Vps Factors Are Required for Efficient Transcription Elongation in Budding Yeast
Naseem A. Gaur, Jiri Hasek, Donna Garvey Brickner, Hongfang Qiu, Fan Zhang, Chi-Ming Wong, Ivana Malcova, Pavla Vasicova, Jason H. Brickner, Alan G. Hinnebusch

Abstract

There is increasing evidence that certain Vacuolar protein sorting (Vps) proteins, factors that mediate vesicular protein trafficking, have additional roles in regulating transcription factors at the endosome. We found that yeast mutants lacking the phosphatidylinositol 3-phosphate [PI(3)P] kinase Vps34 or its associated protein kinase Vps15 display multiple phenotypes indicating impaired transcription elongation. These phenotypes include reduced mRNA production from long or G+C-rich coding sequences (CDS) without affecting the associated GAL1 promoter activity, and a reduced rate of RNA polymerase II (Pol II) progression through lacZ CDS in vivo. Consistent with reported genetic interactions with mutations affecting the histone acetyltransferase complex NuA4, vps15Δ and vps34Δ mutations reduce NuA4 occupancy in certain transcribed CDS. vps15Δ and vps34Δ mutants also exhibit impaired localization of the induced GAL1 gene to the nuclear periphery. We found unexpectedly that, similar to known transcription elongation factors, these and several other Vps factors can be cross-linked to the CDS of genes induced by Gcn4 or Gal4 in a manner dependent on transcriptional induction and stimulated by Cdk7/Kin28-dependent phosphorylation of the Pol II C-terminal domain (CTD). We also observed colocalization of a fraction of Vps15-GFP and Vps34-GFP with nuclear pores at nucleus–vacuole (NV) junctions in live cells. These findings suggest that Vps factors enhance the efficiency of transcription elongation in a manner involving their physical proximity to nuclear pores and transcribed chromatin.

NEWLY synthesized proteins that are transported from the Golgi to the lysosome/vacuole traverse the endosome, as do ubiquitinated proteins that are removed from the plasma membrane by endocytosis en route to the vacuole for degradation. Ubiquitinated cargo proteins progress through early and late endosomes, are concentrated at the outer membranes of multivesicular bodies (MVB), and are then sequestered in intralumenal vesicles (ILVs) of the MVB. Fusion of the MVB with the vacuole delivers cargo proteins to the vacuole lumen for degradation by vacuolar hydrolases. “Class C and D” Vacuolar protein sorting (Vps) factors participate in vesicle fusion at the endosome, while cargo sorting and delivery to the ILVs at the MVB involves class E Vps proteins, including the components of the soluble ESCRT (endosomal sorting complex required for transport) complexes ESCRT-0, -I, -II, and -III (Bowers and Stevens 2005; Hurley and Emr 2006; Raiborg and Stenmark 2009).

It is thought that ESCRT-0 is recruited from the cytoplasm to the endosomal outer membrane by interaction with the phosphoinositide PI(3)P, where it acts to recruit and concentrate ubiquitinated cargo proteins and transfer them to the ESCRT-I complex. ESCRT-I activates the ESCRT-II heterotrimer that, in turn, recruits the ESCRT-III components, which are believed to assemble filaments instrumental in invagination of the MVB membrane. The AAA-ATPase Vps4, recruited by ESCRT-III subunits, functions to pinch off the membrane invaginations to produce ILVs containing cargo proteins and to recycle the ESCRT factors back to the cytoplasm (Raiborg and Stenmark 2009).

There is increasing evidence that certain Vps proteins have additional functions in cytoplasmic signaling pathways that regulate transcription in the nucleus. In budding yeast, ESCRT-III factor Snf7/Vps32 and the subunits of ESCRT-II were first identified genetically by their requirements for robust accumulation of SUC2 mRNA (Tu et al. 1993; Yeghiayan et al. 1995; Kamura et al. 2001). The transcription factor Rim101 is proteolytically activated on recruitment to the MVB outer membrane via ESCRT-III factor Snf7/Vps32 to permit expression of pH-responsive genes (Boysen and Mitchell 2006). The Gα subunit (Gpa1) of a heterotrimeric G protein activates the PI 3-kinase Vps34 (a class D Vps factor) at the endosomal membrane to promote the transcriptional response to mating factors (Slessareva et al. 2006). Activation of genes for utilization of alternative nitrogen sources by Gln3 is enhanced by Vps factors, and it appears that Gln3 must traffic in vesicles containing Vps10 between Golgi and endosome for subsequent nuclear entry (Puria et al. 2008). Recently, evidence was presented that the phosphoinositide PI(3,5)P2, produced at the late endosome promotes assembly of a transcriptional cofactor complex that enhances galactose induction of GAL gene transcription in the nucleus (Han and Emr 2011).

We found previously that robust activation of amino acid biosynthetic genes by yeast transcription factor Gcn4 requires a subset of Vps factors that function at the MVB. The defects in activation of Gcn4 target genes were most pronounced in mutants lacking certain Vps C or D factors, with lesser but still significant defects observed in vps mutants lacking particular ESCRT proteins, including ESCRT-II factors (Snf8/Vps22, Vps25, or Vps36) and ESCRT-III subunits Snf7/Vps32 and Vps20. Gcn4 synthesis is induced at the translational level in response to starvation for any single amino acid (Hinnebusch 2005). In the vps mutants, Gcn4 synthesis was induced properly and Gcn4 could enter the nucleus and bind to upstream activation sequences (UAS), but did not efficiently stimulate preinitiation complex (PIC) assembly at the promoter (Zhang et al. 2008). We hypothesized that a signal transduction pathway operates to dampen the transcriptional response to amino acid starvation by Gcn4 in response to endosome dysfunction (Zhang et al. 2008). More recently, evidence was provided that sterol limitation also down-regulates Gcn4 function in the nucleus, in a manner involving sterol binding protein Kes1 and its ability to inhibit PI(4)P-directed vesicular protein trafficking with an attendant increase in cellular sphingolipids. It appears that elevated sphingolipids provoke a reduction in Gcn4 function in a manner involving the CDK8 module of the transcriptional coactivator Mediator complex (Mousley et al. 2012).

In our previous study, some of the strongest defects in transcriptional activation by Gcn4 were observed in mutants lacking Vps34 and Vps15. Vps34 is the sole kinase in budding yeast that synthesizes PI(3)P in endomembranes, and Vps15 is a protein kinase associated with Vps34 required for its function (Bowers and Stevens 2005). Galactose induction of a GAL1-lacZ reporter was impaired in vps15Δ and vps34Δ mutants, but not as dramatically as seen for Gcn4-dependent reporters, suggesting that the function of Gal4, and possibly other transcriptional activators besides Gcn4, is also down-regulated to a lesser extent in response to endosome dysfunction. However, as shown below, we found subsequently that PIC assembly and transcription initiation occurs normally at the GAL1 promoter in vps15Δ and vps34Δ mutants. Furthermore, vps15Δ and vps34Δ mutations evoked much stronger reductions in expression of lacZ reporters driven by Gcn4 compared to mRNA transcripts of authentic Gcn4 target genes (Zhang et al. 2008). These last observations led us to suspect that the greatly reduced expression of lacZ reporters observed in vps mutants involved a defect in transcription elongation in addition to the defective PIC assembly observed specficially for Gcn4 target genes. Indeed, it is well established by Aguilera and colleagues that efficient elongation through lacZ coding CDS requires a full complement of transcription elongation factors in yeast cells (Chavez et al. 2001; Morillo-Huesca et al. 2006)—a fact which has been exploited to identify novel factors involved in transcription elongation (Tous et al. 2011).

Accordingly, we set out to investigate whether Vps15, Vps34, and various other soluble Vps factors are required for efficient transcription elongation in yeast cells. The results presented below support this possibility, including evidence that the rate of elongation by Pol II through lacZ coding sequences is reduced in vps15Δ and vps34Δ cells. Our findings further suggest that reduced cotranscriptional recruitment of the histone acetyltransferase complex NuA4 to CDSs could be one factor underlying the elongation defect in these mutants. Interestingly, we also obtained evidence that elimination of Vps15 or Vps34 impairs localization of the GAL1 and INO1 genes to the nuclear periphery during transcriptional activation. Unexpectedly, using chromatin immunoprecipitation (ChIP) analysis, we detected cross-linking of these and several other Vps proteins to the CDSs of various genes in vivo in a manner dependent on transcriptional activation. Consistent with this, we detected dynamic association of a fraction of GFP-tagged Vps15 and Vps34 with nuclear pores in live cells under various culture conditions. Together, our results indicate that inactivation of various Vps proteins reduces the efficiency of transcription elongation in vivo and raise the possibility that at least a subset of these proteins might stimulate transcription elongation in a manner involving their physical association with, or proximity to, transcribed chromatin at the nuclear periphery.

Materials and Methods

Yeast strain and plasmid constructions

All strains and plasmids used in this study are listed in Tables 1 and 2, respectively. Wild-type (WT) strain BY4741 and deletion derivatives thereof were described previously (Giaever et al. 2002) and purchased from Research Genetics. The presence of the reported deletion alleles was confirmed by PCR amplification of genomic DNA (Swanson et al. 2003). Strains harboring myc13 epitope tags were generated as described previously (Swanson et al. 2003) and verified by PCR analysis of chromosomal DNA and Western blot analysis with anti-myc antibodies. Strains harboring trp1Δ::hisG were generated using TRP1 knock-out construct pNK1009 as described previously (Alani et al. 1987).

