Weak mitochondrial targeting sequence determines tissue-specific subcellular localization of glutamine synthetase in liver and brain cells

Glutamine synthetase [GS; L-glutamate-ammonia ligase (ADP forming); EC6.3.1.2] catalyzes the ATP-dependent formation of glutamine from glutamate and ammonia and therefore plays an important role in ammonia detoxification and in the recycling of the neurotransmitter glutamate. In neural tissues, GS expression is confined to astrocytes and is cytoplasmic in all examined vertebrates (Kennedy et al., 1974; Linser and Moscona, 1983; Smith and Campbell, 1983; Smith and Campbell, 1987). However, the subcellular localization of GS in liver cells is dependent on the ammonia detoxification system used. In higher vertebrates, such as mammals, which utilize the ureotelic system of ammonia-detoxification, GS is cytoplasmic in cells of both liver and neural tissue (Smith and Campbell, 1988; Wu, 1963). By contrast, in the marine elasmobranchs, such as dogfish shark, which utilize the ureosmotic system of ammonia detoxification, GS is cytoplasmic in neural tissue but mitochondrial in liver cells (Smith et al., 1987). Mitochondrial localization of hepatic GS is also required for the uricotelic system of ammonia detoxification. This system has apparently evolved as a water-conserving mechanism in the dinosaurs and their kin, and is utilized today by several species including birds (Campbell et al., 1987). In this system, intramitochondrially created ammonia is converted to glutamine by the action of GS, and this is followed by synthesis of the excreted final product, uric acid (Vorhaben and Campbell, 1972). The hepatic GS in these species is confined to the mitochondrial matrix of all liver cells (Smith and Campbell, 1987; Vorhaben and Campbell, 1972; Vorhaben and Campbell, 1977). Thus, uricotelic and ureosmotic vertebrates share a common trait of targeting the GS enzyme to the mitochondria in liver cells and cytoplasm in neural tissue.

Proteins targeted to the mitochondrial matrix are often expressed as a pre-polypeptide with an N-terminal extension that contains a mitochondrial targeting signal (MTS). These targeting presequences are commonly between 10 and 80 amino acids long and are capable of forming an amphipathic α-helix that is rich in basic and hydroxyl residues and lacks acidic residues (von Heijne, 1986). Import into the matrix is dependant on two energy sources: the mitochondrial membrane potential (ΔΨ), which exerts an electrophoretic pulling force on the positively charged MTS, and ATP hydrolysis in the matrix (Mokranjac and Neupert, 2008). Once inside the matrix, the MTS sequences are proteolytically cleaved, although in a number of cases the signals remain uncleaved (Neupert, 1997).

In the dogfish shark, which utilizes the ureosmotic system for ammonia detoxification, we have recently found that the differential targeting of GS is achieved by tissue-specific alternative splicing (Matthews et al., 2005). This species contains a single GS gene, but express two isoforms of the GS protein: a large isoform that occurs mainly in the mitochondrial compartment of liver cells, and a smaller isoform in the cytoplasm of brain cells (Laud and Campbell, 1994). The formation of neural and liver isoforms and their differential subcellular localization is controlled by an alternative splicing process, which generates two different GS transcripts (Matthews et al., 2005). The liver transcript contains an alternative exon that is not present in the neural one. This exon leads to acquisition of an upstream in-frame start codon (uAUG), the addition of 29 residues to the N-terminus of the molecule, and formation of a MTS. Therefore, the liver product is targeted to the mitochondria, whereas the neural one is retained in the cytoplasm.

Chicken, which utilize the uricotelic system for ammonia detoxification, also contain a single copy of the GS gene (Patejunas and Young, 1987) and target the GS enzyme to the mitochondria in liver cells and to cytoplasm in neural tissue. Although regulation of GS expression in chicken has been extensively studied in the past (for a review, see Vardimon et al., 1999), the mechanism for differential targeting of GS remained largely unknown. Here we show that in chicken, unlike dogfish shark, tissue-specific subcellular localization is not achieved by alternative splicing, but rather by a novel mechanism that can differentially target an identical protein to the mitochondria in hepatocytes and to cytoplasm in astrocytes. This mechanism relies on the evolutional selection of a ‘weak’ MTS, which is known to depend on a highly negative ΔΨ for import. We show that the ΔΨ in hepatocytes is considerably more negative than in astrocytes and that converting the targeting signal of GS into ‘strong’ MTS abolished the capability to confer tissue-specific subcellular localization. Thus, although the GS gene has been highly conserved during evolution, the mechanism for tissue-specific subcellular localization has evolved independently twice.

