Interplay of heterogeneous transcriptional start sites and translational selection of AUGs dictate the production of mitochondrial and cytosolic/nuclear tRNA nucleotidyltransferase from the same gene in yeast.

ATP (CTP):tRNA nucleotidyltransferase catalyzes the addition of the CCA end to tRNAs. In yeast, nucleotidyltransferase is encoded by the CCA1 gene and is localized to three cellular compartments: mitochondria, nucleus, and cytosol. There are three in-frame ATGs near the 5' end of the CCA1 open reading frame. Primer extension experiments show multiple transcription initiation sites upstream of ATG1 and between ATG1 and ATG2. Fractionation of cells carrying a CCA1-COXIV fusion gene demonstrates that all three in-frame AUGs are used as sites of initiation of translation. Therefore, both transcription of CCA1 mRNA with heterogeneous 5' ends and translation from downstream AUGs in CCA1 mRNAs play a role in the synthesis of three nucleotidyltransferase isozymes. Protein initiating from AUG1 is required for mitochondrial protein synthesis and, like many other proteins targeted to mitochondria, it is processed at the amino terminus upon import into the organelle. The shorter proteins arising from AUG2 and AUG3 provide nuclear/cytosol activity.

Sorting isozymes carry out analogous functions in more than one cellular location yet are encoded by the same gene (1). They have heterogeneous amino-terminal ends. Most commonly, selection of the AUG to be used for translation initiation depends on whether transcripts initiate upstream or downstream of the first ATG (ATG1) coded by the ORF' (for review, see Ref. 2). Translational bypass of an AUG that is in a "poor" context for initiation of translation is the second mechanism allowing the initiation of protein synthesis from more than one AUG in a single transcript (2). For sorting isozymes, the site of translation initiation determines the efficiency with which the protein products are localized to different compartments a n d o r exported from the cell.
Serine:pyruvate aminotransferase, encoded by the rat SPT gene, belongs to a class of sorting isozymes partitioned between mitochondria and peroxisomes. Two mRNAs are synthesized from SPT (7). The longer transcript codes for a n amino-terminal extension that localizes protein to mitochondria. Peroxisomal protein is synthesized from the shorter transcript and lacks the amino-terminal extension (7).
M,M-dimethylguanosine-specific tRNA methyltransferase and human uracil-DNA glycosylase, products of the TRMl and UNG genes, respectively, are partitioned between the nucleus and mitochondria (8)(9)(10)(11) and define a third class. TRMl transcripts, which initiate upstream of the first ATG, provide protein to the mitochondria. Transcripts that initiate between the first and second ATG provide protein to both mitochondria and nuclei (8-10, 12, 13). A well defined nuclear localization signal effectively targets Trmlp to the nucleus where it remains. The mechanism resulting in two forms of Ungp is not known; the form having a 77-residue amino-terminal extension is mitochondrial, but the shorter isozyme is nuclear (11).
A*-Isopentenyl-pyrophosphate:tRNA isopentenyltransferase and ATP(CTP):tRNA nucleotidyltransferase, products of MOD5 and CCAl yeast genes, respectively, belong to a fourth class partitioned between the mitochondria, nucleus, and cytosol (14,15). MOD5 transcripts include both ATGs (16), but the poor context for translational initiation surrounding AUGl as well as short 5"untranslated leaders allow read-through to AUGP (17). The long form of Mod5p is enriched in the mitochondria, but, unlike Trmlp, a pool of it remains in the cytosol (1). Mod5p initiated from AUG2 is in both the nucleus and the cytoplasm (14). ATP(CTP):tRNA nucleotidyltransferase catalyzes the addition of CMP and AMP to the 3' end of tRNA (19). It is required because the CCA end is not coded in the tRNA genes of either nuclei or mitochondria (15,19). In yeast, tRNA precursors in the nucleus have CCA ends (20,21). Approximately 30% of nucleotidyltransferase activity has been shown to be in the nucleus in Xenopus oocytes (22). Nucleotidyltransferase activity is present in the cytosol of rat liver cells (23) and Xenopus oocytes (22), where it repairs tRNA that has lost part or all of its 3'-terminal CCA end. Repair of mature tRNA CCA ends, which presumably occurs in the cytosol, has also been demonstrated in yeast (24).
