Yeast phospholipid biosynthesis is linked to mRNA localization

Control of the localization of mRNA allows spatial control of gene expression and contributes to the generation of cellular asymmetry or synaptic plasticity (Holt and Bullock, 2009; Martin and Ephrussi, 2009). Localized mRNAs are generally transported as RNA–protein complexes (RNPs) to their destination sites, where they become translationally active, allowing site-specific protein synthesis (Besse and Ephrussi, 2008; Martin and Ephrussi, 2009).

In budding yeast, >50 transcripts are selectively transported to the bud (Aronov et al., 2007; Oeffinger et al., 2007; Shepard et al., 2003), among them ASH1 mRNA, which encodes a daughter-cell-specific transcriptional repressor of the HO endonuclease gene (Long et al., 1997). Localization of ASH1 and the other mRNAs requires a set of proteins that includes the RNA-binding proteins She2p (Böhl et al., 2000; Niessing et al., 2004) and She3p (Böhl et al., 2000; Müller et al., 2011), as well as the myosin-V motor Myo4p (Böhl et al., 2000; Krementsova et al., 2011).

Many bud-localized mRNAs encode membrane or membrane-associated proteins (Aronov et al., 2007; Shepard et al., 2003) that are translated at the cytoplasmic face of the ER. Yeast contains two major types of ER – the perinuclear ER, which is continuous with the nuclear envelope, and the cortical ER (cER, also called plasma-membrane-attached ER), which is a meshwork of tubular ER structures underlying the plasma membrane (Du et al., 2004). During budding, tubules migrate from the mother cell into the bud, followed by the spreading of cER along the bud cortex (Du et al., 2004; West et al., 2011). Several observations indicate that the localization of a subset of mRNAs to the bud is coordinated with cER segregation. Firstly, mutations in several genes affect both processes. Mutations in Myo4p and She3p affect both ER inheritance by the daughter cell and mRNA localization (Estrada et al., 2003; Long et al., 1997). Similar observations have been made for genes that are essential for the early steps of ER inheritance, for cER docking at the bud tip (Aronov et al., 2007; Fundakowski et al., 2012) or for the formation of ER tubules (Fundakowski et al., 2012). Furthermore, the RNA-binding protein She2p co-fractionates with ER membranes and can bind to protein-free liposomes with a membrane curvature similar to that of ER tubules, suggesting that this protein connects localized mRNAs to ER (Genz et al., 2013). These findings led to the idea that localized mRNAs encoding membrane or ER proteins travel to the bud by associating with cER tubules. Similar coordinated distribution of ER and mRNAs have also been reported in other biological systems. For example, in ascidians, macho1 and HrPEM mRNAs tightly colocalize with the cER at different stages of redistribution from egg to zygote (Sardet et al., 2005). In the claw frog Xenopus laevis, Vg1 mRNA localization patterns highly overlap with those of ER markers at different stages of the oocyte development (Kloc and Etkin, 1995).

In order to gain more insight specifically into the mechanism and the determinants of ER-dependent mRNA localization to the yeast bud, we aimed to identify novel factors required for this type of mRNA localization. We designed a visual screen to allow us to analyze the localization pattern of two early-expressed mRNAs, WSC2 and EAR1, in a collection of ER mutant strains. Here, we report the identification of several novel genes whose deletion affects the localization of WSC2 and EAR1, and we show how defects in lipid biosynthesis lead to ‘road blocks’ in mRNA transport to the bud on ER tubules.

