The Nuclear Exosome and Adenylation Regulate Post-Transcriptional Tethering of Yeast GAL genes to the Nuclear Periphery

Abstract

Some activated yeast genes, including GAL genes, associate with and remain tethered to the nuclear periphery even after transcriptional shutoff. To identify factors that affect GAL gene-nuclear periphery association after transcriptional shutoff, we designed a plasmid-based visual screen. Although many factors affected tethering during transcription, only a few specifically affected post-transcriptional tethering. Two of these, Rrp6p and Lrp1p, are nuclear exosome components with known roles in retaining RNA near transcription sites (dot RNA). Further experiments showed that an exosome mutation coupled with transcriptional shutoff leads to a post-transcriptional increase in polyadenylated GAL1 3′ ends, which accompanies a loss of unadenylated (pA-) GAL1 RNA, a loss of post-transcriptional gene-periphery tethering as well as a decrease in dot RNA levels. This suggests that the exosome inhibits adenylation of some GAL1 transcripts, which results in the accumulation of pA-RNA adjacent to the GAL1 gene. We propose that this dot RNA, probably via RNP proteins, is an important component of the physical tether that links the GAL1 gene to the nuclear periphery.

INTRODUCTION

Nuclear mRNA undergoes multiple covalent and non-covalent processing events before export to the cytoplasm. Many of these steps occur co-transcriptionally, which include the addition of a 5′ 7-methylguanosine cap, a poly-A tail and in many cases the removal of introns. mRNA is also packaged into a nuclear mRNP (messenger ribonucleoprotein particle), which is also in part co-transcriptional. A surveillance system located close to transcription sites retains or degrades improperly processed or packaged post-transcriptional mRNAs in foci or “dots” near the site of transcription. Successful mRNPs then traverse the nucleoplasm and dock at a nuclear pore prior to cytoplasmic export (Vinciguerra and Stutz, 2004).

Recent studies in S. cerevisiae suggest that some regulated genes (e.g. GAL1, GAL2, GAL10, HSP104, INO1, HXK1) also dock at the nuclear periphery subsequent to activation (Taddei, 2007). This may serve to facilitate mRNA export (gene gating hypothesis; (Blobel, 1985) and may also reflect a functional coupling of transcription and export. Indeed, gene-pore tethering at the nuclear periphery (used in this paper interchangeably with gene-periphery tethering) apparently relies on active transcription. For example, specific cis-acting DNA sequences including the TATA box and UASg elements of GAL1 and the promoter of GAL2 are required to establish a gene-pore interaction (Dieppois et al., 2006; Schmid et al., 2006). Trans-acting transcriptional activators such as SAGA complex components Sus1p and Ada2p are also required (Brickner and Walter, 2004; Cabal et al., 2006).

However, other experiments suggest that the RNA, RNP and/or RNA processing are relevant to gene-pore tethering. RNA-binding proteins such as Mex67p, the Sac3-Thp1-Cdc31-Sus1 complex, and Mlp1p are important (Cabal et al., 2006; Casolari et al., 2005; Dieppois et al., 2006; Drubin et al., 2006). In addition, both GAL1 and HXK1 tethering require their 3′-UTRs (Abruzzi et al., 2006; Taddei et al., 2006). Although this may not reflect a direct effect of RNA or RNP, a coherent view suggests that the gene-pore tethering mechanism exploits nascent RNP-pore contacts as well as transcription factor- and chromatin-pore contacts.

Consistent with a role for transcription-independent tethers, our previous studies as well as those from the Brickner laboratory showed that GAL genes remain at the nuclear periphery well after transcription has been fully repressed by glucose addition (Abruzzi et al., 2006; Brickner et al., 2007). Within 5′ of glucose addition, GAL gene transcription is shutoff, and RNA PolII is no longer associated with DNA (Mason and Struhl, 2005). However, GAL genes remain tethered to the pore for much longer, at least 30′ (Abruzzi et al., 2006) or for generations (Brickner et al., 2007). Chromatin itself might also be relevant, as the histone variant H2A.X is reported to be required for post-transcriptional tethering of the INO1 locus to the pore (Brickner et al., 2007). In addition, a post-transcriptional pool of GAL mRNPs is retained near the site of transcription in a dot that could contribute to a post-transcriptional gene-pore tether (Abruzzi et al. 2006).

