A Molecular Switch for Targeting between Endoplasmic Reticulum (ER) and Mitochondria

Abstract

dAKAP1 (AKAP121, S-AKAP84), a dual specificity PKA scaffold protein, exists in several forms designated as a, b, c, and d. Whether dAKAP1 targets to endoplasmic reticulum (ER) or mitochondria depends on the presence of the N-terminal 33 amino acids (N1), and these N-terminal variants are generated by either alternative splicing and/or differential initiation of translation. The mitochondrial targeting motif, which is localized between residues 49 and 63, is comprised of a hydrophobic helix followed by positive charges (Ma, Y., and Taylor, S. (2002) J. Biol. Chem. 277 ,27328 -27336). dAKAP1 is located on the cytosolic surface of mitochondria outer membrane and both smooth and rough ER membrane. A single residue, Asp³¹, within the first 33 residues of dAKAP1b is required for ER targeting. Asp³¹, which functions as a separate motif from the mitochondrial targeting signal, converts the mitochondrial-targeting signal into a bipartite ER-targeting signal, without destroying the mitochondria-targeting signal. Therefore dAKAP1 possesses a single targeting element capable of targeting to both mitochondria and ER, with the ER signal overlapping the mitochondria signal. The specificity of ER or mitochondria targeting is determined and switched by the availability of the negatively charged residue, Asp³¹.

cAMP-dependent protein kinase (PKA)2x is an important regulator in cells, and it is involved in many cellular functions (1-4). Increasing evidence indicates that PKA fulfills its specific cellular regulation in part by subcellular localization in addition to selective recognition of its substrates (5). Subcellular localization of PKA is typically achieved through the binding of the regulatory subunits to scaffold proteins termed AKAPs (a kinase-anchoring proteins) where different AKAPs are expressed in different cell types. Depending on the targeting domain or motif an AKAP carries, it can bring PKA to different subcellular locations such as the nucleus, plasma membrane, endoplasmic reticulum (ER), Golgi, mitochondria, microtubules, etc. (6-8).

dAKAP1 (also known as AKAP1, S-AKAP84, and AKAP121) is one of the few dual specificity AKAPs that were identified through interaction with RIα for binding to both RI and RII (9). A unique feature of this protein is that it can bind both PKA-I and PKA-II. In addition to binding to PKA, dAKAP1 binds PP1, tubulin, a Myc-binding protein (AMY-1), PPD1, and mRNA (10-15). Several isoforms of dAKAP1 have been identified in mouse, rat, and human varying at both the N terminus and the C terminus, with the PKA-binding domain located within the common core in the center of the molecule (9, 16, 17). The ER/mitochondrial targeting domain of dAKAP1 is located at the N terminus (17, 18). This targeting domain also contains an overlapping mitotic spindle-targeting site (11). Variation in the N terminus can alter the localization of dAKAP1 to either ER or mitochondria, thereby changing the distribution of PKA in the cell (18, 19). The alternative anchoring of PKA on either mitochondria or ER implies that PKA may be involved in important cellular metabolism and regulations related to these two organelles, such as energy production, protein synthesis, Ca²⁺ signaling, and/or apoptosis (8, 14).

dAKAP1 bearing a shorter N terminus (dAKAP1a and 1c) targets exclusively to mitochondria, whereas the molecule bearing 33 additional residues (N1) at the N terminus (dAKAP1b and 1d) targets to ER (18). Similar to many known mitochondria-targeting signals, the mitochondria-targeting motif in dAKAP1 contains a hydrophobic region and positively charged residues to the immediate C terminus (19-25). The N1 segment, however, does not contain an abundance of hydrophobic residues and therefore does not increase the hydrophobicity of the targeting motif, as found in other ER-targeting motifs in the molecules targeting to both ER and mitochondria (24-27). However, how the extra 33 residues convert the mitochondrial targeting motif into an ER-targeting signal is not clear. To understand the role of PKA anchoring to ER and mitochondria and how this anchoring is regulated, we explored the molecular mechanism for dAKAP1 binding to both mitochondria and ER. Here we describe a novel mechanism that dAKAP1 employs to switch the anchoring of PKA between ER and mitochondria.

EXPERIMENTAL PROCEDURES

Reagents—Antisera raised in the lab against PKA-Cat and dAKAP1 were used to detect PKA-Cat and dAKAP1. Other antibodies are from commercial sources: anti-mouse RIIα from Santa Cruz Biotechnology, anti-CytC from BD Biosciences, anti-calnexin (against the cytosol-exposed C-terminal tail), and anti-calreticulin both from Abcam.