Genetic screens to uncover proteins required for mRNA localization have previously been performed using ASH1 mRNA as a substrate (Jansen et al., 1996; Sil and Herskowitz, 1996). However, several proteins needed for proper ER morphology and segregation that did not affect ASH1 did affect the localization of mRNAs encoding membrane-bound proteins, including WSC2, IST2 and EAR1. The differential requirements suggest that genetic screens for mutants affecting ASH1 localization (Jansen et al., 1996; Sil and Herskowitz, 1996) might have missed genes required for regulating the localization of mRNAs encoding membrane proteins. We therefore performed a visual screen for mutants that mislocalize WSC2 or EAR1 mRNAs, which are expressed at early cell cycle stages when ER segregation occurs. To mark the mRNAs, we integrated bacteriophage MS2 loops into their 3′UTRs (producing WSC2-MS2L and EAR1-MS2L). These mRNAs were coexpressed with MS2-binding protein (MS2CP) fused to 3×GFP, leading to specific labeling of the RNP granules that include these mRNAs. We limited the screen to a collection of deletion and hypomorphic mutants of essential and non-essential genes with the gene ontology term ‘endoplasmic reticulum’, i.e. genes that code for proteins that localize to the ER or have a function related to the ER (Fig. 1). A list of the genes that were analyzed (489 in total) is provided (supplementary material Tables S1 and S2). Two parallel screens were performed, which allowed us to validate our results in two different systems and to check the feasibility of a large-scale screen. In the first screen, plasmids carrying either WSC2-MS2L or EAR1-MS2L were co-transformed with a plasmid expressing MS2CP–3×GFP into deletion mutants, followed by manual screening for RNA localization defects (Fig. 1A). In the parallel automated screen, MS2CP–3×GFP under the control of a MET25 promoter was integrated into the yeast genome at the HO locus. After transformation with WSC2-MS2L or EAR1-MS2L plasmids, the yeast were crossed with mutants carrying deletions of ER-related genes, and haploid cells were selected that carried the deletion mutation, the MS2CP-3×GFP gene and the corresponding plasmid (Giaever et al., 2002) (Fig. 1B). These cells were analyzed by automated fluorescence microscopy as described in Materials and Methods. The manual and automated screens revealed 45 and 24 mutants with a potential localization defect. These mutants were subjected to a second manual screen to assess the effect on localization in ≥100 cells per mutant. We observed that several mutants displayed a different mRNP mislocalization phenotype than that of the she mutants. Whereas the latter display an exclusive accumulation of mRNPs in the mother cell (Fig. 2A), the new mutants contained mRNPs in both the mother and the daughter cell. In neither case did we observe an exclusive localization to the bud as seen in wild-type cells. Such symmetric distribution of mRNPs (subsequently referred to as ‘mislocalization’) was observed following the deletion of 13 non-essential genes (Table 1). In order to obtain quantitative data, we divided cells into two classes, cells with exclusively bud-localized mRNPs and cells with symmetric distribution of mRNPs. Throughout this work we display the percentage of cells showing the mislocalization phenotype as defined above. The genes identified in this screen included genes coding for proteins required for vesicle coat formation, like CHC1 (Gurunathan et al., 2002) and SFB3 (Miller et al., 2002), genes affecting lipid metabolism, including phosphatidylethanolamine methyltransferase (CHO2, also known as PEM1; Kodaki and Yamashita, 1989), phospholipase C (PLC1; Flick and Thorner, 1993), phosphatidylinositol 3-kinase (VPS34; Auger et al., 1989), the transcriptional repressor of phospholipid synthesis genes OPI1 (Wagner et al., 1999), the long-chain base-1-phosphate phosphatase (LCB3; Mandala et al., 1998) and an activator of serine palmitoyltransferase, TSC3 (Gable et al., 2000). Furthermore, we found that deletion of ICE2, which is involved in cER inheritance and morphology (Estrada de Martin et al., 2005; Estrada et al., 2003), and of an uncharacterized open reading frame (ORF) (YGL007C-A) also influenced the localization of WSC2-MS2L and EAR1-MS2L. In contrast to cells of the she2Δ strain, of which 83.3% showed WSC2-MS2L mRNP mislocalization and 96.6% showed EAR1-MS2L mRNP mislocalization, the mislocalization was less pronounced in the mutant strains identified in this screen. Because we observed the strongest mislocalization of WSC2 and EAR1 in cells lacking CHO2 (56% and 61% of cells showed mislocalization, respectively) we focused on this gene for further analysis.