The CCAl gene is an essential gene and contains multiple in-frame ATGs at the 5' end of the ORF (25). The definitive experiment that demonstrated the importance of CCAl to mitochondrial function altered these upstream ATGs. CCAl mutants temperature-sensitive for growth on all carbon sources 13361 were transformed with wild-type and mutant genes missing either ATGl or ATGl and ATG2. The cells containing a CCAl gene missing ATGl have decreased mitochondrial nucleotidyltransferase activity and decreased respiratory growth. Cytosolidnuclear enzyme activity was not affected (15). Cells containing a CCAl gene in which both ATGl and ATG2 were altered were respiratory-deficient (15). These experiments demonstrated that translation from the upstream ATGs was important for complementation of the temperature-sensitive enzyme in mitochondria and established Ccalp as a sorting isozyme shared among nuclei, mitochondria, and the cytosol.
We have extended our studies of CCAl to address ATG use and selection in wild-type cells. We report here that all three upstream in-frame ATGs play a role in initiation of translation of Ccalp. Cell fractionation demonstrates that Ccalp synthesized from AUGl is enriched in mitochondria. The shorter Ccalps from downstream AUGs provide cytosolhuclear activity. AUG selection in CCAl results from a combined role of transcription initiation and translational selection. Sites of transcription initiation occur upstream of ATGl and ATG2 so initiation at the first AUG in these CCAl mRNAs will result in synthesis of the longest and intermediate size Ccalps. As there are no transcriptional start sites between ATG2 and ATG3, Ccalp arising from the AUG3 is made from & A s containing upstream AUGs.
Protein Sequencing-F'rotein isolated from whole cells as described in Ref. 28 was transferred to Immobilon-P (29) and sequenced in the ICBR Protein Chemistry core facility a t the University of Florida.
CAGCACTATTCTGCAGTAG was used to alter the ATG at position 28 To construct the double mutant ccal-M2,M3, the oligonucleotide AGof the ORF to CAG. DNA carrying ccal-M2 was then isolated, and the ATG at position 52 was altered to AGA with oligonucleotide AGAAT-TCGTTCTAGATTTCTGAGCAGC. To construct the double mutant ccal-Ml,M3, the ATG at position 1 of the ORF was altered to ATC with oligonucleotide C C G T A G G A T A T C C A . DNA carrying ccal-MI was isolated, and the ATG at position 52 was altered to AGA as described above. The ccul-MI and ccal-M1,M2 mutants are described by Chen et al. (15). The 648-base pair XbaYhruII fragment from the altered CCAl gene, positions 186-4334, was replaced with the 1-kb XbaYHindIII fragment from vector G18 (33). The HindIII restriction site was modified into a blunt end (31). The result is a fusion gene with nucleotides 1-186 of the CCAl ORF fused in-frame to pseudomature C O W coding sequences.
The BamHIISalI fragment can-ying the fusion gene was subcloned into the multicopy shuttle vector pRS426.
RNA Isolation and Primer Extension-Total RNA was isolated from yeast cells (34) carrying either CCAl or CCAl-COXN on the multicopy vector pRS426. Poly(A) RNA was isolated from total cellular RNA with an Oligotex-dT mRNA kit (Qiagen) according to the manufacturer's instructions. The 5' end of CCAl and CCAI-COXN fusion mRNAs were mapped as described by Triezenberg (35). The oligonucleotide CGT-TCAGCAAG'ITACAGATGTTC that is complementary to CCAl mRNA from position 109 to 131 of the CCAl ORF was labeled with [y-32PldATP and used as a primer for extension. The primer was annealed to either 25 pg of total cellular RNA or 0.25-0.5 pg of poly(A) RNA at 60 "C for 45 min and extended with avian myeloblastosis virus reverse transcriptase (Life Sciences) at 50 "C for 45 min. To provide a sizing ladder, the same oligonucleotide was used for sequencing CCAl and CCAl-COXNgenes.