DNA Constructs and Sequence Analysis—Murine dAKAP isoforms 1a and 1b were subcloned into the pCMV-Sport1 vector (Invitrogen) containing a C-terminal FLAG tag (19). The N-terminal regions of dAKAP1a (amino acids 34-175) or dAKAP1b (amino acids 1-175) were subcloned into pCDNA3 (Invitrogen) using EcoRI and XhoI. The proteins were in vitro translated with the quick-coupled T7 TnT system (Promega) as previously described (19). The N-terminal sequences of dAKAP 1a and 1b (amino acids 1-63) were inserted into pEGFP-N1 (Clontech) with PCR. To avoid the interference of charged residues encoded by the polylinker, the sequence between KpnI and AgeI sites in the vector were replaced with a new linker (GGATCCGTGGCTCAGCTCACCGGT) that has a single BlpI recognition site. The sequences of anchoring domain were therefore inserted in between EcoRI and BlpI sites. Most of the mutations in the anchoring domain of dAKAP1 were made using Kunkel methods (28). A few difficult mutants were made by a modified quick change method (29). The mitochondria marker, mtGFP (30), was a gift from Dr. R. Y. Tsien (University of California, San Diego). The ER targeting marker, ER-GFP was a gift from Dr. D. G. Pestov (University of Illinois) (31). The genomic organization of dAKAP1 was generated with online-genomic Blast analysis.

Transfection Assay—10T 1/2 cells were cultured with Dulbecco's modified Eagle's medium containing 10% fetal bovine serum in 6-cm dishes containing 1.2-mm round coverslips (Fisher). The cells at 60-70% confluency were then transferred to a 24-well dish and transfected. Transfection was performed using either CytoFectine (Bio-Rad) or LipofectAmin-plus (Invitrogen) according to the manufacturer's protocols. When cells were co-transfected, different DNA constructs were incubated at an equal molar ratio, adjusted to the same total weight with blank vector. The cells were incubated for 6-7 h post-transfection before fixing and indirect immunofluorescence.

Immunofluorescence and Photo-microscopy—Immunofluorescent labeling was performed as previously described (19, 32). The cells were fixed with 4% paraformaldehyde in phosphate-buffered saline for 10 min followed by permeabilization with 0.3% Triton X-100 in phosphate-buffered saline for 15 min at room temperature. dAKAP1a was detected using a monoclonal αFLAG antibody (clone m2; Sigma) and a rhodamine-conjugated donkey anti-mouse secondary antibody (Jackson Laboratory, Bar Harbor, ME). To detect wild type and mutants of dAKAP1b, a polyclonal antibody against N1 (18) was used n conjunction with a rhodamine-conjugated donkey anti-rabbit antibody (Jackson Laboratory). The final concentration of the antibodies was 10 μg/ml. All of the antibodies were diluted in a solution containing 0.5% Nonidet P-40, 5 mg/ml bovine serum albumin in phosphate-buffered saline, pH 7.2, and incubated with cells at 37 °C. The coverslips were mounted on glass slides with Slowfade medium (Molecular Probes) and analyzed on a Leica microscopy equipped with a Mitsubishi digital camera.

Fractionation of Mitochondria and ER—Mitochondria and ER were isolated and purified from mouse liver according to Pagliarini et al. (33) and Sharma et al. (34). Mouse liver was homogenized with a motor-driven 30-ml Potter-Elvehjem tissue grinder five times in MSHE (210 mm mannitol, 70 mm sucrose, 5 mm HEPES-KOH, pH 7.4, and 1 mm EGTA). Nuclear material and unbroken cells were removed by centrifugation for 10 min at 2300 rpm (600 × g_max) in a Sorval SS34 rotor. Crude mitochondrial fractions were collected by centrifugation for 10 min at 11,200 rpm (15,000 × g_max). This fraction was then further purified by centrifugation on a Percoll-Histodenz gradient. The material between 17.5 and 35% Histodenz was collected as purified mitochondria (31). The post-mitochondria supernatants were fractionated by centrifugation at 100,000 × g for 1 h in a Ti70 rotor to collect crude ER as pellets (microsome) and supernatants (S100) (31). Crude ER was further fractionated on a discontinuous sucrose gradient by centrifugation at 29,000 rpm (100,000 × g_max) in a sw55 rotor for 16 h (34). Smooth ER was collected between 0.25 and 1.3 m sucrose, and rough ER was collected between 1.3 and 2.5 m. Protein (20 μg) from each fraction was analyzed by Western blotting. When fractions were protease treated, 5 μg of trypsin was incubated with the samples (350 μg of protein) for 0-10 min, followed by the addition of soybean trypsin inhibitor 100 μg and Complete (Roche Applied Science) proteinase inhibitors. When the fractions were washed with high salt and/or alkaline solutions, the samples (350 μg of protein) were incubated with either 200 or 500 mm KCl in 5 mm HEPES, pH 7.4, or 0.1 m sodium carbonate, pH 11.5 for 30 min on ice, according to previously published reports (35, 36).

