Bipartite Signals Mediate Subcellular Targeting of Tail-anchored Membrane Proteins in Saccharomyces cerevisiae *

Tail-anchored proteins have an NH 2 -terminal cytoso- lic domain anchored to intracellular membranes by a single, COOH-terminal, transmembrane segment. Sequence analysis identified 55 tail-anchored proteins in Saccharomyces cerevisiae , with several novel proteins, including Prm3, which we find is required for karyogamy and is tail-anchored in the nuclear envelope. A total of six tail-anchored proteins are present in the mitochondrial outer membrane and have relatively hydrophilic transmembrane segments that serve as targeting signals. The rest, by far the majority, localize via a bipartite system of signals: uniformly hydrophobic tail anchors are first inserted into the endoplasmic reticulum, and additional segments within the cytosolic domain of each protein can dictate subsequent sorting to a precise destination within the cell. Tail-anchored proteins have a single transmembrane segment at their carboxyl terminus, and many of the proteins that mediate subcellular traffic and programmed cell death are tail-anchored into select of cells In the Bcl-2 of proteins, key of the programmed cell death pathway in cells, Fluorescence Microscopy— For fluorescence microscopy, cells were visualized directly or after staining with Mitotracker (MitoTracker Red CM-H2X Ros) according to the standard protocol from Molecular Probes. All fluorescence images were captured using a Bio-Rad MRC1024 confocal scanning laser microscope mounted on a Zeiss Ax- ioscop. For the mating studies, a Nikon fluorescence microscope with GFP and DAPI filter sets and a (cid:3) 100 DIC (Nomarski) objective was used. In this case images were captured by a Micromax digital camera with Metamorph imaging software. In preparation for fluorescence microscopy cells were grown to mid-log phase at 25 °C in semisynthetic medium. In assays for nuclear transport, wild-type, rna1-1 , or prp20-1 cells were transferred from 25 °C liquid culture to a 37 °C water bath for 30 min immediately prior to preparation for microscopy. (Biospec Cell centrifugation incubated on with intermittent vortexing. and insol- uble proteins were separated by centrifugation at

Tail-anchored proteins have a single transmembrane segment at their carboxyl terminus, and many of the proteins that mediate subcellular traffic and programmed cell death are tail-anchored into select membranes of eukaryotic cells (1). In the Bcl-2 family of proteins, key regulators of the programmed cell death pathway in animal cells, 12 of the 16 known family members are tail-anchored to either the mitochondrial outer membrane or endoplasmic reticulum, and their membrane location is critical for function (2). The SNAP-receptors (SNAREs) 1 are a family of proteins essential for intracellular membrane fusion, and 19 of the 23 SNAREs in yeast are tail-anchored proteins. Membrane fusion absolutely requires the participation of SNAREs in both the donor and acceptor membrane, and each of the known SNAREs has a restricted location at a defined membrane compartment of the endomembrane system (3,4).
Tail-anchored proteins fold co-translationally and the single hydrophobic segment at their carboxyl terminus allows for post-translational insertion into membranes (5,6). Once membrane is inserted, the amino-terminal domain of the protein is displayed in the cytosol. An elegant study on the tail-anchored SNARE synaptobrevin-1/VAMP-1a found that the protein is inserted into the endoplasmic reticulum and subsequently sorted to presynaptic vesicles (5). Cytochrome b 5 , another tailanchored protein, is inserted into the membrane of the endoplasmic reticulum and maintained there despite some escape to, and retrieval from, the cis-Golgi cisternae (7).
But tail-anchored proteins are also located in the mitochondrial outer membrane, and the precise signal that distinguishes these from those targeted to the endoplasmic reticulum is still not clear. Deletion mutagenesis has shown the signal is contained within the tail segment, and in the few different model proteins examined to date critical determinants have been either the presence of charged residues or in some cases the number of hydrophobic residues (8 -14). An understanding of the precise targeting signals that direct the majority of tail-anchored proteins to the endoplasmic reticulum, but allow some to go exclusively to mitochondria, has been hampered by the relatively small number of model tail-anchored proteins that have been available for study.