Northern Blot Analysis-Total RNA was electrophoretically separated on a formaldehyde-agarose gel (31) and transferred to Zetaprobe (Bio-Rad) according to the manufacturer's instructions. Prewashing, prehybridization, and hybridization steps were carried out as described by Morales et al. (36). Probes complementary to a 1.2-kb HindIII fragment carrying URA3 and to the 1-kb EcoRV/fiuII CCAl DNA fragment were prepared by nick translation using a nick translation kit (Promega).
Cellular Fractionation-Strain WD1 transformed with vector pRS426 carrying wild-type or mutant CCAl-COXN fusion genes was grown in 3 liters of synthetic complete medium (37) lacking uracil with galactose as the carbon source. The cells were harvested in mid log phase and pelleted. The pellet was washed, and spheroplasts were isolated as described previously (28) except that 2.5 mg of yeast lytic enzyme (ICN)/g of cell pellet was substituted for zymolyase. A portion of the spheroplasts was frozen for subsequent Western blot analysis. Remaining spheroplasts were washed and suspended in 2 ml of 0.6 M mannitol, 10 m~ Ms-HCI (pH 7.5), 5 m~ EDTA, 5 ~IM EGTA, 1 nm phenylmethylsulfonyl fluoride, 1 m~ benzamidine, and 10 pg/ml each antipain, leupeptin, pepstatin, and chymostatin (disruption buffer)/g of cell pellet. Spheroplasts were disrupted with a Dounce homogenizer and debris removed by centrifugation for 5 min at 2,000 x g. A portion of cell lysate was then frozen, and the postmitochondrial supernatant obtained by centrifugation for 15 min at 27,000 x g. The mitochondria were washed with disruption buffer followed by a 5-min centrifugation (1,000 x g ) to remove debris. Mitochondria were suspended in disruption buffer and pelleted two more times by centrifugation for 15 min at 27,000 x g . Mitochondria were then suspended in 0.6 M mannitol, 10 nm Tris-HC1 (pH 7.5), 1 m~ EDTA, 1 n m EGTA and collected in a microcentrifuge and stored frozen as pellets at -70 "C.
Western Blot AnaZysisSodium dodecyl sulfate-polyacrylamide electrophoresis was as described previously (38). 100 pg of spheroplast, 100 pg of cell lysate, 100 pg of postmitochondrial supernatant, and 26 pg of mitochondrial extract as determined with the Bradford protein reagent (Bio-Rad) according to the manufacturer's instructions were separated on a 13.5% sodium dodecyl sulfate-polyacrylamide gel and then transferred to Immobilon-P transfer membrane (Millipore) with a Bio-Rad transblotter according to the manufacturer's instructions. Nonspecific binding sites of Immobilon-P were blocked with 20 n m Tris base, 0.14 M NaCl (pH 7.6) (1 x TBS), and 5% Carnation nonfat dry milk for 60 min at room temperature. Incubation with primary antiserum was for 60 min a t room temperature in 1 x TBS, 5% milk, and 0.1% Tween 20.
Antiserum to CoxIVp, kindly provided by Dr. Robert Jensen (Johns Hopkins), was used at a 1:7,000 dilution. Antiserum to Mrpap, kindly provided by Dr. Steven Ellis (University of Louisville), was used at a 1:5,000 dilution. Antiserum to Actlp (Boehringer Mannheim) was used at a 1:5,000 dilution. Following six 5-min washes in 1 x TBS and 0.1% Tween 20, the membrane was incubated for 60 min with a 1:3,300 dilution of blotting grade affinity-purified goat anti-rabbit IgG (H+L) horseradish peroxidase conjugate (Bio-Rad) in 1 x TBS, 5% milk, and 0.1% Tween 20. The membrane was then washed six times, 5 min each, in 1 x TBS and 0.1% Tween 20. Antigen-antibody complexes were detected with reagents from an ECL kit (Amersham Corp.) according to the manufacturer's instructions.

