the potential role of CyP-M in MTP modulation with an independent approach.
Besides giving clues about the interactions of CyP-M with mitochondria, the experiments of Fig. 3 allowed us to devise an efficient protocol for purification of CyP-M in enzymatically active form, which is illustrated in Fig. 4. SMP were prepared from isolated mitochondria in sucrose medium and sedimented by ultracentrifugation. Resuspension of SMP in acidic medium released CyP-M and relatively few other proteins (lanes 1), resulting in a substantial one-step CyP-M enrichment. This  3 and 4, respectively). It can be seen that a substantial fraction of CyP-M (arrow) sediments with the SMP membranes. Panel C, the SMP pellet (corresponding to lanes 3 of panels A and B) was resuspended in identical volumes of isotonic sucrose, pH 7.4 (controls), or as detailed below and spun again at 150,000 ϫ g for 1 h. CyP-M distribution between the membrane (pellets, upper row) and soluble (snts, lower row) fractions was then analyzed by Western blotting with anti-CyP-M antiserum after protein separation by SDS-PAGE and transfer to nitrocellulose. The resuspension medium was isotonic sucrose, pH 7.4 (controls, lanes 1 and 7), isotonic NaCl, pH 7.4 (lanes 2), isotonic sucrose, pH 7.4, supplemented with 50 g⅐ml Ϫ1 CSA (lanes 3) or CSH (lanes 4); isotonic sucrose, pH 5.0 (lanes 5), and isotonic sucrose, pH 5, with DPC present at 10 mM during sonication and at 1 mM thereafter (lanes 6). It must be stressed that after decanting the supernatants, the pellets were gently rinsed and carefully resuspended in the same initial volume of buffer prior to treatment of both supernatants and resuspended pellets with Laemmli gel sample buffer. Identical aliquots were then separated, so that the sum (pellet ϩ supernatant) of each lane represents the total CyP-M initially present in that sample. Note that data from two different SMP preparations are collected (lanes 1-4 and 5-7, respectively). soluble fraction was adjusted to pH 8.2, sequentially passed through an anion-and a cation-exchange matrix (lanes 2 and 3, respectively), and eluted from the latter with 0.1 M NaCl (lanes 4). After dialysis, CyP-M was purified with one passage on a HPLC cation-exchange column, wherefrom it eluted at about 0.1 M NaCl (lanes 5). CyP-M purity was further checked by silver staining of the HPLC eluate, which confirmed the absence of any other proteins (not shown).
CyP-M purified with this protocol exhibits PPiase activity, which could be inhibited by CSA but not by CSH (see the following paragraphs). We therefore decided to study the inhibitory profile of CyP-M PPiase activity and of MTP opening with a panel of CSA derivatives possessing different affinities for CyP-A. Besides giving information on the properties of CyP-M itself, a correlation between the relative potencies at inhibition of CyP-M and of MTP would be a good indication that a functional link between them indeed exists. The CSA derivatives we studied and their structural modifications relative to the native CSA molecule are illustrated in Table I. Fig. 5 shows typical experiments where the concentration dependence of inhibition of the MTP (panels A and B), of PPiase activity of purified CyP-M (panels C and D), and of PPiase activity of human recombinant CyP-A (panels E and F) was studied, allowing accurate determinations of the apparent I 50 for the tested inhibitor. In these representative experiments, the inhibitory compounds were CSA (panels A, C, and E) and MeVal-4- CS (panels B, D, and F). The results of a series of experiments carried out with identical protocols with all derivatives are summarized in Table II, where the indicated I 50 values for PPiase and MTP inhibition were obtained from triplicates of full titration curves. In the case of MTP inhibition, it can be seen that modification of residue 3 was without effect, while a small decrease of affinity followed modification of residues 4, 1, and 2. The most relevant change, with a clear decrease in affinity, was displayed by the CSA derivative modified in position 8. In the case of inhibition of PPiase activity with both CyP-M and CyP-A, the relative affinity was increased for the three (cf. Ref. 38) and four substitutions. Interestingly, the least efficient inhibitor of CyP-M was the 8-substituted derivative, which was nearly as effective at inhibiting CyP-A as CSA itself. On the other hand, CSH was ineffective in all cases.