RESULTS

The 33-residue N1 Fragment Is Generated by Alternative Splicing at the N Terminus of dAKAP1—dAKAP1 was cloned from a mouse embryonic library with multiple isoforms bearing several N and C termini (9). Different clones from sperm, S-AKAP84 and AKAP121, have similar C-terminal variations but lack the N-terminal variant (17). In Fig. 1 we show the exon organization of dAKAP1. Mouse genomic BLAST analysis of these cDNA clones and the newly deposited dAKAP1 alleles AK076216, AK132749, and BC065135 revealed that these various termini match different exons, indicating that these isoforms are generated by alternative splicing (Fig. 1A). For simplicity, only one allele is shown to represent the three. dAKAP1a is the shortest protein. It is encoded by exon 4 and then directly linked exon 5. dAKAP1b is identical to 1a at the 3′ end. Other isoforms have two additional 3′ compositions beyond exon 4: dAKAP1c, 1d, AKAP121, AK076216, AK132749, and BC065135 have exons 4 and 5 plus exons 7-15 at the 3′ end that encode two putative RNA-binding domains, a K homology (KH) domain, and a Tudor domain (Fig. 1, A and B); S-AKAP84 is similar to the above variants except that it uses exon 6 instead of exon 5. S-AKAP84 is similar to dAKAP1a in size, because exon 6 contains an in-frame stop codon that encodes 24 residues, whereas dAKAP1a has 20 residues encoded by exon 5 in the same place. At the 5′ end, however, dAKAP1c is identical to dAKAP1a.

FIGURE 1.

Multiple isoforms of dAKAP1 are generated by alternative splicing. A, exon organization and the splicing combinations in dAKAP1. The exons are shown as boxes or vertical lines, reflecting different sizes. The exons used in different isoforms are linked with thin lines. B, different isoforms of dAKAP1 have different combinations of functional elements. The shaded regions correspond to coding exons; the white boxes correspond to noncoding exons. The positions of the known functional domains are indicated. The localizations of each isoform are indicated on the right (17, 18). N0 corresponds to the 30-residue mitochondrial targeting domain (residues 34-63), and N1 correspond to the ER specific modifier motif (residues 1-33). AKB stands for PKA-binding domain. K homology (KH) and Tudor are the elements homologous to the corresponding RNA-binding domains. C, generation of ER targeting domain in dAKAP1b by exon 2 and exon 4. The junction of exons 2 and 4 is marked as a vertical line, and the sequences from part of exon 2 are underlined. The methionine encoded by exon 2 is marked with 1. The two methionines encoded by exon 4 are marked with 34 and 49, respectively.

There are three longer forms that are derived from differences at the 5′ end: one has exon 1 proceeding exon 4 (in AK076216 and BC065135), the second has exon 2 proceeding exon 4 (in dAKAP1b and 1d), and the third has exon 3 in front of exon 4 (in S-AKAP84, AK132749 and AKAP121). Exon 3 in AK132749 is 12 nucleotides shorter than AKAP121, probably because of strain variation. Because neither exon 1 or 3 have an in-frame initiation codon, translation of AKAP121, S-AKAP84, AK076216, AK132749, and BC065135 would start from exon 4, which corresponds to dAKAP1a and 1c, with the first initiation codon 32 nucleotides within the 5′ boundary of the exon 4. dAKAP1b and 1d, on the other hand, carry an initiation codon in exon 2 that is 68 nucleotides to the 3′ boundary and is in-frame with the protein encoded by exon 4. Therefore, in dAKAP1b and 1d, the merger of 68 nucleotides in exon 2 and the first 31 nucleotides at the beginning of exon 4 generates an extra fragment of 99 nucleotides encoding for 33 residues (Fig. 1C). The production of two isoforms at the N terminus by alternative splicing could therefore provide the primary mechanism to generate two anchoring proteins that target to different organelles in different tissues.