We applied several bioinformatic approaches to identify tail-anchored proteins encoded in the genome of Saccharomyces cerevisiae. Fifteen novel tail-anchored proteins were discovered; we report here on the localization of the previously unrecognized tail-anchored proteins. Analysis of the targeting segments from the 55 tail-anchored proteins in yeast suggests a bipartite system of signals: hydrophobic character in the tail segment determines targeting to the endoplasmic reticulum instead of mitochondria and discrete sorting signals that then direct tail-anchored proteins to their correct subcellular destination.

EXPERIMENTAL PROCEDURES
Plasmids and Yeast Strains-DNA fragments corresponding to each open reading frame were amplified by PCR using primers that generated one in-frame restriction site immediately preceding the start codon and another following the stop codon (oligonucleotide sequences available on request). PCR products were cloned behind GFP-S65T under the control of the MET25 promoter and expressed from a centromeric plasmid (13). In semisynthetic (SD) media, expression from the plasmid is partially repressed.
* This work was supported by a grant from the Australian Research Council (to T. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Fluorescence Microscopy-For fluorescence microscopy, cells were visualized directly or after staining with Mitotracker (MitoTracker Red CM-H2X Ros) according to the standard protocol from Molecular Probes. All fluorescence images were captured using a Bio-Rad MRC1024 confocal scanning laser microscope mounted on a Zeiss Axioscop. For the mating studies, a Nikon fluorescence microscope with GFP and DAPI filter sets and a ϫ100 DIC (Nomarski) objective was used. In this case images were captured by a Micromax digital camera with Metamorph imaging software. In preparation for fluorescence microscopy cells were grown to mid-log phase at 25°C in semisynthetic medium. In assays for nuclear transport, wild-type, rna1-1, or prp20-1 cells were transferred from 25°C liquid culture to a 37°C water bath for 30 min immediately prior to preparation for microscopy.
Membrane Isolation and Analysis-Microsomal membrane fractions were prepared by differential centrifugation. Cells (50 -100 OD 600 units) were suspended in buffer 88 (250 mM sorbitol, 150 mM potassium acetate, 5 mM magnesium acetate, 0.5 mM phenylmethylsulfonyl fluoride, 1.2 g/ml leupeptin, 0.75 g/ml antipain, 0.25 g/ml chymostatin, 1 g/ml pepstatin, 50 mM HEPES, pH 6.8) and disrupted by two bursts (each of 2-min duration) in a mini-beadbeater-8 (Biospec products) using silica/zirconia beads. Cell debris was removed by centrifugation at 500 ϫ g for 5 min. A crude membrane fraction was collected by centrifugation at 16,000 ϫ g for 10 min. Membranes were extracted by resuspension in either 1% Triton X-100 or 100 mM Na 2 CO 3 and incubated for 30 min on ice with intermittent vortexing. Soluble and insoluble proteins were separated by centrifugation at 100,000 ϫ g in a Beckman Airfuge.
Miscellaneous-Published procedures were used for isolation of mitochondria and trypsin shaving, SDS-PAGE, and immunoblot analysis (16). Detailed comparative hydrophobicity analyses of the targeting sequences in each tail-anchored protein made use of the ProtParam site at (expasy.proteome.org.au/cgi-bin/protparam) using a window of 5 amino acids to scan through the predicted transmembrane segment according to the Kyte-Doolittle algorithm.

RESULTS
In addition to the tail-anchored proteins known in yeast, sequence analysis revealed 15 open reading frames that could be expressed as GFP fusions that localize to discrete subcellular membranes. Fig. 1 shows that Fis1 and YFL046w are mitochondrial proteins (Fig. 1A), seven proteins localized generally to the endoplasmic reticulum membrane (Fig. 1B), and four are localized to specific subdomains of the endoplasmic reticu- lum (Fig. 1C). YLR238w and YOR324c are found in clusters within the bounds of the endoplasmic reticulum, and Prm3 is confined to the perinuclear (nuclear envelope) membrane. YPL206c, a protein showing sequence similarity to bacterial glycerophosphodiester phosphodiesterases, was found concentrated in lipid bodies, regions of endoplasmic reticulum specialized for lipid metabolism ( Fig. 1C; Ref. 17). Two proteins closely related in primary structure, YDL012c and YBR016w, are located in the plasma membrane (Fig. 1D). Both YDL012c and YBR016w are localized more intensely in regions of new membrane synthesis, toward the emergent buds of dividing cells and also in the schmoo structure (Fig. 1E, arrowheads) of haploid cells in the presence of mating pheromone.
