Register for Mechanobiology 2016!

Integral membrane proteins Brr6 and Apq12 link assembly of the nuclear pore complex to lipid homeostasis in the endoplasmic reticulum
Christine A. Hodge, Vineet Choudhary, Michael J. Wolyniak, John J. Scarcelli, Roger Schneiter, Charles N. Cole


Cells of Saccharomyces cerevisiae lacking Apq12, a nuclear envelope (NE)-endoplasmic reticulum (ER) integral membrane protein, are defective in assembly of nuclear pore complexes (NPCs), possibly because of defects in regulating membrane fluidity. We identified BRR6, which encodes an essential integral membrane protein of the NE-ER, as a dosage suppressor of apq12 Δ. Cells carrying the temperature-sensitive brr6-1 allele have been shown to have defects in nucleoporin localization, mRNA metabolism and nuclear transport. Electron microscopy revealed that brr6-1 cells have gross NE abnormalities and proliferation of the ER. brr6-1 cells were hypersensitive to compounds that affect membrane biophysical properties and to inhibitors of lipid biosynthetic pathways, and displayed strong genetic interactions with genes encoding non-essential lipid biosynthetic enzymes. Strikingly, brr6-1 cells accumulated, in or near the NE, elevated levels of the two classes of neutral lipids, steryl esters and triacylglycerols, and over-accumulated sterols when they were provided exogenously. Although neutral lipid synthesis is dispensable in wild-type cells, viability of brr6-1 cells was fully dependent on neutral lipid production. These data indicate that Brr6 has an essential function in regulating lipid homeostasis in the NE-ER, thereby impacting NPC formation and nucleocytoplasmic transport.


In eukaryotic cells, all transport between the nucleus and cytoplasm takes place through nuclear pore complexes (NPCs), very large, macromolecular channels that span the inner and outer nuclear membranes (INM and ONM, respectively) (for reviews, see Tran and Wente, 2006, Lim et al., 2008). NPCs have eightfold symmetry perpendicular to the nuclear envelope (NE) and contain multiple copies of approximately 30 proteins called nucleoporins. Relatively little is known about how NPC assembly occurs, which requires fusion of the INM and ONM to create a membrane tunnel within which is embedded an NPC. In metazoan cells, two types of NPC assembly take place: re-assembly late in mitosis as the NE reassembles, and de novo assembly during interphase, when the number of NPCs doubles (for reviews, see Antonin et al., 2008; D'Angelo and Hetzer, 2008). Saccharomyces cerevisiae has a closed mitosis and all NPC biogenesis therefore occurs de novo.

Several factors in addition to nucleoporins have been implicated in NPC biogenesis. These include components of the Ran GTPase system ( Ryan and Wente, 2002), whose role in coordinating assembly and disassembly of karyopherin-cargo complexes during nucleocytoplasmic transport has been well studied (for reviews, see Conti et al., 2006; Cook et al., 2007; Stewart, 2007; Terry et al., 2007). Cells carrying specific mutant alleles affecting these proteins showed mislocalization of nucleoporins as well as accumulation of nucleoporin-containing cytoplasmic vesicles ( Ryan and Wente, 2002; Ryan et al., 2003; Ryan et al., 2007). Proteins involved in endoplasmic reticulum (ER) to Golgi trafficking, including components of the COPII coat, have also been implicated in NPC assembly ( Ryan and Wente, 2002).

Previously, we identified Apq12 in a synthetic genetic array (SGA) screen starting with the rat8-2 allele of DBP5, which encodes a key mRNA export factor ( Scarcelli et al., 2007). Apq12 is an integral membrane protein of the ER and NE, and is required for efficient assembly of NPCs ( Scarcelli et al., 2007). Although Apq12 is not essential for viability, apq12 Δ cells are cold sensitive (16°C) for growth and grow best at elevated (>30°C) temperatures. In the absence of Apq12, a subset of nucleoporins mislocalize to cytoplasmic foci, particularly those comprising the cytoplasmic filaments of the NPC. Electron microscopy revealed that apq12 Δ cells contain many defective pores that contact the INM but not the ONM, where the cytoplasmic filaments are located ( Scarcelli et al., 2007).

Addition of the membrane-fluidizing agent benzyl alcohol to apq12 Δ cells suppressed mislocalization of cytoplasmic filament nucleoporins and defects in mRNA export ( Scarcelli et al., 2007). Interestingly, increasing the amount of benzyl alcohol beyond the level used to suppress apq12 Δ phenotypes causes nucleoporin mislocalization in wild-type cells ( Izawa et al., 2004). This suggests that NPC assembly is sensitive to the fluidity of the NE and that the observed defects in NPC assembly in apq12 Δ cells could result from improper regulation of the lipid composition of the nuclear membrane in response to changes in temperature. Cells respond to changes in temperature and environmental conditions by adjusting the composition of their membranes to maintain the fluidity and flexibility required for optimal function ( Nishida and Murata, 1996). For example, following a shift to lower temperatures, cells adjust the phospholipid composition of their membranes to contain an elevated proportion of unsaturated and shorter acyl chains. Relative levels of other membrane lipid components including sterols also change in response to temperature shifts ( Buttke et al., 1980; Sharma, 2006).

To gain further insight into the interplay between biophysical properties of the NE and NPC assembly, we conducted a screen for dosage suppressors of the cold-sensitive growth defect of apq12Δ. This screen identified BRR6, an essential gene that was originally identified in a screen for cold-sensitive mutants defective in mRNA export ( de Bruyn Kops and Guthrie, 2001). Brr6 encodes an integral membrane protein of the ER and NE, with two predicted transmembrane domains. Cells carrying the cold-sensitive brr6-1 allele are defective for proper localization of some nucleoporins ( de Bruyn Kops and Guthrie, 2001). Interestingly, overexpression of APQ12 partially suppressed the growth defect of brr6-1 cells, indicating that Apq12 and Brr6 share some common function.