Targeting to Different Organelles Can Be Generated by Alternative Initiation of Translation from a Single mRNA—Previously, we identified two N-terminal splice variants of dAKAP1 (18). The first splice variant is found in isoforms 1a and 1c, initiates transcription at exon 4, and encodes a protein that is targeted to the mitochondria. The second splice variant is found in isoforms 1b and 1d, initiates transcription at exon 2, and encodes a protein that is targeted to the ER. In addition to alternative splicing, initiation of translation from different codons provides an alternative mechanism for generating two proteins from a single gene (37). In our initial studies, we wanted to examine whether different initiation codons exist within a single mRNA from dAKAP1. As diagramed in Fig. 2A, in dAKAP1, codons 1, 34, and 49 are all located in a consensus with optimal initiation of translation (38) and can, therefore, potentially encode several dAKAP1 variants by a leaky scanning mechanism (37). In vitro translation was used to determine whether multiple start sites in dAKAP1a and dAKAP1b exist. Fig. 2C shows that full-length dAKAP1b clearly has two distinct bands, one that co-migrates with dAKAP1a and a second slightly larger band, reflecting initiation from at least two positions (Fig. 2C). It was unclear from these experiments whether multiple start sites for dAKAP1a exist. To show the different initiation products more clearly, we analyzed the in vitro translation products with shorter constructs from the first 175 residues of dAKAP1a and 1b (Fig. 2D). In vitro translation of dAKAP1b generated two products from codons 1 and 34 (Fig. 2, C and D), whereas a mutant at Met³⁴ and Met⁴⁹ only produced the longer product (Fig. 2D). The shorter product in dAKAP1b is the same as the longer form of dAKAP1a and therefore should carry a mitochondria-targeting signal (Fig. 2B). Similarly, dAKAP1a can use the same mechanism to produce two products from codon 34 and 49 (Fig. 2D and Ref. 19). Mutation of Met⁴⁹ abolished the shorter product. Both products can target to mitochondria (19), although the shorter one lacks an intact microtubule-binding site (11). The generation of two molecules by differential initiation of translation sites for targeting to two organelles could also provide a mechanism for modulation of PKA distribution to different locations in cells.

FIGURE 2.

Differential initiation of translation of dAKAP1a and 1b generates multiple N termini. A, diagram of the dAKAP1b mRNA showing multiple initiation codons and potential leaky scanning initiation. The mRNA is depicted as dotted line, and the protein products are shown as wavy lines. B, amino acid residues of the N-terminal targeting domain of dAKAP1 showing several methionines that can serve as initiation sites. The functional elements in this domain are highlighted on the sequences, and the targeting motif is indicated below the sequences. C, in vitro translation products of dAKAP1a and dAKAP1b demonstrate multiple transcription start sites. D, in vitro translation of a fragment of dAKAP1 corresponding to the 5′ sequences up to 175 amino acids to facilitate visualization. Transcription from the alternate start sites in dAKAP1a were blocked when Met⁴⁹ was mutated to Leu. Similarly, the two alternate start sites in dAKAP1b were blocked upon mutation of Met³⁴ and Met⁴⁹ to Cys and Leu, respectively.

dAKAP1 Is Located on the Cytosolic Surface of Mitochondria and ER—To explore the mechanism of targeting and the potential role of dAKAP1 function in mitochondria and ER, we next wanted to biochemically characterize the localization of dAKAP1 on these two organelles. Mitochondria, smooth ER, and rough ER were isolated and purified from mouse liver as described under “Experimental Procedures.” dAKAP1 was primarily found in the mitochondria, smooth ER, and rough ER, with only a small amount in the post ER fraction (Fig. 3A). PKA-Cat is also found in all of the fractions. However, most of the PKA-Cat is in the post ER fraction, reflecting its multiple localizations and associations. Cytochrome c (CytC), a mitochondria protein, and calnexin, an ER protein, were used to measure the efficient separation of mitochondria and ER.

FIGURE 3.

