Presenilin 1-associated protein (PSAP) was first identified as a protein that interacts with presenilin 1. It was later reported that PSAP is a mitochondrial protein that induces apoptosis when overexpressed in cultured cells. PSAP is also known as mitochondrial carrier homolog 1 (Mtch1). In this study, we show that there are two proapoptotic PSAP isoforms generated by alternative splicing that differ in the length of a hydrophilic loop located between two predicted transmembrane domains. Using RT-PCR and Western blot assays, we determined that both isoforms are expressed in human and rat tissues as well as in culture cells. Our results indicate that PSAP is an integral mitochondrial outer membrane protein, although it contains a mitochondrial carrier domain conserved in several inner membrane carriers, which partially overlaps one of the predicted transmembrane segments. Deletion of this transmembrane segment impairs mitochondrial import of PSAP. Replacement of this segment with each of two transmembrane domains, with opposite membrane orientations, from an unrelated protein indicated that one of them allowed mitochondrial localization of the PSAP mutant, whereas the other one did not. Our interpretation of these results is that PSAP contains multiple mitochondrial targeting motifs dispersed along the protein but that a transmembrane domain in the correct position and orientation is necessary for membrane insertion. The amino acid sequence within this transmembrane domain may also be important. Furthermore, two independent regions in the amino terminal side of the protein are responsible for its proapoptotic activity. Possible implications of these findings in PSAP function are discussed.
- presenilin 1-associated protein-mitochondrial carrier homolog 1
- Alzheimer's disease
alzheimer's disease is a neurodegenerative disorder characterized at the cellular level by the extracellular accumulation of protein aggregates mainly constituted by amyloid β (Aβ) peptide and intracellular tangles that involve hyperphosphorylated tau protein (25). It is still not clear which event is responsible for the onset of the disease, and several factors have been found to play important roles apart from the formation of protein aggregates. For instance, it has been shown that soluble oligomers of Aβ are toxic for neurons (16). The Aβ peptide is generated by proteolytic processing from its precursor protein Aβ precursor protein (APP). Two protease activities named β-secretase and γ-secretase are responsible for Aβ generation. Presenilins are components of the γ-secretase complex and are assumed to contain the catalytic site of the protease, although this is still a matter of discussion. Presenilins also have other cellular functions apart from their role in processing APP (27).
In some instances, neuronal degeneration was observed before the formation of Aβ plaques, which suggested that cell death could be a factor in the pathogenesis of Alzheimer's disease before Aβ deposition. The first demonstration of the involvement of presenilins in apoptotic cell death came from a report of Vito et al. (31) who showed that ALG3, a mouse protein, could rescue T cells from apoptotic cell death induced by Fas. ALG3 is 98% identical to the 103 carboxyl terminal amino acids of presenilin 2 (PS2). Several other reports have linked the Aβ peptide and presenilins to apoptosis. For instance, Aβ induces release of cytochrome c from isolated mitochondria (12); intracellular Aβ induces neuronal apoptosis (11); presenilin 1 (PS1) interacts with Bcl-2 (1); PS1 and PS2 interact with Bcl-XL (22); PS1 protects against neuronal apoptosis induced by the protein PAG (38); PS2 induces p53-dependent apoptosis (2); mutant PS1 induces apoptosis and downregulates Akt/PKB (33); and both presenilins interact with FKBP38 and promote apoptosis by reducing mitochondrial Bcl-2 (32). The relationship between apoptosis and Alzheimer's disease gets more complicated if we consider that APP, as well as presenilins, are substrates for caspases (19, 20), although the amyloidogenic pathway appears to be independent of APP processing by caspases (26). In a different line of research, Yoo et al. (37) reported changes in voltage-dependent anion channel (VDAC) levels in patients with Alzheimer's disease and Down syndrome. VDAC is a component of the mitochondrial permeability transition pore.
A definitive link between presenilins and mitochondria has been established recently when Hansson and coworkers (8) reported the presence of the major γ-secretase components in mitochondria. This confirmed a previous report of the presence of PS1 in rat mitochondria (3). Xu et al. (34, 35) reported the identification of a presenilin 1-associated protein (PSAP) that localizes to mitochondria and induces apoptosis when overexpressed in tissue culture cells.
In this report we show that PSAP has two ubiquitously expressed isoforms generated by alternative splicing, which is highly conserved in different organisms. Both isoforms induce apoptosis when overexpressed in cells in culture and are integral outer membrane proteins. We also analyzed several deletion mutants of PSAP and found that both proteins are imported into mitochondria using multiple internal targeting signals but require a transmembrane domain in the correct orientation, or with the correct sequence, for insertion into the outer membrane. Both isoforms appear to contain two proapoptotic domains in the amino terminal side of the protein.
MATERIALS AND METHODS
All reagents were of molecular biology grade unless stated otherwise. Restriction enzymes were purchased from Roche, Stratagene, and Invitrogen. Pfu polymerase was from Stratagene. Eco Taq polymerase was from Ecogen. Custom-made primers were from Invitrogen.
PSAP clones for both PSAP isoforms were first identified as est sequences and obtained from the I.M.A.G.E. Consortium through the MRC geneservice. Clone 5314188 corresponds to the larger isoform (PSAPL) and contains the entire coding region. Clone 6428572 corresponds to the shorter one (PSAPS) and is missing the amino terminal region. Sequencing of these clones revealed three nucleotide differences, with respect to PSAP sequences found in GenBank, that were present in both clones. The PSAPL clone also lacked a nucleotide, causing a frame shift, in the center of the coding sequence. None of these nucleotide polymorphisms were found in PSAP est sequences available in GenBank, which is surprising because the two clones were of different origins, PSAPL coming from the hypothalamus and PSAPS coming from a lung epidermoid carcinoma cell line, most likely from different individuals. To avoid potential problems in the characterization of PSAP due to these apparent polymorphisms, we obtained new clones from Invitrogen. Clone CS0DE007YK11 contains PSAPL and clone CS0DI003YM05 contains PSAPS, both in the vector pCMVSport, but both clones are missing the first putative start codon of PSAP, as discussed in discussion. Complete sequencing of these two clones revealed no nucleotide differences with the consensus PSAP sequence (Ref. 34 and GenBank est sequences). These two clones were used to construct new clones to express PSAP-EGFP fusion proteins in eukaryotic cells.
Construction of expression vectors.
Clone pVLG1 contains the full consensus coding region of PSAP in pCMVSport and was made as follows. We first digested Invitrogen clone CS0DE007YK11 (pCMVSport6-PSAPL-ATG2) with XbaI and HindIII (this removed an XhoI site located in between), polished the ends, and religated to obtain clone pCMVSport6-PSAPL-ATG2-DXho. This clone was digested with EcoRI-XhoI (XhoI is located between the second and third ATGs of PSAP), and this region was replaced with the equivalent region from pJAC217. pJAC217 consists of the EcoRI-BamHI fragment from IMAGE clone 5314188 (contains full-length PSAPL, including the 3′-UTR but with a frame-shifting nucleotide insertion and three-point mutations) cloned in pCDNA3.1(−). In this way, we replaced the amino terminal region of PSAP (up to the XhoI site) in the clone from Invitrogen (which was missing the first ATG) with the equivalent region from the IMAGE clone (which contained the first ATG), generating a clone that contains the full coding sequence of PSAP with no nucleotide changes.