View this table:
Table 1 Yeast strains used in this study
View this table:
Table 2 Plasmids used in this study

Strains CRY1605 and CRY1606 were generated by transformation of strains CRY1541 and CRY1581 with plasmid pNSP1-RFP. Strains HQY1584 and HQY1586 were constructed by transforming strains CRY1541 and CRY1581, respectively, with NurI-digested pHQ2061, harboring the SPT4-mCherry::hphMX4 cassette, selecting for growth on YPD containing hygromycin. Replacement of SPT4 with SPT4-mCherry::hphMX4 was confirmed by PCR analysis of chromosomal DNA using primers 972 and 1167 (Table 3), and by Western blot analysis using antibodies against mCherry.

View this table:
Table 3 Primers for strain construction or verification, or preparation of probes for Northern analysis

NSP1 was tagged at its C terminus with mTagBFP in strains HQY1584 and HQY1587 to yield strains CRY1718 and CRY1719, respectively, using a newly constructed tagging plasmid (I. Malcova, unpublished results). The latter was produced by inserting a XhoI–BamHI fragment containing the gene encoding yeast enhanced mTagBFP (ytBFP, a kind gift of Mads Kaern’s laboratory at the University of Ottawa) into pFA6a-natNT2. The ytBFP::natNT2 cassette was PCR amplified from the resulting tagging plasmid with primers Nsp1mbfp5 and Nsp1C2 using Phusion DNA polymerase (NEB) and used to transform HQY1584 and HQY1587 by selecting for growth on YPD medium containing nourseothricin (100 µg/ml, cloNat, Werner Bioagents, Jena). Correct integration of the cassette was verified by PCR analysis using primers Nsp1diaC and mTAGBFPrev, and the size of the protein fusion was checked by Western blot analysis using anti-tRFP antibody (Evrogen).

Plasmid pHQ2061 was constructed by PCR amplification from genomic DNA of a fragment containing the WT SPT4 CDS flanked by HindIII/NruI and BamHI sites using primers 1853/1854 and inserted between the HindIII and BamHI sites of pBS35 (fusing SPT4 CDS in-frame to mCherry CDS); subsequently a fragment containing the SPT4 3′-UTR flanked by SacI–NruI and SpeI sites was PCR amplified from genomic DNA using primers 1855/1856 and inserted between the SacI and SpeI sites of pBS35 to produce pHQ2061. Plasmid pNG18 was constructed by PCR amplification of an ∼0.97-kb URA3 fragment from genomic DNA of WT strain BY4741 by using primers N381/N382. The PCR-amplified URA3 fragment was digested with BamHI/ClaI and cloned into BamHI/ClaI-digested pCM184LAUR.

Strains H1486 (wild type), NGY11 (vps34Δ), and NGY12 (vps15Δ) were transformed with GFP-Lac repressor plasmid pAFS144 digested with NheI (Straight et al. 1996). The resulting strains were transformed with either p6LacO128-GAL1 digested with NruI (Brickner et al. 2007) or p6LacO128-INO1 digested with StuI (Brickner and Walter 2004), giving rise to the following six strains: DBY375 (GAL1:LacO), DBY376 (vps34Δ GAL1:LacO), DBY377 (vps15Δ GAL1:LacO), DBY452 (INO1:LacO), DBY455 (vps34Δ INO1:LacO), and DBY457 (vps15Δ INO1:LacO).

Gene length-dependent accumulation of mRNA assays and Northern analysis

Measurement of Pho5 enzymatic activity for gene length-dependent accumulation of mRNA (GLAM) ratio determinations was carried out as described previously using transformants of the appropriate strains harboring plasmids YCplac33 (empty URA3 vector), pSCH202 (PGAL1-PHO5, URA3) and either pSCH209 (PGAL1-PHO5-LAC4, URA3), or pSCH212 (PGAL1-PHO5-lacZ, URA3) (Morillo-Huesca et al. 2006).

Isolation of total RNA from yeast and Northern blot analysis were carried out as described previously (Ginsburg et al. 2009). DNA probes used were a 0.9-kb EcoRV-digested PHO5 internal fragment isolated from plasmid pSCH202, or the following PCR fragments amplified from the genomic DNA of BY4741 using primers listed in Table 3: 0.45 kb of YAT1 CDS, 0.57 kb of LYS2 CDS, 0.4 kb of IMD2 CDS, and 0.9 kb of the SCR1 gene. Each DNA fragment was radiolabeled with [α32P]-dCTP by using the Megaprime DNA labeling system (Amersham).

Analysis of GAL1 positioning in yeast nuclei

Chromatin localization experiments were performed as described (Brickner et al. 2010) using strains expressing GFP fused to Lac repressor and the GAL1 gene marked with an array of Lac repressor binding sites. Cells were stained with antibodies against GFP and Nsp1.

ChIP analysis

Yeast cell cultures (100 ml) at A600 of 0.5 to 0.6 were mixed with 11 ml of formaldehyde solution [50 mM HEPES-KOH (pH 7.5), 1 mM EDTA, 100 mM NaCl, and 11% formaldehyde] and cross-linked for 20 min at room temperature with intermittent shaking and then quenched with 15 ml 2.5 M glycine. Cells were collected by centrifugation and washed twice with 100 ml ice-cold Tris-buffered saline. The cells were broken by vortexing with glass beads in 500 μl of lysis buffer [50 mM HEPES-KOH (pH 7.5), 1 mM EDTA, 140 mM NaCl, 1% Triton X-100, 0.1% sodium deoxycholate, and protease inhibitors]. Glass beads were removed from the lysates and washed with 500 μl of lysis buffer, and the resulting 1-ml lysates were sonicated to yield DNA fragments of 300–500 bp. Supernatants containing soluble chromatin were obtained by centrifugation at 13,000 × g and stored at −80°. Fifty microliters of chromatin were used for immunoprecipitations, and an identical aliquot was reserved as the “input” sample. Chromatin was immunoprecipitated using Dynabeads Pan Mouse IgG (Invitrogen) coupled with antibodies described below for 2 hr at 4°, recovered immune complexes were washed once with phosphate-buffered saline (PBS) containing BSA (5 mg/ml), twice each with lysis buffer, wash-buffer I [50 mM HEPES-KOH (pH 7.5), 1 mM EDTA, 500 mM NaCl, 1% Triton X-100, and 0.1% Na-deoxycholate] and wash-buffer II [10 mM Tris-HCl (pH 8.5), 250 mM LiCl, 1 mM EDTA, 0.5% NP-40, and 0.5% sodium deoxycholate], and once with TE [10 mM Tris-HCl (pH 8.0) and 1 mM EDTA]. The immunoprecipitated complexes were eluted at 65° for 15 min with 100 μl elution buffer [50 mM Tris-HCl (pH 8.0), 10 mM EDTA and 1% SDS] and for 10 min with 150 μl of elution wash buffer [50 mM Tris-HCl (pH 8.0), 1 mM EDTA, and 0.67% SDS], and the eluates were combined (IP sample). The matched input and IP samples were incubated overnight at 65° to reverse the cross-links. The samples were then treated with proteinase K (Ambion), at 100 μg/250 μl of chromatin, for 2 hr, and DNA was extracted twice with phenol:chloroform:isoamyl alcohol (25:24:1) and once with chloroform:isoamyl alcohol (24:1), ethanol precipitated, and resuspended in 30 μl TE containing RNase (10 μg/ml). Two microliters of resuspended DNA from IP and input samples were used for each PCR reaction in the presence of [33P]-dATP with the appropriate primers (listed in Table 4). The radiolabeled amplified fragments were resolved by PAGE and quantified with a phosphorimager. For each primer set employed, we optimized the PCR conditions to ensure that the amounts of amplified 33P-labeled products being generated are proportional to the amounts of input DNA over the range of DNA concentrations present in the IP or input samples. For the ChIP analysis of Snf7-myc and Snf8-myc in Figure 7, G–I, and of Rpb3 in Figure 3A, POLI CDSs were amplified as a negative control in addition to the specific ARG1 or GAL1 sequences of interest. An intergenic sequence from chromosome V (ChrV-1) was amplified as a negative control for all other ChIP experiments using primers 948/949, with the exception of the Rpb3 ChIP in Figure 4 where primers ExtChrV-1/ExtChrV-2 (ChrV-2) were used instead. Two immunoprecipitations were conducted on at least two chromatin samples isolated from independent cultures and the PCR analysis of IP and input DNA samples was carried out in duplicate or triplicate. Ratios of the amounts of PCR fragments for specific to control DNA sequences generated from IP samples were normalized to the corresponding ratios for input samples to yield occupancy values, and mean occupancies were calculated from replicate experiments.

View this table:
Table 4 Primers for ChIP assays

Antibodies

Unless stated otherwise, 1 μl of the following antibodies was used for each ChIP assay: Mouse monoclonal anti-Rpb3 (Neoclone), mouse monoclonal anti-myc (Roche), and mouse anti-Ser5P-Rpb1 (H14 from Covance). For Western blot analysis of myc13-tagged proteins in whole cell extracts prepared under denaturing conditions (Reid and Schatz 1982), the anti-myc antibody was used at the dilution recommended by the vendor.