Localization of dAKAP1 on cytosolic surface of mitochondria and ER. A, distribution of dAKAP1 and PKA-Cat were analyzed with Western blot in purified fractions from mouse mitochondria, rough, and smooth ER. CytC and calnexin were used as mitochondria and ER markers, respectively. Normalized protein (25 μg) from each fraction was loaded in the gel for analysis. B, purified mitochondria were treated with trypsin or washed with 0.2 or 0.5 m KCl or 0.1 m Na₂CO₃, and the association of dAKAP1, PKA-Cat, and RIIα with mitochondria was examined by Western blotting. Blotting for CytC was used to demonstrate that the mitochondria were intact. C and D, purified rough ER (C) and smooth ER (D) were treated with trypsin, or washed with 0.2 or 0.5 m KCl or 0.1 m Na₂CO₃, and the association of dAKAP1, PKA-Cat, and RIIα was examined by Western blotting. Calnexin was used as a marker for cytosolic facing proteins on ER, and calreticulin was used as marker for ER lumen proteins.

Purified intact mitochondria were treated with trypsin transiently to determine whether dAKAP1 is topologically accessible. dAKAP1 was rapidly degraded, suggesting that it is localized on the cytosolic surface of the outer membrane of mitochondria (Fig. 3B). Treatment with trypsin did not affect CytC, which is located in the intermembrane space. Type II PKA was also degraded in a similar manner as dAKAP1, suggesting that most of the type II PKA on mitochondria co-exist with dAKAP1 (Fig. 3B). We next washed the mitochondria with high concentrations of KCl (0.2 m and 0.5 m) and Na₂CO₃ (0.1 m, pH11.5) to determine whether the association of dAKAP1 to mitochondria is through specific insertion in the membrane (Fig. 3B) (35, 36). dAKAP1 was resistant to all of these washes. Alkaline treatment breaks the outer membrane and released CytC, but it does not wash away dAKAP1. Alkaline treatment removed most of the PKA-Cat but did not wash away RIIα. From this information, we conclude that dAKAP1 is inserted in the outer membrane of mitochondria.

A similar set of experiments was undertaken for the rough ER (Fig. 3C) and smooth ER (Fig. 3D) to determine the localization of dAKAP1. Similar to results obtained for mitochondria, dAKAP1 was sensitive to trypsin and digested. The cytosol-exposed C-terminal tail of calnexin was also digested, but calreticulin, a lumen-associated protein, was not. Therefore, dAKAP1 appears to be located on the cytosolic surface of both smooth and rough ER. Most type II PKA on rough ER was digested away similar to dAKAP1, suggesting that most of the PKA on rough ER is on the cytosolic surface. However, type II PKA was only partially digested away along with dAKAP1 on smooth ER, suggesting that a certain amount of PKA is located in the lumen of smooth ER (Fig. 3D). Washing rough ER and smooth ER with high concentrations of KCl or Na₂CO₃ did not disrupt dAKAP1 association with smooth or rough ER, although alkaline conditions seem to have broken ER membranes and washed away some of PKA-Cat, but not PKA-RIIα. Therefore, we conclude that the association of dAKAP1 with rough ER or smooth ER is mediated through specific insertion of the protein into the membrane. The localization of dAKAP1 to the cytosolic surface of both ER and mitochondria suggest that similarities exist in targeting dAKAP1 to these organelles.

A Negatively Charged Residue Is Required for ER Targeting—As previously established, dAKAP1b and dAKAP1d are alternatively spliced with exon 2 from the dAKAP1 gene and have 33 amino acids added to the N terminus (18, 19). These splice variants are localized to the ER rather than the mitochondria. Thus the extra 33 amino acids contain an element that modifies and converts the mitochondrial targeting signal in dAKAP1a into an ER-targeting signal. Our previous results indicate that deletion of the first 13 amino acids does not affect ER targeting in dAKAP1b (18). Therefore, the modifier element for ER targeting must be within residues 15-33. To determine the key element required for the conversion from mitochondrial to ER targeting, we conducted a series of substitution mutations within this region (Fig. 4). Of one double- and seven triple-amino acid scanning mutations tested, only one mutation region failed to target to ER (amino acids 31-33; Sub31A). As shown in Fig. 4B (panels g-i), this mutant targeted to mitochondria like dAKAP1a (Fig. 4B, panels a-c). To further identify the key element, residues 31-33 were mutated individually to Ala. Only Asp³¹ was found to be required for ER targeting; mutation of D31A targeted to mitochondria rather than ER (Fig. 4B, j-l). Other mutations at position 31 were tested. Mitochondrial localization was observed when aspartic acid 31 was replaced with asparagine, and tyrosine but not glutamic acid. Only the mutation to glutamic acid maintained ER targeting (Fig. 4B, panel m-o); the other mutations abolished ER targeting and targeted to mitochondria (Fig. 4A). The single acidic residue at position 31 is thus required in dAKAP1b to suppress the fused mitochondrial signal and convert it into an ER-targeting signal.