Clone pJAC220B (PSAPL-EGFP) was made to express full-length PSAPL fused to EGFP (enhanced green fluorescent protein) in pEGFP-N1 (BD-Clontech) and was constructed as follows. On one side, IMAGE clone 5314188 was digested with XhoI and BamHI to remove most of the PSAP coding region and the 3′-UTR. On the other side, the same clone was used as template in a PCR with primers PSAPXHO (5′-TCGAGGCTCGAGCTCGCGAT-3′) and PSAPBAM (5′-AGGATCCTCCAGGGCAAAGCATGA-3′) to generate an XhoI-BamHI fragment that replaced the removed one and contained a BamHI site in frame with EGFP (in vector pEGFP-N1) instead of the PSAP stop codon. This clone was named pJAC218 and was used to construct clone pJAC220, by moving its EcoRI-BamHI fragment into pEGFP-N1. Because this clone was derived from the IMAGE clone, it contained the above-mentioned frame-shift mutation and three-point mutations (we realized this when this clone was fully sequenced). The XhoI-BamHI region from pJAC220 was replaced with the equivalent region from clone CS0DE007YK11, using the same PCR approach with the same primers, to generate pJAC220B.
Clone pVLG2 (PSAPL-DN18-EGFP) was made to express PSAPL from the second ATG fused to enhanced green fluorescent protein (EGFP) and was generated as follows. pJAC220B was digested with NheI and HindIII, polished with Pfu polymerase, and religated with T4 DNA polymerase to construct pJAC220B-DXho. This eliminated an XhoI site located in the vector right upstream of PSAP, leaving the construct with just the XhoI site located within PSAP. The EcoRI-XhoI region from this clone was removed and replaced with an equivalent fragment from the Invitrogen clone CS0DE007YK11, which misses the first PSAP ATG, to generate pVLG2.
pJAC225B (PSAPL-DN64-EGFP) was made to express PSAPL from the third ATG fused to EGFP and was constructed by removal of the XhoI-XhoI fragment from pJAC220B and religation.
pVLG4 (PSAPS-EGFP) was made to express PSAPS from the first ATG fused to EGFP as follows. CS0DI003YM05 (PSAPS without the first ATG, in pCMVSport) was used as template in a PCR with primers PSAPXHO and PSAPBAM, the product was digested with XhoI and BamHI, and used to replace the equivalent XhoI-BamHI region in clone pJAC220B-DXho.
pVLG3 (PSAPS-DN18-EGFP) was made to express PSAPS from the second ATG fused to EGFP and constructed in a similar way as pVLG4: pVLG2 was digested with XhoI and BamHI to remove the region located between both sites, which was then replaced with the same XhoI-BamHI fragment used for pVLG4.
pVLG5 (PSAPS-DN64-EGFP) was made to express PSAPS from the third ATG, in a similar way as pVLG3, but using pJAC225B instead of pVLG2.
To construct clones to express deletion mutants, we first generated a pBluescript II SK(+) plasmid without a Hind III site by digesting with HindIII, polishing with Pfu polymerase, and religation, generating a pSKDHind vector. We then moved PSAPL from pJAC220B, digested with EcoRI and BamHI, into this vector to create clone pJAC253. This clone was used as template in a PCR with primers PSAPDL1 (5′-CTAGGATCAAGCTTGATGAAGTGGGCCAGCAG-3′) and PSAPDL2 (5′-CTAGGATCAAGCTTAAGTTCGTGATGGGGATT-3′) (note that these primers point away from each other) to generate a PCR product that was digested with HindIII and ligated to recircularize it, creating clone pJAC254. The EcoRI-BamHI fragment from this clone was moved back into pEGFP-N1 to create clone pJAC257, which expresses PSAPLDLoop fused to EGFP. PSAPLDLoop contains a lysine and a leucine instead of the hydrophilic loop located between both transmembrane domains of PSAP. Similar approaches were used to construct clones that express PSAPLDTM1 (pJAC259, using primers PSAPDTM11, 5-CTAGGATCAAGCTTGAATCCCAGCAGCCCTTC-3′, and PSAPDTM12, 5′-CTAGGATCAAGCTTATGTTCATCAATGCCTACCTG-3; a lysine, a leucine, and a methionine, KLM, replace transmembrane domain 1) and PSAPLDTM2 (pJAC260, primers PSAPDTM21, 5′-CTAGGATCAAGCTTGAACTTGGTATAGCTCCG-3′, and PSAPDTM22, 5′-CTAGGATCAAGCTTATGAACTGCGGGCTGCAAGCT-3′; KLM instead of transmembrane domain 2). PSAPLDN112 (pJAC255) and PSAPLDN168 (pJAC256) were constructed by doing PCR on pJAC220B with primer PSAPDN1 (5′-CTAGGATCGAATTCATGCCCCCCACCCTTGGG-3′) and GFPR (5′-CTTCAGGGTCAGCTTGCCGT-3′) or primer PSAPDN2 (5′-CTAGGATCGAATTCATGAAGAAGGTTTTCCCT-3′) and GFPR, digesting the product with EcoRI and BamHI and cloning into pEGFP-N1. PSAPLDC51 (pJAC258) construct was made by doing PCR with primer GFPF (5′-CATTGACGCAAATGGGCGGT-3′) and PSAPDC1 (5′-CTAGGATCGGATCCTTGTTCACAGCCATGAGG-3′) on pJAC220B and cloning EcoRI-BamHI into pEGFP-N1. PSAPLDI (pJAC269) was constructed by PCR with primers PSAPDI1 (5′-CTAGGATCAAGCTTGCTACCCCGAGTCACAGTAG-3′) and PSAPDI2 (5′-CTAGGATCAAGCTTATGTTCGTTGGATTAATCCCTCA-3′) in a way similar to clone pJAC254. The same approach was used to make construct PSAPLDMCD (pJAC265) using primers (5′-CTAGGATCAAGCTTCTGCATCATCATCTCGTAGG-3′) and (5′-CTAGGATCAAGCTTATGTTGTGGGGCTGTAACCTGCT-3′).
Clone pANT1, which expresses PSAPL without residues 113–168 fused to EGFP (PSAPLDAPO), was constructed using primers PSAPAPOF (5′-CTAGGATCAAGCTTATGCCCCCCACCCTTGGGAC-3′) and PSAPAPOR (5′-CTAGGATCAAGCTTGCTACCCCGAGTCACAGTAG-3′) using a PCR approach similar to the one used for other internally deleted constructs.
Clone pVLG62, which expresses PSAPL without residues 64–168 (PSAPLDAPO2), was constructed by carrying out a PCR with primers GFPF and PSAPAPO2 (5′-CTAGGATCAAGCTTCCTGCGGGCCGAGGGCTGA-3′) on pJAC220B DNA, digesting the product with HindIII and using the resulting fragment to replace the HindIII-HindIII fragment from clone pANT1.
Clone pJAC270, which expresses PSAPL without residues 1–64 and 113–168 (PSAPLDAPO3), was constructed by digesting clone pANT1 with XhoI, eliminating the 158-bp fragment between both XhoI sites, and religating.
Clone pVLG60, which encodes PSAPL with its TM1 replaced by TM2 of Pit-2 (PSAPLTMP2), was constructed as follows. The region encoding the second TM domain of Pit-2 was amplified by PCR from human heart cDNA using primers Pit2-2F (5′-CTAGGATCAAGCTTAGGCAGGCATGCATTTTAGC-3′) and Pit2-2R (5′-CTAGGATCAAGCTTTTTGGCGCCTAGTAACACGGA-3′). The product was digested with HindIII (present in both primers) and the resulting fragment inserted in clone pVLG28 (PSAPLDTM1 in pBluescript II SK+) previously linearized with HindIII. An EcoRI-BamHI fragment from this clone was inserted into pEGFP-N1 to generate clone pVLG60.
Clone pVLG61, which encodes PSAPL with its TM1 replaced by TM3 of Pit-2 (PSAPLTMP3), was constructed exactly as clone pVLG60 but by using PCR primers Pit3-3F (5′-CTAGGATCAAGCTTATGGCTGGGGAAGTTAGTGC-3′) and Pit3-3R (5′-CTAGGATCAAGCTTCAGGAAGGAAGCAATCAGCT-3′).