Here, we report that cytoplasmic filaments but not nuclear basket nucleoporins are mislocalized in brr6-1 cells at low temperatures (16°C-23°C), in a similar fashion to apq12 Δ cells. In contrast to apq12 Δ, however, brr6-1 cells are hypersensitive to agents that increase membrane fluidity such as benzyl alcohol and oleic acid, and to drugs that inhibit sterol or fatty acid synthesis. Consistent with this drug sensitivity, brr6-1 displays strong genetic interactions with mutants that have defects in the late part of the sterol biosynthetic pathway or in fatty acid elongation. Biochemical analyses revealed that brr6-1 cells contain dramatically elevated levels of total sterols (particularly steryl esters and episterol) and triacylglycerols (TAGs). Moreover, overproduction of these non-essential storage lipids is functionally important, because brr6-1 cells are non-viable if neutral lipid synthesis is blocked by deletion of the respective biosynthetic genes. The data indicate that Apq12 and Brr6 are likely to have an important role in regulating lipid homeostasis in the ER, thereby impacting NPC biogenesis, nuclear transport and nuclear mRNA metabolism.


Genetic interactions between BRR6 and APQ12

To understand better the function of Apq12, we conducted a genetic screen to identify dosage suppressors of the cold-sensitive growth phenotype of apq12Δ. We found that high-copy-number plasmids containing BRR6 restored growth of apq12 Δ at 16°C to a similar degree as did direct complementation with wild-type APQ12 ( Fig. 1A). Conversely, overexpression of APQ12 partially suppressed the growth defect of brr6-1 at both low and high temperatures ( Fig. 1A). A double mutant carrying brr6-1 and apq12Δ was unable to grow at all temperatures tested (data not shown). brr6-1 cells show a modest (30% of cells) mRNA export defect at 30°C, and this was also partially suppressed by overexpression of APQ12 ( Fig. 1B). Interestingly, overexpression of BRR6 completely suppressed the mRNA export defect of apq12 Δ cells ( Fig. 1B).

To determine whether overexpression of BRR6 could restore proper localization of nucleoporins in apq12 Δ cells, we examined the subcellular distribution of GFP-tagged Nup82 and Nup188. Control apq12 Δ cells carrying an empty vector displayed aberrant localization of both Nup82-GFP and Nup188-GFP, whereas apq12Δ cells harboring the APQ12 plasmid displayed normal nucleoporin localization ( Fig. 1C). Overexpression of BRR6 completely restored Nup188-GFP localization, whereas Nup82-GFP localization was partially restored. In the latter case, most cells showed Nup82-GFP at both the NE and in cytoplasmic foci, many of which were located near the nuclear periphery ( Fig. 1C, and data not shown). Overexpression of BRR6 also restored Nup159/Rat7 localization to about the same extent as that of Nup82-GFP (data not shown). Overexpression of APQ12 partially suppressed mislocalization of nucleoporins in brr6-1 cells ( Fig. 1C). At 30°C and18°C, the fraction of brr6-1 cells with mislocalized Nup82-GFP was approximately 25% and 60%, respectively. At 18°C, the fraction was reduced to about 25% in cells overexpressing APQ1. Taken together, these results indicate that Apq12 and Brr6 probably have related functions that affect growth at low temperature, mRNA export, and NPC assembly and distribution.

Fig. 1.

Dosage suppression of growth, mRNA export and nucleoporin mislocalization of brr6-1 and apq12Δ cells by APQ12 and BRR6. (A) Serial dilutions of apq12Δ and brr6-1 cells overexpressing BRR6 and APQ12 were spotted onto plates and incubated at the temperatures shown. (B) In situ hybridization assay to analyze the effect on the mRNA-export defects of apq12Δ and brr6-1 cells of over-expressing BRR6 and APQ12 by using multicopy plasmids. (C) Direct visualization of Nup82-GFP and Nup188-GFP in apq12Δ and brr6-1 cells overexpressing BRR6 and APQ12. Nup82-GFP and Nup188-GFP were expressed from the NUP82 and NUP188 chromosomal loci. Scale bars: 5 μm.

Fig. 2.

Mislocalization of cytoplasmic filament nucleoporins in brr6-1 cells. (A) Wild-type and brr6-1 cells expressing Nup60-GFP or Nup82-GFP were shifted to 20°C overnight and live cells directly visualized. (B) Wild-type and brr6-1 cells grown either at 30°C or shifted to 18°C overnight before visualization. Note the heterogeneity of the distribution of Nup82-GFP in brr6-1 but not in wild-type cells (right). Scale bars: 5 μm.

brr6-1 affects the assembly of NPC filaments

BRR6 is essential for viability and encodes a 197 amino acid protein; this protein, similarly to Apq12, is an integral membrane protein of the NE-ER, with two predicted transmembrane domains. The brr6-1 allele (R110K) is both cold (16°C) and heat (37°C) sensitive. However, even at the optimal permissive temperature (30°C), brr6-1 cells have defects in mRNA export and two core nucleoporins, Nsp1 and Nup188, are mislocalized ( de Bruyn Kops and Guthrie, 2001). Because cytoplasmic filaments but not nuclear basket nucleoporins are mislocalized in apq12 Δ cells ( Scarcelli et al., 2007), we extended earlier studies by localizing additional nucleoporins in brr6-1 cells. This revealed extensive mislocalization of Nup82-GFP, a cytoplasmic filament nucleoporin, to cytoplasmic and perinuclear foci ( Fig. 2A), but no mislocalization of Nup60-GFP, a nuclear basket nucleoporin, at any temperature (data not shown). At all temperatures, Nup82-GFP was localized properly in some cells (with all or most Nup82-GFP at the nuclear periphery), whereas in others, all or most Nup82-GFP was cytoplasmic with many bright foci. Nup159, another cytoplasmic filament nucleoporin, was also extensively mislocalized (data not shown), with phenotypes ranging from normal or nearly normal localization to mislocalization of most or all Nup159-GFP. We hypothesize that the heterogeneity seen in the localization of cytoplasmic filament nucleoporins in the brr6-1 background reflects the different numbers of copies of plasmid-borne brr6-1 in different cells. Because we were unable to replace BRR6 in the genome with brr6-1, a single copy of brr6-1 appears to be insufficient to support growth.

Fig. 3.

Electron microscopy reveals NE abnormalities in brr6-1 cells. Wild-type and brr6-1 cells were grown at 30°C or shifted to 18°C overnight and processed for thin-section electron microscopy. (A) Wild-type cells, 18°C; (B) wild-type cells, 30°C; (C-E) brr6-1 cells, 18°C; (F,G) brr6-1 cells, 30°C. Arrows and arrowheads: electron-dense structures and inclusions [arrowheads (A,B,G), arrow (D)], sometimes clustered [arrows (E,F)] or clustered in additional double membranes [black arrows (C,G)]; extensive white expansions of the NE and ER [white arrows (C)]. n, nucleus.