The absolute I 50 values of the CSA derivatives for MTP inhibition cannot be directly compared with those obtained for inhibition of PPiase activity, because the former values were derived from studies on whole mitochondria and may therefore be influenced by membrane partitioning. It is more instructive to compare the potency of the inhibitor under study relative to that of CSA. Table III illustrates the results of this analysis, where the IC 50 of all CSA derivatives has been divided by that displayed by CSA. The data have been arranged so that the series reflects decreased affinity for inhibition of the MTP. It can be appreciated that within each group the derivatives appear in the same order for inhibition of PPiase activity, both with CyP-M and CyP-A. The only exception is for substitution in positions 1 and 2, where the difference is, however, negligible. When the absolute value of the ratios is considered, on the other hand, a striking difference emerges for the CSA derivative substituted in position 8. Indeed, the affinity is decreased over 10-fold for MTP, over 5-fold for CyP-M, and less than 2-fold for CyP-A. In other words, for this derivative the fit is extremely good between inhibition of MTP and CyP-M but not of CyP-A. DISCUSSION In this paper we have (i) characterized CSA-binding proteins associated with isolated mitochondria, showing that the 20-kDa CyP-M is the main and possibly the only specific mitochondrial CSA binding protein; (ii) studied the interactions of CyP-M with SMP and their modulation by potent inhibitors of the MTP, a CSA-sensitive mitochondrial channel; (iii) devised a new protocol for CyP-M purification; and (iv) studied the inhibitor profile of highly purified CyP-M and of MTP with a series of CSA derivatives with graded affinity for CyP-A. Our results support the idea that CyP-M association with the MTP is essential for channel opening, suggest that MTP inhibition by CSA and protons may be due to CyP-M unbinding from its putative site on the MTP, and show that calcineurin inhibition is not involved in MTP inhibition by CSA.
How Many CSA-binding Mitochondrial Proteins?-The first relevant result of this paper is reported in Fig. 1. Although three CSA-binding proteins of molecular masses 18, 20, and 22 kDa can be purified by affinity chromatography from a conventional mitochondrial preparation isolated by differential centrifugation, only the 20-kDa species is selectively localized to mitochondria, while the 18-and 22-kDa species, which are both present in a postmitochondrial 150,000 ϫ g particulate fraction (25), are most likely derived from endoplasmic reticulum vesicles, which are associated with the outer mitochondrial membrane. A second point of interest is that no other proteins could be specifically eluted from the CSA affinity matrix (Fig. 1), even after previous removal of CyPs (results not shown). These data suggest that CyP-M is the main and possibly the only mitochondrial receptor for CSA, which makes it the most likely candidate for MTP inhibition by CSA.
Recently, Andreeva and Crompton (23) have exploited the synergistic effects of ADP and CSA at MTP inhibition (36,37) to identify a 10-kDa protein in a rat liver mitochondrial fraction that could be photolabeled with a tritiated, 8-substituted CSA derivative in the presence of ADP after soluble components had been removed by sonication at pH 8.1 and centrifugation (23). It was proposed that the 10-kDa species is an integral membrane protein, and that it most likely represents the target for CSA on the pore (23). Subsequently, the same group has reported photolabeling also of a 22-kDa protein possessing CSA-sensitive PPiase activity (K i for CSA 5 nM) from rat heart mitochondria, and of a 18-kDa protein from rat liver mitochondria (39) and suggested that the 10-kDa liver species may in fact be a proteolytic product of the 18-kDa protein rather than the CSA membrane receptor (39). In the absence of structural information on these proteins, it is hard  to lanes 3, panels A and B of Fig. 3) was resuspended in isotonic sucrose, pH 5.0, and spun at 150,000 ϫ g for 1 h. The supernatants were brought to pH 8.2, and CyP-M was purified to homogeneity by sequential chromatographic steps on Q-Sepharose, S-Sepharose, and HR 5/5 mono-S HPLC columns as described in detail under "Materials and Methods." to make predictions about their possible relationships with CyP-M, and with the MTP regulatory mechanism proposed here. However, we note that (i) due to the tight association of the endoplasmic reticulum with mitochondria, both the 22-kDa CyP-B and the 18-kDa CyP-A copurify with the 20-kDa CyP-M in CSA affinity-based protocols (Fig. 1); (ii) the protocol used for membrane preparation by Andreeva and co-workers (sonication at pH 8.1; Refs. 23 and 39) is expected to remove a large fraction of CyP-M (Fig. 3); (iii) the endoplasmic reticulum is expected to retain its integrity in sonication protocols designed for disruption of mitochondria and thus to represent a major component of the vesicular fraction sedimented by ultracentrifugation after sonication; and (iv) the ion-exchange chromatog-   Fig. 3). Based on these considerations, on their apparent molecular weights, and on their physico-chemical properties, we suspect that the larger proteins labeled by Andreeva et al. (39) in heart and liver subcellular fractions enriched in mitochondria are the 22-and 18-kDa endoplasmic reticulum-associated CyP-B and CyP-A, respectively.