To express PSAP without tags, we used clone pVLG1, described above, for PSAPL from the first ATG, clones from Invitrogen for PSAPL and PSAPS from the second ATG, and clone pJAC283 for PSAPS from the first ATG. This one was constructed by introducing an EcoRI-XhoI fragment from pJAC220B together with an XhoI-BamHI fragment from Invitrogen PSAPS clone in pEGFP-N1 (since PSAP conserves its original stop codon and 3′ UTR there is no GFP fusion made).
As control for mitochondrial targeting of EGFP, we used a cDNA corresponding to the first 72 amino acids of Xenopus pol gamma B cloned into pEGFP-N1 (GB-EGFP, Carrodeguas and Bogenhagen, unpublished observations), which works well in human cells (23).
All PCR reactions used to make subclones of PSAP were carried out by using Pfu polymerase from Stratagene, under conditions suggested by the manufacturer. When the 5′ region of PSAP was included in the amplification, the reactions were supplemented with 2.5% DMSO due to the GC richness of this region. Reactions were initiated by a 2-min denaturation step at 94°C followed by 30 cycles consisting on 94°C for 30 s, 55°C for 30 s, and 72°C for 1 min/kb, ending with a 10-min extension at 72°C, using an MJ Research Minicycler.
Reverse transcription-polymerase chain reaction assays.
RT-PCR assays were carried out using the Multiple Tissue cDNA panels I and II from Clontech. Two microliters of each cDNA was used as template in a 50-μl PCR reaction containing 16 mM (NH4)2SO4, 67 mM Tris·HCl (pH 8.8), 0.1% Tween-20, 2 mM MgCl2, 0.4 μM of each primer, and 5 units of Ecotaq DNA polymerase (Ecogen). Primers used were PSAP1 (5′- AACCTGCTGGCCCACTTCATCAAT-3′) and PSAPECO (5′-TGAATTCAGGTTACTCCAGGGCAAAGCA-3′). PCR cycles were as follows: an initial denaturation step at 94°C for 2 min followed by 35 cycles of 94°C for 15 s, 60°C for 15 s, and 72°C for 1 min. Glyceraldehyde-3-phosphate dehydrogenase (G3PDH) was used as an internal standard with primers supplied by Clontech, either in singleplex or multiplex mode together with PSAP primers. To determine nonsaturating PCR conditions, 10-μl fractions of each PCR reaction were removed after 22, 27, 30, and 33 cycles and analyzed on 2% agarose-TBE gels to separate amplified products of 370 bp for PSAPL, 319 bp for PSAPS, and 983 bp for G3PDH. All samples were run in the same gel, which was stained with ethidium bromide and photographed using a Bio-Rad Gel Doc 1000 apparatus and Molecular Analyst software. The best results (good intensity without saturation) were obtained after 27 cycles in both singleplex and multiplex reactions.
Membrane topology prediction.
We used the following computer programs to predict the membrane topology of PSAP: SOSUI (classification and secondary structure prediction of membrane proteins, expert protein analysis system, Mitaku group, Dept. of Biotechnology, Tokio University of Agriculture and Technology), HMMTOP (prediction of transmembrane helices and topology of proteins, ExPASy topology prediction) (29, 30), TMHMM (14), TMpred (prediction of transmembrane regions and orientation) (10), and TopPred (topology prediction of membrane proteins) (9, 6).
Cell culture and transfection.
Cells were cultured in Dulbecco's modified Eagle medium supplemented with 2 mM glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin sulfate, and 10% fetal calf serum (all reagents from Invitrogen) at 37°C in a 5% CO2 atmosphere. Tissue culture dishes and flasks were from TPP or Nunc. Transient transfections were carried out with lipofectamine 2000 (Invitrogen) by following manufacturer's instructions. For microscopy, cells were grown on round, 1-cm diameter coverslides on 24-well plates and transfected at a confluence of ∼50%, higher for Western blot analyses. We used mainly HeLa cells (from human cervical cancer) and HEK293 cells (from human embryonic kidney).
Mitochondrial localization and detection of apoptosis.
Cells were treated with MitoTracker Red CMX-Ros or MitoTracker Red 580 (Molecular Probes) at concentrations of 25 and 200 nM, respectively, for 30 min, and the medium was replaced with fresh one and incubated for an additional 30 min at 37°C. Hoechst 33342 (Molecular probes) was added to the cells at a concentration of 1 μg/ml 15 to 30 min before processing. Cells were fixed by incubation in 3.7% paraformaldehyde in PBS for 15 min at room temperature or in 3.7% formaldehyde in culture medium at 37°C and mounted for microscopy using 90% glycerol and 0.25% 1,4-diazabicyclo[2.2.2]octane (DABCO, as antifading). For immunocytochemistry with anti-Tom20, cells were fixed in a solution containing 50% DMEM and 50% of a 3.7% paraformaldehyde solution in PBS, followed by a solution of 3.7% paraformaldehyde in PBS and then blocked and permeabilized in a solution containing 1% BSA and 0.1% Triton X-100 (TX-100). Anti-Tom 20 from Santa Cruz Biotechnology was used at a dilution of 1:400 and an Alexa Fluor 568-conjugated goat anti-rabbit secondary antibody (Molecular Probes) was used for detection. Microscopy was carried out using a Zeiss Axiovert 200 M fluorescence microscope and a Zeiss Pascal and Leica TCS SP2 AOBS confocal microscopes. With the Axiovert, high magnification images at ×640 were obtained using a ×40 objective combined with an optovar (×1.6). Low magnification images were obtained at ×160, using the ×10 objective and the optovar. For EGFP, a 470/525 filter was used. For Hoechst, we used a 365/420 filter. A triple filter 400/460, 495/530, and 570/610 was used for EGFP, MitoTracker and Hoechst visualization together. With the Zeiss confocal, the pinhole was conserved at 1 for all pictures, the excitation laser was at 488 for green, and at 543 for red. The emission filters were BP505–530 and LP560, respectively. Acquisition time was 9 s, and the scan was averaged by two line scans. With the Leica confocal microscope, the laser emission lines used were 488 and 568 nm. Files were analyzed by Adobe Photoshop and images were merged using RVB format.
For the detection of poly(ADP-ribose)polymerase (PARP) and green fluorescent protein in Western blot analyses, cells grown in 24-well plates and transfected with 0.8 μg of DNA and lipofectamine 2000 (Invitrogen) were trypsinized 24 h posttransfection, centrifuged at 200 g for 8 min, and resuspended directly in SDS-loading buffer (Laemmli). Samples were boiled, DNA was broken by passage of the lysate through a 25-gauge syringe, and proteins were separated by SDS-PAGE with Laemmli buffers and transferred onto BioTrace polyvinylidene fluoride (PVDF) membranes. We used an anti-PARP antibody from BD Pharmingen (mouse 7D3-6) at 1 μg/ml, anti-green fluorescent protein antibodies (mixture of clones 7.1 and 13.1) from Roche at 0.4 μg/ml, and a polyclonal anti-actin from Sigma at 1:500 dilution in a blocking solution containing 0.5% Tween-20 and 5% nonfat dry milk in phosphate-buffered saline (PBS). The secondary antibodies were an horseradish peroxidase (HRP)-linked sheep anti-mouse from Amersham or an HRP-linked goat anti-rabbit from Sigma, used at 1:5,000 dilution in the same blocking solution. Blots were developed using the ECL Plus kit and Hyperfilm MP from Amersham.