Live-cell imaging by fluorescence microscopy

Distributions of fusion proteins in living cells were analyzed with an oil immersion 100×/1.4 objective using the Olympus Cell R detection and analyzing system based on the motorized Olympus IX-81 inverted microscope, Hammamatsu Orca/ER digital camera and the following highly specific mirror units: (i) EGFP filter block U-MGFPHQ, excitation (exc) max 488 nm, emission (em) max 507 nm; (ii) RFP filter block U-MFRFPHQ, exc max 558 nm, em max 583 nm; and (iii) BFP filter block U-MFBFPHQ, exc max 390 nm, em max 460 nm). The Cell R system enables us to obtain several optical sections through the cell. Images were processed, merged, and analyzed using Olympus Cell R, Imaris, National Institutes of Health ImageJ and Adobe CS5 software. Images from selected optical layers were presented.

Results

Elimination of Vps15 or Vps34 confers sensitivity to 6-azauracil and mycophenolic acid and dampens IMD2 induction by these inhibitors

The strong reductions in lacZ reporter expression observed in vps15Δ and vps34Δ mutants (Zhang et al. 2008) led us to suspect that the efficiency of transcription elongation was impaired in cells lacking these Vps proteins. To gain additional evidence for a transcription elongation defect, we examined sensitivity of vps15Δ and vps34Δ cells to the drugs 6-azauracil (6-AU) and mycophenolic acid (MPA). These drugs lower GTP pools by inhibiting IMP dehydrogenase (IMPDH) encoded by IMD3 and IMD4, which catalyzes the rate-limiting step in de novo guanine nucleotide biosynthesis. Numerous mutations affecting transcription elongation factors confer sensitivity to both 6-AU and MPA (Exinger and Lacroute 1992; Jenks and Reines 2005 and references therein). Interestingly, vps15Δ and vps34Δ cells display sensitivity to these drugs comparable in degree to that exhibited by mutant cells lacking the elongation factor Spt4 (Hartzog et al. 1998), and these 6-AU-sensitive (6-AUS) and MPAS phenotypes were complemented by the corresponding WT VPS alleles on plasmids (Figure 1A). Significant, albeit reduced, sensitivity to MPA and 6-AU was also evident in the pep7Δ/vps19Δ, pep12Δ/vps6Δ, and vps45Δ mutants, and sensitivity to an even smaller degree was evident in the vps4Δ and snf7Δ/vps32 strains as well (Figure 1, B and C). Like vps15Δ and vps34Δ, these other five vps mutants are also defective for vesicular protein transport from the Golgi to the vacuole (Bowers and Stevens 2005). Vps10 is an integral membrane protein that functions as the receptor for carboxypeptidase Y to mediate its trafficking from the late Golgi to late endosome (Bowers and Stevens 2005); and it was recently implicated in promoting nuclear entry of Gln3 (Puria et al. 2008); however, eliminating the VPS10 gene conferred no sensitivity to 6-AU or MPA (Figure 1C).

Figure 1

Vps factors are required for WT resistance to 6-AU and MPA and for robust induction of IMD2 transcription by these elongation inhibitors. (A) Serial 10-fold dilutions of transformants of the indicated genotypes harboring an empty URA3 vector (YCplac33) or URA3 plasmids containing VPS15 (pVPS15; row 4) or VPS34 (pVPS34; row 6) spotted on SC −Ura or SC −Ura containing 15 µM MPA or 100 μg/ml 6-AU and incubated for 3 days (SC) or 5 days (MPA and 6-AU) at 30 °C. Transformants of WT (BY4741), spt4Δ (6986), vps15Δ (3236), and vps34Δ (5149) strains were analyzed. (B and C) Conducted as in A using transformants of WT (BY4741), spt4Δ (6986), vps15Δ (3236), vps34Δ (5149), snf7Δ (1580), vps4Δ (5588), pep7Δ (3682), pep12Δ (1812), vps10Δ (3043), and vps45Δ (4462) strains except that cells were incubated for 6 days on plates containing 75 μg/ml 6-AU. (D and E) Northern analysis of IMD2 and SCR1 transcripts in strains of the indicated genotypes cultured to mid-exponential phase in SC −Ura at 30 °C and incubated with 6-AU at 100 μg/ml for 2 hr. Total RNA was extracted and subjected to blot-hybridization analysis with probes for IMD2 and SCR1 transcripts. A representative blot is shown in D, in which successive lanes for each strain contain amounts of the same RNA samples differing by a factor of 2 (indicated by the ramps), and the mean ratios of IMD2 to SCR1 intensities quantified by phosphorimaging analysis of multiple experiments are plotted in E with the SEM shown as error bars. Empty vector transformants of WT (BY4741, lanes 1 and 2), spt4Δ (6986, lanes 3 and 4), thp1Δ (1764, lanes 5 and 6), and vps34Δ (5149, lanes 7 and 8) strains, a pVPS34 transformant of strain 5149 (lanes 9 and 10), and a pVPS34N736K transformant of strain 5149 (lanes 11 and 12) were analyzed. (F and G) Same as D and E except, using 15 µM MPA instead of 6-AU.

Sensitivity to 6-AU and MPA does not necessarily indicate a transcription elongation defect; however, mutations in bona fide elongation factors have been found to reduce transcriptional induction of IMD2, encoding an MPA-resistant form of IMPDH (McPhillips et al. 2004), in response to 6-AU or MPA treatment (Riles et al. 2004). It is significant, therefore, that the vps34Δ mutation reduces IMD2 mRNA abundance in the presence of 6-AU or MPA to an extent similar to that given by mutations that eliminate Spt4 or the Thp1 subunit of the elongation/mRNA export complex TREX-2 (Figure 1, D–G). Furthermore, complementation of the IMD2 expression defect in vps34Δ cells by episomal VPS34 was impaired by a mutation (N736K) that inactivates its PI 3-kinase activity (Slessareva et al. 2006) (Figure 1, D–G). These findings are consistent with the possibility that inactivation of the PI 3-kinase activity of Vps34 compromises the efficiency of transcription elongation in vivo.

Elimination of Vps15 or Vps34 impairs expression of lacZ reporters

We determined that vps34Δ and vps15Δ cells display another prominent phenotype of yeast elongation mutants, of inefficient elongation through the GC-rich or long CDS of bacterial lacZ and fungal LAC4 (Chavez et al. 2001; Morillo-Huesca et al. 2006). First, we employed an in vivo reporter described by Aguilera and colleagues in which a lacZ-URA3 translational fusion is expressed from the Ptet promoter in ura3 auxotrophic strains. Mutations in known elongation factors impair expression of the URA3 portion of the reporter and confer poor growth on (−Ura) medium lacking uracil (Jimeno et al. 2002). To control for possible effects of mutations on Ptet promoter activity, we constructed a matched Ptet-URA3 reporter missing the lacZ CDS and compared growth of vps mutants harboring the Ptet-lacZ-URA3 vs. Ptet-URA3 reporters on −Ura medium. In the otherwise wild-type ura3 strain, the Ptet-lacZ-URA3 transformants grew more slowly and exhibited a lower plating efficiency, on SC −Ura compared to the corresponding Ptet-URA3 transformants [Figure 2A, WT, lacZ vs. control (C)], indicating reduced expression of URA3 CDS from the Ptet-lacZ-URA3 reporter. The reduction in plating efficiency of Ptet-lacZ-URA3 vs. Ptet-URA3 transformants was clearly exacerbated in the vps15Δ, vps34Δ, pep7Δ, and pep12Δ mutants, and possibly also in the snf7Δ strain, but not in the vps4Δ mutant (Figure 2A, SC −Ura, cf. lacZ and C transformants). These results, combined with the sensitivity of the vps mutants to 6-AU and MPA shown above, suggest that eliminating particular Vps proteins reduces the efficiency of transcription elongation, with relatively stronger defects conferred by the absence of Vps15 or Vps34 compared to other Vps factors, e.g., Vps4, that are equally critical for vesicular protein transport to the vacuole.

Figure 2

Vps factors are required for efficient transcription elongation through long and GC-rich coding sequences. (A) Serial 10-fold dilutions of transformants of the indicated genotypes harboring pCM184-LAUR (lacZ) containing Ptet-lacZ-URA3 or control plasmid pNG18 (C) containing Ptet-URA3 spotted on SC −Leu or SC −Leu −Ura and incubated for 5 days at 30 °C. The Ptet-lacZ-URA3 and Ptet-URA3 constructs are depicted schematically at the top. Transformants of WT (NGY1), snf7Δ (NGY3), pep7Δ (NGY4), pep12Δ (NGY5), vps15Δ (NGY6), vps34Δ (NGY7), and vps4Δ (FZY810) strains were analyzed. (B) Ratios of Pho5 enzymatic specific activity in transformants of the indicated genotypes harboring plasmids containing PGAL1::PHO5-LAC4 (pSCH209) or PGAL1-PHO5 (pSCH202) (both shown schematically above), i.e., GLAM ratios, were measured. Strains also contained either empty LEU2 vector YCplac111 (lanes 1–3 and 5) or plasmids containing VPS15 (pNG9, lane 4) or VPS34 (pNG11, lane 6). The appropriate transformants of WT (BY4741), spt4Δ (6986), vps15Δ (3236), or vps34Δ (5149) strains were grown to mid-exponential phase in SC −Ura −Leu with 2% galactose as carbon source at 30 °C before measuring Pho5 activity in whole cell extracts (WCE). (C and D) Northern analysis of transcripts from PGAL1-PHO5 (top two panels) and PGAL1-PHO5-LAC4 (bottom two panels) in strains of the indicated genotype grown to mid-exponential phase in SC −Ura −Leu containing 2% galactose as carbon source at 30 °C. Total RNA was extracted and subjected to blot-hybridization analysis with probes for the PHO5 CDS or SCR1. A representative blot is shown in C and the mean ratios of PHO5-LAC4 (long) to PHO5 (short) mRNA levels, each normalized to SCR1, was quantified by phosphorimaging analysis of multiple experiments and plotted in D with the SEM shown as error bars. Empty vector transformants of WT (BY4741), spt4Δ (6986), thp2Δ (2861), vps15Δ (3236), and vps34Δ (5149) strains (lanes 1–4 and 7), a pNG9 transformant of strain 3236 (lane 5), a pNG10 transformant of strain 3236 (lane 6), a pNG11 transformant of strain 5149 (lane 8), and a pNG12 transformant of strain 5149 (lane 9) were analyzed. (E and F) Same as in C and D except that PGAL1-PHO5-lacZ (pSCH212) was analyzed instead of PGAL1-PHO5-LAC4.