We showed earlier that apq12 Δ cells are defective in NPC biogenesis but not in NPC stability ( Scarcelli et al., 2007) and wanted to determine whether this is also the case for brr6-1 cells. Two factors complicate this analysis. First, whereas the apq12Δ cell population is genetically homogeneous (all cells lack Apq12), brr6-1 cells are not. The brr6-1 allele is carried on a CEN plasmid (with the genomic copy of BRR6 deleted), and therefore the number of copies of the mutant gene is not the same in all cells. Second, in apq12Δ cells there is no detectable mislocalization of any nucleoporins at 30°C and extensive mislocalization at 16°C ( Scarcelli et al., 2007). This makes it straightforward to observe and quantify changes in nucleoporin localization when apq12Δ cells are temperature shifted. By contrast, nucleoporins mislocalized in brr6-1 cells are mislocalized to some extent at all temperatures, although the defect is more severe at lower temperatures. Fig. 2B shows a montage of brr6-1 cells grown at 30°C or shifted to 18°C overnight. The heterogeneity of phenotypes is readily apparent and indicates that even under optimal growth condition (30°C), Nup82-GFP is mislocalized in many cells. Because cells are able to grow and divide at this temperature, the nucleoporin-mislocalization phenotype indicates that brr6-1 cells are defective in NPC assembly. Therefore, it is difficult to determine whether NPCs are also unstable in brr6-1.

To define the NPC mislocalization and NE phenotype of brr6-1 at the ultrastructural level, cells were fixed and analyzed by electron microscopy ( Fig. 3). In wild-type cells at both 18°C and 30°C, NPCs appear as electron-dense material that spans the lumen of the NE (arrowhead in Fig. 3A,B). Essentially all NPCs from wild-type cells had this normal appearance. By contrast, in brr6-1 cells at both 18°C ( Fig. 3C-E) and 30°C ( Fig. 3F,G) severe abnormalities were seen. The overall number of NPCs or related structures seen in each section was approximately half that seen in wild-type sections. Of these, about one-third appeared to be normal. In brr6-1 cells, there were many electron-dense inclusions beneath the INM or between the INM and ONM (black arrow in Fig. 3D, white arrowheads in Fig. 3G), often roughly spherical, larger than NPCs and sometimes clustered (arrows in Fig. 3E,F). At 30°C, there were about half as many of these inclusions as normal-appearing NPCs, whereas in cells maintained at 18°C, dense inclusions in these locations were about twice as common as normal-appearing NPCs. Interestingly, membrane blebs, most containing electron-dense inclusions, were seen in brr6-1 cells at 30°C (about as many blebs near the NE as there were normal-appearing NPCs), but very rarely in cells maintained at 18°C overnight. We cannot tell whether these are partially-assembled NPCs, aggregates of nucleoporins or other material. In about one-third of thin sections, electron-dense structures were clustered together in additional double-membranes (black arrows in Fig. 3C,G), often extending linearly a considerable distance away from the NE ( Fig. 3C). Although these structures resemble NPCs, we do not believe that they are NPCs or NPC-related structures because GFP-tagged nuclear basket nucleoporins and some core nucleoporins were detected only at the nuclear periphery in brr6-1 cells, and never in these projections (data not shown). The composition of these structures is not known. The extensive white expansions of the NE and ER (white arrows in Fig. 3C) might be lipids, reflecting abnormalities in lipid metabolism (see below). These were observed in approximately 30% of thin sections of brr6-1 cells, but in only about 5% of sections of wild-type cells, and those seen in wild-type cells were much smaller than those in brr6-1 cells. In apq12Δ cells, there were very few normal NPCs and a large number of what appeared to be partial NPCs extending from the INM part of the way across the lumen of the NE ( Scarcelli et al., 2007). Partial NPCs were rare in brr6-1 cells and very few normal-appearing NPCs were present. This might indicate a more severe NPC assembly defect in these cells compared with apq12 Δ cells. These aberrant phenotypes were seen very rarely in wild-type cells.

brr6-1 is hypersensitive to treatments that affect membrane fluidity and to drugs that inhibit lipid biosynthesis

A key finding made earlier was that the growth of apq12 Δ cells in medium containing benzyl alcohol partially suppressed defects in nucleoporin localization and mRNA export ( Scarcelli et al., 2007). The growth of brr6-1 cells was inhibited by benzyl alcohol at all temperatures ( Fig. 4A). Since the cold-sensitive phenotype conferred by the brr6-1 and apq12 Δ mutations might reflect a loss of ability to make proper adjustments in membrane fluidity, we wondered whether supplementing the medium with the unsaturated fatty acid oleic acid could rescue the defect. Surprisingly, 5 mM oleic acid was detrimental to the growth of both brr6-1 and apq12 Δ cells ( Fig. 4A).

Fig. 4.

Effect of compounds that alter membrane properties or inhibit lipid biosynthetic pathways on growth of wild-type, brr6-1 and apq12Δ cells. Serial dilutions of wild-type, brr6-1 and apq12Δ cells were grown on YPD plates or on plates containing: (A) the membrane fluidizers benzyl alcohol or oleic acid, or (B) drugs that inhibit sterol synthesis (terbinafine and ketoconazole), or fatty acid synthesis (cerulenin) and incubated at the temperatures shown.

We examined the effects on brr6-1 and apq12 Δ cells of terbinafine (5-50 μg/ml) and ketoconazole (1-2 μg /ml), which are compounds that inhibit different enzymes in the ergosterol biosynthesis pathway ( Petranyi et al., 1984; Sheets and Mason, 1984; Meredith et al., 1985). The two drugs had little or no effect on growth of wild-type or apq12 Δ cells ( Fig. 4B). Strikingly, the brr6-1 strain was hypersensitive to both. Cerulenin is an inhibitor of fatty acid synthase ( Omura, 1981). Growth of both brr6-1 and apq12Δ was inhibited by cerulenin under conditions where growth of wild-type cells was not ( Fig. 4B). We conclude that the brr6-1 mutation confers sensitivity to inhibitors of multiple lipid biosynthetic pathways.