CHO2 encodes phosphatidylethanolamine N-methyltransferase (Enzyme Commission number 2.1.1.17), a key enzyme in phosphatidylcholine biosynthesis (Kodaki and Yamashita, 1989; Summers et al., 1988). Δcho2 cells have WSC2-MS2L or EAR1-MS2L mRNPs in the mother cell and bud, indicating that only a fraction of the corresponding mRNA pool is correctly localized (Fig. 2A). By contrast, wild-type cells exclusively accumulate MS2L-tagged mRNP particles in the bud, whereas cells lacking She2p (Fig. 2A), She3p or Myo4p (Bertrand et al., 1998; Fundakowski et al., 2012) accumulate mRNPs only in the mother cell. In order to rule out the possibility that mRNA mislocalization is due to defective actin cytoskeleton cables stretching from mother cells to the bud, we expressed an actin-binding protein, Abp140–mCherry (Yang and Pon, 2002) in cho2Δ cells (Fig. 2B). No difference in Abp140–mCherry distribution between cho2Δ and wild-type cells could be observed.

Because we have previously found a correlation between the loss of ER segregation or altered ER morphology and RNA mislocalization (Fundakowski et al., 2012; Schmid et al., 2006), we tested whether deletion of CHO2 results in changes in ER structure. A previous analysis suggests that this is not the case (Thibault et al., 2012). However, in this study a general luminal ER protein (Kar2p) was used as an ER marker that stains mainly the perinuclear ER. In order to specifically assess potential morphological changes in the cortical ER we analyzed the distribution of three ER marker proteins, Scs2-TMD–RFP, Rtn1–mCherry and Tgb3–mCherry, all of which are membrane proteins and constituents of cER (Fig. 2C). Scs2p is a tail-anchored integral ER protein that connects cER to the plasma membrane, and its transmembrane domain (TMD) is sufficient for targeting red fluorescent protein (RFP) to ER (Loewen et al., 2007). Rtn1p belongs to the reticulon family of proteins and is involved in shaping ER tubules that are preferentially found in cortical ER (Voeltz et al., 2006). Tgb3 is a movement protein of plant potexvirus that was found to specifically bind to highly curved cER tubules in yeast (Lee et al., 2010). All three cER marker proteins were mislocalized in cho2Δ cells (Fig. 2C). Scs2-TMD–RFP was present in cER and perinuclear ER in wild-type cells, and a strong accumulation of Scs2-TMD–RFP could be seen at the tip of the bud, indicating the presence of cER in the bud. The cER localization of Scs2-TMD–RFP in cho2Δ cells was visible as several large aggregate-like structures in the cytoplasm and at the cell periphery, whereas perinuclear ER localization remained unchanged. A similar cER redistribution was seen for Rtn1–mCherry. Tgb3–mCherry was detected in patches at the cortex of wild-type mother cells and buds, but the peripheral patches disappeared in Δcho2 cells. We conclude that loss of Cho2p results in the redistribution of cortical ER markers, severe changes to ER morphology and partial loss of mRNA localization.

Several localized mRNAs, including WSC2, require proper cER formation and segregation into the bud for their proper localization (Fundakowski et al., 2012). We therefore investigated whether mislocalization of WSC2-MS2L is due to the release of the mRNA from ER or whether it is caused by association of the mRNA with ER aggregates that cannot move into the bud. For colocalization studies, MS2L-tagged WSC2 mRNA was expressed in cells carrying Rtn1–mCherry as the ER marker. Wild-type cells showed a colocalization of WSC2-MS2L mRNPs and cortical ER at the bud tip, indicating that both are transported to the bud (Fig. 3A, top row). In cho2Δ cells, WSC2-MS2L mRNPs still colocalized with the Rtn1–mCherry ER marker; however, colocalization was also seen within the intracellular aggregates in the mother cell (Fig. 3A, bottom row).

The RNA-binding protein She2p can associate with membranes in vitro and co-purifies with ER from cell extracts (Aronov et al., 2007; Genz et al., 2013; Schmid et al., 2006). In order to test whether the ER-localized WSC2-MS2L mRNPs in the mother cell are still associated with She2p, we co-expressed a GFP-tagged She2p with WSC2-MS2L and detected the mRNA by means of a fusion of the MS2 coat protein to five mCherry molecules (see Materials and Methods). The majority of WSC2-MS2L mRNPs in the buds of wild-type cells (100%, n = 12), as well as those in the mother cells of cho2Δ mutants (83%, n = 29), contained She2p (Fig. 3B). We therefore conclude that the intracellular cER structures observed in cho2Δ cells bind to WSC2-MS2L mRNPs that still contain She2p and that this capturing prevents the movement of both ER and localized mRNPs into the bud.