Differential targeting of chicken GS was microscopically observed following immunostaining of monolayer cultures of primary hepatocytes and astrocytes with anti-GS antibody and with MitoTracker (a mitochondrial marker). In agreement with previous studies (Smith and Campbell, 1983), GS was mitochondrial in hepatocytes and cytoplasmic in astrocytes (Fig. 1A). Similar results were also obtained by western blot analysis of mitochondrial and cytoplasmic fractions of liver and brain tissues, using antibodies against the mitochondrial protein mHsp60 and the cytoplasmic protein tubulin as controls. Here too, accumulation of GS was mitochondrial in liver cells and cytoplasmic in the brain (Fig. 1B).

As described above, differential targeting of GS in liver and neural tissue in marine elasmobranchs such as dogfish shark is achieved by an alternative splicing: the liver cell splice product encodes a MTS, whereas the neural tissue splice product lacks this sequence (Matthews et al., 2005). To determine whether this GS targeting mechanism has been evolutionary conserved, we examined the possible presence of differentially spliced GS transcripts in chicken liver and brain. Northern blot analysis revealed no size difference between GS transcripts from the two tissues (Fig. 2A). Similarly, RT-PCR products obtained by using primers for exon 1 and 7, the two end exons of the GS gene, displayed no difference in size (Fig. 2B). This is consistent with a previous report that demonstrated that sequences of two cloned GS cDNAs from chicken liver and brain were identical (Campbell and Smith, 1992). Yet, considering that a cDNA clone represents a single transcript, this finding does not exclude the possibility that liver cells contain other GS transcripts that differ from those of brain. Signals that target proteins to the mitochondrial matrix are usually encoded by the 5′ region of the transcript. This is the case in dogfish shark, where GS contains the initiator AUG codon in exon 2 and an alternative exon of only 95 bases, which leads to acquisition of a mitochondrial targeting signal in intron 1 (Matthews et al., 2005). Because the initiator AUG codon of chicken GS is also located in exon 2, we examined the possible presence of an upstream alternative exon. To this end, we sequenced intron 1 of the GS gene (GenBank accession no. EU369427) using the chicken genomic GS clone pGS116 (Vardimon et al., 1986), and subjected the sequence to analysis by MTS prediction programs (PsortII, MitoprotII, Predotar, Mitopred and TargetP). Several potential MTS regions were identified within intron 1 (Fig. 2C). PCR or RT-PCR analysis, using primers for sequences upstream or inside the potential MTS regions, generated products of anticipated sizes on the genomic GS clone (Fig. 2D, lanes 2-6), but not on liver or brain RNA (Fig. 2D, lanes 8-12 and 14-18). Consistently, RT-PCR analysis using primers for exon 1 and 2 resulted in a single fragment in both liver and brain (Fig. 2D, lanes 7 and 13, indicated by an arrow) that corresponded to the size predicted from splicing out of the whole intron 1 region. Taken together, these findings suggest that chicken liver and brain cells contain an identical GS transcript and that tissue-specific subcellular localization of chicken GS is not achieved by a mechanism of alternative splicing as found in elasmobranchs.

Differential targeting of a protein product might also be achieved by alternative initiation of translation, which leads to the production of proteins that differ in their N-terminal domain and molecular mass. However, Western blot analysis revealed no size difference between liver and brain GS (Fig. 3A). Considering that the N-terminus of mitochondrial proteins is often cleaved on entry into the mitochondrial matrix, the difference between liver and brain GS might be diminished and not detectable by western blot analysis. Therefore, we compared sequences of the GS protein from brain and liver using mass spectrometry. Liver and brain GS were isolated by immunoprecipitation (Fig. 3B), cleaved with trypsin, AspN or GluC and subjected to analysis by mass spectrometry (Fig. 3C). We obtained coverage of 68% of the primary GS sequence (Fig. 3C, underlined) including the first 85 residues at the N-terminal of the protein. The results revealed that the N-terminus of liver GS is not cleaved on mitochondrial translocation and that the detected amino acid sequences of liver and brain GS are identical. In both tissues, the initial methionine was removed and, apart from acetylation of alanine at position 2, no other post-translational modifications in the sequenced fragments were detected.