To learn more about gene-pore tethering mechanisms and specifically post-transcriptional tethering, we designed a two-plasmid system for quick and easy monitoring of gene-pore localization in a variety of genetic backgrounds. Both the GFP-bound LacO array and the GAL gene were on a single copy plasmid, which replaced the GFP-bound array of Lac or Tet operators adjacent to a GAL gene in the chromosome (Brickner and Walter 2004; Abruzzi et al, 2006; Cabal et al 2006; Dieppois et al, 2006; Schmid et al, 2006; Taddei et al, 2006). Gene activation caused the plasmid to localize to the nuclear periphery essentially indistinguishably from a chromosomal GAL locus.

We then conducted a small scale visual screen of the viable yeast knockout collection to identify factors important for tethering. Many genes (23) were identified that decrease tethering, but only three specifically affected post-transcriptional tethering. Of these, two encode the nuclear exosome components Rrp6p and Lrp1p. Further experiments suggest that the nuclear exosome promotes post-transcriptional gene-pore tethering by inhibiting polyadenylation. Repression of polyadenylation contributes to dot formation, and enhanced polyadenylation due to the lack of full exosome function destabilizes post-transcriptional dots as well as the gene-periphery tether. The experiments provide insight into the previously described role of the nuclear exosome in dot formation (Hilleren et al., 2001) as well as the mechanism of gene-pore tethering.

RESULTS

Two-plasmid based gene-nuclear periphery tethering assay

One plasmid contains 256 LacO repeats and a GAL-GFP-GALpA reporter gene (pVS2; pVS1 is identical but contains no GAL gene; Figure 1A). The reporter was previously characterized and contains a GFP open reading frame flanked by the GAL1 promoter and GAL1 3′UTR (Abruzzi et al., 2006; Dower et al., 2004). The other plasmid expresses both LacI-GFP and Nup49-GFP, to mark the first plasmid as well as the nuclear periphery with GFP (pKA67; Fig. 1A). We transformed wild-type yeast (BY4741) with the two plasmids (or pVS1 instead of pVS2) and examined LacI-GFP localization in glucose (GLU) or galactose (GAL)-grown cells (Fig. 1B). LacI-GFP was scored as internal or peripheral, i.e., away from the rim or touching/overlapping the Nup49-GFP signals.

Figure 1. A plasmid-borne GAL gene localizes to the nuclear periphery upon transcription activation.

A) Schematic diagram of the two-plasmid system used to monitor GAL gene localization. pVS2 contains 256 LacO repeats, the GAL-GFP-GALpA reporter gene, LEU2 selective marker and a centromere. pKA67 contains Nup49-GFP, LacI-GFP, a URA3 selectable marker and a centromere (see Materials and Methods for details). B) Cells containing either a LacO plasmid expressing GAL-GFP-GALpA (pVS2) or a control plasmid lacking the GAL reporter gene (pVS1) were grown in either glucose (GLU) or galactose (GAL). The intra-nuclear position of the plasmid was monitored using de-convolution microscopy and determined to be either internal (gray) or peripheral (adjacent to or overlapping the nuclear periphery; black). When transcription is activated (GAL), the GAL reporter-containing plasmid (pVS2) becomes associated with the nuclear periphery. C) Cells containing the GAL-reporter containing lacO plasmid (pVS2) were grown in galactose and then glucose was added to repress transcription. Cells were fixed for microscopy at various time points after transcription shutoff and the localization of the plasmid was determined to be either internal (gray) or peripheral (black). As previously observed for a chromosomal GAL reporter gene, the GAL-reporter containing plasmid persists at the nuclear periphery for at least 30′after transcription shutoff.