FIGURE 4.

A single acidic residue in dAKAP1b required for ER targeting. A, diagram of mutations made on the N-terminal 33 residues of dAKAP1b. Black boxes and letters denote mitochondrial-targeting mutations, and yellow boxes and red letters denote ER-targeting mutations. B, representative images of cells showing the localization of various dAKAP1 constructs. dAKAP1a (panel a) was labeled against the C-terminal FLAG tag with indirect immunofluorescence. dAKAP1b wild type and the mutants sub31A, D31A, D31E and S25D/D31A (panels d, g, j, m, and p, respectively) were labeled against the N-terminal sequences N1. Subcellular localization markers (green) and the merged images are labeled on the panels next to each dAKAP1 constructs. C, diagram of mutations on the potential phosphorylation site (Ser²⁵) in dAKAP1b and their effect on targeting. The substitution mutations are shown with the new residues over the site of changes. The full-length dAKAP1b is depicted as a wavy line.

The Negatively Charged Residue Asp³¹ Is a Separate Element From the Mitochondria-targeting Motif—It has previously been shown that the hydrophobic region and the positive charged residues are integral functional elements for mitochondria targeting (21). Because the negatively charged Asp³¹ is only two residues away from the first residue of dAKAP1a, we wanted to determine whether Asp³¹ integrates with the existing mitochondrial targeting signal to retarget dAKAP1 to the ER. To detect the integrity of the ER-targeting signal, two or four residues were inserted at the N-terminal junction between dAKAP1a and 1b (Fig. 4A). These insertions push the negative charge further away from the mitochondrial targeting sequence and introduce a half or one potential helix turn if an α-helix is formed near the junction. As summarized in Fig. 4A, neither insertion affected ER targeting, suggesting that the distance between the negative charge and the mitochondrial targeting signal is not critical and that there is no well defined structural element in this region (data not shown). Furthermore, these results indicate that no α helix is required to correctly position the negative charge, and the negatively charged residues are not integrated with the remainder of the targeting domain.

A putative PKA phosphorylation site exists at Ser²⁵, six residues N-terminal to Asp³¹. To determine whether a negative charge in this position can support ER targeting, we generated acidic amino acid substitutions, S25D and S25E, in dAKAP1b to mimic the phosphorylated state individually. Fig. 4B shows that neither S25D nor S25E had any noticeable effect on ER targeting. However, when the negative charge was removed in a double mutation (D31A/S25D or D31A/S25E), either Asp or Glu at residue 25 was able to suppress the mitochondrial targeting of D31A and restore ER targeting (Fig. 4, C and B, panels p-r). The newly introduced negative charge compensated for the missing negative charge in D31A and rescued the ER signal. These data agree with the insertion experiments, indicating that the positioning of the negative charges is not critical for ER targeting. It is likely that the element in N1 for ER targeting is not part of an integrated structure with the mitochondria targeting domain. The single negative charge in the N1 segment could therefore be a separate switch motif that serves as a suppressor to the mitochondrial-targeting signal.

The C-terminal Positively Charged Residues of the Targeting Domain Are Critical for Both ER and Mitochondria Targeting—It has been reported that the balance of hydrophobic and positively charged residues in the targeting domain is important for both ER and mitochondria localization. More hydrophobic residues promote ER targeting, whereas more positive charges prefer mitochondria targeting (21, 26, 39). One role of the positive charged element is to suppress binding of the hydrophobic segments to ER and promote mitochondria targeting (22-24, 26, 35). We wanted to determine the role of the positively charged motif in ER targeting and whether Asp³¹ promotes ER targeting through muting this suppressive positive charged element in the C terminus of the mitochondria-targeting motif (residues Arg⁶¹-Lys⁶²-Lys⁶³ and Arg⁶⁵). To test this, we made double mutations at Asp³¹ and at the positively charged residues in dAKAP1b (Fig. 5). We first generated the mutations in the full-length dAKAP1b. In these constructs, the run of five charged residues RKKDR(61-65) was replaced with three alanines (Ala4). All these mutations target to ER, independent of whether Asp³¹ was present or not (Fig. 5A). D31A mutation is unable to cause mitochondrial targeting without the positively charged element. However, if we only changed residues 61-63 to Alanine (SubA), we obtained a mitochondrial-targeting signal. It appears that this positively charged element is required for mitochondria targeting and acts as a suppressor of an ER-targeting signal, although this ER-targeting signal could be nonspecific (23, 24, 26, 39-41). This element has to be on the C-terminal side of the targeting domain. When we removed the charged residues at position 61-65 and placed two positively charged residues (RK) at the N terminus of dAKAP1a, or in front of the 15AA minimal targeting sequence to generate Ins34 and Ins49, respectively, mitochondria targeting were not restored (Fig. 5A).