RESULTS
CCAl lFanscript Copy Number Is Low in S, cerevisiae"T0 detect CCAl mRNA, Northern blot analysis was performed using RNA from wild-type cells and cells transformed with the CCAl gene on a multicopy plasmid. CCAI-specific and URA3specific radiolabeled probes were used. Two transcripts are detected in total RNA isolated from wild-type cells (Fig. l, Lane l ).  small ribosomal RNA and thus must be about 1,600 bases. In wild-type cells (lane 11, the amount of CCAl message is low and appears comparable to the low abundance URA3 mRNA (39).
Panscripts from the CCAl Gem Initiate Upstream of ATGl and between ATGl and ATG2"Primer extensions were performed to locate the 5' ends of CCAl transcripts. Because of the low abundance of CCAl transcripts, cells containing a multicopy plasmid carrying CCAl were used. Total cellular RNA and poly(A) RNA gave the same result in primer extension studies. As the large number of apparent transcript ends were always seen, they are not likely a result of degradation but may arise from a secondary structure that interferes with extension. Regardless, no 5' mRNA ends fall between ATGB and ATG3 of the CCAl ORF, although a small proportion of the total transcripts was found to end just 3' to ATG3 (Fig. 2). Even if these result from premature termination and their real 5' ends fall between ATGB and ATG3, their contribution to the total population of mRNA would be quite small. As translation initiation in yeast usually occurs at the first AUG in a mRNA, this result suggested that AUG3 might not be used in wild-type cells. However, we have shown previously that AUG3 can serve as a site of translation initiation in CCAl genes without ATGl and ATG2 (15). As described below, two separate approaches were used to examine further AUG usage in CCAl transcripts.
Protein Sequence Analysis of the Amino-terminal End of Ccalp-As one approach to understanding AUG usage in the CCAl gene, we isolated (28) and sequenced Ccalp (29) from whole cell extract. The sequence obtained, TNSNFVLN, begins 1 amino acid following the methionine coded by AUG3. One interpretation is that synthesis starts at AUG3, and the initiating methionine is removed. Another is that translation begins a t all of the AUGs, and physiological processing or proteolysis during purification produces a unique end. As shown below, in addition to AUG3, synthesis initiates from AUGl and AUG2, but the resulting proteins are not processed between the methionine coded by AUG3 and the following threonine. Since only a small amount of protein (-16 pmol) was available for sequencing, we do not know whether the other termini are not detected because they are blocked or whether we did not begin with enough protein to see them.
Altering ATG Codons Differentially Alters the Ability of CCAl to Complement Growth on Fermentable and Nonfermentable Carbon Sources-Wild-type and mutant CCAl genes (Table I) with one or more altered methionine codons were cloned into a single copy vector. The effects of the mutations were tested by transforming them into a temperature-sensitive tRNA nucleotidyltransferase mutant. All strains grow at the permissive temperature because the chromosomal CCAl gene provides functional Ccalp (Fig. 3). At the nonpermissive temperature, functional Ccalp must be provided by a plasmid.  lanes 2, 4, and 5 ) . This shows that Ccalp missing the first 9 or the first 17 amino acids can meet the cytosolidnuclear needs of the cell. The same cells, however, grow very slowly on glycerol medium at 37 "C (Fig. 30, compare  lane 1 with lanes 2 , 4 , and 5). That they do grow is clear from a comparison with cells transformed with plasmid alone (lane 6). Thus, Ccalp initiated from AUGl is required for normal mitochondrial function. We also found that after several days, papillae appeared among cells carrying the ccal-MI gene (Fig.  30, lane 2). This indicates that suppression of the mitochondrial defect may occur.
Cells transformed with the CCAl gene missing ATGS and ATG3 (ccal-M2,M3) grow very slowly on glucose medium a t 37 "C in comparison with cells transformed with CCAl (Fig.  3B, compare lanes 1 and 3). Clearly, translation from AUGl cannot meet the needs of cytoplasmic tRNA biosynthesis. Perhaps not enough Ccalp is produced to meet the needs of the cell, or perhaps the majority of Ccalp from AUGl is sequestered in mitochondria and not available for cytosolidnuclear needs. Again, after several days, papillae were observed, indicating that suppression of this phenotype may also occur. Given that the cytosolidnuclear needs are not met, it is not surprising that cells carrying ccal-M2,M3 are also unable to grow on glycerol medium (Fig. 30, lane 3).