Cho2p catalyzes the first step during the conversion of phosphatidyl-N-ethanolamine (PtdEtn) to phosphatidylcholine (PtdCho) in the cytidine diphosphate-diacylglycerol (CDP-DAG) methylation pathway (Fig. 4A), the primary route for the synthesis of PtdCho in the absence of exogenous choline (Kodaki and Yamashita, 1989; Summers et al., 1988). Loss of Cho2p results in a severe imbalance of PtdEtn and PtdCho levels, with PtdEtn levels increasing from 15–20% to 40–50%, whereas PtdCho levels can drop from 40–45% to 10% of total cellular lipids (Fig. 4B; Summers et al., 1988). Reduction in the amount of PtdCho is also seen in mutants lacking Opi3p, the yeast phosphatidyl-N-methylethanolamine N-methyltransferase (Kodaki and Yamashita, 1989; McGraw and Henry, 1989), which converts the product of Cho2p, phosphatidyl-N-monomethylethanolamine (PMME), to PtdCho (Fig. 4A,B; Carman and Henry, 1989). Consequently, we tested whether opi3Δ mutants affect mRNA localization as well. Surprisingly, although we observed the reported strong decrease in the amount of PtdCho and an increase in PMME (Fig. 4B), deletion of OPI3 had little impact on the localization of WSC2-MS2L mRNA (14.7±7% of cells displayed mislocalization, compared with 56±14.1% and 83±16.6% of cells displaying mislocalization in cho2Δ and she2Δ cells, respectively; ±s.d.) (Fig. 4C; Table 1). PtdCho levels in mutants that are defective in Cho2p or Opi3p can be increased by the addition of choline to the medium, which is metabolized to PtdCho in the Kennedy pathway (Fig. 4A; Carman and Henry, 1989). Addition of 1 mM choline to the growth medium of cho2Δ cells partially improved WSC2-MS2L localization (27.7±15.7% of cells displayed mislocalization; Fig. 4C). This partial improvement can be explained by the reduced activity of PSD1 in the presence of choline. This gene encodes the major phosphatidylserine (PtdSer) decarboxylase that converts PtdSer into PtdEtn (Carson et al., 1984). Our results indicate that a general reduction in the amount of PtdCho is not the main cause of mRNA mislocalization and suggest a specific defect due to the accumulation of the Cho2p substrate PtdEtn. To test this idea, we overexpressed Opi3p in cho2Δ cells, because overexpression of Opi3p can suppress the PtdEtn methylation defect of cho2 mutants by substituting for Cho2p in the conversion of PtdEtn to PMME (Preitschopf et al., 1993). Overexpression of Opi3p in cho2Δ cells reduces PtdEtn levels (Fig. 4B) and rescues WSC2 localization (Fig. 4C; 8.3±8.2% mislocalization), indicating that the Cho2p-mediated conversion of PtdEtn to PMME is required for correct mRNA localization. Finally, in order to create a similar PtdEtn and PtdCho imbalance to that observed in cho2Δ cells, we tested mRNA localization in a strain lacking Opi3p (leading to a reduced PtdCho level) and overexpressing Psd1p (leading to an increased PtdEtn level). In this strain, WSC2 localization is severely impaired (49±6.9% of cells display mislocalization). In summary, our analysis of mutants of phospholipid biosynthesis suggests that the observed mRNA mislocalization in cho2Δ cells is due to a failure to convert PtdEtn to PtdCho, thereby creating an imbalance of these lipids. This analysis reveals a prominent role for PtdEtn levels in regulating mRNA localization.