Results were very similar to those previously reported for a chromosomally integrated GAL-GFP-GALpA reporter gene (Abruzzi et al., 2006). The GAL-GFP-GALpA plasmid (pVS2) was almost equally scored (50:50, peripheral: internal) in glucose-grown cells, i.e., when the GAL gene is silent (Fig. 1B). However, galactose growth caused peripheral localization to increase to ~80%. The control vector, pVS1, showed no change between galactose and glucose (Fig. 1B).

Our previous experiments showed that a chromosomally integrated GAL-GFP-GALpA reporter remains associated with the nuclear periphery for ~30 min after transcriptional repression by glucose addition (Abruzzi et al., 2006). A very similar result was observed for the plasmid-borne GAL-GFP-GALpA (Fig 1C). 15′ after transcriptional repression by glucose addition, the plasmid distribution remained unchanged with ~80% peripheral tethering. Even after 30′, ~60% of the plasmids were localized to the nuclear periphery. Approximately 60′ after transcriptional shutoff, the GAL reporter apparently lost its gene-pore tether (Fig.1C), as previously observed for the chromosomal version of this reporter gene (Abruzzi et al., 2006)

Visual screen to identify genes important for GAL gene peripheral localization

To identify new genes involved in GAL gene-nuclear periphery tethering, we utilized a sub-library of strains from the yeast knock-out deletion collection (Open biosystems). It included genes involved in mRNA processing/degradation, transcription, chromatin modifications, galactose regulation and nuclear pore components (Chekanova et al., 2007; see Materials and Methods). We transformed this sub-library as well as the wild-type control strain with the LacO-reporter plasmid (pVS2) and the plasmid expressing LacI-GFP and Nup49-GFP (pKA67). Cells were analyzed microscopically after growth in glucose (GLU; repressed), galactose (GAL; active) and 15′ after transcriptional repression by glucose addition (SHUTOFF). The strategy was designed to identify two classes of factors involved in gene-pore tethering: those important for the establishment or maintenance of tethering during active transcription (GAL) and those required for the maintenance of gene-pore tethering after transcriptional shutoff. 46 different strains from the deletion sub-library were examined, and the results are summarized in Fig. 2A.

Figure 2. A visual screen for genes that affect gene-pore localization.

A visual screen was performed on a sub-library of the yeast deletion collection (Open Biosystems). A) Of the 46 deletion strains tested, approximately half of the mutants affected the localization of the GAL gene when transcription was active. Interestingly, only three mutants specifically affected the post-transcriptional tethering of the GAL-reporter gene to the nuclear periphery. B) Mutations in two nuclear exosome components/cofactors, Lrp1p and Rrp6p, are required for the post-transcriptional tethering of the GAL-reporter plasmid to the nuclear periphery. Localization of the LacO-GAL-reporter plasmid was examined in wild-type, lrp1Δ, and rrp6Δ cells grown in glucose (GLU; gray), galactose (GAL; black) or 15′ after transcription was repressed by the addition of glucose (SHUTOFF; striped). The percentage of plasmids that localize to the nuclear periphery was determined (see Materials and Methods). When transcription is active (galactose media), the localization of the GAL reporter plasmid in rrp6Δ and lrp1Δ cells is indistinguishable from that of wild-type (compare black bars). In contrast, when transcription is repressed, the GAL reporter plasmid is no longer retained at the nuclear periphery in rrp6Δ and lrp1Δ cells (compared striped bars).

Surprisingly, many strains showed a complete lack of plasmid localization to the nuclear periphery during active transcription (Fig. 2A and Table 1). These included mutants in a wide variety of processes, including transcription elongation, mRNA processing and chromatin modification. Interestingly and in contrast, only three strains had an effect on tethering maintenance after shutoff. They included the two non-essential nuclear exosome components, RRP6 and LRP1 (Fig. 2B). In rrp6Δ and lrp1Δ strains, the GAL plasmid localized normally to the pore when transcription was active (compare WT and mutant GAL conditions), but it was rapidly delocalized after glucose addition (Fig. 2B). Similar results were observed with a third exosome mutant, the temperature sensitive rrp4-1 strain and only at the non-permissive temperature (data not shown). The results indicate the importance of the nuclear exosome to post-transcriptional tethering.