Membrane Interactions of CyP-M and Inhibition of the MTP-The effect of acidic pH on the MTP is well documented (12). We have been able to track the inhibitory effect of protons to a matrix site (9) and to show that it is mediated by reversible protonation of histidyl residues (2). The demonstration that acidification can release CyP-M from SMP in a DPC-sensitive fashion and that the effect of H ϩ can be mimicked by CSA but not CSH (Fig. 2), together with our previous findings on the role of histidyl residues in MTP activity (2), acquires a particular meaning in the light of recent structural and functional studies of the interactions of immunosuppressant drugs with their intracellular receptors, the immunophilins (CyPs and FKBPs). Yu and Fesik (40) have shown that His-126 of CyP-A, which is in close proximity to the CSA binding site, has a pK a of 6.34 in the uncomplexed protein, while in the CSA-complexed form, the pK a is shifted to 4.65. His-126 has been implied in ligandsubstrate interactions and appears to contribute to the hydrophobic pocket, which is essential for substrate binding (41) and for enzyme function (42). A similar situation occurs with FKBP, the intracellular target for the immunosuppressant drugs FK506 and rapamycin, which is structurally unrelated to CyPs (43,44) yet possesses PPiase activity and shares calcineurin as the target for immunosuppression (45,46). Binding of the FK506 analogue ascomycin to FKBP shifted the pK a of His-87, which is again located close to the ligand binding pocket, from 5.92 to 4.86 (40).
It must be stressed that although His-126 of CyP-A appears not to be involved in catalysis (40), it is essential for enzyme function since a His-126 3 Gln mutant of CyP-A displayed less than 1% of the activity of the wild-type (42). It appears likely that a critical histidyl residue with similar properties will be found in the substrate-binding pocket of CyP-M, and based on the results of Fig. 3, it seems plausible that CSA binding, or protonation of this residue, may cause similar conformational effects on CyP-M, leading to its dissociation from the putative binding site on the MTP.
Immunophilins and Regulation of Ion Channels-A role for immunophilins in the regulation of ion channels has been clearly documented in four cases. A first observation is that expression of homo-oligomeric ligand-gated ion channels in Xenopus oocytes is inhibited by CSA, and this effect can be reversed by overexpression of CyP-A (47). The effect is posttranslational, and is most likely exerted at the level of homooligomerization of the polypeptide subunits, suggesting that CyP-A-channel associations are required for proper conformation (47). A second case is the identification of a specific CyP-B binding protein, which appears to be expressed in intracellular vesicular compartments (48). This protein, which interacts specifically with CyP-B but not CyP-A or CyP-C, encodes an integral membrane protein involved in intracellular Ca 2ϩ signaling, and its overexpression is able to overcome the Ca 2ϩ requirement of several T cell activators acting through nuclear factor of activated T cells in a CSA-sensitive fashion (48). A third case is represented by a 12-kDa FK506-binding immunophilin (FKBP12), which copurifies with sarcoplasmic reticulum membranes (49). Extraction of FKBP12 profoundly alters the Ca 2ϩ diffusion properties through the SR-CRC and its response to caffeine and abolishes activation by FK506 (50). FKBP12 stabilizes SR-CRC channel gating to two major subconductance states, while in its absence the channel populates a large number of subconductances (51). A fourth case is the tight association of FKBP12 with the IP 3 R, which is structurally and functionally related to SR-CRC (52). Disruption of IP 3 R-FKBP12 interactions by FK506 increased Ca 2ϩ flux through IP 3 R, and this effect could be reversed by added FKBP12, suggesting that FKBP12 exerts a regulatory role on IP 3 R-mediated Ca 2ϩ fluxes (52).
A number of analogies link the MTP to these observations. Single-channel recordings of MTP in rat liver mitoplasts show a number of subconductance states (53), which are reminescent of those of FKBP-stripped SR-CRC (51). It is tempting to speculate that this MTP behavior depends largely on variable degrees of CyP-M extraction because CyP-M is largely removed by the hypotonic shock procedure needed for mitoplast preparation. 2 A second analogy between modulation of MTP and SR-CRC by CSA and FK506, respectively, is that in neither case does the immunosuppressant need inhibition of calcineurin activity. Indeed, MeVal-4-CS is as effective as CSA itself at pore inhibition (Ref. 25 and Table II), while the effects of FK506 on the SR-CRC can be mimicked by rapamycin, which likewise does not inhibit calcineurin (51).