The previous analysis demonstrates how an increase in the amount of PtdEtn provokes mRNA mislocalization. We have previously shown that the mRNA localization factor She2p can directly bind to protein-free liposomes (Genz et al., 2013). In order to test whether PtdEtn influences She2p binding, we generated 80-nm diameter liposomes with increasing PtdEtn contents ranging from 0% to 50% (see Materials and Methods), incubated them with recombinant She2p and performed a flotation analysis to determine the amount of She2p that fractionated with (and thus bound to) liposomes (Genz et al., 2013). We observed an increase in the amount of She2p binding as PtdEtn levels increased from 0% to 40%. At higher PtdEtn contents She2p binding decreased again, although only by 20% of its maximum level (Fig. 5A). This decrease correlated with a major collapse in the liposomes from vesicles to large clumps, as revealed by dynamic light scattering (Fig. 5B). PtdEtn belongs to the class of so-called type II lipids that have an overall conical molecular shape resulting from the comparatively small cross-sectional area of the polar head group and, thus, preferentially accumulate in curved structures. Therefore the large structures that appear with high PtdEtn levels likely represent aggregates or abnormal membrane structures, because negative curvature stress due to high PtdEtn levels can lead to malformed biological membranes (Gruner, 1985). These larger lipid-containing aggregates might be an in vitro counterpart of the intracellular clusters seen with cER marker proteins (Fig. 2). Previous studies have shown that an increase in membrane curvature of liposomes results in better binding of She2p, consistent with the model that She2p links localized mRNAs to tubular ER, which has a highly curved structure. The increase in She2p binding to liposomes with increasing PtdEtn contents can therefore be explained by the preference of She2p for curved membranes, which are likely to contain high PtdEtn levels.

In order to test whether She2p is required for the association of WSC2 mRNP with ER aggregates, we determined the degree of WSC2-MS2L mRNP colocalization with the ER (represented by Rtn1–mCherry) in the presence and absence of She2p. In cho2Δ she2Δ mutants, only 12±7.3% (±s.d.) of cells displayed colocalization of WSC2-MSL2 mRNPs with ER clusters, in contrast to 52±31% of cho2Δ cells (Fig. 5C). We frequently observed mRNPs in the proximity of the ER but not colocalizing with it. These experiments suggest that She2p mediates the association of WSC2-MS2L mRNPs with the intracellular ER aggregates. If the high content of PtdEtn in cho2Δ mutants led to a reduction in the overall functional amounts of She2p due to the capturing of the latter on the ER surface, one would also expect a reduction in the correct localization of ASH1 mRNA in cells with high concentrations of PtdEtn. This is indeed the case (Fig. 5D). Whereas ASH1-MS2L mRNPs are localized exclusively to the bud in >90% of wild-type cells (Fundakowski et al., 2012), in 30% (n = 83) of cho2Δ cells these mRNPs are detectable in the mother cell.

In a screen for new proteins involved in regulating the mRNA localization of membrane protein transcripts in budding yeast, we revealed the dominant role of phospholipids, especially PtdEtn, in enabling the correct localization of such mRNAs. Of the mutants that we identified in our screen, 7 out of 13 are deleted for genes encoding enzymes (CHO2, LCB3, PLC1, TSC3, ICE2 and VPS34) or a transcription factor (OPI1) involved in phospholipid or sphingolipid metabolism. A closer analysis of the phenotype of cho2Δ cells showed that an imbalance of the PtdEtn∶PtdCho levels results in a malformed cortical ER network and mislocalization of the two tested mRNAs, WSC2 and EAR1. Other genes identified in our screen might also influence mRNA localization through effects on the levels of PtdEtn versus PtdCho. Because OPI1 controls the expression of several phospholipid synthesis genes, including CHO2, its downregulation could lead to similar defects to those observed for cho2Δ (Carman and Han, 2011). Furthermore, the phospholipid biosynthesis phenotype observed in cho2Δ mutants is aggravated by deletion of ICE2 (Tavassoli et al., 2013).