HEK293 cells were grown in T-75 flasks to a confluency of ∼80% and transfected using lipofectamine. Twenty four hours posttransfection, floating cells were combined with trypsinized attached cells. Around 107 cells were lysed and fractionated by differential centrifugation as previously described (21). Different aliquots of the mitochondrial fractions were subjected to treatments with protease K, digitonin, and/or Triton X-100. Protease K (Sigma) was used at 0.1 μg/μl of mitochondrial suspension and was inhibited by addition of phenylmethylsulfonyl fluoride (PMSF) before the different aliquots were analyzed by Western blot analysis. For the detection of mitochondrial proteins, we used antibodies against Tom20 (outer membrane) from Santa Cruz Biotechnology at 0.2 μg/ml, cytochrome c (intermembrane space) from BD Pharmingen at 0.3 μg/ml, cytochrome oxidase subunit I (inner membrane) from Molecular Probes at 0.2 μg/ml, and heat shock protein 60 (matrix) from Sigma at 1:2,000 dilution. Conditions of incubation with antibodies and detection were as described above.
Sodium carbonate extraction of mitochondrial proteins.
Untransfected or transfected HEK293 cells (18 h posttransfection) from two 75-cm2 flasks were collected by gentle centrifugation of the culture media at 300 g for 5 min and washed twice with PBS. Attached cells were washed twice with PBS, collected with a cell scraper, mixed with previously collected floating cells, and centrifuged at 300 g for 5 min. The cell pellet was resuspended in 1 ml homogenization buffer (10 mM HEPES, pH 7.4, 0.2 M mannitol, 0.07 M sucrose, 1 mM EDTA, and 0.2 mM PMSF). Cells were broken by 20 strokes in a dounce homogenizer with a loose-fitting pestle followed by 20 strokes with a tight-fitting pestle, followed by centrifugation at 800 g for 5 min to remove nuclei and unbroken cells. The resulting supernatant was centrifuged at 5,100 g for 10 min at 4°C to pellet mitochondria.
The mitochondrial pellet was resuspended in 1.7 ml of 0.1 M Na2CO3 (pH 12) and incubated on ice for 20 min. A small aliquot was removed (total mitochondria) and the remaining was centrifuged at 144,000 g for 1 h at 4°C. The supernatant was transferred to a new tube, and the pellet was resuspended in PBS with 1% TX-100 using a Pasteur pipette followed by a 20-gauge needle and then a 27-gauge needle. Twenty microliters of each fraction were analyzed by SDS-PAGE.
Antipeptide antibodies were prepared against peptides MGASDPEVAPWARGGAAG (from now on referred to as MGAS) and EARARDPPPAHRAH (referred to as EARA). The amino terminal end of peptide MGAS was blocked by acetylation, and amino acids lysine and cysteine were added at the carboxyl terminus to conjugate it to the carrier protein using the sulfhydril group of the cysteine. Peptide EARA was blocked in its carboxyl terminal side by amidation and was conjugated to the carrier protein through the free amino group at the amino terminal amino acid. Peptides were conjugated to hemocyanin from Limulus poliphemus (LPH) for immunization and to CNBr-activated sepharose for affinity purification. Two rabbits were immunized with each conjugated peptide using standard immunization protocols. Purified antibodies were stored at −20°C in Tris-glycine buffer, pH 7.5, with 250 mM NaCl and 0.02% thimerosal.
PSAP has two isoforms generated by alternative splicing.
During our first Basic Local Alignment Search Tool (BLAST) analyses of PSAP, we found that some est clones were missing an internal 51-bp region with respect to the sequence reported by Xu et al. (34). This was confirmed by reviewing the annotated human genes at the National Center for Biotechnology Information (NCBI) and by RT-PCR assays using human cDNA as a template. PSAP is annotated in GenBank as the mitochondrial carrier homolog 1 (Mtch1) and has two isoforms, Mtch1a (the shorter isoform) and Mtch1b (the larger isoform); in this report we refer to them as PSAPS (short) and PSAPL (large), respectively. The missing region in PSAPS corresponds to the last 51 bp of exon 8. This sequence starts with a GT dinucleotide, conforming to the GT/AG rule. This indicates that splicing of intron 8 can take place in two different ways: a canonical one that removes the whole intron and an alternative one that removes the last 51 nucleotides of exon 8 attached to intron 8 (Fig. 1B). Furthermore, Mtch1 has a putative start codon located upstream of that described by Xu et al. (34). The sequence named CGI-64 also corresponds to PSAP but starting at a third putative start codon located further downstream (Fig. 1A).
To determine which start codon is used in vivo, we developed antibodies against a peptide comprising the first possible 18 amino acids of PSAP (anti-MGAS) and another one against a peptide comprising residues 37–50 (anti-EARA), located between the second and the third possible start codons. Anti-EARA was able to detect both PSAP isoforms in Western blots using mitochondria from different cells and tissues (see Fig. 2B) but anti-MGAS was not. Both were able to detect the recombinant protein expressed in cells in culture, although anti-MGAS showed a slightly lower sensitivity than anti-EARA in titration assays (not shown). As another approach to solve this matter, we run recombinant proteins (with no tags) in parallel with endogenous ones. The results, shown in Fig. 2C, indicate that both PSAP isoforms starting at the first possible start codon migrate slower than the endogenous proteins, whereas recombinant PSAPL and PSAPS starting at the second ATG showed the same migration as the endogenous proteins. This suggests that the second ATG is used as start codon in vivo. Curiously, alignment of the residues encoded between the first and the second ATG (first possible 19 amino acids of PSAP) from several species (Fig. 2D) indicated that they are very well conserved, which suggests that the first ATG might still be used under some circumstances. This issue is further discussed in discussion.
It is also interesting to note that an internal 17-residue difference (between PSAPL and PSAPS) produces a much larger migration difference than an 18-residue difference in the amino terminus of the protein (between the first and the second ATG).
Although these results are presented here for convenience, they were not obtained until anti-PSAP antibodies were available, which was after the experiments with deletion mutants shown below had been carried out. This is the reason we made several PSAP constructs starting at the first possible ATG, which was the most likely one to be used in vivo at that time. Also the results presented above do not completely rule out the possibility that the first ATG is used as start codon under some circumstances. In any case, starting at either the first or second ATG does not make any difference with respect to mitochondrial localization and induction of apoptosis, as we show below, and therefore we refer to the recombinant proteins used numbering the residues starting at the first ATG, which also matches PSAP sequences available in gene databases.
The two PSAP isoforms are highly conserved in mammals.
The same two isoforms found in humans are also present in other mammals such as mouse (PSAPL, GenBank NM_019880; PSAPS, AB040292), rat (PSAPL, XM_215358; PSAPS fragment, CV115767), pig (PSAPL fragment, BW962845; PSAPS fragment, DN103176), bovine (PSAPL fragment, XM_582757; PSAPS fragment, DT859797), and dog (PSAPL fragment, XM_846846; PSAPS fragment, XM_859620). Apart from having the same alternative splicing sites, PSAP is highly conserved among these organisms, for instance, mouse PSAPL (GenBank NM_019880) is 98.7% similar and 98.2% identical to human PSAPL and rat PSAPL (XM_215358) is 98.5% similar and 98.2% identical to its human counterpart. The fact that the splicing sites are perfectly conserved among these organisms strongly suggests that the removal of 51 nucleotides, corresponding to 17 amino acids, from the larger isoform is responsible for an important functional difference between both isoforms.
Splicing removes a loop located between two putative transmembrane regions.
We carried out membrane topology predictions using five different programs. All five programs predicted two transmembrane domains located around amino acid positions 253–275 and 316–338 (slight differences in the limits of both segments were observed depending on the program used). Three programs (TMpred, HMMTOP, and TopPred) predicted a third putative transmembrane domain located around positions 82–102, although with a lower probability. TMpred also predicted a fourth putative transmembrane segment between residues 11 and 31, with the lowest probability of all four segments. Since transmembrane segments located around amino acid positions 253–275 and 316–338 are predicted by all the programs used, we refer only to them from now on.