Most mitochondrial matrix proteins carry N-terminal targeting sequences, termed MTS, although some might contain MTS sequences in the C-terminal domain (Folsch et al., 1998; Lee et al., 1999). To probe the existence of an N- or C-terminal mitochondrial targeting signal in chicken GS, we fused the first 50 N-terminal or the last 50 C-terminal residues of the protein upstream (pCh-50N-EGFP) or downstream (pEGFP-Ch-50C) to the reading frame of the reporter EGFP, respectively. As controls, we used the reporter construct that contains the MTS sequence of dogfish shark GS fused in frame to EGFP [pDf-MTS-EGFP (Matthews et al., 2005)], and the EGFP reporter alone (pEGFP-N1). Monolayer cultures of hepatocytes and astrocytes were transfected with the different constructs, and the subcellular localization of the chimeric proteins was assayed. The cells were stained with MitoTracker to identify mitochondrial localization. It should be noted that not all cells are transfected and consequently more cells are seen stained with MitoTracker than express EGFP. In cells transfected with the pEGFP-N1 construct, EGFP was homogenously distributed in the cytoplasm and nucleus of both hepatocytes and astrocytes (Fig. 4C). A similar pattern was observed in cells transfected with the chimeric C-terminal construct, p-EGFP-Ch-50C (Fig. 4A). There too, the chimeric EGFP was homogenously distributed in hepatocytes and astrocytes, indicating that the attached GS residues did not contain a localization signal. By contrast, transfection of the N-terminal pCh-50N-EGFP construct resulted in a cell-type-specific distribution of EGFP (Fig. 4D): similarly to the endogenous GS, this Ch-50N-EGFP was mitochondrial in hepatocyte cells, but mainly cytoplasmic in astrocytes. This cell-type-specific distribution was not due to an intrinsic failure of astrocytes to transport EGFP to the mitochondria because pDf-MTS-EGFP (harboring the dogfish MTS) was mitochondrial in both hepatocytes and astrocytes (Fig. 4B).

Next, we examined whether GS can be imported into liver mitochondria in a standard in vitro import reaction. Purified chicken liver mitochondria were isolated and incubated with in vitro translated GS. Import into mitochondria was determined by resistance to externally added proteinase K. We detected essentially no import of GS or of GS-DHFR (GS N-terminal 50 amino acid residues fused to DHFR) (Fig. 5A, bottom and top left panels, respectively). By contrast, Su9-DHFR, a known control of mitochondrial matrix targeting that is cleaved upon entry (Karniely et al., 2006) (Fig. 5A, top right panel), or Su9-GS (Fig. 5A, bottom right panel) were imported efficiently. Similar results were obtained when import was assayed under coupled translation and import conditions (Knox et al., 1998), in the presence of elevated concentrations of NADH (8 mM), ATP (4 mM) or succinate (4 mM) or following denaturation with urea (not shown). These findings suggest that the N-terminal residues of chicken GS establish a non-conventional MTS that requires for import mitochondrial properties and/or cellular components not available in an in vitro system.

Tissue-specific modification of MTS residues represents an attractive mechanism for differential targeting of GS. Considering that phosphorylation of MTS residues has been shown to affect mitochondrial import (Amutha and Pain, 2003; Anandatheerthavarada et al., 1999; Robin et al., 2003), we decided to examine the possible involvement of tissue-specific phosphorylation in differential targeting of GS. Inspection of the GS N-terminal 50 amino acid sequence, using a program that predicts Ser/Thr/Tyr phosphorylation sites (NetPhos), revealed two main candidates for phosphorylation: Tyr17 and Ser7. Substitution of Tyr17 with the phosphor-mimetic Glu residue (pCh-50N-EGFP_Y17E) was indeed sufficient to impair mitochondrial import: the chimeric protein was cytoplasmic not only in astrocytes, but in hepatocytes as well (Fig. 5B). By contrast, substitution of Tyr17 or Ser7 with Ala (pCh-50N-EGFP_S17A or pCh-50N-EGFP_Y17A, respectively), an amino acid residue that cannot accept a phosphor group, had no effect on the subcellular localization of the chimeric protein: in both cases EGFP was mitochondrial in hepatocytes and cytoplsmic in astrocytes (Fig. 5B). These finding suggested that tissue specificity is not facilitated by differential phosphorylation of these residues. In agreement, MS/MS analysis did not detect phosphorylated residues in brain or liver GS and no phosphorylation was observed by western blot analysis of immunoprecipitated GS molecules using anti phospho-tyrosine Abs (not shown).

Another possible mechanism for differential targeting of GS is based on the fact that translocation of mitochondrial proteins across the inner mitochondrial membrane is dependent on the magnitude of ΔΨ (Huang et al., 2002; Martin et al., 1991). ΔΨ is created by pumping protons from the matrix to the inter-membrane space in conjunction with electron transport through the respiratory chain, and facilitates import by exerting an electrophoretic effect on the positively charged MTS. We examined whether the magnitude of ΔΨ of hepatocytes is different from that of astrocytes. We performed live-cell imaging of hepatocytes and astrocytes stained with tetramethyl rhodamine methyl ester (TMRM), a mitochondria-specific fluorescent cation that accumulates in the mitochondrial matrix according to ΔΨ. The acquired images were analyzed by a computer-assisted method for automated quantification of ΔΨ (Koopman et al., 2008). Remarkably, the results showed that TMRM intensity of hepatocytes is about 40% higher than that of astrocytes and, therefore, that the ΔΨ in astrocytes is substantially less negative than that in hepatocytes (Fig. 6A,B). Further quantification of the TMRM images also revealed that hepatocyte mitochondria are more branched, larger and longer than those of astrocytes (Fig. 6C—E). Analysis by electron microscopy showed that hepatocyte mitochondria are internally more complex, containing intricate folding of the mitochondrial inner membrane (Fig. 6F). Given that the inner membranes contain the respiratory chain components, it is possible that the magnitude of ΔΨ is a reflection of mitochondrial morphology (Benard and Rossignol, 2008).