Table 1. Genes important for gene-nuclear periphery tethering.

Strains which showed only a post-transcriptional effect are in bold

Complex/ORF name	Functional significance
THO complex (MFT1,HPR1,THO2,THP2)	Complex required for transcription elongation
PAF complex (HPR1,LEO1,PAF1,CCR4)	Same as above
CBP20	Small subunit of the hetero-dimeric cap binding complex
REF2	Component of large CPF complex
TEX1	Component of TREX complex
LSM12	Protein of unknown function
MIG1	Transcription factor involved in glucose repression
ESC1	Protein localized to the nuclear periphery and required for telomere silencing
ISW1	Member of ISW1 class of ATP-dependent chromatin remodeling complex.
ISW2	As above
SPT8	Component of SAGA
ADA2	As above
SUS1	As above
IOC2	Member of Isw1b complex
SPP1	Subunit of COMPASS histone methylase complex
SNF1	AMP-activated serine/threonine kinase
Nuclear exosome (RRP6, LRP1)	Component of nuclear exosome

RNA requirement for post-transcriptional gene-nuclear periphery tether

The importance of the nuclear exosome suggests that tethering may be RNA- or RNP-mediated, as previously proposed (Abruzzi et al., 2006; Chekanova et al., 2007). However, an experiment from the Stutz lab showed that removing the open reading frame (and therefore the majority of the RNA) from GAL2 did not eliminate nuclear peripheral localization upon transcriptional activation (Dieppois et al., 2006). To address the same question but in the context of post-transcriptional tethering, we performed a similar experiment in our system and compared localization between cells grown in steady-state galactose, steady-state glucose and 15′ after transcriptional repression by glucose addition.

As shown previously for GAL2, removal of the GAL1 ORF (GAL1 ORFLESS) did not affect peripheral tethering under galactose conditions. However, GAL1 ORFLESS was not stably maintained at the periphery after transcriptional shutoff, in contrast to the wild-type GAL1 gene (Fig. 3). This provides further evidence that post-transcriptional gene-tethering is RNA- or RNP-mediated.

Figure 3. GAL1 mRNA is important for post-transcriptional gene-pore tethering.

To determine whether the GAL1 mRNA is important for post-transcriptional tethering, we compared the localization the GAL1 gene (black) with a GAL1 gene lacking the ORF (GAL1 ORF less; gray; see Materials and methods.) Cells were isolated after growth in glucose (GLU; transcription off), galactose (GAL; transcription activated) or 15′ after transcriptional repression (SHUTOFF) and the intra-nuclear location of the GAL loci were determined. As shown previously for GAL2 (Dieppois et al., 2006), the removal of the GAL1 ORF does not affect the ability of this gene to localize to the nuclear periphery when transcription is active (GAL). However, when transcription is repressed (SHUTOFF), the GAL1 ORF is no longer associated with the nuclear periphery strongly suggesting that mRNA is important for the post-transcriptional tethering of the gene to the nuclear periphery.

The nuclear exosome is required for post-transcriptional dots

Our previous experiments showed that GAL-GFP-GALpA mRNAs accumulate in discrete foci that are adjacent to the transcription site even in wild-type strains. Additional experiments indicated that foci or “dots” correlate with gene tethering to the nuclear periphery (Abruzzi et al., 2006; Chekanova et al., 2007). Taken together with a previously established relationship between the nuclear exosome and heat shock dots (Hilleren et al., 2001), the experiments shown above suggested that the exosome modulates the gene-nuclear periphery tethering by affecting dot RNA. To assay the role of the nuclear exosome in the formation or stability of plasmid-derived nuclear RNA, we used fluorescent in-situ hybridization (FISH) with message-specific GFP probes to examine GFP mRNA localization in wild-type, rrp6Δ, and lrp1Δ cells grown in glucose, galactose or 15′ after transcriptional shutoff by glucose addition.