As altering ATGs could change the use of other ATGs in the

FIG. 3. Growth characteristics of a cca-1-containing strain transformed with wild-type and mutant
CA1 alleles. Equal amounts of cells were spotted on glucose medium a t 25 "C (panel A) and 37 "C (panel B ) and on glycerol medium a t 25 "C (panel C) and 37 "C (panel D ) . The mutant CCAl genes were missing eitherATGl,ATG2, or ATG3 or a combination thereof from the ORF as described in Table I. same ORF, we sought to learn whether all three ATGs are used in the wild-type gene. We have been unable to find electrophoretic conditions that will separate translation products made, in vitro, from genes containing only ATG1, ATG2, or ATG3 and thus cannot directly measure the level of each protein in cellular extracts. As an alternate strategy, we have constructed CCAl-COXN fusion genes that express Ccal-CoxIV fusion proteins from the CCAl promoter. The smaller size of the Ccal-CoxIV protein allows proteins translated from each AUG to be separated and detected with an antibody to CoxIVp. Mitochondrial import of the fusion protein could also be assessed in vivo by complementation of a COXN mutant strain.

CCAl-COXW Fusion Proteins Complement a COXW Mutant when Expressed from a Multicopy Plasmid-CCAl-COXN fu-
sion genes with the CCAl promoter were unable to complement the C O X N mutant when present on a single copy vector. This is consistent with the low level of CCAl mRNA observed by Northern analysis. Therefore, all studies were done with wildtype and mutant CCAl-COXNfusion genes on multicopy plasmids (Table I) (ccal-cozN-Ml) or ATGl andATG2 (ccal-coxZVMl,M2) each grow slower on glycerol medium than cells carrying the wild-type fusion gene (CCAl-COXW) (Fig. 4 B ,   compare lane 1 with lanes 2 and 5). As with the wild-type gene, the amino-terminal amino acids of the CCAl ORF are important to the mitochondrial pool of the protein. Lack of these sequences, however, does not completely abolish the complementation. The fusion gene ccal-codV-MI is able to complement the coxN growth defect better than ccal-coxN-Ml,M3, which shows a dramatic decrease in growth on glycerol (Fig.   4B, lane 4 ). This comparison suggests that protein is initiating from both AUG2 and AUG3 of ccal-coxN-MI.

FIG. 5. Immunoblot of cell extracts from a COXN-deficient strain transformed with wild-type and mutant
CCAl-COXN alleles. Cell extracts from strain WDl carrying wild-type fusion protein or protein in which one or more of the first three ATGs of the ORF have been altered as described in Table I (Table I) revealed two proteins in cells transformed with CCAl-COXW (Fig. 5, lane 1 ) and two similar sized proteins in cells transformed with ccal-coxN-MI (compare Fig. 5, lanes 1 and 4 ). To determine if the shortest protein arises from processing of a larger protein rather than from translation initiation a t different AUGs, a gene retaining only the second ATG (ccal-coxN-MI, M3) was constructed. Only one protein, that arising from AUG2, is observed. Its mobility is similar to the largest protein observed in cells carrying CCAl-C O X N or ccal-coxN-MI (Fig. 5, compare lane 3 with lanes 1  and 4 ) . The mobility of the smallest protein corresponds to a protein originating from AUG3 in ccal-coxN-M1,M2 extracts (compare Fig. 5, lane 5 with lanes 1 and 4). Therefore, unless changing the methionine encoded by ATG3 abolishes the putative processing event, translation initiation from AUGB accounts for synthesis of the smallest protein produced from CCA1-COXW and ccal-coxW-MI.