To provide additional evidence for elongation defects in these vps mutants, we employed the GLAM assay developed by Morillo-Huesca et al. (2006) involving a PGAL1-PHO5-LAC4 construct harboring LAC4 CDS inserted into the 3′-UTR and a matched PGAL1-PHO5 reporter with the same promoter but without LAC4 CDS. The LAC4 sequences add ∼3 kb to the transcript length but are not translated, such that both reporter transcripts produce the same Pho5 protein. A variety of mutants lacking known elongation factors express reduced steady-state amounts of the long reporter mRNA and, hence, decreased GLAM ratios of Pho5 enzyme activity produced from the long vs. short construct (Morillo-Huesca et al. 2006). In agreement with previous results for these reporters, the spt4Δ mutant exhibits a GLAM ratio that is only ∼20% of the WT value (Figure 2B). By comparison, the GLAM ratios in the vps15Δ and vps34Δ mutants were 50 and 40% of WT, and these defects were largely (vps15Δ) or completely (vps34Δ) complemented by the cognate WT alleles (Figure 2B).

We also conducted Northern analysis to quantify the GLAM ratios, and comparing mRNA expression from the PGAL1-PHO5-LAC4 and PGAL-PHO5 constructs revealed that deletions eliminating Spt4 or the Thp2 subunit of the Transcriptional defect of Hpr1 by Overexpression (THO) elongation complex (Chavez et al. 2000) confer strong reductions in the long/short mRNA ratio (Figure 2, C and D), as expected from previous results (Morillo-Huesca et al. 2006). Northern analysis of the vps15Δ and vps34Δ mutants confirmed the occurrence of ∼60% reductions in the ratio of long to short mRNAs produced by these two reporters and further revealed that complementation of these defects by the cognate plasmid-borne alleles was abolished by mutations that impair the protein kinase activity of Vps15 or PI 3-kinase activity of Vps34 (Slessareva et al. 2006) (Figure 2, C and D). Similar results were obtained using a PGAL1-PHO5-lacZ reporter that contains lacZ instead of LAC4 CDS in the 3′-UTR (Morillo-Huesca et al. 2006) (Figure 2, E and F). These findings support the conclusion that the efficiency of elongation is reduced in vps15Δ and vps34Δ cells, albeit not to the same extent observed in the known elongation mutants spt4Δ and thp2Δ.

The analyses of mRNA expression from the PGAL1-PHO5 constructs in the vps15Δ and vps34Δ mutants shown above suggested that these mutations do not significantly affect PGAL1 promoter activity (Figure 2, C and E, PGAL1-PHO5 blots), such that the reductions in expression of the long PGAL1-PHO5-LAC4 and PGAL1-PHO5-lacZ reporters in these mutants likely involve defects in transcription elongation. To confirm that these vps mutations do not affect PIC assembly at the PGAL1 promoter, we conducted chromatin immunoprecipitation (ChIP) analysis of the Pol II subunit Rpb3 at the native GAL1 promoter. As expected, elimination of the transcriptional activator Gal4 in the gal4Δ mutant reduces Rpb3 occupancy at the GAL1 TATA box and CDS under galactose induction to the low, background levels observed in WT cells under noninducing conditions (Figure 3A). By contrast, the vps15Δ and vps34Δ mutants exhibit essentially WT levels of Rpb3 occupancy in the promoter region on galactose induction, confirming that PIC assembly is unaffected by elimination of the Vps15 and Vps34 proteins. A moderate reduction in Rpb3 occupancy in the CDS was observed in the vps15Δ strain, which would be consistent with a transcription elongation defect.

Figure 3

Vps factors do not affect PIC assembly at GAL1 but are required for efficient transcription of several PGAL1-driven CDS. (A) Occupancies of Rpb3 at the TATA region in the promoter and 3′ORF of chromosomal GAL1 were measured by ChIP analysis of strains of the indicated genotype cultured in SC medium containing 2% raffinose and either induced by adding galactose to 2% for 30 min (IN) or kept untreated (UI). Cross-linked chromatin was immunoprecipitated with anti-Rpb3 antibodies and DNA from immunoprecipitated (IP) and input samples was subjected to PCR (using primers listed in Table 4) in the presence of [33P]-dATP to amplify radiolabeled fragments of the GAL1 TATA region, 3′ORF , or POL1 CDS (Table 4) analyzed as a control, and the PCR amplicons were quantified by phosphorimaging analysis. Ratios of GAL1 TATA and 3′ORF to POL1 CDS signals for IP samples were normalized to the corresponding ratios for input samples to yield occupancy values. Immunoprecipitations were conducted in duplicate on at least two chromatin samples prepared from two replicate cultures and the PCR analysis of precipitated DNA sequences was conducted in duplicate for each immunoprecipitated sample. Error bars represent the standard error of the mean (SEM). WT (BY4741), gal4Δ (1044), vps15Δ (3236), and vps34Δ (5149) strains were analyzed. (B) Northern analysis of transcripts expressed from PGAL1-YAT1 (shown schematically and contained on pSCH247) in strains of the indicated genotypes grown at 30 °C to mid-exponential phase in SC −Ura −Leu containing 2% raffinose and induced with 2% galactose for 2 hr. Total RNA was extracted and subjected to blot-hybridization analysis with probes to the YAT1 CDS or SCR1. A representative blot is shown on the left and the mean ratios of YAT1 mRNA to SCR1 RNA levels were quantified by phosphorimaging analysis of multiple experiments and plotted here with the SEM shown as error bars. Empty vector transformants of WT (BY4741), spt4Δ (6986), thp2Δ (2861), vps15Δ (3236), and vps34Δ (5149) strains (lanes 1–4 and 6), a pNG9 transformant of strain 3236 (lane 5), a pNG11 transformant of strain 5149 (lane 7), and a pNG12 transformant of strain 5149 (lane 8), all harboring pSCH247, were analyzed. (C) Northern analysis of transcripts from PGAL1-LYS2 in strains of the indicated genotype conducted as in B except that the strains harbored pSCH227 vs. pSCH247, a pNG10 transformant of strain 3236 (lane 6) was analyzed, and probes of LYS2 CDS and SCR1 were employed. (D) GLAM ratios were measured as in Figure 2B using the appropriate transformants of WT (BY4741), spt4Δ (6986), vps15Δ (3236), vps34Δ (5149), snf7Δ (1580), snf8Δ (2826), vps4Δ (5588), pep7Δ (3682), and vps10Δ (3043) strains, harboring pSCH202 or pSCH209, and normalized to the GLAM ratio of the WT strain.

It has been shown that the long CDS of the native gene LYS2 and the GC-rich CDS of YAT1 exhibit enhanced requirements for elongation factors for efficient mRNA production (Chavez et al. 2001). To determine whether the vps15Δ and vps34Δ mutations also reduce the efficiency of transcription elongation through these native CDS, we examined mRNA expression from plasmid-borne PGAL1-YAT1 and PGAL1-LYS2 constructs harboring the YAT1 or LYS2 CDS. On induction with galactose, these constructs produce transcript levels far in excess of the endogenous YAT1 or LYS2 transcripts, and in a manner strongly dependent on the Hpr1 subunit of the THO elongation complex (Chavez et al. 2001) and Spt4 (Rondon et al. 2003a). As shown by the Northern analyses in Figure 3, B and C, the vps15Δ and vps34Δ mutations reduce mRNA production from the PGAL1-YAT1 and PGAL1-LYS2 constructs by 40–60%, comparable to that given by the spt4Δ and thp2Δ mutations for the PGAL1-LYS2 construct, but much less than that observed for the PGAL1-YAT1 construct in spt4Δ cells.

As shown above, deletions of various VPS genes besides VPS15 and VPS34 conferred sensitivity to 6-AU and MPA (Figure 1, A–C) and also appeared to reduce transcription elongation through the Ptet-lacZ-URA3 reporter (Figure 2A), albeit to lesser extents than observed in vps15Δ and vps34Δ cells. Hence, we characterized additional vps mutants in the GLAM assay using the PGAL1-PHO5-LAC4 and PGAL1-PHO5 constructs described above. With the exception of the vps10Δ and snf7Δ deletions, all of the VPS deletions we tested reduce the GLAM ratio significantly, but to different extents. Among the defective mutants, the snf8Δ/vps22Δ strain exhibits the smallest effect, reducing the ratio by only 21%, whereas vps4Δ and pep7Δ decrease the ratio by ∼40%. Consistent with its lack of 6-AU and MPA sensitivity (Figure 1C), the vps10Δ mutant displays a WT GLAM ratio (Figure 3D). Although the vps15Δ and vps34Δ mutants display the strongest reductions in the GLAM ratio among the vps mutants tested, a defect in elongation is exhibited to different extents by other vps mutants using LAC4 (Figure 3D) and lacZ reporters (Figure 2A). However, because Snf7, Snf8, and Vps4 are required for vesicular protein trafficking from the MVB to vacuole, it appears that disruption of this process per se is not sufficient to confer the relatively stronger elongation defects displayed by vps15Δ and vps34Δ mutants.