The magnitude of ΔΨ required for mitochondrial import has been shown to depend on the length and net positive charge of the targeting sequence: proteins with a short or less positively charged MTS require a higher electrical potential for import than do proteins with a longer or a more positively charged MTS (Huang et al., 2002; Martin et al., 1991). To identify more precisely the targeting signal of GS, we inspected the N-terminal 50 residues by secondary structure prediction programs (PSIPRED, HNN, Jpred, SCRATCH and PredictProtein). The results revealed that, similarly to human GS, whose crystal structure has recently been resolved (Krajewski et al., 2008), the N-terminal residues of avian GS comprise three structural elements: an α-helix at the most N-terminal region and two extended strands downstream to the α-helix. Helical wheel projection of the predicted α-helical region (residues 2-19) showed that, similarly to known MTSs, this region can form a positively charged amphipathic α-helix (Fig. 7A). We examined whether GS N-terminal residues, which encompass the α-helix structure, constitute a functional MTS. Analysis of chimeric EGFP constructs that contain the first 37 (pCh-N37-EGFP), 23 (pCh-N23-EGFP) or 14 (pCh-N14-EGFP) residues of GS or a 50 amino acid segment spanning between residues 35 and 84 (pCh-35-84-EGFP), revealed that the first 23 residues, which include the complete α-helix structure, are indeed sufficient to confer mitochondrial localization in hepatocytes and cytoplasmic localization in astrocytes (Fig. 7B, see also supplementary material Fig. S1). This short targeting sequence constitutes a weak MTS: it has particularly low scores from MTS prediction programs (MitoProtII, PsortII, TargetP) (e.g. with MitoProtII it scores 0.136 compared to the MTS of dogfish GS that scores 0.692). It contains only three positive residues, all of which are lysine, and two histidines that have a pKa of approximately 6.0 and probably do not contribute positive charges under physiological conditions.

Considering that the ΔΨ in hepatocytes is more negative than in astrocytes it is possible that the weak MTS provides a tool for differential targeting of GS: In hepatocytes, the highly negative ΔΨ might exert a sufficiently strong pulling force to drive translocation of GS, whereas in astrocytes, ΔΨ might be below the threshold required for GS import. We examined whether converting the targeting signal of GS into a strong MTS, by elevating the net positive charge or extending the size of the targeting sequence, would compensate for a less-negative ΔΨ and facilitate mitochondrial targeting also in astrocytes. Indeed, substitution of the histidine residues with arginine (pCh-50N-EGFP_H15R and pCh-50N-EGFP_{H8R, H15R}), a positively charged residue under physiological pH, resulted in mitochondrial localization in both hepatocytes and astrocytes (Fig. 7C, see also supplementary material Fig. S1). Similarly, extending the size of the MTS by including two copies of the targeting sequence (pCh-N23 × 2-EGFP) facilitated mitochondrial targeting also in astrocytes (Fig. 7D, see also supplementary material Fig. S1). These findings suggest that avian GS has evolved to include a weak MTS, which allows tissue-specific targeting of GS by taking advantage of the differential magnitude of ΔΨ.

The most common means to target a protein to two different subcellular localizations is to maintain two copies of the gene. This is the case in Drosophila melanogaster, which contains two distinct GS genes (Caizzi et al., 1990): one encodes a GS isoform that contains a MTS and is targeted to the mitochondria, whereas the other encodes an isoform that lacks a MTS and is retained in the cytoplasm (Caizzi et al., 1990). Marine elasmobranchs and birds, which utilize the ureosmotic and uricotelic systems for ammonia detoxification, respectively, contain a single GS gene (Laud and Campbell, 1994; Patejunas and Young, 1987). This gene apparently arose from a gene sharing common ancestry with the Drosophila melanogaster cytoplasmic GS (Pesole et al., 1991). Nevertheless, in these two species, the GS enzyme is targeted to the mitochondria in liver cells and cytoplasm in neural tissue. Analysis of the molecular mechanism that underlies the differential targeting of GS in these two species revealed that although the GS gene has been highly conserved during evolution (Kumada et al., 1993), two distinct mechanisms for differential targeting of GS have evolved independently.