Synthetic genetic interactions between lipid biosynthesis mutants and brr6-1 and apq12 Δ

We described earlier the genetic interaction between apq12 Δ and acc1-7-1 ( Scarcelli et al., 2007). ACC1 encodes acetyl-CoA carboxylase, the rate-limiting enzyme in fatty acid synthesis. The apq12Δ acc1-7-1 double-mutant strain grew considerably less well than either single mutant at lower temperatures, and considerably better at 34°C than the temperature-sensitive acc1-7-1 mutant alone. We extended this analysis by testing for synthetic interactions of both apq12 Δ and brr6-1 with deletions of non-essential genes whose products act in diverse lipid biosynthetic pathways. The results are summarized in Table 1. brr6-1 displayed extensive synthetic lethality, with mutations affecting sterol and fatty acid metabolism (e.g. erg2,erg3,erg5,erg6; elo1,elo2,elo3; mga2, spt23). These genetic interactions are consistent with the hypersensitivity of brr6-1 cells to inhibitors of sterol and fatty acid biosynthesis. Although there were fewer cases where apq12 Δ showed synthetic lethality or growth defects when combined with lipid biosynthetic mutants, we did observe synthetic growth defects between apq12Δ and erg2Δ.

View this table:
Table 1.

Genetic interactions of apq12Δ and brr6-1 with mutations affecting lipid biosynthesis

Nem1 and Spo7 are subunits of a heterodimeric phosphatase ( Siniossoglou et al., 1998; Santos-Rosa et al., 2005) that regulates indirectly several phospholipid biosynthetic genes by modulating the phosphorylation state of Pah1, a transcriptional repressor whose targets are phospholipid biosynthetic genes ( Siniossoglou et al., 1998; Santos-Rosa et al., 2005). The morphology of the NE is abnormal in cells lacking Spo7, Nem1 or Pah1/Smp2, which is also the case in brr6-1 and apq12 Δ cells. brr6-1 displayed synthetic lethality with spo7 Δ but not with nem1 Δ. This is discussed further below.

apq12 Δ and brr6-1 cells have aberrant lipid profiles

The observations that brr6-1 and, to a lesser extent, apq12 Δ cells, are sensitive to drugs that inhibit lipid biosynthetic pathways, together with the fact that these mutants exhibit synthetic interactions with mutations affecting multiple lipid biosynthetic pathways, suggest that Apq12 and Brr6 might directly or indirectly regulate membrane lipid composition and/or lipid synthesis. We examined the phospholipid and neutral lipid composition of apq12Δ and brr6-1 cells by labeling with [3H]palmitic acid for 6 hours at 20, 24 or 37°C. Lipids were extracted and analyzed by thin layer chromatography (TLC). This revealed that wild-type and mutant strains had the same phospholipid composition (data not shown). By contrast, there was a twofold increase in the level of neutral triacylglycerols (TAGs) in brr6-1 cells at 37°C ( Fig. 5A). This hyper-accumulation was rescued completely by the presence of wild-type BRR6, indicating that Brr6 function is important for neutral lipid homeostasis at 37°C.

Drug sensitivity and genetic interactions also suggested a possible role for Brr6 in sterol function and/or metabolism. To examine whether apq12 Δ and brr6-1 affect sterol metabolism, lipids were extracted from cells grown overnight at 24°C, and the sterol profile was determined by gas chromatography and mass spectrometry (GC-MS). This revealed a sevenfold increase in the level of total sterols in brr6-1 compared with wild-type cells ( Fig. 5B), which was due primarily to increased levels of steryl esters, because the levels of free sterols in the brr6-1 strain were not significantly different from the wild type (data not shown). In addition, this analysis revealed a three- to fourfold increase in the level of the sterol episterol in both apq12 Δ and brr6-1 cells ( Fig. 5C). Episterol is an intermediate in ergosterol biosynthesis, but is normally present at a very low level. This phenotype was also rescued by the presence of plasmid-borne APQ12 (in apq12Δ cells) or BRR6 (in brr6-1 cells). Taken together, these results indicate that apq12 Δ and brr6-1 cells have defects in sterol synthesis and that brr6-1 cells possess a strong defect in neutral lipid homeostasis, resulting in the accumulation of both triacylglycerols and steryl esters.

Fig. 5.

Analysis of TAGs and sterols in apq12Δ and brr6-1 cells. (A) brr6-1 cells accumulate triacylglycerols. Cells were labeled with [3H]palmitic acid for 6 hours at 37°C. Lipids were extracted, separated by thin layer chromatography and levels of radiolabeled fatty acids in the different lipid classes were quantified by radio-scanning. (B) brr6-1 cells have elevated levels of steryl esters. (C) brr6-1 and apq12 Δ cells accumulate the sterol precursor episterol. Cells were cultivated overnight at 24°C. Lipids were extracted and sterols were analyzed and quantified by GC-MS, using cholesterol as an internal standard. These analyses were performed independently two (B,C) or three (A) times and results shown are means ± s.e.m.

Overexpression of Brr6 restores normal episterol levels to apq12 Δ cells

We next tested how the altered sterol levels of apq12 Δ and brr6-1 cells were affected by overexpression of BRR6 or APQ12, respectively. The data in Fig. 6A show that overexpression of BRR6 restored the normal level of episterol in apq12Δ cells. By contrast, only a modest reduction in the level of episterol was seen when APQ12 was overexpressed in brr6-1 cells ( Fig. 6B) and this was not statistically significant. Most likely, the difference in the ability of Apq12 and Brr6 to suppress the defects in episterol levels reflects the fact that Brr6 is essential and Apq12 is not. This could also underlie the partial suppression of the brr6-1 growth defect by overexpression of APQ12, in contrast to the complete suppression of the apq12 Δ growth defect by BRR6. Importantly, the suppression of both the growth defects and elevated episterol level of apq12 Δ by BRR6 supports the hypothesis that Apq12 and Brr6 have direct roles in modulating or sensing lipid levels or membrane properties.

Fig. 6.

Overexpression of BRR6 suppresses the elevated level of episterol in apq12Δ cells. apq12Δ and brr6-1 cells containing an empty vector or overexpressing BRR6 (A) or APQ12 (B), respectively, from multicopy plasmids, were cultivated overnight at 24°C. Lipids were extracted and sterols were analyzed and quantified by GC-MS. These analyses were performed independently two times and results shown are means ± s.e.m.