The Tail Segment Is Necessary for Targeting to the Endoplasmic Reticulum-Comparative sequence analysis of the carboxyl-terminal segments of the 41 tail-anchored proteins targeted to the endoplasmic reticulum (including the 17 proteins sorted to other membranes of the secretory pathway) revealed no obvious motifs in the primary structure that might serve as a common targeting signal. However, hydropathy analysis through the transmembrane segments suggests regions of high (Ͼ3.3 units) hydropathy score in each protein ( Fig. 2A). Conversely, hydropathy analysis of the predicted transmembrane domain from Fis1 (Fig. 2B) and for YFL046w, Tom5, Tom6, Tom7, and Tom22 (data not shown) suggests this segment of each polypeptide is more amphipathic than for proteins targeted to the endoplasmic reticulum. Previous work has shown that the transmembrane segments of the translocase subunits Tom5, Tom6, Tom7, Tom22, and Fis1 are necessary and sufficient for targeting mitochondria (13,14,18). 2 To test whether the character in the transmembrane segment of Fis1 distinguished it as a protein destined for mitochondria, site-directed mutagenesis was used to replace hydrophilic residues with leucines, such that the hydrophobicity 2 T. Beilharz and T. Lithgow, unpublished observations.  (37) is indicated (arrows), as is the perinuclear endoplasmic reticulum (N denotes the nucleus). C, yeast cells transformed with a plasmid encoding GFP fused to the carboxyl-terminal 35 amino acids of YOR324c ("YOR324tail") or transformed with a plasmid encoding GFP-fused to the carboxyl-terminal 35 amino acids of Bos1 were viewed by fluorescence microscopy (D). E, a crude membrane fraction was isolated from yeast cells expressing GFP-YORtail, solubilized, and centrifuged to separate solubilized proteins ("S") from insoluble material ("P"). Proteins were then analyzed by immunoblotting after SDS-PAGE. approached that of proteins targeted to the endoplasmic reticulum (Fig. 2B). When yeast cells expressing these mutant proteins as GFP fusions were analyzed by fluorescence microscopy, the Fis1 mutant, Fis1(L4), is targeted to the endoplasmic reticulum (Fig. 2C). Similar results were found with mutations made in the transmembrane segment of Tom22 (data not shown). Fluorescence is sometimes also observed in the lumen of the vacuole, perhaps reflecting turnover of the inappropriately targeted proteins (Fig. 2C, "V").

Bipartite Signals for Targeting and Sorting Proteins in Intracellular Membranes-Tail-anchored proteins in the endoplasmic reticulum can display distinct patterns of localization.
For example YOR324c is found in discrete clusters throughout the endoplasmic reticulum (Fig. 3A), whereas Bos1 is a SNARE found sparingly in the endoplasmic reticulum but enriched in the Golgi (Fig. 3B, arrows). The hydrophobic tail segments from each of these proteins are sufficient to target GFP to the endoplasmic reticulum, where it remains generally distributed (Fig.  3, C and D). The proteins are integrated into the membrane as judged by their resistance to extraction by sodium carbonate treatment of isolated membranes ( Fig. 3E; data not shown).
We identified a single protein, Prm3, whose distribution was restricted to the nuclear envelope and sought to understand whether its targeting and sorting were also mediated by bipartite signals. A series of deletion mutants were constructed (Fig.  4A) and tested for subcellular localization. Deletion of 91 amino acids, leaving only the carboxyl-terminal segment of Prm3 (GFP-Prm3⌬1-91), targets the GFP reporter to the endoplasmic reticulum (Fig. 4B), which is continuous with the outer membrane of the nuclear envelope (19). A second signal is present in Prm3(⌬1-68), causing the fusion to be localized specifically to the nuclear envelope (Fig. 4C). Sequence analysis suggested a nuclear localization sequence (NLS) between Pro 68 and Lys 75 .