How could an increase in PtdEtn relate to inhibition of mRNA localization? To compensate for an imbalance in the PtdEtn∶PtdCho ratio, cells alter their proteome and increase the synthesis of proteins involved in stress-response pathways, including the unfolded protein response, the ER-associated degradation pathway and the induction of heat-shock proteins (Thibault et al., 2012). However, we do not believe that these proteome changes are directly related to mRNA mislocalization, as loss of Opi3p, the second methyltransferase required for PtdCho biosynthesis, leads to very similar proteome changes (Thibault et al., 2012) but not to a defect in WSC2 localization, indicating a specific role of Cho2p and of its substrate, PtdEtn. The stronger binding of She2p to liposomes with an increasing PtdEtn content supports a more specific role of PtdEtn. Previous experiments have not indicated a role for specific lipids in She2p binding (Genz et al., 2013); however, this conclusion was based on the fact that the omission of phospholipids with negative net charge, like PtdSer and phosphatidylinositol, did not alter She2p binding. However, PtdEtn was not tested in this system. The variations in She2p binding to liposomes of different PtdEtn contents as recorded in our in vitro assays suggest that the membrane association of She2p and associated mRNAs requires a regulated content of the curvature-supporting lipid PtdEtn, which fits with previous observations of She2p binding to highly curved membranes (Genz et al., 2013). Although we have not been able to determine whether the cER aggregates seen in cho2Δ cells are enriched in PtdEtn, it is likely that they are, because the total cellular PtdEtn level in cho2Δ mutants can reach to up to 50%, a level at which aggregation of liposomes was detectable by dynamic light scattering (DLS) (Fig. 5). Thus, the intracellular structures detected with cER markers could represent membrane aggregates that have collapsed owing to their high PtdEtn content (Gruner, 1985). These membrane structures could entrap She2p and associated localized mRNAs. In addition, extensive She2p capturing could also result from proliferation of the ER. Such proliferation is, for example, seen in conditional sec24-11 mutants with a non-functional COPII coatomer complex (Peng et al., 2000). The sec24-11 allele genetically interacts with SFB3, another gene that we have identified in our screen (Table 1). Other coatomer components have also been implicated in mRNA localization, including the COPII factors Sec21p, Sec23p and the small GTPase Arf1p (Trautwein et al., 2004). However, in contrast to the phenotype that we have observed in cho2Δ, where a fraction of mRNPs remain in the mother cell, mutations in these genes have been reported to disrupt late stages of mRNA localization, such as anchoring or association with the cell cortex. Nevertheless, these results underscore a direct link between membrane trafficking or membrane composition and mRNA localization.

Because two additional genes whose loss reduces mRNA localization (LCB3 and TCB3; Gable et al., 2000; Mandala et al., 1998) encode enzymes that are involved in sphingolipid metabolism, it remains to be determined in the future how other lipids, including sphingolipids, can influence the membrane interaction of She2p and mRNA localization. More generally, our results indicate how dynamic the process of mRNA granule targeting to the bud is; the correct localization of mRNAs requires that ER tubules acting as transport vehicles are kept unclogged and in pristine structural condition.

All yeast strains in this study were derived from a BY4741 background. Deletion mutants in various ER-related genes were obtained from the EUROSCARF yeast deletion collection. Gene tagging or deletion of genes in specific backgrounds (e.g. to create double mutants) was performed by standard PCR-based transformation methods (Janke et al., 2004). All plasmids used in this study are listed in supplementary material Table S3. Oligonucleotides for tagging, gene disruption or cloning are listed in supplementary material Table S4, and yeast strains are listed in supplementary material Table S5.