Consistent with previous assays of chromosome-derived RNA (Abruzzi et al., 2006), plasmid-derived GAL-GFP-GALpA mRNAs accumulated in dots independent of the nuclear exosome, i.e. the dots that form in wild-type, rrp6Δ, and lrp1Δ galactose-grown cells are virtually indistinguishable (Fig. 4A). However, transcriptional repression caused dot RNA to disappear rapidly only in rrp6Δ and lrp1Δ cells. This extends the correlation between dot RNA and gene-nuclear periphery tethering.

Figure 4. The nuclear exosome is required for dot persistence in the absence of transcription.

Message-specific fluorescence in-situ hybridization (FISH) was performed to examine whether GAL-GFP-GALpA mRNAs are retained adjacent to the site of transcription when the nuclear exosome is inactivated. A) Wild-type, rrp6Δ, and lrp1Δ cells were examined after growth in glucose, galactose or 15′ after transcriptional repression. GFP message-specific FISH shows that dots accumulate in all three strains when transcription is activated (GAL). 15′ after transcription is shutoff, the dots persist in wild-type cells, but are no longer detectable in both rrp6Δ and lrp1Δ cells (SHUTOFF). B) To rule out the possibility that the dot disappearance was an indirect affect, we examined the kinetics of dot persistence in a temperature sensitive nuclear exosome mutant, rrp4-1. rrp4-1 cells were grown in galactose at 25°C until they reached mid-log phase and shifted to the non-permissive temperature (37°C). At the same time, transcription was repressed by the addition of glucose. FISH was performed on cells fixed at various times after transcription shutoff and exosome inactivation, and the percentage of cells containing dots was determined. A rapid loss of dots upon exosome inactivation and transcription shutoff suggests that the effect is direct.

To help distinguish between direct and indirect effects of the exosome mutant strains, we also examined the kinetics of dot disappearance in rrp4-1, a conditional allele of a nuclear exosome component. rrp4-1 cells were shifted to the non-permissive temperature, and transcription was repressed simultaneously by glucose addition; time points were taken and the number of dot-containing cells was determined (Fig. 3B). Dots disappeared in approximately 60% of the cells after only 5 min, indicating that exosome effect on dot RNA occurs rapidly after exosome inactivation. This suggests that the nuclear exosome may act directly to generate or stabilize dot RNA.

A GAL-GFP-ribozyme reporter is not affected by nuclear exosome mutations

This paradoxical role of the nuclear exosome on dot RNA has been previously observed (Hilleren et al, 2001). Because some studies also indicate a relationship between the nuclear exosome and 3′-end formation (Burkard and Butler, 2000; Milligan et al., 2005; Torchet et al., 2002), we tested whether the loss of post-transcriptional dots and gene-nuclear periphery tethering is related to 3′ cleavage and polyadenylation. To this end, we assayed the exosome mutant effects on a very similar GAL reporter gene - except that the usual GAL1 3′-UTR was replaced by a self-cleaving hammerhead ribozyme (GAL-GFP-RZ). This reporter generates mRNA that is not polyadenylated in vivo (Dower et al., 2004). We compared the gene localization and dot persistence of GAL-GFP-RZ with that of the normal GAL-GFP-GALpA control in both wild-type and rrp6Δ strains (Fig. 5).

Figure 5. A GAL-GFP-RZ reporter that cannot be cleaved and polyadenylated is unaffected by mutations in the nuclear exosome.