FIGURE 5.

Dissection of functional elements in the targeting domain by mutagenic analysis. A, triple-alanine substitution mutations at the C-terminal positively charged region of the targeting domain. B, mutations expanded into the hydrophobic region of the targeting domain. C, GFP fusion constructs of the targeting domain mutations. Anti-N1 antibody was used along with GFP to detect the fusion proteins starting from Met¹. The pattern of diagram is the same as Fig. 4C. The internal deletion mutations are shown with a dotted line, and the GFP protein is shown as an oval in the targeting domain-GFP fusion constructs. A few representative images of the mutations are shown on the right of each panel to show the effects of the mutations on targeting.

The positively charged residues also have some different effects on ER and mitochondria signals. Deletion of basic amino acids 61-63 (RKR) retained ER targeting; however, longer deletions (amino acids 58-63 or 34-63) completely abolished ER targeting (Fig. 5B). Reducing the net positive charges by substituting residues 61-63 with three Glu (Sub(E)) does not affect targeting in dAKAP1a, whereas this same acidic mutation in dAKAP1b abolished all targeting (Ref. 19 and Fig. 5B).

To study how Asp³¹ generates an ER-targeting signal with the positively charged element in the targeting domain, we put the double mutations into constructs that contained the N-terminal targeting domain fused to GFP. Asp³¹ was replaced with Ala and RKK (61-63) was replaced with three alanines (Ala₃). We also removed all charged residues from the linker region of the vector, pEGFP, to avoid possible interference. An anti-N1 antibody was also used along with GFP to detect the translation products starting from Met1. As expected, the wild type targeting domain fusion protein targets to ER and the mutant 1-63 (D31A) targets to mitochondria (Fig. 5C). Ala₃, however, destroyed all targeting capability, regardless of whether Asp³¹ was present. Thus the positive charges appear to be necessary for both ER targeting and for mitochondrial targeting as shown previously (19). The presence of Asp³¹ muted the mitochondria-targeting signal and formed a new ER-targeting signal. Both positive and negative charged elements are required for ER targeting. dAKAP1 will target to mitochondria if a negative charged element is missing from the N1 segment.

DISCUSSION

dAKAP1 has several isoforms that can be generated by alternative splicing. All have a PKA-binding element in the center, with variations at the N and C termini. At the N terminus, two isoforms (dAKAP1b and 1d) have exon 2 and generate a 33-residue modifier segment that is essential for ER targeting. dAKAP1a and 1c lack exon 2 and target to mitochondria. At the C terminus, dAKAP1 has three forms. The longer form includes two RNA-binding domains, a K homology and a Tudor domain, whereas the shorter form does not have these RNA-binding elements. A putative caspase site exists at the junction between the shorter and longer forms (Asp⁵⁶⁹ in dAKAP1d) and may lead to the conversion of an RNA-binding form into a non-binding form post-transcriptionally. In the previous study, we have found that a helix (residues 49-57) is essential for targeting to both mitochondria and ER (19). For mitochondria targeting an amphipathic helix is sufficient, whereas a fully hydrophobic helix is required for ER targeting. In dAKAP1a and 1c, the targeting domain is composed of this hydrophobic core with a positively charged patch at the immediate C terminus. The addition of 33 additional residues in 1b and 1d suppresses the mitochondrial targeting signal and converts it into an ER-targeting signal. In this study, we found that dAKAP1 is located on the cytosolic surface of the mitochondria outer membrane and ER membrane. We also took a mutagenic approach to study the N-terminal bi-functional targeting motif of dAKAP1 for its role in mitochondria and ER targeting (summarized in Fig. 6). We demonstrate here that a single amino acid, Asp³¹, is critical for this mitochondria-ER switch. With this unique targeting signal, the suppression by Asp³¹ provides an organelle-specific mechanism for a subcellular switch and therefore provides an easy way to regulate the switch of the localization of PKA.