Vps15 and Vps34 stimulate the rate of Pol II elongation through lacZ coding sequences in vivo

To gain further insight into the nature of the elongation defect displayed by the vps15Δ and vps34Δ mutants, we explored whether eliminating Vps15 or Vps34 reduces the rate of transcription elongation in vivo. To this end, we analyzed the kinetics of Pol II elongation through the lacZ coding sequences of the PGAL1-PHO5-lacZ reporter. On glucose addition to cells growing with galactose, Pol II recruitment to the GAL1 promoter is blocked and preexisting elongating Pol II molecules finish transcribing the CDS. The kinetics of Pol II run-off during this last wave of elongation can be determined by ChIP analysis of Rpb3 occupancy at various times after adding glucose, providing a measure of the elongation rate in vivo (Mason and Struhl 2005).

As expected, after addition of glucose to WT cells, Pol II vacated the promoter and 5′ end of the PGAL1-PHO5-lacZ CDS more rapidly than from the 3′ end of the CDS. Thus, after 2 min, there was a large decline in Rpb3 occupancy at the promoter, progressively smaller reductions at locations 1.52 kb and 2.87 kb downstream from the promoter, and no reduction 4.06 kb from the promoter, whereas by 3 to 4 min, most of the Pol II had cleared the entire lacZ CDS (Figure 4B). Interestingly, the rate of Pol II run-off was significantly lower in the vps15Δ and vps34Δ mutants. This point is appreciated by noting that the decreases in Rpb3 occupancies at the 2.87-kb and 4.06-kb locations between 2 and 3 min in glucose in the mutant cells (Figure 4, C and D) were considerably smaller than the reductions observed between 2 and 3 min at the corresponding locations in WT cells (Figure 4B). Even after 5 min, the Pol II run-off was incomplete in both vps mutants (Figure 4, C and D). These findings suggest that elimination of Vps15 or Vps34 evokes a reduced rate of elongation by Pol II through the GC-rich lacZ CDS. The fact that in vps15Δ cells the Rpb3 occupancies were actually higher at the 3′ end of the CDS after 2 min in glucose than in galactose medium (Figure 4C) might indicate that the elongation defect in this mutant is exacerbated by the switch from galactose to glucose, leading to a modest build-up of Pol II toward the 3′ end of the CDS. Finally, ChIP analysis of Rpb3 occupancies across the lacZ CDS under steady-state inducing conditions (exponential growth in galactose medium), revealed no reductions in Pol II occupancy at the 3′ end relative to the 5′ end of the CDS in the vps15Δ and vps34Δ mutants compared to the Pol II occupancies seen at the corresponding locations in WT cells (data not shown). Thus, elimination of these Vps factors does not seem to provoke dissociation of Pol II from the template DNA despite a reduction in the elongation rate during transcription of the lacZ CDS.

Figure 4

vps15Δ and vps34Δ mutations reduce the transcription elongation rate through lacZ CDS in vivo. (A–D) ChIP analysis of Pol II (Rpb3) occupancies across the PGAL1-PHO5-lacZ construct. pSCH212 transformants of WT (BY4741) (B), vps15Δ (3236) (C), or vps34Δ (5149) (D) strains were grown in SC medium containing 2% raffinose at 30 °C, induced by adding galactose to 2% for 1 hr, and repressed by adding glucose to 4% for the indicated times (minutes) before cross-linking. ChIP was performed with anti-Rpb3 antibodies and PCR amplification in the presence of [α33P]-dATP using the primers shown schematically in A, and listed in Table 4, located in the PGAL1 promoter or in lacZ CDS at the indicated distances downstream of the transcription start site, and the control ChrV-2 primers (that amplify an intergenic region on chromosome V). Rpb3 occupancies were calculated as described in Figure 3A, and the occupancies in glucose (2–5 min) were normalized to those in galactose (0 min) and set to unity for time zero. Mean relative occupancies and SEM (error bars) were calculated from immunoprecipitations conducted in duplicate on at least two chromatin samples prepared from two replicate cultures and the PCR analysis of precipitated DNA sequences was conducted in duplicate for each immunoprecipitated sample.

Cotranscriptional recruitment of NuA4 to CDS in vivo

A genome-wide study of genes directly involved in histone acetylation or deacetylation revealed an unexpected enrichment of genetic interactions with genes involved in endosome/vacuole functions. In particular, vps15 and vps34 mutations were found to be synthetically lethal with a mutation in Epl1, a subunit of the histone acetyltransferase (HAT) complex NuA4, and a snf8 mutation was synthetically lethal with mutations in the Yng2 and Esa1 subunits of NuA4 (Lin et al. 2008). We recently presented evidence that NuA4 is cotranscriptionally recruited to CDS and that a conditional mutation affecting Esa1, the HAT catalytic subunit of NuA4, reduces the rate of Pol II elongation in vivo (Ginsburg et al. 2009). These findings led us to consider whether eliminating the Vps15 or Vps34 proteins from cells would reduce NuA4 recruitment to CDS as one aspect of the impairment of transcription elongation in vps15Δ and vps34Δ cells. To this end, we conducted ChIP analysis of a functional Myc-tagged version of the NuA4 subunit Epl1 at the GAL1 gene. In agreement with previous results (Ginsburg et al. 2009), induction of GAL1 with galactose leads to increased Myc-Epl1 occupancy of both the 5′ and 3′ ends of the GAL1 CDS [Figure 5B, cf. WT uninduced (UI) vs. WT induced (IN)]. Interestingly, the Myc-Epl1 occupancy under inducing conditions is reduced at both locations in the GAL1 CDS in vps34Δ cells to nearly the same levels seen in WT cells under noninducing conditions. In addition, the vps15Δ mutation reduces the galactose induction of Myc-Epl1 at the 3′ end of the GAL1 CDS. Highly similar results were observed in ChIP analysis of Myc-Yng2, another subunit of NuA4 (Figure 5C). As shown above in Figure 3A, Rpb3 occupancy in the 3′ end of the GAL1 CDS was reduced somewhat in these strains. However, the small reduction in Pol II occupancy clearly cannot account for the strong decrease in NuA4 occupancy seen in vps34Δ cells, and the reduction for NuA4 exceeds that of Pol II even in vps15Δ cells. Thus, we conclude that cotranscriptional recruitment of NuA4 to GAL1 CDS is greatly compromised in cells lacking Vps34 and is also significantly impaired in the absence of Vps15.

Figure 5

vps15Δ and vps34Δ mutations reduce cotranscriptional recruitment of NuA4. (A–C) ChIP analysis of myc-tagged NuA4 subunits Epl1 (B) and Yng2 (C) at GAL1 CDS using primers depicted in A. In B, WT (NGY117), vps15Δ (NGY102), and vps34Δ (NGY103) strains, harboring EPL1-myc13, were grown in SC medium containing 2% raffinose at 30 °C (UI), and induced with 2% galactose (IN) for 30 min before cross-linking. ChIP analysis was performed as described in Figure 3A except using anti-myc antibodies and the GAL1 CDS primers depicted in A and control ChrV-1 primers (that amplify an intergenic region on chromosome V) listed in Table 4. In C, ChIP was performed as in B except using the following YNG2-myc13 strains: WT (NGY118), vps15Δ (NGY119), and vps34Δ (NGY120). Immunoprecipitations were conducted in duplicate on at least two chromatin samples prepared from two replicate cultures and the PCR analysis of precipitated DNA sequences was conducted in duplicate for each immunoprecipitated sample. Error bars represent the standard error of the mean (SEM).

Localization of activated GAL1 and INO1 at the nuclear periphery requires Vps15 and Vps34

The GAL1 gene, when transcriptionally active, localizes at the nuclear periphery in association with the nuclear pore complex (Casolari et al. 2004; Cabal et al. 2006; Brickner et al. 2007). Having found that Vps15 and Vps34 are required for efficient transcription elongation through various CDS driven by the GAL1 promoter, we next asked whether these proteins are also required for targeting of active GAL1 to the nuclear periphery. An array of Lac repressor binding sites was integrated downstream of the GAL1 gene in wild type, vps15Δ and vps34Δ strains expressing a GFP fusion to Lac repressor (Straight et al. 1996; Brickner et al. 2007). Cells were shifted to galactose medium for 2 hr and the fraction of cells in the population in which GAL1 colocalized with the nuclear periphery (marked with the nuclear pore protein Nsp1; Figure 6A) was scored (Brickner et al. 2010). A random distribution results in ∼25–30% colocalization (Brickner and Walter 2004). In agreement with previous findings (Brickner et al. 2007), in the WT strain, GAL1 localized at the nuclear periphery in 35 ± 2% of cells cultured in glucose and 65 ± 5% of cells grown with galactose. In the vps15Δ and vps34Δ mutant, we observed a strong defect in the targeting of GAL1 to the nuclear periphery in galactose (32 ± 2% and 40 ± 4%, respectively) (Figure 6B). The INO1 gene also is targeted to the nuclear periphery in WT cells cultured without inositol, which derepresses INO1 transcription (Brickner and Walter 2004). Using a set of strains with Lac repressor binding sites integrated at INO1, we found that both vps15Δ and vps34Δ evoke marked reductions in INO1 localization under derepressing conditions (Figure 6C). Thus, Vps15 and Vps34 are required for normal targeting of the activated GAL1 and INO1 genes to the nuclear periphery.