In marine elasmobranchs, such as dogfish shark, previous studies have shown that differential localization is achieved by tissue-specific alternative splicing that generates two different GS transcripts (Matthews et al., 2005). The liver transcript contains an upstream alternative exon that is not present in the neural one and leads to the formation of MTS. Here, extensive RT-PCR analysis of the chicken GS transcript excluded the presence of an upstream alternative exon, and sequence analysis of the GS protein showed that the amino acid sequence of liver and brain GS is the same. Furthermore, analysis of chimeric constructs that contain various regions of the GS protein revealed that the capability to confer tissue-specific subcellular localization is confined to the first 23 N-terminal residues of chicken GS, which are sufficient to target a chimeric EGFP construct to the mitochondria in hepatocytes and to the cytoplasm in astrocytes. These findings indicate that tissue-specific subcellular localization of avian GS is achieved by a novel mechanism that can differentially localize an identical protein in liver and brain cells.

Differential targeting of a single translation product might be achieved by several possible mechanisms, one of which is post-translational modification. Protein modification, such as phosphorylation, might affect the accessibility of a targeting sequence by altering its folding or ability to interact with another protein or by directly modulating its targeting properties (Karniely and Pines, 2005). Protein phosphorylation activates, for example, the cryptic MTS of the cytochrome P450 family member CYP2B1 (Anandatheerthavarada et al., 1999), enhances the mitochondrial import of the glutathione S-transferase protein (Robin et al., 2003), but inhibits the mitochondrial import of the yeast protein YNK1 (nucleotide diphosphate kinase 1) (Amutha and Pain, 2003). The possibility that differential targeting of GS involves the function of a tissue-specific kinase was examined by amino acid substitutions. Our results showed that although substitution of Tyr17 with the phosphor-mimetic Glu residue was sufficient to impair mitochondrial import in hepatocytes, substitution of Tyr17 and Ser7 with Ala, did not facilitate mitochondrial import in astrocytes. In addition, MS/MS analysis revealed that apart from acetylation of Ala at position 2 in GS from both liver and brain, there are no other post-translational modifications in the sequenced fragments. These findings suggest that tissue-specific subcellular localization is not achieved by post-translational modification of the GS protein.

Translocation of proteins into the mitochondrial matrix is ultimately dependent on a sufficiently large electrochemical proton gradient across the mitochondrial inner membrane. It has been suggested that the electrophoretic effect produced by ΔΨ on MTS leads to an active pulling mechanism that includes catalyzed unfolding of protein domains (Huang et al., 2002; Shariff et al., 2004). The magnitude of ΔΨ required for import is dependent on the length and/or net positive charge of the mitochondrial targeting signal. Proteins with a ‘strong’ signal, which is characterized by the presence of basic amino acids and the absence of acidic ones, can be imported at a moderately negative ΔΨ, whereas proteins with a ‘weak’ signal, characterized by a low content of positive charged residues, require higher ΔΨ (Martin et al., 1991). In addition, proteins with a long targeting signal (i.e. 40-50 residues) are not dependant on a highly negative ΔΨ for import, possibly because they reach the matrix at the initial interaction with the import machinery and become unfolded by the mitochondrial Hsp70 (Huang et al., 2002; Shariff et al., 2004). Recently, we have shown that dual-targeted mitochondrial proteins tend to have a weaker MTS than exclusive mitochondrial proteins (Dinur-Mills et al., 2008). GS is an example in which the weak MTS has functional significance. Analysis of the GS N-terminal residues revealed that the targeting signal is relatively short and is encompassed within the first 23 residues. This region forms, according to prediction programs and to the crystal structure of the highly homologous human GS, an α-helical structure. Imposition of this region onto an α-helical wheel projection shows that it has the properties of an amphipatic α-helical MTS with a hydrophilic face that contains some positive and polar residues and lacks negative ones. This short targeting sequence established a weak MTS, as judged by the particularly low scores given by MTS prediction programs and by the low number of positively charged residues. In contrast to the canonical MTS of Su9, this weak MTS was incapable of directing mitochondrial import in a standard in vitro import reaction. Considering that a weak MTS is dependant on a highly negative ΔΨ for import, we decided to examine the magnitude of ΔΨ of liver and brain cells. Remarkably, our results revealed that ΔΨ in hepatocytes is considerably more negative than in astrocytes. This finding represents, to the best of our knowledge, the first example of cell-type-specific differences in ΔΨ. Changes in ΔΨ have been previously observed in neurological disorders (Abou-Sleiman et al., 2006; Mortiboys et al., 2008), aging cells (Sugrue and Tatton, 2001) and tumors. A more negative ΔΨ has been detected in a variety of carcinomas (Chen, 1988; Fantin et al., 2002; Fantin et al., 2006) and in chemically induced and oncogene-induced malignant transformation in various cell types (Liang et al., 1999; Zarbl et al., 1987). The more negative ΔΨ in tumor cells has been attributed to the shift in glucose metabolism: normal cells produce most of the ATP from glucose through oxidative phosphorylation whereas many cancer cells exhibit lower oxidative phosphorylation activity and produce ATP by conversion of glucose to lactate. This change in glucose metabolism is causatively related to the more negative ΔΨ in tumor cells (Fantin et al., 2006). Liver is the major organ involved in glucose homeostasis by means of gluconeogenesis (i.e., glucose production from precursor compounds such as lactate) and glycolysis. A highly negative ΔΨ in liver cells might be functionally related to the complex metabolic functions exerted by the cells, but might also reflect liver mitochondria ultrastructure, which exhibits a larger surface area (Benard and Rossignol, 2008).