FIGURE 6.

Schematic diagram of the targeting domain of dAKAP1. Asp³¹ suppresses the mitochondrial targeting signal and converts it into an ER-targeting signal in dAKAP1b. The sequences required for mitochondria targeting are the C-terminal positively charged residues and a few hydrophobic residues on the opposite side of an α-helix. It requires more hydrophobic residues for ER localization. In addition to the C-terminal positively changed residues, the targeting signal requires a run of hydrophobic residues plus a negatively charged residue to the N-terminal side. The essential elements are highlighted with lines and boxes, and the essential residues identified are marked with dots and small boxes.

This negative charged switching element is in a separate motif from the mitochondria-targeting signal. It can be moved further apart and be substituted by a different negative charged residue at another location. The existence of a negatively charged residue helped convert a mitochondria signal into an ER signal. It seems that the ER signal uses a novel mechanism to preferentially associate with ER in cells that does not allowing the molecule to access mitochondria. Whether there are any interacting proteins that may help ER targeting for dAKAP1 needs to be confirmed.

Several molecules that have overlapping targeting domains are known to target to both ER and mitochondria including Bclx, Cytb5, TOM 20, TOM 70, VAMP1, and cytochrome P-450, etc. (22, 25, 26, 35, 41, 42). Binding to signal-recognition particle (SRP) on ER is mediated by the hydrophobic regions co-translationally (22, 35, 42). These molecules can also change their preferences from ER to mitochondria and are transported to mitochondria post-translationally but with distinct mechanisms (22, 35). Several isoforms of cytochrome P-450 have an N-terminal “chimeric signal” domain for ER and mitochondria targeting, which is similar to dAKAP1 (27, 42, 43). Switching from ER to mitochondria is achieved by proteolysis in P4501A1 to remove an N-terminal segment and generate a shorter mitochondria signal (27, 43). However, P4502A1 and P4502E1 use cAMP-dependent phosphorylation on Ser¹²⁸ or Ser¹²⁹, respectively, to increase the affinity to mitochondria (42, 44). The phosphorylation can activate a “cryptic” mitochondrial targeting signal embedded within the targeting domain and increased the affinity of the molecules to chaperones, and this promotes mitochondria binding (42, 44). Although Ser²⁵ in dAKAP1b is a potential PKA phosphorylation site, the effect of the phosphorylation on Ser²⁵ is unknown. Treatment of the cells with forskolin has no effect on the targeting of dAKAP1b (data not shown). Treatment of the in vitro translation products of dAKAP1b or mutant D31A with PKA-Cat also has no effect in an in vitro mitochondria binding assay (data not shown). How dAKAP1 attaches to mitochondria and ER and whether phosphorylation by PKA has any effect on the affinity to the organelles needs to be determined.

It has been shown that the positively charged element can suppress nonspecific signal-recognition particle binding mediated by hydrophobic motifs and promote mitochondria targeting (22, 35). Such a nonspecific ER targeting element seems to exist C-terminal to the positively charged motif in certain dAKAP1 mutants (such as Ala4) when the suppressive positive motif is removed. However, this downstream ER element may not contribute to ER targeting in wild type dAKAP1, because the targeting signal is fully contained within the N-terminal 63 residues (Fig. 5). There are no residue requirements that are C-terminal to residue 63 for ER targeting. In addition, the positively charged motif is required for ER targeting and is probably a part of the ER signal. Kaufmann et al. (24) proposed that the positive element could also prevent the molecule from leaking out during transport to mitochondria and in this way enhance mitochondria targeting. We postulate that Asp³¹ does not negate this effect to reduce mitochondria targeting and enhance ER targeting, because we do not see much change of mitochondria targeting in vitro. Our data suggest that Asp³¹ altered the mitochondria-targeting signal and transformed it into an ER signal.

In summary, the production of two isoforms at the N terminus, by alternative splicing, provides a mechanism to generate two PKA-anchoring proteins that target to different organelles. Generation of several dAKAP1 molecules with different initiation sites for translation could be an alternative way to produce multiple AKAPs that target to ER, mitochondria, and microtubules. The mechanism of switching from mitochondria to ER targeting by adding a single negatively charged residue is an economic way to modulate different targeting locations, and this can be further augmented by phosphorylation. This switch mechanism alters the targeting domain but does not need to encode an entirely new protein.