Production of neutral lipids is essential for viability of brr6-1

To examine whether the increased levels of TAGs and steryl esters in brr6-1 are of functional importance, we tested the viability of brr6-1 and apq12 Δ cells in combination with deletions of ARE1 and ARE2, which encode the two yeast sterol acyltransferases, or LRO1 and DGA1, which encode two acyltransferases that synthesize triacylglycerol. Because an are1 Δ are2 Δ double mutant is viable ( Yang et al., 1996; Yu et al., 1996), synthesis of steryl esters is dispensable in an otherwise wild-type background. Similarly, because a lro1 Δ dga1 Δ double mutant is viable, synthesis of triacylglycerides is also dispensable ( Oelkers et al., 2002; Sorger and Daum, 2002). However, combining are1 Δ are2 Δ with brr6-1 caused a strong synthetic growth defect ( Fig. 7A), indicating that conversion of free sterols to steryl esters is important for growth of brr6-1 cells. Synthetic lethality was seen when brr6-1 was combined with lro1Δ dga1Δ ( Fig. 7B), indicating that TAG synthesis is essential for growth of brr6-1 cells.

The fact that brr6-1 cells accumulate elevated levels of steryl esters and require steryl ester formation for survival could indicate that these cells have elevated levels of free sterols in the ER that might result from a defect in export of free ergosterol from the ER. We monitored the subcellular distribution of sterols using the fluorescent sterol analogue NBD-cholesterol, which we have previously shown to reflect the known natural subcellular distribution of ergosterol ( Reiner et al., 2006). Mutant cells deficient in heme production (to block endogenous sterol synthesis and allow uptake of exogenous sterols) were incubated with NBD-cholesterol for 1 hour at 24°C and its distribution was analyzed by fluorescence microscopy. This revealed prominent ring-like staining of the plasma membrane and punctuate intracellular staining of lipid droplets in wild-type cells ( Fig. 8A). By contrast, the brr6-1 mutant displayed substantially elevated staining of aberrantly large lipid droplets that were frequently in close proximity to or possibly associated with the nuclear ER ( Fig. 8A, arrowheads), as revealed by visualization of the ER with the ER luminal marker Kar2-mRFP-HDEL ( Gao et al., 2005). Endogenously produced sterols showed similar localization, as revealed by staining with filipin (data not shown). Lipid droplets are dedicated storage organelles for neutral lipids and contain primarily steryl esters and triacylglycerides. These observations, together with the observed elevated levels of steryl esters, indicate that brr6-1 hyperesterifies sterols that are synthesized endogenously, as well as sterols taken up from the outside.

Fig. 7.

Production of steryl esters and triacylglycerides is essential in brr6-1 cells. (A) Analysis of brr6-1 synthetic interactions with are1 Δ are2 Δ. ARE1 and ARE2 were deleted in a brr6-1 background to produce the are1 Δ are2 Δ brr6-1 triple mutant. Growth of the indicated strains on YPD plates is shown. (B) Analysis of brr6-1 synthetic interactions with lro1 Δ dga1 Δ. LRO1 and DGA1 were deleted in a brr6-1 strain containing a wild-type copy of BRR6 on a URA3 plasmid. Cells were plated on YPD plates and then replicated to YPD plates and to plates containing 5-fluoro-orotic acid (5-FOA) to select for colonies that have lost the URA3 BRR6 plasmid. Plates were incubated for 4 days at 24°C.

To monitor the sterol hyperesterification phenotype more directly, heme-deficient cells were incubated with [14C]cholesterol and the conversion of free cholesterol to steryl esters was analyzed by TLC analysis of radiolabeled sterols. This revealed that brr6-1 cells accumulated approximately one-third more free sterol ( Fig. 8B) than wild-type cells. Furthermore, brr6-1 cells had a strongly elevated rate of steryl ester formation as they produced three times as many steryl esters than wild-type cells after 8 hours of incubation in medium containing [14C]cholesterol ( Fig. 8B).

Fig. 8.

brr6-1 cells accumulate free sterols and steryl esters. (A) Subcellular distribution of NBD-cholesterol. Heme-deficient cells of the indicated genotype and expressing the ER marker Kar2-mRFP-HDEL were incubated with the fluorescent sterol analog NBD-cholesterol for 1 hour at 24°C and the distribution of NBD-cholesterol was analyzed by fluorescence microscopy. White arrowheads indicate staining of the NE-ER. The numbers indicate the proportion of NBD-cholesterol-stained lipid droplets that colocalize with the Kar2-mRFP-HDEL-stained NE-ER (n=100 cells). Scale bar: 5 μm. (B) Heme-deficient cells of the indicated genotype were incubated with [14C]cholesterol. Samples were removed at the time points indicated, lipids were extracted and the levels of free and esterified cholesterol were quantified. Results shown are means ± s.e.m.

The observation that brr6-1 has strongly elevated levels of the two neutral lipids, steryl esters and TAGs, and the fact that steryl ester and TAG formation is essential in brr6-1 point to a possible function of Brr6 in neutral lipid homeostasis. The enzymes that synthesize neutral lipids (Are1, Are2, Dga1 and Lro1) are localized in ER membranes (for a review, see Czabany et al., 2007). The resulting neutral lipids are then deposited into lipid droplets. These lipid droplets are believed to be formed from the ER membrane, from which they might bud off to become independent organelles ( Murphy and Vance, 1999). It is interesting to note that apq12Δ was identified as an mld (many lipid droplets) mutant in a screen of the yeast knockout strains to identify mutants with altered lipid-droplet morphology ( Fei et al., 2008).

To examine whether brr6-1 displayed an aberrant lipid-droplet phenotype, lipid droplets were visualized using Erg6-RFP as a marker protein. Erg6-RFP was mislocalized to the NE in brr6-1 cells, and lipid droplets were frequently seen encircling the NE of these cells ( Fig. 9, arrowheads). Quantification of this lipid droplet phenotype revealed that in 72.5% of brr6-1 cells, lipid droplets were arranged in a circular pattern around the DAPI-stained nucleus. In wild-type or apq12 Δ cells, by contrast, lipid droplets were observed in a circular pattern around the nucleus in fewer than 30% of cells. These observations suggest that the overproduction of neutral lipids in brr6-1 results in over-accumulation of neutral lipids in the NE and perinuclear ER, where the enzymes for TAG and steryl ester synthesis are located, and that these neutral lipids are not efficiently partitioned into lipid droplets. Accumulation of neutral lipids in the ER, as suggested by ultrastructural analysis ( Fig. 3C), could then result in altered membrane properties and defects in NPC assembly.