Point mutations in the putative NLS caused degradation of the full-length Prm3 fusion protein, and so to test whether the putative NLS is functional, we analyzed a soluble version, GFP-Prm3(⌬109 -133). GFP-Prm3(⌬109 -133) is localized exclusively within the nucleus, and confocal sectioning revealed it distributed throughout the nucleoplasm (Fig. 4D). To be certain the information for nuclear localization is contained within the amino acid sequence P 68 GRVRKHK 75 , site-directed mutagenesis was used to convert Lys 73 to Ala 73 , and the mutant protein GFP-Prm3(⌬109 -133)* is compromised in import, with much of the protein located in the cytosol (Fig. 4E).
If Prm3 has a bipartite targeting sequence, with the P 68 GRVRKHK 75 truly acting as an NLS, the general nuclear import machinery of the cell should be required to maintain the full-length, membrane-embedded Prm3 within the nucleus. Ran is a small GTP-binding protein that plays a critical role in transport of proteins into the nucleus (20,21). A gradient of Ran-GTP across the nuclear membrane is established by the differential localization of two enzymes: the Ran-GTP exchange factor (Ran-GEF) in the nucleus and the Ran-GTP-activating protein (Ran-GAP) in the cytosol, and mutations in either of these two enzymes leads to collapse of nuclear traffic.
Temperature-sensitive mutants for Ran-GEF (prp20-1) and Ran-GAP (ran1-1) were transformed and the distribution of Prm3 measured by fluorescence microscopy. At the permissive temperature of 25°C, the localization of Prm3 in the Ran-GEF mutants (Fig. 4F) and the Ran-GAP mutants (Fig. 4G) is nuclear. However, after shift to the non-permissive temperature of 37°C, Prm3 is found throughout the membranes of the endoplasmic reticulum, suggesting that without a functioning Ran cycle the protein cannot be sorted from the outer to inner membrane of the nuclear envelope.
Since Prm3 is the first protein described in yeast that might be localized to the inner membrane of the nuclear envelope, we sought to determine its function by analysis of ⌬prm3 mutant cells. Haploid yeast cells of opposite mating types, either wild type (Mata) with wild type (Mat␣) or ⌬prm3 (Mata) with ⌬prm3 (Mat␣), were mixed and incubated on rich medium for 3-4 h, then fixed and stained with DAPI to visualize DNA. In wild-type crosses, nuclear DNA is usually mixed by the time the first diploid bud appears near the junction of the two parent cells of the mating pair (Figs. 5A). However, the cytoductants that arise from ⌬prm3 ϫ ⌬prm3 crosses are unable to undergo normal karyogamy, and nuclear material remains distinct but tightly juxtaposed at the point where the nuclei simultaneously squeeze through the bud neck, whether staining for nucleic acid (Fig. 5B) or nucleoplasmic protein (Fig. 5C).
If, as suggested by this microscopy, Prm3 is required for the fusion of nuclear envelopes, then ⌬prm3 cells should remain haploid through subsequent generations. To test this possibility, ten zygotes from mating mixtures like those described in Fig. 5 were isolated by micromanipulation. The progeny from wild-type crosses are diploid: in the course of mitosis they develop bud scars at each end of the ovoid cells (Fig. 5D), they cannot mate with haploid cells of either mating type, and they will undergo meiosis if placed on sporulation media (data not shown). By contrast, the progeny from ⌬prm3 ϫ ⌬prm3 matings were exclusively haploid: the cells that arise mitotically from these cytoductants have a single, spiralled pattern of bud scars (Fig. 5D), cannot be sporulated, but can mate with (wild-type) cells of the opposite mating type (data not shown). The colonies that initially arise from these cytoductants show a distinct furrow, with cells isolated from one FIG . 5. Prm3 is essential for karyogamy. A, wild-type haploid cells of opposite mating types were mixed and incubated for 3-4 h on YPAD agar, fixed in 70% ethanol for 10 -20 min, washed, and resuspended in 1 mg/ml DAPI and viewed by fluorescence microscopy. Arrows denote the bud emerging from the parent cells of the mating pair. B, ⌬prm3 haploid cells of each mating type were mixed and analyzed as described above. C, ⌬prm3 haploid cells of each mating type, transformed with a plasmid encoding GFP-Prm3(⌬109 -133) were mixed and analyzed by fluorescence microscopy. The periphery of the budding zygotes was highlighted with the cell wall-specific stain concanavalin A-rhodamine B isothiocyanate. D, cells recovered from mating mixtures described above were allowed to grow to mid-log phase and then stained with Calcofluor. Bud scars are seen at either end of the diploid cells from the wild-type matings, but a characteristic whorl of scars is found on one end only of the cells derived from the ⌬prm3 x ⌬prm3 cytoductants. side of the furrow exhibiting mating type a and cells from the other side being of mating type ␣. 3 Transcript profiling has shown that the PRM3 gene is induced in response to both pheromone stimulation (22) and during sporulation (23), and both processes culminate in karyogamy (see review by Rose (24)). Because Prm3 localization is dependent on a NLS and the Ran-GTPase cycle, we propose Prm3 as the first component of the cellular karyogamy machinery to function in membrane fusion at the level of the inner membrane of the nuclear envelope. DISCUSSION In addition to the discovery and localization of a large set of novel tail-anchored proteins, analysis of the sequence data provided by these proteins suggests a general model for the targeting of tail-anchored proteins to each intracellular membrane.