When the nuclear exosome is inactivated, the GAL-GFP-GALpA gene exhibits neither post-transcriptional tethering to the nuclear periphery nor post-transcriptional dot persistence. To test whether cleavage and/or polyadenylation is required for these effects, we repeated the experiments using a reporter, GAL-GFP-RZ that cannot be polyadenylated. This reporter is identical to GAL-GFP-GALpA except that the GAL1 3′UTR is replaced by a self-cleaving hammerhead ribozyme. A) The localization of GAL-GFP-GALpA and GAL-GFP-RZ in both wild-type and rrp6Δ cells was examined over time after transcription repression. As shown previously, both GAL-GFP-GALpA and GAL-GFP-RZ persist at the nuclear periphery after transcription is repressed (blue and red;_Abruzzi et al., 2006). When the exosome is inactivated, the GAL-GFP-GALpA gene is no longer retained at the nuclear periphery after transcription shutoff (light blue). In contrast, the GAL-GFP-RZ reporter gene remains at the nuclear periphery for at least 60′ in the absence of transcription (pink). B) Message specific FISH was performed on cells expressing either GAL-GFP-GALpA or GAL-GFP-RZ in wild-type or rrp6Δ cells. Cells were fixed after growth in galactose (GAL) or various times after being shifted into glucose to repress transcription. As reported earlier, GAL-GFP-GALpA and GAL-GFP-RZ dots persist in the absence of transcription (dark blue and red; Abruzzi et al., 2006) When the exosome is inactivated by the deletion of RRP6, GAL-GFP-GALpA dots disappear rapidly while GAL-GFP-RZ dots remain stable (compare light blue and pink).

In both backgrounds, the two genes localized to the nuclear periphery upon transcriptional activation (Fig. 5A, Time 0). Both genes also remained at the periphery for at least 15′ after transcriptional repression by glucose addition (Fig. 5A; compare dark blue and red, (Abruzzi et al., 2006). However, a deletion of RRP6 had no effect on post-transcriptional tethering of the GAL-GFP-RZ reporter (Fig. 5A; compare red and pink). This contrasts with the rapid delocalization of the GAL-GFP-GALpA reporter from the nuclear periphery by ~15 minutes after transcription shutoff. This indicates that post-transcriptional gene-pore tethering of GAL-GFP-RZ is insensitive to loss of the exosome, suggesting that cleavage and polyadenylation of the GAL-GFP gene play a role in the exosome-sensitive loss of tethering.

To relate this gene-tethering insensitivity to dot persistence, we examined GAL-GFP-RZ dots in the rrp6Δ strain. They were insensitive to transcriptional inhibition (Fig. 5B; compare light blue and pink), like the insensitivity of the gene-pore tethering described above (Fig. 5A). This contrasted with the rapid disappearance of GAL-GFP-GALpA dots in rrp6Δ cells (Fig. 3) but was similar to wild-type cells, in which both GAL-GFP-GALpA and GAL-GFP-RZ dots persist for up to 60′ after transcriptional repression (Fig. 5B; red and dark blue; see Materials and Methods). These differences indicate that the nuclear exosome impacts neither GAL-GFP-RZ dot persistence nor its gene-periphery tethering. As the GAL-GFP-RZ reporter is not a substrate for cleavage and polyadenylation, this strengthens the connection between the nuclear exosome and 3′ end formation in dot persistence and further suggests that dot RNA is relevant to gene-periphery tethering.

A post-transcriptional increase in a GAL1 poly A+ RNA species

To test directly whether the nuclear exosome mutations might impact cleavage and/or polyadenylation, we compared GAL1 mRNAs from wild-type and rrp6Δ cells as a function of time after transcriptional repression. To this end, we assayed mRNA levels and 3′ ends (by 3′ RACE) and compared cells grown in glucose, galactose and at different times after transcriptional shutoff (Fig. 6A).

Figure 6. In rrp6Δ cells, GAL1 mRNA undergoes a novel post-transcriptional polyadenylation event.