Fig. 9.

brr6-1 cells have aberrant lipid droplets. Cells were transformed with a plasmid expressing the lipid droplet marker protein Erg6-RFP and analyzed by fluorescence microscopy. White arrowheads indicate lipid droplets that encircle the NE-ER. The numbers indicate the proportion of cells that display circular arrangement of Erg6-RFP marked lipid droplets around the nucleus (n=100 cells). DNA was revealed by staining with DAPI. Scale bar: 5 μm.


Genetic interactions and similar mutant phenotypes for brr6-1 and apq12Δ

We identified BRR6 as a dosage suppressor of apq12 Δ. brr6-1 and apq12Δ are synthetically lethal, and APQ12 and BRR6 were able to suppress the cold-sensitive growth defects of each other ( Fig. 1). Although the phenotypes seen in brr6-1 cells are similar to those of apq12Δ, brr6-1 defects were generally much stronger. brr6-1 was more sensitive to chemical and pharmacological agents that impact lipid biosynthesis and membrane fluidity, showed more extensive alterations in the levels of neutral lipids, and was synthetically lethal with more lipid biosynthetic mutants than was apq12Δ. These differences are consistent with a more important role for Brr6 in the maintenance of membrane fluidity, than for Apq12, which is not essential.

A third gene, encoding a Brr6-related protein, BRL1 (BRR6-like), was identified as a suppressor of a temperature-sensitive xpo1 (exportin 1) mutant ( Saitoh et al., 2005). Brl1 is essential, has substantial homology to Brr6 (24% identical and 46% similar) and is also found in the NE. Proteins containing the domain shared between Brr6 and Brl1 are present in fungi that undergo closed mitoses, including Candida glabarata and Cryptococcus neoformans. brl1 mutants share phenotypes of nucleoporin mislocalization and mRNA-export defects with brr6-1 and apq12Δ ( Saitoh et al., 2005). BRR6 was able to suppress the growth defect of brl1 mutants, which were synthetically lethal with brr6-1. Brl1 and Brr6 interacted with each other in a two-hybrid analysis ( Saitoh et al., 2005). However, we found that BRL1 could not suppress apq12 Δ (unpublished results). Future studies will investigate the possibility that these three proteins function together.

Normal nuclear morphology is dependent upon membrane composition

Many earlier studies indicated that the shape of the nucleus is related to lipid composition of cellular membranes. Nem1 and Spo7 are the catalytic and regulatory subunits, respectively, of an ER-NE-associated protein phosphatase ( Siniossoglou et al., 1998; Santos-Rosa et al., 2005). The target of this enzyme is Pah1/Smp2 ( Santos-Rosa et al., 2005), which acts as a transcriptional repressor of several phospholipid biosynthesis genes and is a homologue of mammalian Lipin, expressed at a high level in adipose tissue ( Peterfy et al., 2001). Pah1 has phosphatidic acid phosphatase activity and thereby controls the synthesis of triacylglycerols ( Han et al., 2006). NEM1, SPO7 and PAH1 are non essential, but strains lacking any one have abnormal nuclear morphology including extensions of the NE that contain NPCs ( Siniossoglou et al., 1998; Tange et al., 2002). Overproduction of Pah1 restores normal nuclear membrane structure to nem1 Δ and spo7 Δ cells. In cells lacking Pah1, defects in lipid metabolism affect multiple classes of lipids including triacylglycerols ( Han et al., 2006). These studies demonstrate that nuclear shape and the distribution of NPCs depends upon proper regulation of lipid biosynthesis. The finding that brr6-1 is synthetically lethal with spo7Δ but not nem1Δ suggests that deletions of NEM1 and SPO7 affect certain aspects of TAG synthesis differently from each other. Spo7 is the regulatory subunit for the Spo7/Nem1 heterodimer ( Siniossoglou et al., 1998; Santos-Rosa et al., 2005) and could be important for localized activation of Pah1. This might explain why brr6-1 is lethal when combined with spo7 Δ but viable when combined with nem1 Δ. We also found that PAH1 is essential in our wild-type strain background. PAH1, SPO7 and NEM1 have also been linked genetically to NUP84 ( Siniossoglou et al., 1998; Santos-Rosa et al., 2005). Together, these results are consistent with our finding that apq12 Δ and brr6-1cells have alterations in lipid metabolism, NE structure and shape, and nucleoporin localization.

Other studies have suggested that the lipid composition of NE itself is important for NPC biogenesis. A mutation in ACC1/MTR7, which encodes acetyl CoA carboxylase, was isolated in a screen for mRNA-export mutants ( Schneiter et al., 1996). Acc1 is essential, and is responsible for synthesis of malonyl CoA, the key building block for synthesis of fatty acids. mtr7-1 acc1-7-1 cells have been shown to have abnormal NPCs and mislocalized nucleoporins, and were defective in the formation of very-long-chain fatty acids, something not observed with other acc1 alleles ( Schneiter et al., 1996). This suggested that very-long-chain fatty acids might have a role in NPC assembly, perhaps by helping to bring together the INM and ONM at points where fusion is to occur, as NPCs are assembled ( Schneiter et al., 1996; Schneiter et al., 2004).

brr6-1 and apq12 Δ might affect the functioning of a membrane-fluidity sensor

An interesting possibility to account for our results is that the apq12 Δ and brr6-1 mutations impact the known ability of cells to sense environmental changes that normally trigger modifications in membrane composition needed to maintain membrane homeostasis (for reviews, see Murata and Los, 1997; Los and Murata, 2004; Zhang and Rock, 2008). Little is known about how fluidity and other biophysical properties of membranes are sensed and how this information is transduced to adjust lipid synthesis and membrane composition. Fluidity sensors are thought to respond to decreases in temperature, in part through activating fatty acid desaturases, which introduce double-bonds into acyl chains of fatty acids and phospholipids, and thereby have a large impact on the melting temperature of membranes. To function properly, cells must be able to sense the need to induce changes in membrane composition and to prevent changes of too great a magnitude.

Yeast has a single fatty acid desaturase, Ole1 ( Stukey et al., 1989), whose production is tightly regulated. Mga2 and Spt23, two homologous membrane-associated transcription factors related to mammalian NF-κB, are released from membrane association by proteolysis when cells sense the need to activate OLE1 ( Hoppe et al., 2000). Neither Mga1 nor Spt23 is essential, but cells lacking both are not viable ( Zhang et al., 1999). brr6-1 showed synthetic lethality with both mga2 Δ and spt23 Δ ( Table 1). Morphological defects in the NE were seen following a shift of an mga2 Δ spt23-ts mutant to non-permissive temperature ( Zhang et al., 1999). Cells lacking Ole1 are able to grow when supplemented with unsaturated fatty acids (UFAs) but the membrane abnormalities seen following depletion of UFAs are substantially more severe than those seen in mga2 Δ spt23-ts cells ( Zhang et al., 1999). One possibility to account for this is that membrane anchoring of Mga2 and Spt23 might be sensitive to changes in membrane fluidity, such that changes in fluidity could promote proteolytic activation of Mga2 and Spt23, thereby leading to inappropriate activation of OLE1.