Interestingly, the 51 bp removed by alternative splicing encode 17 amino acids located between both predicted transmembrane segments, leaving a hydrophilic loop of 23 amino acids in PSAPS with respect to the predicted 40-amino acid loop in PSAPL.
Both isoforms are ubiquitously expressed.
Xu et al. (34) analyzed the tissue levels of PSAP mRNA using a Multiple Tissue Northern Blot from BD-Clontech and concluded that PSAP levels are higher in the brain. Their Northern blot included mRNA from eight different human tissues and did not differentiate between both PSAP isoforms (51 nt difference in 1.9 kb mRNA). We wanted to find out whether any of both isoforms was tissue specific, which could give a hint about their possible differential function. We used BD-Clontech Multiple Tissue cDNA (MTC) (Fig. 2A, panels I and II) to determine whether the mRNAs for both PSAP isoforms were present in all 16 human tissues included in these panels or if any of them was tissue specific. Panel I contains cDNAs derived from the same eight tissues included in the Northern blot analysis of Xu and coworkers (34). We used primers located on each side of the alternatively spliced region to obtain a 370-bp PCR product for PSAPL and a 319-bp one for PSAPS, which could be separated easily by electrophoresis in 2% agarose-TBE gels (Fig. 2A). Controls with cloned PSAPL and PSAPS separately produced each of the expected products (not shown). We used primers for G3PDH as an internal control. The same results were obtained with singleplex PCR (primers for PSAP and for G3PDH in different tubes) as well as with multiplex PCR (all four primers in a single tube). Note that the cDNA amounts from tissues included in MTC panel I are slightly lower than the amounts of cDNAs included in panel II (Fig. 2A, more intense G3PDH product in tissues from panel II). Since the method used is only semiquantitative, we will not draw conclusions about the relative levels of both isoforms in different tissues, but from our data it can be confidently concluded that both isoforms are expressed in all tissues studied, with apparently the lowest levels in skeletal muscle. There are also subtle differences in the levels of the mRNAs for both isoforms in a given tissue.
To confirm that both protein isoforms were indeed present in cells, we carried out Western blot assays with anti-EARA antibodies in mitochondria from human cultured cells and from rat tissues. Figure 2B shows the results obtained with mitochondria from the rat heart, brain, and kidney. Both isoforms were also detected in mitochondria from human HEK293 cells (see Fig. 3) and HeLa cells (not shown).
Both PSAP isoforms are integral mitochondrial outer membrane proteins.
PSAPL has been previously shown to be a mitochondrial protein (35). We carried out subcellular fractionation to find out whether both isoforms were located in mitochondria. Figure 3A shows the results obtained with untransfected HEK293 cells using anti-EARA antibody to detect both PSAP isoforms by Western blot assays. Both isoforms are present in the mitochondrial fraction, although under these conditions they cannot be detected in the total cell lysate. Since PSAP has homology with mitochondrial inner membrane carriers and two likely transmembrane domains, we reasoned that it could be located in the inner membrane. To find out whether this was the case, we carried out protease K protection assays with mitochondria purified from HEK293 cells. Both PSAP isoforms disappeared upon treatment of intact mitochondria with protease K, as happened with Tom20, whereas solubilization of the membranes was necessary for digestion of cytochrome c (Fig. 3B). No fragments of PSAP could be detected in these assays, indicating that PSAP is an outer membrane protein with its amino terminal side (peptide EARA) exposed on the cytosolic side of the membrane.
We next studied whether both PSAP isoforms were attached to the outer membrane or were integral to it, by extracting purified mitochondria with sodium carbonate. Both PSAP isoforms were precipitated by this method, appearing in the pellet after centrifugation (Fig. 3C), as did VDAC, an outer membrane integral protein. Cytochrome c, an intermembrane space protein loosely attached to the inner membrane, but not integral to it, appeared in the supernatant, as expected. Therefore, both PSAP isoforms are integral mitochondrial outer membrane proteins.
Both PSAP isoforms induce apoptotic cell death when overexpressed in cells in culture.
To find out whether both PSAP isoforms induce apoptotic cell death when overexpressed in cultured cells, we transfected cells with six constructs expressing PSAPL or PSAPS starting at either the first, second, or third ATGs. When cells are analyzed a few hours after transfection, transfected cells can be observed that still show mostly intact nuclei and a distinct punctate pattern corresponding to mitochondria, although the EGFP signal is still weak. At longer times after transfection, i.e., 30 h, most cells appear rounded, nuclear chromatin is condensed or fragmented, and mitochondria appear rounded and clustered, which are typical signs of apoptosis. The same results were obtained with cells transfected with all six constructs. Figure 4A shows the results obtained by imaging HeLa cells 30 h posttransfection with the two constructs having the greatest size difference, either full-length PSAPL (Fig. 4A, panels A and B) or PSAPS-DN64 (Fig. 4A, panels C and D). The majority of transfected cells appear rounded, with shrunk nuclei and punctiform mitochondria, which tend to aggregate next to the nucleus. Most cells transfected with EGFP preceded by a mitochondrial targeting sequence (GB-EGFP), as a control, present filiform mitochondria (Fig. 4A, panel E) and rounded nuclei (Fig. 4A, panel F), typical of healthy cells. To corroborate these results by using a different technique, we analyzed PARP cleavage in transfected cells. PARP is a substrate of caspase 3 and is generally used as a marker of apoptosis, since during apoptosis it is cleaved from its 116-kDa intact form into 85- and 25-kDa fragments. Analysis of HEK293 cells transfected with both full-length PSAP isoforms by Western blot analysis using an anti-PARP antibody revealed that both proteins induced PARP cleavage, since its 85-kDa fragment was present in these cells (Fig. 4B). PARP cleavage induced by PSAPDN64 is shown as part of ⇓Fig. 6B.
These results indicate that both PSAP isoforms, even lacking the first 64 amino acids, are able to induce apoptotic cell death.
PSAP is imported into mitochondria using an internal targeting signal.
We studied the mitochondrial import of both PSAP isoforms by fluorescence microscopy using EGFP-tagged PSAP variants. We first constructed vectors encoding PSAP that lacked 18 (PSAP-DN18) or 64 (PSAP-DN64) amino terminal amino acids, fused to EGFP. Both deletions were generated in PSAPL and PSAPS. PSAPL-DN18 corresponds to the protein studied by Xu et al. (34, 35) and starts at the second possible ATG of PSAP. PSAPL-DN64 starts at the third ATG of PSAP, having the same amino terminus as GenBank sequence CGI-64. We used different cell lines in these assays, including HeLa, HEK293, IMR90, and pig LLCPK cells (from proximal renal tubule) and obtained similar results with all of them, although HEK293 tend to round up and detach from the culture surface sooner than the other cell types. All of these cell types showed the cytoplasmic punctate pattern typical of mitochondrial localization when transfected with PSAP-EGFP fusion constructs (not shown). In fact, PSAP with larger amino terminal deletions was still capable of being imported into mitochondria (as shown in Fig. 6B) (N112 and N168). These results indicated that PSAP is not imported into mitochondria using a cleavable amino terminal targeting sequence. Fusion of EGFP at the carboxyl terminus precluded the use of a less common carboxyl terminal cleavable targeting signal (17). These data indicate that PSAP uses an internal mitochondrial targeting signal.
Mitochondrial import and proapoptotic activity of deletion mutants.