The synthetic gene array (SGA) query strain RJY3864 was created as follows. In order to visualize the mRNP on the background of the yeast deletion collection, the MS2 coat protein (MS2CP), fused to GFP and in conjugation with the NatMX4 cassette, was genomically integrated into the HO locus of strain RJY3863. This was achieved by fusion PCR: MS2-CP-GFP was amplified from RJP1486 with primers RJO3992 and RJO3993. The NatMX4 cassette was amplified from plasmid RJP1873 with primers RJO3990 and RJO3991. Primers 3991 and 3992 contain overlapping sequence. The two PCR products were fused by PCR using primers 3990 and 3993, which contain overhangs homologous to the HO gene. The fused PCR product was transformed into strain RJY3863 and integrated into the HO locus. The resulting strain, RJY3864, was further transformed with RJP1773 or RJP1815, containing WSC2-MS2L or EAR1-MS2L. Plasmids used in this work are listed in supplementary material Table S3. Plasmids expressing MS2-tagged WSC2 and EAR1 were created as follows. WSC2-MS2L was amplified by PCR from the genome of strain RJY3626 using primers RJO3761 and RJO3762 and ligated into YCplac33 or YEplac195 (Gietz and Sugino, 1988). The resulting plasmids were named RJP1767 and RJP1773, respectively. EAR1-MS2 was amplified by PCR from the genome of strain RJY3624 using primers RJO3988 and RJO3989 and was ligated into Yeplac195. The resulting plasmid was named RJP1815. RJP1817 and RJP1888 were created by amplifying OPI3 and PSD1 from yeast genomic DNA with primers RJO4187, RJO4188, RJO4556 and RJO4557, and cloning the products by the sequence- and ligation-independent cloning (SLIC) method into plasmid YEplac181 (Li and Elledge, 2007). RJP1890 was created by digestion of ABP140-2×mCherry (first 17 amino acids of ABP140; Kilchert and Spang, 2011) with KpnI and SacI from plasmid 1841 and ligating into plasmid YEplac181. Plasmid RJP1889-pMS2-CP-5×mCherry was created as follows. MS2-CP, including the MET25 promoter, was amplified from plasmid 1486 using primers RJO4577 and RJO4578 and was cloned by the SLIC method into plasmid pRS313. A single mCherry sequence was amplified using primers RJO4581 and RJO4582 from plasmid RJP1423, digested with BamHI and BglII and ligated into the BamHI site downstream of the MS2-CP. An additional mCherry unit was inserted the same way. Single mCherry and a transcriptional terminator was amplified using primers RJO4581 and RJO4595 from plasmid 1423 and was ligated into the BamHI site downstream of the 2×mCherry. The destroyed BglII site was converted to a BamHI site by site-directed mutagenesis, using primers RJO4598 and RJO4599. The 3×mCherry was then released with BamHI and cloned into a plasmid with 2×mCherry, thereby creating 5×mCherry.

A sub-library was created from the yeast deletion mutant collection (Giaever et al., 2002) by picking 318 non-essential genes that, according to the S. cerevisiae Genome Database (SGD), code for an ER component or are involved in lipid biosynthesis (see supplementary material Tables S1 and S2). In addition, we added a selection of ORFs coding for proteins with unknown function and additional ORFs with unrelated function for negative control. This collection was transformed with plasmids expressing either WSC2-MS2L or EAR1-MS2L together with a plasmid expressing the MS2 coat protein fused to GFP. Yeast cells from a fresh selective plate were scraped, resuspended in 2 ml of synthetic complete (SC) medium, grown for 3–4 h at 30°C and dropped onto a thin agarose layer of SC medium with reduced methionine concentration (44 mg/l) for induction of the MET25 promoter controlling MS2-CP–GFP expression. The agarose layer was covered with a coverslip and cells were incubated at 30°C for 30 min before images were captured on a Zeiss CellObserver Z1 fluorescence microscope operated by Axiovision 4.8 software (Zeiss). For cER structure visualization, 40 Z-stack images spaced at 0.25 µm were captured and deconvolved using a theoretical point spread function, autolinear normalization and automatic z-correction provided by the Axiovision 4.8 software package.

The construction of the SGA query strain is described above. The SGA strain containing the mRNA visualization system was crossed against two sub-libraries using SGA methodology (Cohen and Schuldiner, 2011; Tong et al., 2001). The first sub-library was created from the yeast deletion collection (Giaever et al., 2002) and consisted of 379 non-essential genes. The second sub-library was created from a decreased abundance by mRNA perturbation (DAmP) library and consisted of 323 hypomorphic alleles of essential genes (Breslow et al., 2008; Schuldiner et al., 2005). The genes were selected according to the SGD description as related to ER or their GFP-tagged protein was reported to localize to the ER.