CCAl-COXW alleles
Although we know that Ccalp initiating from AUGl is required for mitochondrial function, we cannot detect Ccal-CoxIV fusion protein originating from AUGl in immunoblots. Either protein initiated from AUGl migrates similarly to that initiated from AUG2 or it is processed to a shorter form. Ccal-coxN-M2,M3 was constructed to clarify the situation. Two proteins are found in cells carrying ccal-corN-M2,M3 (Fig. 5, lane  2). The shorter migrates slightly faster than the protein originating from AUGB in cells carrying ccal-coxlV-Ml,M3 (Fig. 5,  compare lanes 2 and 3 ) . Although there is much less of the larger protein, it exhibits the mobility expected from a protein  (lanes 1 , 4 , 7, and lo), postmitochondrial supernatant (lanes 2,5,8,11), and mitochondria (lanes 3,6,9,12) from strain WD1 carrying the CCAl-COXNfusion gene or CCAl-COXN in which one or more of the first three ATGs of the ORF have been altered as described in Table I Table I. Fractionation of the known cytosolic protein Actlp, and the known mitochondrial protein Mrp2p, were followed to monitor the effectiveness of the procedure. Total extract from cells transformed with ccal-coxN-M2,M3 has two proteins (Fig. 6, lane 4 ) . These proteins are almost nondetectable in the postmitochondrial supernatant (Fig. 6 , lane 51, but they are enriched in the mitochondrial fraction (Fig. 6, lane 6). Thus, the fusion protein initiating from AUGl is processed as indicated by two protein forms.
A substantial amount of protein initiating from AUG2 and AUG3 remains in the postmitochondrial supernatant (Fig. 6, compare lanes 7 and 8). Neither is enriched in mitochondria (Fig. 6, lane 9). Total extract from cells carrying CCAl-COXN contains more protein initiating from AUG3 than from AUGl and AUG2 combined (Fig. 6, lane 1 ) .
mRNA Dunscripts from CCAl and CCA1-COXW Genes Have the Same 5' Ends-Primer extension to compare the 5' ends of mRNA Erom cells transformed with the CCAl gene and the CCAl-COXN genes were done. Primer extension experiments show that, as for the CCAl gene, the 5' ends of the CCAl-COXN transcripts map upstream of ATGl and between ATGl and ATGB (Fig. 2). DISCUSSION The CCAl gene codes for a housekeeping enzyme required for the biosynthesis of all cellular tRNAs. Northern analysis of CCAl transcripts indicates a low level of expression of this gene, and primer extension shows that not all transcripts extend to the upstream most ATG in the ORF. Experiments that change the three 5' ATGs in the longest ORF in both the wildtype and fusion gene constructs demonstrate the importance of the translation initiation site to cellular location of the enzyme.
In S. cerevisiae, synthesis of isozymes from the same ORF is accomplished either because transcriptional initiation sites occur upstream and downstream of the first ATG or because the environment surrounding the first AUG in the mRNA has a primary or secondary structure unfavorable for translation initiation (for review, see Ref. 2). Both play a role in the synthesis of yeast Ccalp isozymes. Primer extension experiments (Fig. 2) show that CCAl transcripts are heterogeneous. Transcripts initiate 5' to the first ATG and between the first and second ATGs. Immunoblots of Ccal-CoxIV fusion proteins show that protein synthesis originates from each of the three AUGs (Fig.  6). Consequently, Ccalp arising from AUGl is synthesized from the longest messages. Protein originating from AUGB is probably made from the shorter messages, but it is also possible that the long messages contribute to translation from AUG2. Translation from AUG3 must occur from mRNAs having upstream AUGs.
Of the 131 yeast genes analyzed by Cigan and Donahue (411, 95% adhere to the scanning mechanism, with initiation occurring at the first AUG encountered by the translational machinery (40). Comparative analysis of yeast genes indicates the presence of the consensus sequence, 5'-ANMlUM+3UCU-3' surrounding the initiating methionine codon (41,42). In higher eukaryotes translation initiation is influenced by nucleotide sequence surrounding the AUG codon, by secondary structure upstream or downstream of the AUG, and by the distance between the 5' end of the transcript and the AUG (for review, see Ref. 2). Although severely affected by mRNA secondary structure (42,43), translation initiation in yeast is considerably less sensitive than higher eukaryotes to poor sequence context and mRNA leader length (2,41,42,44,45). However, Slusher et al. (17) demonstrated that the translation initiation region surrounding AUGl of MOD5 is very sensitive to nucleotide context. Upon altering the nucleotides surrounding AUGl to match the yeast consensus sequence they observed a 5-10-fold increase in expression from AUGl as well as a concomitant decrease in expression from AUG2 (17). Creation of a longer mRNA leader also increases expression of the long form of Mod5p (17).