To identify regions in PSAP that are important for mitochondrial import and induction of apoptosis, we first generated six different deletion mutants derived from PSAPL (Fig. 5). PSAPLDN112 and PSAPLDN168 contain amino terminal deletions that remove the first 112 and 168 amino acids of PSAPL, respectively. PSAPLDTM1 and PSAPLDTM2 lack transmembrane domain 1 (residues 253–275) and transmembrane domain 2 (316–338), respectively. PSAPLDLoop is missing the whole loop located between the two transmembrane domains (278–313). PSAPLDC51 lacks the carboxyl terminal side, beyond the second transmembrane domain (339–389). See Fig. 5 for diagrams of these mutants. These mutants were fused to EGFP to analyze transfected cells by fluorescence microscopy. In general, we chose HeLa cells for microscopy because they attach better to the culture surface and HEK293 cells for Western blot analyses due to their higher transfection efficiency. HeLa cells were initially transfected with the first nine constructs shown in Fig. 5 and analyzed 8–12 h or 22–24 h after transfection, either in vivo or fixed. We used the mitochondrial-targeting control mentioned above, GB-EGFP. Before fixation, cells were treated with MitoTracker Red to stain mitochondria and with Hoechst 33342 to stain the nuclei. Figure 6A shows confocal microscopy images of HeLa cells 8 h after transfection (PSAPLDN18 is not included in this figure because it behaved as PSAPLDN64). The same results were obtained when an anti-Tom 20 antibody was used to label mitochondria followed by detection with an Alexa Fluor 568-conjugated secondary antibody (not shown). Colocalization of EGFP-fused PSAP and MitoTracker was clearly seen for all constructs except for DTM1, which presented a diffuse cytoplasmic green staining that overlapped only partially with MitoTracker, indicating that mitochondrial import is altered in this mutant. When transfected HeLa cells were observed microscopically, chromatin condensation was evident for all mutants except for DN168 and DTM1, where most transfected cells still showed round nuclei 24 h posttransfection (not shown). Also, whereas most cells were rounded and detached 24 h posttransfection, cells transfected with DN168 and DTM1 mutants were mostly attached and had normal morphology when observed under phase-contrast microscopy (not shown). This suggests that apoptotic activity is reduced or missing in both DN168 and DTM1 mutants.
To corroborate these results, we analyzed transfected HEK293 cell lysates by Western blot analysis with an anti-PARP antibody. We included cells transfected with PSAPL, PSAPS, PSAPSDN64 and the six deletion mutants described above, as well as GB-EGFP. As expected, the results indicated that DN168 and DTM1 mutants showed much less of the 85-kDa PARP fragment (Fig. 5B, top; results obtained with PSAPS are not shown but were similar to the ones obtained with PSAPL; see Fig. 4). Actin was used as a loading control (Fig. 6B, bottom). To find out whether these differences in apoptosis could be due to different expression levels of different mutants, we reincubated the same blot with anti-GFP antibodies (Fig. 6B, middle). The blots shown in Fig. 6B are representative of three individual experiments. We did find differences in expression of different constructs. Surprisingly, DN168 and DTM1 mutants were expressed at higher levels than other mutants or wild-type proteins. Our interpretation of these results is that, since cells transfected with DN168 and DTM1 are not undergoing apoptosis, these cells can accumulate higher levels of these proteins. Curiously, it was also evident that DN112 and DTM2, and, to a minor extent, DLoop, showed reduced expression levels with respect to wild-type protein (PSAPL). Fluorescent microscopic examination of transfected cells using the same integration time for all constructs indicated that green fluorescence was weaker in cells expressing DN112, DTM2, and DLoop (not shown). Nevertheless, the approximate proportion of transfected cells was similar with all mutants, indicating a similar transfection efficiency. These could indicate that DN112, DTM2, and DLoop can kill the cells at lower protein levels than other mutants or wild-type proteins, although dose-effect studies would be necessary to address this point.
These results indicate that deletion of TM1 impairs mitochondrial import. DN168 is still targeted to mitochondria, but its apoptotic activity is seriously impaired, indicating that this deletion affects a proapoptotic domain not involved in mitochondrial import.
The mitochondrial carrier domain is not required for mitochondrial targeting and apoptotic activity.
The results presented so far suggested that the mitochondrial carrier domain (MCD) might be involved in mitochondrial import, since deletion of TM1, which partially overlaps the MCD, affected mitochondrial import. To find out whether the MCD was required for mitochondrial import and/or apoptotic activity, we generated a new construct to express PSAPL lacking the MCD (PSAPLDMCD). We also constructed another vector to express PSAPL without the region comprising residues 169 to 252 (PSAPLDI), i.e., the region located between that deleted in DN168 and the one deleted in DTM1 (see scheme in Fig. 5). We analyzed PARP cleavage and mitochondrial localization using HEK293 and HeLa cells transfected with these two new mutants. To our surprise, both mutants were able to localize to mitochondria and induce apoptosis, as shown in Fig. 6, A and C. This indicated that the MCD was not required for mitochondrial import and apoptotic activity. Furthermore, deletion of the MCD plus the preceding 37 residues (region deleted in PSAPLDI) did not affect these two activities either.
TM1 does not contain the mitochondrial targeting information but is required for insertion into the mitochondrial outer membrane.
The results shown so far suggested that TM1 contained the mitochondrial targeting information, since DTM1 was the only mutant that was not efficiently imported into mitochondria. Nevertheless, some overlapping of EGFP-tagged DTM1 with MitoTracker could be seen in the confocal microscopy images shown in Fig. 6A. To determine the mitochondrial association of DTM1 using other techniques, we first confirmed that recombinant proteins localized to the same compartment as the endogenous ones.
To do this, several aliquots of a mitochondrial preparation from HEK293 cells overexpressing PSAPL-EGFP were treated or untreated with protease K, with different concentrations of digitonin or with Triton X-100 and analyzed by Western blot analysis (Fig. 7A). PSAPL-EGFP was detected with anti-GFP antibodies. Antibodies against proteins from each mitochondrial compartment were used as controls. Untreated mitochondria showed the expected 69-kDa protein, which was missing in protease-treated mitochondria. As expected, Tom20 was slightly reduced in size, whereas cytochrome c was protected from digestion until digitonin was added. Note that, since lower concentrations of protease K were used in this experiment with respect to the one shown in Fig. 3B, Tom20 is first reduced in size before completely disappearing in Fig. 7A. These results indicated that recombinant PSAPL is imported into the outer mitochondrial membrane.
We then carried out subcellular fractionation followed by sodium carbonate extraction experiments to determine which fraction of DTM1 protein was associated with mitochondria. We fractionated cells overexpressing PSAPL or DTM1 in mitochondrial and cytosolic fractions and found a lower portion of DTM1 associated with the mitochondrial fraction with respect to PSAPL (Fig. 7B). When we extracted these mitochondrial preparations with sodium carbonate, our results indicated that a large part of DTM1 protein associated with mitochondria was not integral to the membrane, since it could be released by this treatment, appearing in the supernatant (S in Fig. 7C). PSAPL, on the other hand, was mostly present in the pellet (P), indicating that it is integral to the membrane. Furthermore, since DTM1 mutant did not colocalize with calnexin, an endoplasmic reticulum marker (data not shown), this indicates that this mutant does not associate with cellular membranes in a nonspecific manner, but rather specifically with mitochondria, suggesting that it can be recognized by mitochondrial import receptors but it cannot be inserted into the membrane.