Figure 6

Localization of activated GAL1 at the nuclear periphery requires Vps15 and Vps34. Strains DBY375 (WT), DBY376 (vps34Δ), and DBY377 (vps15Δ), containing an array of Lac repressor binding sites integrated at GAL1 and expressing a GFP fusion to Lac repressor, were grown overnight in glucose-containing medium and then shifted to galactose medium for 2 hr before fixing in methanol and processing for immunofluorescence using antibodies against GFP-Lac repressor and Nsp1. (A) Representative cells stained for GFP (green) and Nsp1 (red) that were scored as nucleoplasmic or peripheral. (B) Average peripheral localization of GAL1 from three biological replicates. For each replicate, ≥30 cells were counted. The mean percentage of cells in which the Lac repressor dot colocalized with the nuclear envelope (Nsp1) is displayed ±SEM. The dynamic range of this assay is from 20 to 80%, and the hatched line represents the distribution of the URA3 gene with respect to the nuclear envelope. (C) Strains DBY452 (WT), DBY455 (vps34Δ), and DBY457 (vps15Δ), harboring an array of Lac repressor binding sites integrated at INO1 and expressing GFP-tagged Lac repressor were cultured overnight in SC lacking inositol either containing or lacking 100 µM myo-inositol and analyzed as in B.

Vps15 and Vps34 are cotranscriptionally cross-linked to the CDS of various yeast genes

Having found that vps15Δ and vps34Δ mutations reduce the efficiency of transcription elongation, cotranscriptional recruitment of NuA4, and targeting of several activated genes to the nuclear periphery, we wondered whether Vps15, Vps34, and other Vps factors could be physically associated with sites of transcription in yeast cells. This possibility was stimulated by reports (discussed below) that Vps factors from other species can be found in the nucleus in association with chromatin, or copurified with transcription elongation factors. To test this possibility in yeast, we first conducted ChIP analysis of functional myc-tagged versions of Vps15, Vps34, Snf7, and Snf8 at the Gcn4 target genes ARG1 and ARG4 in response to induction of Gcn4 by treatment with sulfometuron (SM), which inhibits biosynthesis of isoleucine and valine (Ilv). Our previous work has shown that Ilv starvation with SM evokes a rapid increase in Gcn4 occupancy of UAS elements, followed by recruitment of coactivators Mediator and SwItch/Sucrose NonFermentable (SWI/SNF) to the UAS and the HAT complexes Spt-Ada-Gcn5-acetyltransferase complex (SAGA) and NuA4 to both the UAS and CDS of various Gcn4 target genes (Govind et al. 2005; Qiu et al. 2005; Govind et al. 2007; Ginsburg et al. 2009).

Interestingly, we found that Vps15, Vps34, Snf7, and Snf8 can all be detected in association with both ARG1 and ARG4 on transcriptional induction by Gcn4, and the occupancies of all four Vps factors is consistently higher in the 3′ end of the CDS vs. the UAS or promoter (TATA) regions (Figure 7, A–H, cf. IN vs. UN for WT strains). In the case of Vps15 and Vps34, we established that their increased occupancies at ARG1 and ARG4 on treatment with SM were completely dependent on Gcn4, being abrogated in isogenic gcn4Δ strains (Figure 7, C–F, cf. WT IN vs. gcn4Δ IN). We observed that Snf7, Vps15, and Vps34 are also associated with the GAL1 gene, strictly during its induction by Gal4 in galactose medium, and again exhibiting higher occupancies in the CDS vs. UAS or promoter regions (Figure 8, A–D, cf. Gal vs. Raf). Furthermore, Vps15 and Vps34 were detected in association with the ADH1 and PMA1 CDS, two genes transcribed constitutively in amino acid and glucose-containing medium (Figure 8, E–G).

Figure 7

Transcription-dependent recruitment of Vps factors at Gcn4 target genes ARG1 and ARG4. (A–D) ChIP analysis of myc-tagged Vps15 in VPS15-myc13 strains NGY26 (WT) and NGY28 (gcn4Δ) at ARG1 (A and C) and ARG4 (B and D), using the primers depicted schematically in A and B, respectively, and control primer ChrV-1 listed in Table 4. (E and F) ChIP analysis of Vps34-myc in VPS34-myc13 strains NGY27 (WT) and NGY29 (gcn4Δ) at ARG1 (E) and ARG4 (F) using the primers depicted in A and B and control primer ChrV-1 listed in Table 4. (G–I) ChIP analysis of Snf7-myc (G and I) at ARG1 in SNF7-myc13 strain FZY610 and vps4Δ SNF7-myc13 strain FZY642, and of Snf8-myc in SNF8-myc13 strain FZY644 and vps4Δ SNF8-myc13 strain FZY657 (listed in Table 1) using the primers depicted in A and the control primers POL1CDS listed in Table 4. For all experiments, cells were cultured in SC −Ilv and either treated with 0.5 µM sulfometuron methyl (SM) for 30 min to induce Gcn4 (IN) or left untreated (UI). Cross-linked chromatin was immunoprecipitated with anti-myc antibodies and analyzed as in Figure 3A. Immunoprecipitations were conducted in duplicate on at least two chromatin samples prepared from two replicate cultures and the PCR analysis of precipitated DNA sequences was conducted in duplicate for each immunoprecipitated sample. Error bars represent the standard error of the mean (SEM).

Figure 8

Recruitment of Vps factors to the CDS at GAL1, ADH1, and PMA1. (A–D) ChIP analysis of Vps15-myc in VPS15-myc13 strain NGY26 (B), Vps34-myc in VPS34-myc13 strain NGY27 (C), and Snf7-myc in SNF7-myc13 strain FZY610 (D) conducted for GAL1 using the primers depicted in A and control primer ChrV-1 for B and C, POL1 CDS for C listed in Table 4. Cells were cultured and ChIP analysis was conducted as described in Figure 3A. (F and G) ChIP analysis of Vps15-myc in NGY26 and Vps34-myc in NGY27 conducted for ADH1 and PMA1 using the primers depicted in E and ChrV-1 as control listed in Table 4. Cells were grown to mid-log phase in SC −Ilv and induced with 0.5 µm of SM for 30 min. ChIP analysis was performed as in Figure 3A. Immunoprecipitations were conducted in duplicate on at least two chromatin samples prepared from two replicate cultures and the PCR analysis of precipitated DNA sequences was conducted in duplicate for each immunoprecipitated sample. Error bars represent the standard error of the mean (SEM).

The fact that occupancies of Vps factors are consistently higher in CDS vs. UAS or promoter, suggests that their recruitment requires elongating Pol II and not merely activator binding at the UAS. Supporting this idea, deleting the TATA element at ARG1, which reduces Pol II in the CDS, likewise reduces Vps15 and Vps34 occupancies at ARG1 (Figure 9, A–D, cf. WT vs. TATAΔ). We showed previously that this arg1-TATAΔ mutation does not reduce the UAS occupancy of Gcn4 itself (Qiu et al. 2006).

Figure 9

Cotranscriptional recruitment of Vps15-myc and Vps34-myc is stimulated by the Pol II CTD kinase Cdk7/Kin28. (A–D) ChIP analysis at ARG1 conducted for Vps15-myc in WT (NGY26) and arg1-TATAΔ (NGY56) strains harboring VPS15-myc13 (B), and for Vps34-myc in WT (NGY27) and arg1-TATAΔ (NGY57) strains harboring VPS34-myc13 (C), and for Rpb3 in WT (NGY27) and arg1-TATAΔ (NGY57) strains (D), using primers depicted in A and B and control primer ChrV-1 listed in Table 4. Cells were grown to mid-log phase in SC −Ilv and induced with 0.5 µm of SM for 30 min. Occupancies of Vps34-myc (C) and Rbp3 (D) were measured in the same chromatin extracts. (E–H) ChIP analysis of Rpb1 phosphorylated on Ser5 of its CTD (Ser5P, E), Rpb3 (F) and Vps34-myc (G and H) was conducted on the same chromatin extracts from KIN28 VPS34-myc13 strain NGY69 (WT), kin28-as VPS34-myc13 strain NGY71 (kin28-as), and gcn4Δ VPS34-myc13 strain NGY29. Cells were grown in SC −Ilv and treated with 1-NA-PP1 [1- (1, 1- dimethylethyl)- 3- (1- naphthalenyl)- 1H- pyrazolo[3, 4- d]pyrimidin- 4- amine] at 12 μM for 42 min and SM 0.5 µm for the last 30 min of the NAPP1 treatment. ChIP analysis was conducted as in Figure 3A and values obtained for the KIN28 and kin28-as strains were normalized to those obtained for the gcn4Δ strain to produce the plotted Gcn4-dependent occupancies in E–G. H presents the ratios of occupancies in G and F. (I–K) ChIP analysis conducted as in E–G except using KIN28VPS15-myc13 strain NGY68, kin28-as VPS15-myc13 strain NGY70, and gcn4Δ VPS15-myc13 strain NGY28. For all ChIP experiments, immunoprecipitations were conducted in duplicate on at least two chromatin samples prepared from two replicate cultures and the PCR analysis of precipitated DNA sequences was conducted in duplicate for each immunoprecipitated sample. Error bars represent the standard error of the mean (SEM).