If brr6-1 or apq12 Δ cells were defective in their ability to induce modifications in membrane composition in response to a shift to a lower temperature, it might be expected that supplementing the medium with oleic acid would at least partially suppress the defects. The double bond in the oleate acyl chain introduces a kink that disrupts acyl chain packing in the lipid bilayer. An increase in oleate incorporation into phospholipids therefore can increase membrane fluidity ( Hazel, 1995). Suppression of brr6-1 and apq12Δ by oleic acid was not observed. Rather, both strains showed sensitivity to oleic acid at 23°C and brr6-1 cells were also sensitive to oleic acid at 30°C ( Fig. 4A). This suggests that the observed defects might reflect an excess of membrane fluidity at lower temperatures, as opposed to a deficit, which is consistent with the hypersensitivity of brr6-1 cells to benzyl alcohol. If this were the case, then the primary defect might be the inability of cells to sense when proper membrane composition and dynamics had been restored following a temperature shift.

Lipid esterification is essential for viability of brr6-1 mutant cells

Our finding that the viability of brr6-1 cells depends on esterification of sterols ( Fig. 7) is particularly intriguing and suggests that these cells accumulate elevated levels of free sterols in the ER that need to be neutralized by esterification so that they do not adversely affect ER and NE function. Elevated levels of free cholesterol in the ER of animal cells result in depletion of ER calcium stores, induction of the unfolded protein response (UPR) and ultimately apoptosis ( Feng et al., 2003). Mechanistically, this has been explained by an inhibition of the normal conformational freedom of ER by the cholesterol-induced increase in membrane lipid order ( Li et al., 2004). In yeast, Arv1, which is a conserved ER integral membrane protein, is required for viability of cells that cannot esterify free sterols. Previous work has shown that arv1 Δ mutant cells accumulate sterols in the ER and display elevated levels of steryl esters ( Tinkelenberg et al., 2000). Arv1 is also required for the maturation of glycosylphophatidylinositol (GPI) anchors, which are attached to the C-terminus of proteins in the ER lumen. The alterations in sphingolipid and sterol levels observed in arv1 Δ have been proposed to be a secondary consequence of a reduced flux of lipids out of the ER ( Swain et al., 2002; Kajiwara et al., 2008). Our observation that arv1 Δ is synthetically lethal with brr6-1 indicates that Brr6 provides an overlapping function with Arv1 required for lipid homeostasis in the ER. Although we did not observe defects in ER function in apq12 Δ cells ( Scarcelli et al., 2007), growth of brr6-1 is sensitive to tunicamycin (our unpublished results), a characteristic of mutants with defects in ER function.

Because brr6-1 cells contain elevated levels of TAGs and steryl esters, they would be expected to be more resistant to cerulenin than wild-type cells are. Fatty acids can be released from TAGs and steryl esters to bypass and compensate for the block in fatty acid synthesis caused by cerulenin. Because brr6-1 cells are more sensitive than the wild type to cerulenin, they appear to depend more strongly on ongoing fatty acid synthesis. This might be needed to move another lipid (free sterol, diacylglycerol, or phosphatidic acid) into the neutral lipid pool, where any excess would be less harmful than in a non-acylated form. The finding that brr6-1 is synthetically lethal with elo1Δ, elo2Δ and elo3Δ, supports the idea that the ability to make a range of fatty acids has increased importance when cells are mutant for brr6.

The data presented here provide strong support for the hypothesis that Apq12 and Brr6 have a direct role in lipid and membrane homeostasis and that the observed defects in NPC biogenesis and mRNA export are consequences of the resulting altered membrane properties. Because many strains with defects in mRNA export (e.g. rat8-2/dbp5, mex67-5, prp20-1) have normal NPCs and NE ( Aebi et al., 1990; Segref et al., 1997; Tseng et al., 1998), an mRNA-export defect cannot be the underlying cause of the NE and lipid metabolism abnormalities seen in apq12Δ and brr6-1 cells. Defects in NPCs and the NE are seen in some strains that lack a non-essential nucleoporin or produce a mutant form of an essential nucleoporin. This might indicate that unsuccessful attempts to assemble NPCs can lead to defects in the NE, perhaps resulting from abnormal interactions between the NE and integral membrane nucleoporins or nucleoporins that have a role in curvature of the NE at sites of NPC insertion. However, the fact that neither Apq12 nor Brr6 are components of NPCs makes it unlikely that the primary direct defect from mutations affecting these proteins is in NPC assembly. It will be interesting to see whether there are defects in lipid metabolism in these nucleoporin mutant strains.

It is striking that the NE and NPC defects seen in brr6-1 and apq12 Δ cells are considerably more severe than in cells lacking individual non-essential lipid metabolism genes or those treated with a range of compounds that affect membrane properties and lipid biosynthesis. Further studies will be required to understand the mechanism by which Brr6 and Apq12 affect lipid metabolism and the fluidity of the nuclear membrane, and how this impacts NPC biogenesis.

Materials and Methods

Yeast strains and plasmids

The strains and plasmids used in these studies are listed in supplementary material Table S1. All strains were grown and media prepared using standard methods ( Sambrook et al., 1989). For growth assays, strains were grown overnight, then diluted back to OD600=0.3. Strains were then serially diluted 1:10, and 3 μl of each dilution was plated.


Live-cell fluorescence microscopy was performed using cells grown and mounted in SCD medium. Images were acquired using a Nikon TE2000-E microscope fitted with a Nikon ×100 Plan Apochromat oil objective (NA 1.4), Orca-ER CCD camera (Hamamatsu, Bridgewater, NJ) and Phylum Live Software, version 3.5.1 (Improvision, Lexington, MA) or a Zeiss Axioplan 2 microscope (Carl Zeiss, Oberkochen, Germany) equipped with an AxioCam CCD camera and AxioVision 3.1 software. GFP, Cy3 and Cy5 were visualized by using an X-cite 120 UV lamp and Chroma filter sets. Images were processed using Adobe Photoshop (Adobe Systems, San Jose CA). The fluorescence in situ hybridization assay (FISH) was performed as described previously ( Cole et al., 2002).