To further explore the importance of TM1 for mitochondrial targeting and/or insertion in the membrane, we replaced TM1 with either of two transmembrane domains from the plasma membrane protein Pit2, a phosphate carrier (24). To determine whether the orientation of the transmembrane domain was important in this experiment, we used TM domains 2 (P2, inside-out orientation) and 3 (P3, outside-in orientation) from Pit2. These two TM domains were used to replace TM1 of PSAPL, constructing two vectors that encode PSAPLTMP2 (second TM of Pit2) and PSAPLTMP3 (third TM of Pit2) mutants, as shown in the scheme in Fig. 5. HEK293 and HeLa cells were transfected with both vectors and analyzed for apoptosis induction and mitochondrial localization of the encoded proteins. Data shown in Fig. 8A indicate that, whereas PSAPLTMP2 was mostly dispersed in the cytosol (although a portion of it colocalized with mitochondria), PSAPLTMP3 localized to mitochondria. None of them colocalized with calnexin (not shown). Furthermore, PSAPLTMP3 was able to induce apoptotic cell death, whereas PSAPLTMP2 was not (Fig. 8B). Sodium carbonate extraction of mitochondria from cells overexpressing PSAPLTMP2 indicated that the fraction of this mutant associated with mitochondria is not integrated into the membrane (Fig. 8C). Since we find it unlikely that the third transmembrane domain of Pit2 contains mitochondrial targeting information, the most plausible explanation for these results, together with the results shown in Fig. 7C, is that a TM domain with the correct membrane orientation and/or sequence is required for efficient mitochondrial insertion of the protein. Sequence conservation between TM1 of PSAP and TM3 of Pit2 was not higher than between TM1 of PSAP and TM2 of Pit2 (not shown).
Therefore, TM1 of PSAP does not contain the mitochondrial targeting information but is required for insertion of the protein into the outer membrane. Furthermore, since no other mutant showed impaired mitochondrial import, we must conclude that PSAP does not contain a unique mitochondrial targeting sequence but rather several motifs for recognition by the import machinery dispersed along the protein, as has been described for other mitochondrial outer membrane proteins (see discussion).
Two regions in the amino terminal side of PSAP are involved in apoptosis.
Results shown in Fig. 6 indicated that the region comprised between residues 113–168 was important for the apoptotic function of PSAP (see also scheme in Fig. 5), since removal of these 56 residues from mutant DN112 rendered the protein nonapoptotic. To confirm whether these residues were responsible for the apoptotic activity of PSAP, we constructed a new mutant that lacked only these amino acids. Curiously, this protein, PSAPLDAPO, was still able to induce apoptotic cell death, as shown in Fig. 9. This indicated that, although residues 113–168 were necessary for apoptosis in mutant DN112, they were not required for this function when the rest of the amino terminal region was present, i.e., there must be another region responsible for induction of apoptosis in PSAP when these residues are not present. To find out whether this was the case, we generated two new constructs that combined deletion of region 113–168 with deletion of regions 65–112 (PSAPLDAPO2) or 1–64 (PSAPLDAPO3, see scheme in Fig. 5). Deletion of region 65–168 rendered the protein nonapoptotic, whereas addition of residues 65–112 to mutant DN168 (PSAPLDAPO3) made it apoptotic. Seen from another point of view, addition of residues 65–112 or residues 113–168 to mutant PSAPLDN168 converts this nonapoptotic mutant into an apoptotic one. These results indicate that residues 65–112 (present in PSAPLDAPO3) or 113–168 (present in PSAPLDN112) in PSAP are required for apoptotic activity; this is, PSAP has two different and apparently independent domains involved in apoptosis.
Xu and coworkers (34, 35) described PSAP as a proapoptotic mitochondrial protein that interacts with PS1, but they carried out their work with only one PSAP isoform. We have shown here that PSAP has at least two isoforms, which we named PSAPL and PSAPS. Since human PSAP is encoded by a single gene (located in 6p24.1), these two isoforms must be generated by alternative splicing. Furthermore, the putative donor and acceptor splicing sites conform to the common GT/AG rule. The splicing of PSAP is perfectly conserved in several mammals, suggesting that it is responsible for an important functional difference between both isoforms. Secondary structure and topology analyses of PSAP indicated that it is a likely integral membrane protein with two transmembrane domains. Interestingly, the spliced-away region in the shorter version of PSAP (PSAPS) removes 17 amino acids from the hydrophilic loop located between the two transmembrane domains. RT-PCR analyses of both PSAP mRNAs in 16 human tissues indicated that PSAP is ubiquitously expressed, with apparent low levels in skeletal muscle. This suggests that both PSAP isoforms have a housekeeping function.
The PSAP reported by Xu et al. (34) had 371 amino acids. We have shown here that there is an in-frame start codon located further upstream; nevertheless, our results using anti-PSAP specific antibodies suggest that this upstream ATG is not used as start codon in the cells and conditions used in our assays. This is surprising, since the nucleotides surrounding this ATG conform to the Kozak sequence better that the ones surrounding the second ATG [GCGCCATGG in the first ATG vs. CGGGGATGG in the second ATG, with the consensus Kozak sequence being CC(A/G)CCATGG, where ATG is the start codon; (13)]. Also, the 19 amino acids from ATG one to ATG two are conserved in several mammals: human, chimp, pig, and dog are identical in this region; macaque and mouse differ from the former in one amino acid; rat, opossum, and cow show two amino acid differences in this region. This degree of amino acid conservation is usually seen among functional regions of proteins; therefore, it is intriguing why the protein appears to be synthesized starting at the second ATG, as our results suggest. Nevertheless, since these sequences are also relatively well conserved at the nucleotide level, with 72% identity among the above-mentioned species, it is also possible that the conservation reflects a role in translation regulation, for instance.
In any case, whichever start codon is used for translation of PSAP, the first or the second, it is irrelevant with respect to mitochondrial import and apoptotic activity of PSAP, since we have reported that mutants lacking the first 18 amino acids behave in the same way in these two aspects. Furthermore, a mutant lacking amino acids 19–64 with a COOH terminal FLAG tag was also apoptotic (not shown).
With respect to how PSAP is imported into mitochondria, our results indicate that a transmembrane domain in the position of TM1 appears to be essential for the insertion of PSAP into the outer membrane, whereas TM2 is likely inserted secondarily, not being necessary for the correct membrane orientation of PSAP in terms of its apoptotic activity. PSAP is likely recognized for import by the outer membrane import machinery receptors Tom70 and/or Tom20, which have been described to bind to several regions in proteins with inner mitochondrial targeting signals (4, 5, 28). This is consistent with our data, since no single deletion, except for the one that removes TM1, affects the mitochondrial localization of PSAP mutants, indicating that enough Tom70/Tom20-binding sites remain in all mutants. Since TM1 can be replaced by an unrelated TM domain (TM3 of Pit2) and still be functional, this indicates that this domain does not contain the specific mitochondrial targeting signal. Nevertheless, because TM2 of Pit2 is not able to functionally replace TM1 of PSAP, this indicates that the sequence or membrane polarity of the TM are important for correct PSAP insertion into the outer membrane. Our results suggest that most of the fraction of PSAPLDTM1 that colocalizes with MitoTracker is likely a fraction bound to the outer membrane import receptor but not anchored into the membrane, since it can be released from the membrane fraction by sodium carbonate extraction. The small fraction that remains bound to the membrane could perhaps be retained by anchoring through TM2 or through another hydrophobic region of the protein.
It is noteworthy that TM1 partially overlaps a conserved mitochondrial carrier domain (MCD, residues 206–266) and, therefore, we first assumed that the MCD was involved in mitochondrial targeting. To find out whether this was the case, we generated a construct in which this domain was deleted. To our surprise, this mutant was still able to be imported into mitochondria and also to induce apoptosis. Furthermore, since deletion of the MCD eliminates residues up to position 266 and does not affect mitochondrial import, this indicates that either amino acids 267–275 are sufficient to span the membrane or that surrounding hydrophobic amino acids are recruited as part of the TM domain in protein PSAPLDMCD (for instance, mostly hydrophobic residues F, I, N, A, Y, L on the right side of TM1).
When we generated the several deletion mutants used in this work, we were expecting that several of them would be unable to induce apoptosis. A hypothesis we were considering for the induction of apoptosis by PSAP was that it could participate in the formation of a channel by way of its two transmembrane domains (either oligomerizing or interacting with other membrane proteins). We first considered the idea that shortening the hydrophilic loop between both transmembrane domains, as seen in PSAPS, could prevent its proapoptotic activity. However, PSAPS, which is missing almost half of the loop with respect to PSAPL, is still able to induce apoptotic death. Furthermore, deletion of the whole loop did not prevent induction of apoptosis. Also, PSAP lacking the second transmembrane domain is still proapoptotic, which reduces the probability that its apoptotic activity depends on a channel-forming ability.