The recruitment of histone-modifying enzymes, elongation and termination factors, and mRNA processing factors that function during transcription elongation is stimulated by the heptad repeats (Tyr1Ser2Pro3Thr4Ser5Pro6Ser7) in the C-terminal domain of Pol II subunit Rpb1 (Pol II CTD), dependent on phosphorylation of the CTD repeats on Ser2, Ser5, or Ser7 by various cyclin-dependent kinases (Phatnani and Greenleaf 2006; Buratowski 2009). Cdk7 (Kin28 in yeast) is the enzyme responsible for the majority of Ser5 and Ser7 CTD phosphorylation in yeast cells (Tietjen et al. 2010; Bataille et al. 2012). As expected, we observed that the occupancy of Rpb1 harboring Ser5-phosphorylated CTD repeats (S5P) is higher at the promoter and 5′ end vs. the 3′ end of the CDS at ARG1, and inhibiting an analog-sensitive (as) version of Kin28 with an ATP analog reduced the occupancy of the S5P form of Rpb1 (Figure 9E), while having little effect on levels of total Pol II (Rpb3) at ARG1 (Figure 9F). Interestingly, the occupancy of Vps34 was reduced at the 3′ end of ARG1 on inhibition of Kin28-as (Figure 9G), leading to a reduced Vps34:Rpb3 ratio at this location (Figure 9H). Similar results were obtained for Vps15 on inhibition of kin28-as (Figure 9, I–K). This finding suggests that S5P or S7P enhances, directly or indirectly, the cotranscriptional association of these Vps factors at promoter-distal locations. We conclude that association of the Vps15 and Vps34 with CDS is coupled to transcription elongation and stimulated by CTD phosphorylation by Kin28.

Interestingly, we found that vps4Δ cells display reduced association of ESCRT-II factor Snf8 and ESCRT-III component Snf7 with the ARG1 CDS (Figure 7I). It is known that cells lacking the AAA-ATPase Vps4 fail to produce ILVs and to recycle ESCRT factors to the cytoplasm, causing accumulation of ESCRT-III component Snf7 at the MVB (Babst et al. 1998). The fact that association of Snf7 and Snf8 with the ARG1 CDS is diminished in vps4Δ cells is consistent with the idea that these ESCRT factors are partitioned between endosome- and nucleus-associated pools and that their impaired recycling from the MVB to the cytoplasm in vps4Δ cells decreases their association with chromatin.

We went on to examine association of Myc-tagged versions of several other Vps factors with ARG1 or GAL1 CDS and observed induction-dependent association of Vps4, Vps27, and Vps10 at ARG1 and Vps45, Vps10 and Pep7 at GAL1 at levels comparable to those described above for Vps15, Vps34, Snf7, and Snf8 (Figure 10, A–D). In parallel, we analyzed two different functional myc-tagged protein synthesis initiation factors, eIF4E/Cdc33 and eIF3a/Tif32, that we expected to be primarily, if not exclusively cytoplasmic. We observed twofold or less increases in association of Cdc33-myc and Tif32-myc occupancy on transcriptional induction at ARG1 (Figure 10C), and also at ARG4 and GAL1 (data not shown). Thus, although induction-dependent chromatin association of the translation initiation factors is not negligible, it is significantly lower in magnitude than that seen for the Vps factors.

Figure 10

Vps factors exhibit higher occupancies of transcribed ARG1 CDS than do translation factors eIF4E and eIF3a. (A and C) ChIP analysis of Vps27-myc and Vps4-myc in strains NGY98 and NGY99, respectively (A), and Vps34-myc, Vps10-myc, eIF4E/Cdc33-myc, and eIF3a/Tif32-myc in strains NGY27, NGY116, H3881, and H3879, respectively (C), all conducted for ARG1 as described in Figure 7. Except control primer for Cdc-33my and Tif32-myc was POL1CDS (B) ChIP analysis of Pep7-myc, Vps45-myc, and Vps10-myc in NGY100, NGY101, and NGY116, respectively, conducted for GAL1 as described in Figure 3A. Immunoprecipitations were conducted in duplicate on at least two chromatin samples prepared from two replicate cultures and the PCR analysis of precipitated DNA sequences was conducted in duplicate for each immunoprecipitated sample. Error bars represent the standard error of the mean (SEM).

Fractions of Vps15 and Vps34 colocalize with nucleoporin Nsp1 at edges of nucleus–vacuole junctions

To examine further the possibility that fractions of Vps15 and Vps34 are physically associated with transcribed genes in the nucleus, we visualized GFP fusions to these proteins in living cells that also contain RFP or BFP fusions to the nucleoporin Nsp1 (Aitchison and Rout 2012) or an mCherry fusion to transcription elongation factor Spt4. Consistent with previous results (Huh et al. 2003; Obara et al. 2006), both Vps15-GFP and Vps34-GFP were found primarily in cytoplasmic punctae reflecting their association with endosomes and the outer membrane of the vacuole. Nsp1-RFP was found in punctae localized at the nuclear rim in the manner expected for a nucleoporin, as observed previously (MacKinnon et al. 2009). Under conditions where the Vps proteins were found cross-linked to transcribed chromatin, including 30 min cultivation in SC −galactose medium, we did not observe any obvious accumulation of Vps34 or Vps15 inside the nucleus. However, a fraction of the Vps-GFP punctae colocalized with Nsp1-RFP at the edges of junctions formed between the nucleus and vacuole, from which Nsp1 is largely excluded. This pattern was observed in essentially all cells we examined in cultures prepared in SC with glucose, raffinose, or galactose as carbon source (Supporting Information, Figure S1 and Figure S2). It is known that NV junctions are mediated by interactions between vacuolar membrane protein Vac8 and outer nuclear-membrane protein Nvj1 and are devoid of nuclear pores (Pan et al. 2000). Our time-lapse microscopic observations suggest that Vps34-GFP or Vps15-GFP punctae at the vacuolar membrane are dynamic and transiently associate with nuclear pores (Nsp1-RFP) at the edges of NV junctions (Movies M1–6 in File S1, File S2, File S3, File S4, File S5, and FileS6). Interestingly, when cells were cultivated overnight on rich glucose (YPD) agar medium and then investigated under agarose strips cast in minimal galactose medium, >50% of them displayed obvious overlapping signals of Vps34-GFP or Vps15-GFP with Nsp1-RFP (Figure S3).

Spt4-mCherry, examined as a nuclear marker, showed relatively homogenous nuclear localization in the manner expected for a transcription elongation factor distributed throughout the nucleoplasm in exponentially growing cells (Huh et al. 2003). We found that in the cells shifted from YPD plates to minimal galactose medium before microscopy, Spt4-mCherry usually did not occupy the entire nuclear space defined by the Nsp1-BFP nuclear rim. In rare instances a portion of Vps15 appeared to be located inside the nucleus of these cells (Figure S4A), and analysis of optical sections revealed that the Vps15-GFP signal was surrounded by Spt4-mCherry material within the nuclear rim defined by Nsp1-BFP (see layer 2). Similarly, as shown in Figure S4B, the Vps15-GFP signal could be observed as an inclusion inside the Spt4-mCherry domain (layer 2). Together, our microscopic findings indicate that a fraction of Vps15 and Vps34 are in proximity to nuclear pores at the edges of NV junctions, and that rarely, the Vps15-GFP fusion protein can be found within the nucleus under certain culture conditions.

Discussion

We demonstrated previously that eliminating certain VPS genes whose products are involved in Golgi-to-vacuole vesicular protein trafficking reduced the ability of transcriptional activator Gcn4 to stimulate PIC assembly and transcription initiation in vivo. Based on the results of assaying a GAL1-lacZ reporter gene, it appeared that the function of Gal4 was also impaired by the vps15Δ and vps34Δ mutations, which reduced GAL1-lacZ expression by 60 and 80%, respectively (Zhang et al. 2008). We showed here, however, that Gal4 function in PIC assembly appears to be normal in vps15Δ and vps34Δ cells and presented several lines of evidence that the reductions in Gal4-dependent lacZ reporter expression observed in these mutants is engendered by a defect at the elongation stage of transcription. Indeed, Aguilera and colleagues have established that reduced mRNA production from lacZ reporters and certain native genes with long or GC-rich CDS are hallmarks of mutations in various elongation factors in yeast (Chavez et al. 2001; Rondon et al. 2003a, 2004; Morillo-Huesca et al. 2006; Tous et al. 2011). We consistently observed decreased mRNA expression from several such reporters, all driven by the GAL1 promoter, in vps15Δ and vps34Δ cells. Employing a ChIP analysis of the kinetics of Pol II elongation in vivo (Mason and Struhl 2005), modified to include lacZ CDS, we also obtained evidence that elimination of Vps15 or Vps34 reduces the rate of Pol II elongation through lacZ CDS in vivo.

We further identified defects in cotranscriptional recruitment of the HAT complex NuA4 to the GAL1 CDS in vps15Δ and vps34Δ cells. Considering that conditional inactivation of the HAT subunit of NuA4 (Esa1) also impairs the rate of Pol II elongation in vivo (Ginsburg et al. 2009), the reduced NuA4 occupancy in CDS we observed in vps15Δ and vps34Δ mutants could contribute to their transcription elongation defects. We additionally demonstrated that vps15Δ and vps34Δ cells are defective for localization of activated GAL1 and INO1 to the nuclear periphery (gene positioning), a process that requires a number of nuclear pore complex (NPC) proteins or associated factors and likely facilitates coordination of transcription and mRNP biogenesis with mRNA export at nuclear pores (Egecioglu and Brickner 2011). Indeed, the nucleoporin Nup2 is required for targeting of both GAL1 and INO1 to the nuclear periphery (Brickner et al. 2007), and the TREX-2/THSC (nuclear pore-associated complex composed of Sac3–Thp1–Cdc31) complex is located at nuclear pores and is required for gene positioning of GAL1 (Cabal et al. 2006) and efficient mRNA export (Fischer et al. 2002; Lei et al. 2003).