Total RNA isolated from cells after growth in glucose, galactose or at various times after transcriptional repression was analyzed using 3′-RACE. A) A 3′-RACE was performed on RNA from rrp6Δ and wild-type cells. As expected from previous studies, we observe two different GAL1 transcripts resulting from polyadenylation sites that are approximately 60bp and 170bp downstream of the stop codon. In wild-type cells both the short and long forms of the RNA gradually disappear after transcription is repressed as would be expected for normal RNA turnover. However, in rrp6Δ cells, while the larger form gradually turns over, the amount of the smaller species actually increases 10-15 minutes after transcription shutoff. B) To verify that the increase in the smaller species is not due to a technical artifact during the PCR step of the 3′-RACE (the two species share the 5′ primer), we eliminated the larger RNA from the assay using oligo-directed RNAse H cleavage (see Materials and methods). We observe the same increase in the short form of GAL1 mRNA as seen in part A suggesting that this effect is not due to a technical artifact. C) The amount of the short GAL1 transcript that persists in rrp6Δ and wild-type cells after transcriptional shutoff was quantitated. In rrp6Δ cells, there is an approximate 5-fold increase in the amount of the short GAL1 transcript approximately 15′ after transcription is shutoff by the addition of glucose. In contrast, the short GAL1 transcript decreases by approximately 2-fold over the same time period. D) Since the 3′-RACE assay only recognizes polyadenylated RNA, we hypothesized that the increase in the short form of the GAL1 mRNA could be due to a novel polyadenylation event by the poly-A polymerase, Pap1p. We performed a 3′-RACE experiment on rrp6Δ cells as well as rrp6Δ cells containing a conditional allele of the poly-A polymerase, pap1-1. These cells were grown at the permissive temperature (25°C) in galactose then shifted to the non-permissive temperature at the same time that glucose was added to repress transcription. In rrp6Δ cells, there is a marked increase in the level of the smaller GAL1 RNA species after 15′ of transcriptional shutoff. When poly-A polymerase is inactivated, we no longer observe this increase in the smaller GAL1 mRNA indicating that Pap1p is required for the novel post-transcriptional polyadenylation event.

The results show two bands, which reflect the two GAL1 3′ UTRs. The sizes indicate that cleavage occurs 60bp and 170bp after the stop codon (Fig. 6A, arrows), entirely consistent with the literature (Sadhale and Platt, 1992). Moreover, the levels of both species decrease after transcriptional shutoff (Fig. 6A; right) as expected. This is also the case for the larger species in the rrp6Δ strain, although turnover may be somewhat slower in this genetic background (Fig. 6A; left). Remarkably, the smaller species increases substantially in level 10 to 15 minutes after transcriptional shutoff in rrp6Δ cells (Fig. 6A; left). The same effect is observed with RNA transcribed from the GAL-GFP-GALpA reporter (data not shown).

To verify that this increase is not due to PCR competition between the two species during amplification, we cleaved the longer species via oligo-directed RNase H cleavage between the two primers to prevent its amplification; the smaller species in not complementary to the cleavage oligonucleotide and is amplified normally. There is still a comparable increase in the level of the smaller species (Fig. 6B). The identical result was obtained four times, one of which was quantitated. The smaller band shows a 5-fold increase, which peaks at 15 min after glucose inhibition (Fig. 6C). Note that RNA species of the appropriate sizes can be detected with an RNase-H Northern blot assay (Supplemental Fig. 1). Moreover, a dramatic increase in the levels of an RNA species with the appropriate properties for the post-transcriptionally adenylated RNA is also visible (Supplemental Fig. 1, lanes 10-11).

To test the hypothesis that the canonical nuclear poly-A polymerase generates this novel post-transcriptional polyadenylation event, we repeated the glucose shut-off 3′ RACE experiment in rrp6Δ cells that also contain a temperature sensitive mutation in the poly-A polymerase, Pap1p. rrp6Δ, pap1-1 cells were grown in galactose at the permissive temperature (25°C) and then shifted to the non-permissive temperature concomitant with glucose addition. There was no detectable increase in the smaller GAL1RNA species after transcriptional shutoff at the non-permissive temperature (Fig. 6D), suggesting that Pap1p is indeed responsible for the post-transcriptional generation of this RNA. Although the pA tail is somewhat short for the canonical co-transcriptional polyadenylation by Pap1p (ca. 40 nts; data not shown), this may reflect subsequent cytoplasmic deadenylation - even at 5 min. Moreover, dots disappear in rrp6Δ cells (~20% have dots) following transcriptional repression, but they remain stable in rrp6Δ, pap1-1 cells at the non-permissive temperature (~80% have dots; Supplemental Fig. 2). Our observations taken together suggest that the dot RNA is polyadenylated post-transcriptionally by Pap1p (Torchet et al., 2002).