The possibility that a weak MTS provides a crucial tool for differential targeting of GS was assayed by converting the targeting signal of GS into a strong MTS. Our results clearly showed that elevation of the net positive charge of the targeting sequence, by substitution of one or two histidine residues with arginine, or by extending the size of the MTS through including two copies of the targeting sequence, abolished the capability to confer tissue-specific subcellular localization. Under these conditions, the chimeric protein was mitochondrial in both liver and brain cells. Our results suggest that uricotelic species have evolved by the selection of a weak MTS to the otherwise highly conserved GS enzyme. This targeting sequence allows taking advantage of the tissue-specific differences in ΔΨ and directing the GS protein to the mitochondria in liver cells and to cytoplasm in brain. The functional link between MTS properties and the magnitude of ΔΨ might provide a mechanistic basis for the redirection of cellular protein under physiological conditions in which ΔΨ is altered.

Chimeric EGFP plasmids were constructed by using first strand cDNA generated from RNA of chicken liver cells. pEGFP-Ch-50C was constructed by PCR amplification of chicken liver cDNA using the following primers (underlined residues indicate restriction sites): 5′-TGTAAGCTTAGCATCCGCATCCCACG-3′ and 5′-ATGGATCCTACGGGGAGCACGGGG-3′ and cloning into pEGFP-C3 (Clontech). All other chimeric EGFP plasmids were constructed by cloning of PCR products into pEGFP-N1 (Clontech). The following primers were used to amplify chicken cDNA: For pCh-50N-EGFP 5′-TGTAAGCTTGGAGCCGAGCGTGGGAG-3′ (primer A) and 5′-ATGGATCCTCGTGGTCCAGAGTGCGG-3′ (primer B); for pCh-37N-EGFP primer A and 5′-TAGGATCCCCAGTCCCGTCGATCCA-3′; for pCh-23N-EGFP primer A and 5′-TAGGATCCCCCTGCGGCAGCTTC-3′; for pCh-14N-EGFP primer A and 5′-ATGGATCCTTGATGGCTTTGCTCAGGTG-3′; for pCh-35-84-EGFP primer A and 5′-TAGGTACCCATGGCTGCGGAGGTC-3′ as well as primer B and 5′-TAGGTACCGGGGAGCACCTCCGCTG-3′. For plasmid pCh-23×2N-EGFP, oligonucleotides spanning residues 2-23 were inserted into BamHI site of plasmid pCh-23N-EGFP. For single amino acid substitution a two-step method was employed: in the first step two PCR products were generated using pCh-50N-EGFP as template and, in the second step, the PCR products were amplified using primers A and B. The following primers were used for the first round of amplification: For pCh-50N-EGFP_(Y17E) primer A and 5′-GATGGCTTATCAAGCACATGGAGATGAAGCTGCC-3′ as well as primer B and 5′-GGCAGCTTCATCTCCATGTGCTTGAT-3′; for pCh-50N-EGFP_(S7A) primer A and 5′-GGCGAGCGCCCACCTGAG-3′ as well as primer B and 5′-CTCAGGTGGGCGCTCGCC-3′; for pCh-50N-EGFP_(Y17A) primer A and 5′-GATGGCTTATCAAGCACATGGCCATGAAGCTGCC-3′ as well as primer B and 5′-GGCAGCTTCATGGCCATGTGCTTGAT-3′; for pCh-50N-EGFP_(H8R) primer A and 5′-GCGAGCTCCCGCCTGAG-3′ as well as primer B and 5′-CTCAGGCGGGAGCTCGC-3′. For plasmid pCh-50N-EGFP_{H8R, H15R}, with double amino acid replacement, the PCR fragments were generated using pCh-50N-EGFP_(H8R) as template with primer A and 5′-GCAAAGCCATCAAGCGCATGTAC-3′ as well as primer B and 5′-GTACATGCGCTTGATGGCTTTGC-3′. Plasmid pDf-MTS-EGFP is described elsewhere [pL-GS-EGFP (Matthews et al., 2005)]. Plasmid pGS, which contains the entire GS coding sequence under the control of Sp6, was created by PCR amplification of first strand cDNA from chicken liver using primers A and 5′-TGTAAGCTTGCATGCCTGCAGGTC-3′ and cloning into pGEM4 (Promega). Plasmid pSu9-GS was generated by replacing a NcoI/BglII fragment in plasmid pGS with a PCR fragment obtained by amplification using pSu9-DHFR as template with the primers 5′-TACCATGGACAAAATGGCCTCCACTCG-3′ and 5′-TAAGATCTCCGTGGAAGAGTAGGCG-3′. Plasmid pGS-DHFR was generated by replacing a AgeI/NotI fragment in plasmid pCh-50N-EGFP with a PCR fragment obtained by amplification using pSu9-DHFR as template with the primers 5′-ATACCGGTTCGACCATTGAACTGCATCG-3′ and 5′-ATGCGGCCGCCTGGGTATTTTGG-3′. The resulting plasmid was used as template to generate a PCR fragment with primer A and 5′-TGTAAGCTTCAGGGGGAGGTGTGGGAGG-3′, which was cloned into pGEM4. The cloning details can be obtained upon request. All primers were purchased from Sigma. In all plasmids the cloned region was confirmed by sequencing. Plasmid DNA was prepared using the NucleoBond PC500 (Macherey-Nagel) plasmid preparation kit.