Electron microscopy

Electron microscopy was performed as previously described ( Scarcelli et al., 2007). In brief, the cells were grown to an OD600 of 0.5-1.0 in YPD medium, pelleted and resuspended in 0.1 M cacodylate buffer, pH 6.8. Primary fixation was performed with 3% glutaraldehyde, 1% paraformaldehyde and 0.1% tannic acid in 0.1 M cacodylate buffer, pH 6.8, at room temperature for 1 hour and then overnight at 4°C. Cells were washed twice with 0.1 M cacodylate buffer, pH 6.8, and twice with 0.1 M phosphate buffer, pH 7.5, and treated with zymolyase 100T (10 mg/ml) to produce spheroplasts. After washing with phosphate buffer, pH 7.5, and cacodylate buffer, pH 6.8, the cells were retreated with 3% glutaraldehyde, 1% paraformaldehyde and 0.1% tannic acid in 0.1 M cacodylate buffer, pH 6.8, for 1 hour at room temperature, washed three times with 0.1 M cacodylate buffer, pH 6.8, and embedded in low melting temperature agarose (SeaPrep; FMC). Post fixation was performed with 2% osmium tetroxide in 0.1 M cacodylate buffer, pH 6.8, for 1 hour on ice. Subsequently, the cells were washed in cacodylate buffer, pH 6.8, and deionized water, en bloc stained with 0.5% uranyl acetate overnight, dehydrated with ethanol, and embedded in Spurr's resin (medium grade). Thin sections were cut on an ultramicrotome (MT5000; Sorvall) with a section thickness of 100 nm. Sections were poststained with uranyl acetate and Venable and Coggleshell's lead citrate and examined on a transmission electron microscope (JEM 1010; JEOL) at 100 kV.

High-copy apq12 Δ suppressor screen

A Yep13 2 μ LEU2 library ( Rose and Broach, 1991) was utilized to identify genes that could suppress apq12 Δ cold-sensitivity at 16°C. The library was transformed into apq12 Δ cells, and cells were plated on leucine dropout medium and incubated at 16°C for 14 days. Candidate suppressor colonies were retested for growth at 16°C. apq12 Δ cells carrying an empty LEU2 vector were included as a negative control.

Drug treatment

All compounds were purchased from Sigma and dissolved in dimethyl sulfoxide (DMSO) prior to addition to media. Terbinafine from Neil Ryder (Novartis Research Institute, Vienna, Austria) was used in some experiments. With the exception of oleic acid and palmitic acid, the appropriate compound was added to YPD medium following autoclaving. Oleic acid and palmitic acid were added before autoclaving along with 0.5% Tween-40 to facilitate dissolving. Compounds were used at the following concentrations: 0.2% benzyl alcohol, 5 mM oleic acid, 5 μg/ml terbinafine, 1 μg/ml ketoconazole, 2.5 μg/ml cerulenin. For growth analysis in the presence of these agents, strains were diluted tenfold four times across rows of a 96-well plate, and 4 μl samples from the original and diluted samples were spotted onto YPD plates supplemented with the appropriate agent. Plates were grown for 5 days at 23°C, 30°C and 37°C and photographed after 3 and 5 days or at 16°C for 14 days and photographed after 7 and 14 days.

Lipid analysis

Total lipids were analyzed following [3H]palmitic acid labeling of cells. Cells were cultivated in YPD at 24°C, diluted in fresh medium and incubated at 20°C, 24°C or 37°C for 1 hour before addition of 10 μCi/ml of [9,10-3H]palmitic acid (10 mCi/ml; American Radiolabeled Chemicals, St Louis, MO). After 6 hours of labeling, cells were collected, lipids were extracted with chloroform and methanol (1:1, v/v), lipids were dried under a stream of nitrogen and aliquots of the lipid extract containing equal counts were analyzed by thin layer chromatography (TLC) using silica gel 60 plates (Merck, Darmstadt, Germany). Phospholipids were separated using the solvent system chloroform, methanol and potassium-chloride (0.25%) (55:45:5, v/v/v). Neutral lipids were separated using petroleum ether, diethyl ether and glacial acetic acid (70:30:2, v/v/v). Plates were then analyzed by radio-TLC scanning (Tracemaster 20, Berthold Technologies, Bad Wildbad, Germany) and exposure to a phosphorimaging screen (Bio-Rad).

For sterol analysis, cells were cultivated in SC medium overnight at 24°C and processed essentially as described ( Quail and Kelly, 1996) using cholesterol as internal standard. Free sterols and base-hydrolyzed total sterols were derivatized by TMS [N′O′-bis-(trimethylsilyl)-trifluoroacetamide] and analyzed by GC-MS on a Voyager Trace 2000 Series GC-MS (Thermo Fisher Scientific, Waltham, MA) equipped with a Zebron ZB-35 capillary column (35% phenyl-methyl polysiloxane; dimensions: 30 m × 0.25 mm × 0.25 μm film thickness). Sterols were identified based on their mass fragmentation pattern and by comparison to commercially available standards. Staining cells with NBD-cholesterol, sterol uptake and esterification were performed as previously described ( Reiner et al., 2006).


  • Supplementary material available online at

  • * These authors contributed equally to this work

  • Present address: Department of Biology, Hampden-Sydney College, Hampden-Sydney, VA 23943, USA

  • § Present address: Tetragenetics, Inc., One Broadway, 14th Floor, Cambridge, MA 02142, USA

  • We thank Catherine Heath for excellent technical assistance, Christine Guthrie and Anne de Bruyn Kops for a brr6-1 mutant strain, and Susan Wente, Charlie Barlowe and T. Y. Chang for valuable discussions and advice on the manuscript. This work was supported by grants from the National Institute of General Medical Sciences, National Institutes or Health (GM33998) to C.N.C. and from the Swiss National Science Foundation (3100-120650) to R.S. M.J.W. and J.J.S. were supported by a training grant from the National Institute of Arthritis and Musculoskeletal and Skin Disorders, National Institutes of Health. M.J.W. also received support from an undergraduate science education grant to Dartmouth College from the Howard Hughes Medical Institute. Deposited in PMC for release after 12 months.

  • Accepted October 20, 2009.


View Abstract