From analysis of the 41 tail-anchored proteins now known to localize to membranes of the secretory pathway, the only common property we could identify was the uniformly hydrophobic nature of the residues within the transmembrane segment. Previous studies on individual tail-anchored proteins have each suggested that the tail-segment contains targeting information (8 -14). The hydrophobic tail segments of Bos1, Prm3, and YOR324c are sufficient to target each of these proteins to the endoplasmic reticulum, and these truncated fusions remain uniformly distributed through the perinuclear endoplasmic reticulum (i.e. the outer membrane of the nuclear envelope) and peripheral endoplasmic reticulum.
A segment of lower hydrophobic moment, containing glycine, serine, and threonine residues, is found in the six tail-anchored proteins targeted to the mitochondrial outer membrane. Replacement of these residues with leucines, to alter the hydrophobicity and perhaps other structural features such as the rigidity of the helix it can form, prevents targeting to mitochondria. The TOM translocation machinery mediates protein insertion into the outer membrane (25)(26)(27), as well as the import of soluble proteins into mitochondria (28 -31). The recent threedimensional structure of the TOM complex receptor subunit, Tom20, suggests that it has only a very shallow binding groove on its surface, into which binds the hydrophobic face of the targeting sequences found attached to soluble proteins targeted to mitochondria (32,33). Tom20 has been implicated as the import receptor for tail-anchored proteins, and it will be of interest to determine the structural details of how the receptor binds the targeting segments of tail-anchored proteins.
In contrast to insertion of tail-anchored proteins into mitochondria, the nature of the machinery that mediates tail-anchored protein insertion into the endoplasmic reticulum is still unclear. Kutay et al. (5) suggested that the general post-translational Sec machinery might mediate insertion of tail-anchored proteins like VAMP-1a. Whatever the machinery, it is able to insert tail-anchored proteins with hydrophobic tailsegments; even a simple, artificial tail-segment consisting entirely of leucine residues was a suitable substrate for in vitro insertion into membrane vesicles derived from the endoplasmic reticulum (34), and the same tail segment directs GFP to the endoplasmic reticulum when expressed in yeast cells. 3 Distinct regions in the cytosolic domain of VAMP-1a assist its subsequent transport from the endoplasmic reticulum to the presynaptic vesicles. Interactions mediated by distinct regions in the cytosolic domains are also required for Bos1 trafficking to the Golgi, Prm3 targeting to the nuclear envelope, and YOR324c clustering in punctate zones of the endoplasmic reticulum. Thus two independent processes, targeting to the endoplasmic reticulum and subsequent sorting to the correct membrane, are mediated by bipartite targeting and sorting signals. For the Golgi and post-Golgi membranes, sorting would be mediated through vesicular traffic. In the case of Prm3 and any other proteins that might exist in the inner membrane of the nuclear envelope, we suggest that this occurs via the lipid rivulets of pore membrane present in the nuclear pore complex (35,36) that would link the outer and inner membrane bilayers and is catalyzed by Ran and the karyopherins that also drive the import of soluble nuclear proteins.