Liver or brain tissues of chicken embryos (E18) were excised, suspended in isotonic HIM buffer (200 mM mannitol, 70 mM sucrose, 1 mM EGTA, 10 mM HEPES-KOH pH 7.4) including 0.1% BSA and homogenized in a Dounce homogenizer by five strokes with pestle A and two strokes with pestle B. The homogenate was centrifuged at 1000 g for 10 minutes. The supernatant was centrifuged at 12,000 g for 15 minutes to obtain the mitochondrial pellet and the cytoplasmic supernatant. The mitochondrial pellet was washed twice at 12,000 g and resuspended in HIM. The cytoplasmic supernatant was centrifuged at 120,000 g for 60 minutes and the pellet was discarded. Total protein extract was obtained by homogenization in Passive Lysis Buffer (Promega) by five strokes with pestle A and 20 strokes with pestle B, sonication for 5 minutes and centrifugation at 20,000 g for 15 minutes. All steps were carried out at 4°C. Equal portions of total, cytoplasmic and mitochondrial fractions were resolved on 10% SDS-PAGE gel. For western blot analysis, antibodies against tubulin (DM 1A; Sigma), mHsp60 (LK-2; Sigma), GS (Gorovits et al., 1997) or phosphotyrosine (PY20; Santa Cruz Biotechnology) were used. The corresponding horseradish-peroxidase-conjugated secondary antibodies were used, and the cross-reactivity was visualized by the enhanced chemoluminescence (ECL) procedure (Pierce).

For immunoprecipitation, protein extracts (0.5 mg) were precleared by incubation for 16 hours at 4°C with pre-immunserum bound to protein-A-Sepharose (Amersham). Cleared extracts were immunoprecipitated with protein-A-Sepharose bound to anti-GS antibodies overnight at 4°C. The immunoprecipitated proteins were separated on 10% SDS-PAGE and analyzed by western blotting or stained with Coomassie blue. The Coomassie-blue-stained gel slices, containing the GS protein band, were incubated with either trypsin, AspN or GluC, and MS carried out with Qtof2 (Micromass, England) using a nanospray attachment. Data analysis was done using the biolynx package (Micromass, England) and database searches were performed with the Mascot package (Matrix Science, England). Similarity searches of sequences, determined via manual analysis, were carried out with the GCG Wisconsin Package Version 10.3 (Accelrys, San Diego, CA) and Blast search in the NCBI data bank (http://www.ncbi.nlm.nih.gov).

Total RNA was prepared using TriPure isolation reagent (Roche) and first strand cDNA was synthesized following treatment with DNase I (Fermentas), using the Iscript cDNA synthesis kit (Bio-Rad). For RT-PCR analysis, first strand cDNA was amplified by PCR (30 cycles) using the following primers: 5′-TGCCCGCAGCCCAGCCCA-3′ (primer 1), 5′-GTGTCTGTGGGCACGATGCC-3′ (primer for exon 7), 5′-CTCTTGGGTTCGTGGTCCA-3′ (primer 2), 5′-CCCGAAGCTCACCCCACTG-3′ (primer 3), 5′-GGGGTCACATGAAGGGGTT-3′ (primer 4), 5′-TTTCAAGGCTATCAGCACG-3′ (primer 5), 5′-CGGGCTCAGAAGGTGTTA-3′ (primer 6) and 5′-ATGAAGGCTGTTGCTTGGC-3′ (primer 7). The PCR products were fractionated in 1.2% agarose gel and visualized with ethidium bromide. For northern blot analysis, RNA was denatured by heating at 60°C for 10 minutes in 2.2 M formaldehyde and 50% formamide, and fractionated in 1.2% agarose gel containing 2.2 M formaldehyde. The fractionated RNA was transferred to a nitrocellulose filter, hybridized with a GS probe labeled with [³²P] by the random primer DNA labeling mix (Biological Industries, Israel) and visualized by autoradiography.