The submitochondrial localization of CyP-M has not been studied in detail, but its copurification with matrix components (21) 3 indicates that it is a soluble matrix enzyme. Yet, as shown here, CyP-M can associate with SMP. These observations suggest that only a fraction of CyP-M may participate in MTP regulation at any given time and that MTP opening-closure may be dependent on modulation of CyP-M binding to the membrane. Consistent with this idea, it has been shown that phenylarsine oxide, perhaps the most potent MTP agonist (1,9,54), increases the affinity of binding of CyP-M to mitochondrial membranes (55). Based on the findings of the present work and because of the striking analogies with SR-CRC and IP 3 R mentioned above (see Ref. 19 for a general discussion), we predict that the MTP⅐CyP-M complex will prove to be the mitochondrial homologue of the SR-CRC⅐FKBP12 and IP 3 R⅐FKBP12 complexes.
Inhibition of CyP-M PPiase and MTP Activities by CSA Derivatives-Independent evidence for a role of CyP-M in MTP regulation comes from the experiments with CSA derivatives, which were selected solely on the basis of their graded affinity for CyP-A. Our main findings (Tables II and III) can be summarized as follows. (i) Inhibition of MTP by CSA and its analogues correlates well with inhibition of PPiase activity of both CyP-M and CyP-A. (ii) In all cases, the best fit is observed between MTP and CyP-M. (iii) At variance of the requirements for immunosuppression, inhibition of MTP opening by CSA does not require inhibition of calcineurin since MeVal-4-CS retains the ability to inhibit the pore but not that of inhibiting calcineurin (56). (iv) Substitutions in position 8 interfere most with inhibition of MTP and of PPiase activity of CyP-M, while interactions with CyP-A are affected only minimally; this suggests that the structural differences between the two CyP isoforms go beyond the N-terminal sequences. (v) CSH does not inhibit either the MTP or PPiase activity of CyP-M.
The match between the structural requirements for inhibition of MTP and of CyP-M PPiase catalytic activity by CSA is of particular relevance when the mechanism of CyP inhibition by CSA in aqueous solution is considered. In an apparent paradox, a CSA derivative with marginal affinity for CyP-A retained a considerable immunosuppressive activity (57), and the x-ray crystallographic maps at 1.86 Å resolution of CyP-A complexed with CSA or with a 4-substituted CSA derivative were virtually identical despite a 4-fold higher affinity for CyP-A of the latter (58). This led to the proposal that CSA conformer equilibria in aqueous solution rather than that differences in three-dimensional architecture govern binding to CSA (57). This hypothesis has recently been confirmed by direct measurements showing that CyP-A only recognizes a well defined conformation of CSA that exists in water prior to binding and that the rate-limiting step in complex formation between CSA and CyP-A is the rate of cis-trans isomerization of CSA in position 9,10 (38). These observations imply that kinetic factors play a major role in the determination of apparent affinity constants for CSA derivatives. Indeed, the time required for equilibrium binding of CSA to CyP-A was as high as 1 h, and it dropped to less than 5 min (the earlier time point that could be taken) with two 3-substituted CSA analogs mimicking the conformation of CyP-Acomplexed CSA (38).
Based on these findings, it appears likely that the match between inhibition of CyP-M and MTP would be even better with the purified channel, where the complexities arising from CSA partitioning between membrane and aqueous mitochondrial phases can be overcome.
Conclusions and Perspectives-We conclude that CyP-M is the main and probably the only mitochondrial receptor for CSA, that CyP-M interactions with the mitochondrial inner membrane in SMP are consistent with a key role for this immunophilin in the regulation of the MTP, and that CSA inhibition of the MTP does not require interactions with calcineurin. Mitochondrial (dys)function is now increasingly considered as a key event in a variety of forms of cell death, ranging from ischemia (59 -61) to excitotoxic neurodegeneration (62) to oxidant-induced stress (63) to apoptosis (64 -67). Because of its exquisite sensitivity to Ca 2ϩ ions, to the proton electrochemical gradient, and to oxidative stress (68), the MTP appears as a likely target on which many pathological agents or conditions may converge. The involvement of CyP-M in MTP modulation and the dissociation of MTP inhibition from the immunosuppressive effects of CSA demonstrated here offer great promise for the development of new conceptual and pharmacological tools aimed at therapeutic intervention.