Our results indicate that the region located between residues 65 and 168 is important for the apoptotic activity of PSAP. It is noteworthy that mutant DN112 is still able to induce apoptosis but DN168 is not, which led us to think initially that a single apoptotic domain was located between residues 113–168. Actually, this is true for mutant DN112, with apoptotic activity that is eliminated upon removal of 56 amino acids from its amino terminal end (up to position 168). Nevertheless, when the same 56 amino acids are removed from the whole PSAPL protein (i.e., residues 1–112 are present and residues 113–168 are missing -mutant PSAPLDAPO), this protein is still able to induce apoptosis, indicating that another region is responsible for apoptosis in this mutant. Removal of residues 65–168 keeping residues 1–64 prevents apoptosis, but addition of residues 65–112 is sufficient to recover apoptotic activity. The most plausible explanation for these results is that there are two independent regions responsible for apoptotic activity. One would be located between residues 65–112 (responsible for apoptosis in mutants PSAPLDAPO and PSAPLDAPO3) and another one located between residues 113 and 169 (responsible for apoptosis in mutant PSAPLDN112). Both regions are present in all other mutants, although they are unable to induce apoptosis in the mutants that are not efficiently inserted into the outer mitochondrial membrane, PSAPLDTM1, and PSAPLTMP2.
Therefore, insertion into the outer mitochondrial membrane and presence of at least one of two apoptotic domains are necessary for PSAP to induce apoptosis. Analysis of the sequence of both apoptotic regions did not reveal any significant homology between them or similar hydrophobicity profiles.
Grinberg et al. (7) reported that Mtch2 is part of an outer membrane protein complex that recruits tBid and Bax upon treatment of FL5.12 cells with TNF-α. Mtch2 is 48% identical to PSAP/Mtch1, and these names reflect the fact that they contain a mitochondrial carrier domain. Mtch2, unlike PSAP/Mtch1, is not proapoptotic. Our experiments involving protease K digestion of mitochondria from transfected cells indicated that PSAP/Mtch1 is also located in the mitochondrial outer membrane.
As reported by Grinberg et al. (7), Mtch2 appears to have three transmembrane domains, whereas the programs we had used to analyze PSAP did not predict the first of these transmembrane domains. Actually, the PredictProtein server, used by Grinberg and coworkers, is the only program we found to predict a transmembrane domain comprising residues 207 to 224, corresponding to the first transmembrane domain reported for Mtch2. If PSAP has a third transmembrane domain like Mtch2, it would not be involved in the induction of apoptosis or in mitochondrial targeting, since that region is deleted in mutants DI and DMCD, as stated above.
During revision of this manuscript, it came to our attention an article by Yerushalmi et al. (36) describing that Met-hepatocyte growth factor/scatter factor (HGF/SF) induces Mimp (Met-induced mitochondrial protein). Mimp is the same protein as Mtch2. More recently, Leibowitz-Amit et al. (18), reported that induction of Mimp inhibits Met-HGF/SF-induced scattering and tumorigenicity by altering Met-HGF/SF signaling pathways. Taken together with the data reported by Grinberg et al. (7), these results indicate that Mtch2/Mimp participates in more than one signal transduction pathway, since, on one side, it mediates induction of the mitochondrial apoptotic pathway initiated by death receptor ligands, and, on the other hand, it is involved in signal transduction events mediated by tyrosine kinase receptors. Given the sequence homology between PSAP/Mtch1 and Mtch2/Mimp, the localization in the same cellular compartment (the mitochondrial outer membrane) and the fact that PSAP is able to induce apoptosis, it is likely that PSAP/Mtch1 is also involved in signal transduction pathways related to the control of cell growth and death.
Although both PSAP/Mtch1 and Mtch2/Mimp conserve a mitochondrial carrier domain, suggesting that they have a common evolutionary origin with some inner membrane carriers, it appears that these two proteins play different roles in the outer membrane than their homologs do in the inner membrane. Even when the known activities of both proteins differ, they have in common their participation in the apoptotic process, although clearly in different ways. Mtch2/Mimp is not proapoptotic by itself but it appears to be necessary for the induction of apoptosis by tBid in mitochondria. PSAP/Mtch1 is able to induce apoptotic cell death when overexpressed, but the endogenous protein is not causing apoptosis under normal conditions, indicating that some mechanisms are involved in regulating its apoptotic activity. It is unlikely that both PSAP/Mtch1 and Mtch2/Mimp will be involved in the transport of metabolites across the outer membrane in a similar way to carriers of the inner membrane, since these carriers are made up of three repeats of a region, each containing two transmembrane segments, which together form the substrate binding domain (15) and these repeats are not present in either PSAP/Mtch1 or Mtch2/Mimp. Furthermore, the outer membrane is permeable to many metabolites that need specific carriers to cross the inner membrane, so similar carriers are not necessary to transport those metabolites across the outer membrane.
Understanding the meaning of the interaction of PSAP with PS1 will be very interesting, particularly if we consider that all the components of γ-secretase, including APH-1, PEN-2, and nicastrin, as well as PS1 and PS2, have been recently reported to localize in mitochondria (8). Furthermore, if PS1 is located in the inner mitochondrial membrane, as reported by Ankarcrona and Hultenby (3), then its interaction with the outer membrane PSAP opens the possibility that both proteins are involved in some type of communication between the cytosol and the matrix.
With respect to how PSAP induces apoptosis, it will be necessary to carry out further research to identify possible PSAP-interacting proteins, involved in either transient interactions or as components of protein complexes. Structural information about the amino terminal region of PSAP could also give a hint about possible ways of inducing the apoptotic process.
In conclusion, our data indicate that PSAP has two proapoptotic isoforms ubiquitously expressed in human tissues and localized in the outer mitochondrial membrane. These isoforms use multiple internal signaling motifs for import into mitochondria and require the presence of a correctly oriented, or with a specific sequence, transmembrane domain for efficient insertion into the outer membrane. Insertion into the outer membrane and the presence of at least one of two independent domains are required for the apoptotic activity of PSAP. Experiments are needed to determine how PSAP induces apoptosis and how it interacts with presenilin 1.
V. Lamarca is a predoctoral fellow of the Diputación General de Aragón (B056/2004). J. A. Carrodeguas was a researcher of the Ramón y Cajal program. This work was funded by Grants BMC-2003-05265 from the former Ministerio de Ciencia y Tecnología, BFU2006-07026 from the Ministerio de Educación y Ciencia, by FEDER and by Grant UZ-2002-BIO-01 from the University of Zaragoza to J. A. Carrodeguas.
The authors declare that patent applications have been filed concerning the anti-PSAP antibodies and apoptotic domains described in this work.
We thank María Jesús Sancho for technical help. We thank Dr. Amancio Carnero for gift of HEK293 cells, Dr. Isabel Marzo for gift of HeLa cells and careful reading of the manuscript, Dr. Víctor Sorribas for gift of LLCPK cells and help with the Zeiss confocal microscope, and Dr. Yuri Lazebnik for gift of IMR90 cells. We thank Dr. Pilar Zaragoza for access to the Zeiss confocal microscope and Dr. Mario Barac for critical reading. We thank Drs. Daniel Bogenhagen and José A. Enríquez for critical reading and valuable advice.
Current address of A. Sanz-Clemente: National Institute of Neurological Disorders and Stroke, Receptor Biology Unit, NINDS Porter Neuroscience Research Center, Building 35, Room 2C-905, 35 Convent Dr., MSC 3704, Bethesda, MD 20892-3704.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2007 the American Physiological Society