Liver and brain tissues were isolated under sterile conditions from chicken embryos (E18) and primary cultures of astrocytes and hepatocytes were prepared according to published protocols (Mayo et al., 2008; Tarlow et al., 1977). For astrocytes, cerebral cortices were dissociated with trypsin and cultured for 10 days on coverslips coated with poly-L-lysine (0.1 g/l) in DMEM supplemented with 10% FCS. For hepatocytes, the liver tissue was dissociated with collagenase and cultured in 1:1 ratio of H12 medium and DMEM supplemented with 10% FCS and 10% essential and non-essential amino acids in the presence of 5 μg/ml insulin. The cells were transfected with DNA (1 μg per 5×10⁵ cells) using the jetPEI (Polyplus transfection) according to the manufacturer's instructions. Then, cells were stained after 48 hours by incubation with 250 nM MitoTracker Red (Molecular Probes) for 15 minutes at 37°C and fixed with 4% paraformaldehyde, according to published protocols. Confocal imaging was performed using Zeiss R510 confocal laser scanning microscope. Excitation was performed with an argon laser set to 488 nm and emission was detected with a 525±15 nm band-pass barrier filter. Red fluorescence for mitochondria was examined using 568 nm excitation light, and emission was detected with a 580-625 nm filter. For immunostaining, the MitoTracker-stained and fixed cells were incubated with anti-GS antibodies and subsequently with FITC-conjugated goat anti-rabbit antibodies (Jackson ImmunoResearch). Confocal imaging was performed as described above.

For electron microscopy, hepatocytes and astrocytes were harvested and fixed with 2.5% glutaraldehyde in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 2 mM KH₂PO₄, pH 7.4) and post-fixed in 1% OsO₄ in PBS. After dehydration in graded ethanol solutions, the samples were embedded in glycid ether (Serva). Ultrathin sections (~ 0.1 μm) were stained with uranyl acetate and lead citrate and examined in a Jeol 1200 EX TEM. For live-cell imaging, hepatocytes and astrocytes grown on coverslips were incubated with 100 nM of tetramethyl rhodamine methyl ester (TMRM; Molecular probes) for 25 minutes in a dark humidified incubator at 37°C. After incubation, the coverslips were washed with PBS and mounted on an imaging chamber filled with HT solution (132 mM NaCl, 4.2 mM KCl, 1 mM MgCl₂, 1 nM CaCl₂, 5.5 mM D-glucose, 10 mM HEPES, pH 7.4). Confocal images were acquired using Zeiss R510 confocal laser scanning microscope. Excitation was performed using 568 nm excitation light, and emission was detected with a 560 nm long-pass barrier filter. The confocal images were analyzed by computer-assisted methods as described before (Koopman et al., 2008). The parameters examined were mitochondrial branching, area, aspect ratio (AR, which reflects the ratio of mitochondrial length to width) and levels of TMRM fluorescence.

The pGEM4 constructs were used as templates for in vitro SP6 transcription/translation carried out with a TNT Coupled Reticulocyte Lysate System (Promega) in the presence of [³⁵S]methionine. In some reactions, labeled proteins were denatured in 8 M urea and 20 mM HEPES-KOH pH 7.4, at room temperature for 2 hours before the import reaction. Coupled translation and import was performed as described before (Knox et al., 1998) using GS mRNA transcribed from pGS. Import reactions into isolated mitochondria (100 μg per reaction) were carried out in 200 μl SI buffer (0.6 M mannitol, 50 mM HEPES-KOH pH 7.4, 80 mM KCl, 10 mM MgAc, 2 mM KH₂PO₄, 2 mM EDTA, 2 mM MnCl₂ and 3% BSA) for 60 minutes at 30°C. The reaction was started by the addition of buffer B (2 mM ATP, 4 mM NADH, 100 μg/μl creatine kinase and 4.5 mM creatine phosphate) and 2 μl of labeled protein. Import was stopped by diluting the reaction fivefold in ice-cold SHKCL medium (0.6 M sorbitol, 80 mM KCl and 50 mM HEPES-KOH pH 7.4), which included 0.1 μg/ml valinomycin, with or without 50 μg/ml proteinase K. The protease was inactivated by the addition of 20 mM PMSF for 5 minutes on ice. The samples were centrifuged at 12,000 g for 20 minutes. The mitochondrial pellet was resuspended in SHKCL buffer and centrifuged at 12,000 g for another 20 minutes. The final mitochondrial pellet was dissolved in sample buffer, boiled for 5 minutes at 95°C and subjected to SDS-PAGE. Signals of radiolabeled proteins were detected by autoradiography.

We thank Abdussalam Azem for helpful suggestions and fruitful discussions and Ariel Gaaton for MS analysis. This research was supported by the Basic Research Grant of the Tel Aviv University. G.D.M. is a recipient of the Segol Fellowship of Tel Aviv University.