Western blot analysis of Ccall-CoxIV fusion protein shows protein synthesis from AUG3, yet, all messages start upstream of ATG2. Judging from the yeast consensus sequence, AUGl and AUGB of the Ccalp ORF are very weak translation initiation sites, neither site possessing the highly conserved purine in position -3. Only the nucleotide sequence surroundingATG3 has the very highly conserved adenine at position -3. Western blot analysis also shows much more protein is synthesized from AUG3 than from AUG2. As a result of these observations we propose that leaky ribosome scanning of AUGB and possibly AUGl results in protein initiation at AUG3.
Mitochondrial proteins encoded in the nucleus are generally synthesized with an amino-terminal extension of 20-80 amino acids that carry information necessary for targeting protein to mitochondria. Although there is no known consensus amino acid sequence for mitochondrial targeting, the amino termini of mitochondrial targeted proteins are enriched in basic and hydroxylated amino acids with arginine being favored over lysine (46,47). Statistical (47) and experimental data (48,49) show that amphiphilic secondary structure is important for mitochondrial targeting. Often, targeting signals are composed of two domains; the amino-terminal domain is usually an a-helical amphipathic structure with a relatively high hydrophobic moment, and the carboxyl-terminal domain is also amphiphilic but not necessarily a-helical (50).
The amino terminus of Ccalp does not share these characteristics. Although rich in hydroxylated amino acids, it is not particularly rich in basic residues. Protein sequence analysis Yeast Nucleotidyltransferase Expression programs (51,521 do indicate several short a-helical structures from residues 1 to 30. However, the hydrophobic moment of 18-amino acid windows at the amino terminus is lower than for most reported mitochondrial targeting sequences (47). Like ModSp, a functional pool of Ccalp initiating from AUGl appears to be present in the cytosolhucleus, as mal-1 cells transformed with ccal-M2,M3 do grow at 37 "C, albeit very poorly, on fermentable carbon sources. Thus, not all of the long form is sequestered in mitochondria.
The targeting signal of many mitochondrial bound proteins is cleaved by matrix proteases (53). Some sorting isozymes are unusual in that their mitochondrial targeting signals are not cleaved (1, 4, 8, 12). Those of Htslp (18) and Ccal-CoxIVp, however, are. We infer that Ccalp is also processed, although it does not fit a consensus motif identified in mitochondrial targeting signals cleaved by matrix proteases (53). This motif is characterized by arginine residues at position -2 and -10 from the cleavage point. A hydrophobic residue such as phenylalanine a t position -8 and a serine at position -5 have also been observed (53). If, as immunoblots indicate, the cleavage site for Ccalp is just downstream of AUG2, then it may occur at the -10 arginine which is one of the few arginines at the amino terminus.
In summary, ATP (CTP):tRNA nucleotidyltransferase is a sorting isozyme that shares some features with other such proteins but that differs as well. Like ModBp, it is found not only in two compartments but in three: the mitochondria, nucleus, and cytosol. Like other genes coding for sorting isozymes, CCAl contains more than one in-frame ATG, but it is unusual in having three. There are multiple transcription start sites in the CCAl gene which fall upstream of all ATGs as well as between ATGl and ATG2 so that, in part, translation start sites are dictated by the 5' ends of the mRNAs. However, to produce protein from AUG3, upstream AUGs in CCAl mRNAs must be bypassed. Like other genes coding sorting isozymes where one destination is mitochondria, protein initiated from AUGl of the ORF is targeted to mitochondria more efficiently than the other products of the CCAl gene. Some enzyme synthesized from AUGl remains in the cytosolhucleus providing a small amount of activity. The cytosolidnuclear pool of the longest form of the protein is presumably not processed as is the majority which is imported into the organelle. Although we have a good understanding of the mechanisms that provide nucleotidyltransferase to mitochondria many questions about the partitioning of CCAl gene products between nuclei and cytosol remain.