Our ability to demonstrate an elongation defect for particular genes with long or GC-rich CDS, such as lacZ, does not rule out a general requirement for Vps factors for efficient elongation, as transcription of most genes in yeast is unaffected in mutants lacking only a single elongation factor. The elongation process appears to be overdetermined (Mason and Struhl 2005), such that only a few specialized genes require a full complement of cofactors for efficient transcription elongation in vivo (Chavez et al. 2001). Whereas the elongation defect described here likely applies broadly to many genes, the impairment of transcription initiation we observed previously in vps mutants (Zhang et al. 2008) is restricted to a subset of activators that includes Gcn4, but not Gal4, and seems to involve a signaling pathway that responds to sterol limitation (Mousley et al. 2012). Hence, we now interpret the dramatic reductions in expression of Gcn4-dependent lacZ reporters we observed previously in vps15Δ and vps34Δ mutants (Zhang et al. 2008) to be the compound effect of reduced PIC assembly by Gcn4 and impaired transcription elongation through lacZ CDS.

Given their well-established functions in vesicular protein trafficking, we anticipated that Vps15 and Vps34 would promote transcription elongation by an indirect mechanism. Hence, we were surprised to obtain evidence by ChIP assays for physical association of these and several other functionally related Vps proteins with transcribed chromatin. The Vps factors consistently displayed higher occupancies of the CDS vs. UAS or promoter regions of the examined genes, consistent with a widespread role in transcription elongation. The CDS occupancies of the Vps proteins were lower than we observed previously for canonical elongation factors, including Spt4, Spt5, Bur2, and the Paf1C complex (Qiu et al. 2006, 2009), but they were comparable to the occupancies of subunits of SAGA (Govind et al. 2007), NuA4 (Ginsburg et al. 2009), and histone deacetylase complexes (Govind et al. 2010), and they were significantly higher than the twofold or less enrichment of two different protein synthesis factors we analyzed at Gcn4 and Gal4 target genes. Importantly, as observed for conventional elongation factors, the association of Vps proteins with ARG1 or GAL1 CDS was strongly dependent on target gene transcription, occurring at high levels only under conditions where the relevant transcriptional activators are induced (Gcn4) or functional (Gal4). In addition, the TATA promoter element is required for high-level Vps15 and Vps34 occupancies at ARG1 under inducing conditions, and Vps15 and Vps34 occupancies of ARG1 CDS were stimulated by the Pol II CTD kinase Cdk7/Kin28—a characteristic of numerous factors involved in cotranscriptional histone modifications, mRNA processing or nuclear export, and the elongation or termination phases of transcription (Phatnani and Greenleaf 2006; Govind et al. 2007; Pascual-Garcia et al. 2008; Ginsburg et al. 2009; Govind et al. 2010; Qiu et al. 2012).

The specific association of Vps factors with transcribed coding sequences raises the possibility that they function in association with the chromatin to promote transcription elongation rather than influencing the process indirectly at the endosome. In this view, certain Vps factors would be partitioned between cytoplasm and nucleus and have dual functions in protein trafficking and transcription in these two compartments. Further evidence for this partitioning is provided by our finding that association of Snf7/Vps32 and Snf8/Vps22 with ARG1 chromatin is diminished in vps4Δ cells, in which Snf7 accumulates on the MVB outer membrane owing to a defect in ILV formation and recycling of ESCRT-III factors back to the cytoplasm (Babst et al. 1998). By fluorescence microscopy of living cells, we also observed association of a fraction of Vps15-GFP and Vps34-GFP fusions with an RFP fusion to nucleoporin Nsp1 at the edges of NV junctions. This pattern of vacuole/nucleus interaction was observed frequently under various growth conditions. Our time-lapse microscopy revealed that association of Vps punctae with the edges of NV junctions is highly dynamic, such that Vps15/Vps34 colocalization with nuclear pores is transient and observed only in particular optical layers. These findings raise the possibility of a dynamic interaction of Vps15/Vps34 proteins with actively transcribed genes at nuclear pores, which could be consistent with their requirement for efficient gene localization to the nuclear periphery. It is unclear, however, whether an association of Vps15/Vps34 with nuclear pores that is restricted to the borders of NV junctions could account for their widespread cross-linking to transcribed genes and their apparently general effect on transcription elongation. It is worth noting that the Vps34 homolog in Caenorhabditis elegans is concentrated in a perinuclear location, although this cellular location has been connected so far only with its role in vesicle budding from the outer nuclear membrane directed toward the cell periphery (Roggo et al. 2002).

Despite our evidence for physical association of a fraction of Vps15 and Vps34 with transcribed chromatin and nuclear pores, we have been unable to observe a defect in Pol II elongation in whole cell extracts of vps34Δ cells using a DNA template containing two G-less cassettes flanking lacZ CDS (data not shown). This assay was employed previously to provide evidence for direct roles of various factors in transcription elongation, including Spt4 and subunits of the THO complex (Rondon et al. 2003a,b and 2004). Accordingly, the elongation defect we observed in vps15Δ and vps34Δ cells presumably involves an aspect of nuclear structure or nucleus–cytoplasm interaction that cannot be recapitulated in homogenized cell extracts. However, we cannot eliminate the possibility that the elongation defect we documented in vps15Δ and vps34Δ mutants represents an indirect consequence of a defect in vesicular protein trafficking in the cytoplasm that is incidental to the association of Vps15 and Vps34 with transcribed chromatin and nuclear pores.

Notwithstanding this last reservation, one interesting hypothesis would be that Vps15/Vps34 function at the nuclear periphery to promote transcription elongation of activated genes localized to nuclear pores. Vps34 generates PI(3)P at endosomal membranes, which recruits ESCRT-0 factor Vps27 via its FYVE (Fab1, YGL023, Vps27, and EEA1) domain (Raiborg and Stenmark 2009). To our knowledge, there is no prior evidence that Vps27 or any other known yeast FYVE protein (Pep7, Fab1, Pib1, and Pib2) functions in the nucleus. (Although the FYVE factor Fab1 regulates assembly of the Cti6/Cyc8/Tup1 transcriptional cofactor complex, it performs this function at endosomes (Han and Emr 2011). Interestingly, however, plant homeodomain (PHD) fingers can bind various phosphoinositide phosphates (PIPs), including PI(3)P (Gozani et al. 2003), and are found in many chromatin and transcription-related proteins, including the Yng2 subunit of NuA4 (http://smart.embl.de/). Given our finding that transcription-coupled recruitment of NuA4 to the GAL1 CDS is diminished in vps15Δ and vps34Δ cells, it could be proposed that PI(3)P produced by Vps15/Vps34 on the nuclear membrane, possibly at nuclear pores, stimulates recruitment of NuA4 and other PHD-containing transcription cofactors to enhance elongation by Pol II. A precedent for such a mechanism is provided by evidence indicating that mammalian Ing2, an histone deacetylase complex (HDAC) component, is recruited to chromatin by a nuclear pool of PI(5)P via its PHD finger, where it regulates p53 acetylation and apoptosis in response to DNA damage (Gozani et al. 2003). Indeed, there is increasing evidence for phosphoinositides in the inner nuclear periphery and for isoforms of phosphoinositide kinases being located inside the nucleus (Barlow et al. 2010).

If PI(3)P is produced by Vps15/Vps34 on the nuclear membrane in the manner just proposed, this might account for our ability to detect transcription-dependent chromatin association of the FYVE-containing proteins Vps27 and Pep7 by ChIP assays. Moreover, the similar ChIP results observed for Vps factors Snf7, Snf8, Vps4, Vps45, and possibly Vps10, might reflect their association, direct or indirect, with Vps27 or Pep7. However, it seems more difficult to account for our evidence that these other Vps proteins also contribute to efficient transcription elongation, albeit to a lesser extent than do Vps15 and Vps34. One possibility is that impairing vesicular protein trafficking in the cytoplasm by deletions of various Vps factors reduces the nucleus-associated pools of Vps15 and Vps34 and thereby impairs transcription elongation indirectly.

There are other indications of nuclear functions for certain Vps proteins, including the fact that mammalian ESCRT-II was first purified in association with Pol II elongation factor Eleven-Nineteen Lysine-rich Leukemia (ELL) (Shilatifard 1998), and a report that plant Vps34 colocalizes with sites of transcription in plant cell nuclei (Bunney et al. 2000). In addition, the mammalian homolog of the ESCRT-III-related protein Vps46/Did2, when overexpressed, was found to enter the nucleus and locally condense chromatin and to produce a gene-silencing phenotype in Xenopus embryos (Stauffer et al. 2001). Clearly more work is required to determine at the molecular level exactly how Vps factors promote transcription elongation in yeast cells and whether the mechanisms involved depend on their physical interaction with chromatin and nuclear pores.

Acknowledgments

We thank Sebastián Chávez, Andrés Aguilera, Henrik Dohlman, Steven Hahn, Kristine Willis, Roger Tsien, and the Yeast Resource Center at the University of Washington for gifts of plasmids, and Thomas Dever for useful suggestions. This work was supported in part by the Intramural Research Program of the National Institutes of Health. J.H. was supported by P305/12/0480 and RVO61388971, and D.G.B. and J.H.B. were supported by a W. M. Keck Young Scholars in Medical Research Award.

Footnotes

  • Communicating editor: K. M. Arndt

  • Received September 28, 2012.
  • Accepted January 9, 2013.

Literature Cited

View Abstract