The cellular localization of WSC2-MS2L and EAR1-MS2L was then visualized in these mutant strains using a high-throughput microscopy setup (Cohen and Schuldiner, 2011). Briefly, cells were moved from agar plates into liquid 384-well polystyrene growth plates using the RoTor arrayer. Liquid cultures were grown overnight in SD-URA medium in a shaking incubator (LiCONiC Instruments) at 30°C. A JANUS liquid handler (Perkin Elmer) connected to the incubator was used to back-dilute the strains into plates containing the same medium, after which plates were transferred back to the incubator and were allowed to grow for 4 h at 30°C to reach logarithmic growth. The liquid handler was then used to transfer 50 µl of strains into glass-bottomed 384-well microscope plates (Matrical Bioscience) containing 10 µl of Calcofluor White (final concentration of 10 µg/ml) and coated with concanavalin A (Sigma-Aldrich) to allow formation of a cell monolayer. Wells were washed twice with medium to remove unconnected cells, and plates were transferred into an automated inverted fluorescent microscopic ScanR system (Olympus) using a swap robot (Hamilton). The ScanR system is designed to allow autofocus and imaging of plates in 384-well format using a 60× air lens and is equipped with a cooled CCD camera. Images were acquired at an excitation time of 10 ms for the DAPI channel (Calcofluor White stain) and 1500 ms for the GFP channel. After acquisition, images were imported to the Axiovision software, processed and manually reviewed. Deletion strains that were found to show mislocalization of either WSC2 or EAR1 were rescreened using the manual method described. In total, ≥50 cells were counted to assess mRNP localization.

Purification of recombinant She2p was performed as described previously (Müller et al., 2009). Wild-type She2p was expressed as a GST fusion protein in Escherichia coli BL21(DE3)/pRIL (Invitrogen). Purification to >95% homogeneity was achieved by using standard protein purification techniques. The GST tag was removed by cleavage with tobacco etch virus (TEV) protease (Invitrogen). For storage, glycerol was added to final concentration of 20%. She2p was quickly cooled in liquid nitrogen and stored at −80°C.

Artificial liposomes were prepared as described previously (Genz et al., 2013). Briefly, egg yolk L α phosphatidylcholine (PtdCho) and L α phosphatidylethanolamine (PtdEtn) from Sigma were dissolved in chloroform and mixed at an appropriate ratio to obtain phospholipid mixes containing 0%, 20%, 30%, 40% and 50% PtdEtn. A lipid film was prepared by rotation and evaporation of the organic solvent under N₂ atmosphere. Membranes were dissolved to a final total lipid concentration of 10 mg/ml in degassed liposome buffer (20 mM HEPES pH 7.4, 100 mM NaCl). To create unilamellar liposomes, the emulsion was passed 21 times through a polycarbonate filter membrane with 80-nm pore size mounted in a mini extruder (Avanti Polar Lipids). Liposome size distribution was verified by dynamic light scattering in a Zetasizer Nano ZS (Malvern Instruments, Herrenberg, Germany).

For the flotation assay, 100 µl of liposome solution was mixed with 50 pmol of She2p in 190 µl of binding buffer (50 mM HEPES-KOH, 150 mM potassium acetate, 1 mM magnesium acetate, 1 mM EDTA, 1 mM DTT) and incubated for 15 min on ice. A total of 40 µl of the sample was kept as an input control, and 200 µl was mixed with 3 ml of binding buffer containing 70% sucrose and added to the bottom of an SW40 polycarbonate tube. The sample was covered with three cushions of 3 ml of binding buffer containing 50%, 40% and 0% sucrose. After centrifugation to equilibrium (70,000 g for 4 h at 4°C) the liposome-containing interface between the 40% and 0% sucrose cushions was harvested, precipitated by using trichloroacetic acid (TCA) and dissolved in 45 µl of SDS sample buffer. Flotation samples and input controls were analyzed as described above. She2p signals were analyzed densitometrically using ImageJ.

Yeast cell lysates were prepared by enzymatic disruption of yeast cell walls as described previously (Daum et al., 1982). A membrane fraction containing mitochondria and microsomes was isolated by centrifugation (200,000 g, 4°C, 1 h) and resuspended in SEM buffer using a glass Dounce homogenizer. For thin layer chromatography of lipids, phospholipids were extracted from membranes and analyzed according to a published procedure (Vaden et al., 2005). Thin layer chromatography was performed using HPTLC silica gel 60 F254 plates. Phospholipids were stained by spraying the plate with Molybdenum Blue.

We would like to thank Chao-Wen Wang (Academia Sinica, Taipei, Taiwan) for the TGB3-mCherry expression plasmid and Anne Spang (Biozentrum, Basel, Switzerland) for ABP140-mCherry. We are grateful to Ulrike Thiess (Interfaculty Institute of Biochemistry, Tübingen, Germany) for performing secretion assays.