What is the substrate for post-transcriptional polyadenylation in the rrp6Δ strain?

To characterize the putative substrate for the post-transcriptional polyadenylation reaction, we performed LM-3′RACE specifically on the poly A- pool of GAL1 mRNA. A four step approach was used. First, full length GAL1 mRNA was purified from wild type and rrp6Δ using two biotinylated oligonucleotides complementary to the region at the beginning of the ORF (Figure 7A; step 1, oligos shown in red). This RNA pool was then passed over an oligo-dT resin to remove pA+ RNA and enrich for polyA-GAL1 mRNA (Fig. 7A; step 2). A 3′ blocked oligonucleotide with a 5′ P was ligated to the 3′ end of the selected RNA (Figure 7A; step 3, oligo in green), which was followed by RT-PCR for 3′ end analysis (Figure 7A; step 4, oligos in blue; (Salles and Strickland, 1999). The PCR analysis shows only a single, discrete band at the expected size of the first cleavage site in both wild type and rrp6Δ cells grown in glucose (Figure 7B; lanes 1 and 6). Note that the wild-type band is detectable for at least 60 minutes after transcriptional repression with glucose (lanes 3 and 4) despite some intensity decrease (lanes 2-4). In contrast, the discrete band is no longer detectable in rrp6Δ after 15 minutes in glucose (lanes 7, 8), consistent with disappearance of dot RNA. Quantitative dilutions from two independent RNA preparations indicated at least a 50-fold difference between the two 15 min time points (wild-type vs rrp6Δ; data not shown).

Figure 7. LM-3′RACE reveal the presence of a species of GAL1 mRNA cleaved 60 bp 3′ of the stop codon.

A) Diagram showing the steps used to purify and analyze the poly A- GAL1 mRNA via LM-3′RACE. In step 1, two biotinylated oligonucleotides complementary to the region at the beginning of the ORF (red) were used to isolate full length GAL1 mRNA. This total GAL1 mRNA was then passed through oligo dT resin (Ambion) to generate poly A- pool of GAL1 mRNA (Step 2). A 3′ blocked oligonucleotide with a 5′ P was ligated to the 3′ end of the selected RNA (Step 3; green), which was followed by RT-PCR for 3′ end analysis (Step 4; blue). B) LM-3′RACE analysis of poly A- pool of GAL1 mRNA. Total RNA isolated from cells after growth in glucose, galactose and various times after transcriptional repression were used to purify poly A- pool of GAL1 mRNA (see A). LM-3′Race was performed and the products were separated on a 2% agarose gel. In wild type cells we see a discrete band around the size of the first cleavage site (lanes 2, 3 and 4) that persists for >60 min after transcription shutoff. In rrp6Δ, this species is visible when transcription is active (lane 6) and disappears 15 min after transcriptional repression (lane 7). C) Sequencing analysis was performed on both the PCR products as well as the major band from the LM-3′RACE. Sequencing revealed an unpolyadenylated species of GAL1 mRNA cleaved precisely at the first cleavage site. The underlined sequence is the primer VS27.

Sequence analysis of the discrete gel-purified PCR product from wild-type RNA after 15 min in glucose (Fig, 7B, lane 3) revealed that the major mRNA 3′ end is precisely at the first of the two GAL1 cleavage and polyadenylation sites (Fig 7C). The identical 3′ end was found for six of six clones from PCR products that were not gel purified (same primers and ligation; 3 clones from pA-RNA and 3 clones from GAL1 oligo-selected and then pA-RNA; data not shown).

The results suggest more generally that Pap1p is inhibited by the nuclear exosome. This creates a pool of unadenylated nuclear GAL1 mRNA generated by 3′ cleavage at the proximal polyA site (see Discussion). The RNA remains tethered near the site of transcription and also links the gene to the nuclear periphery. We suggest that a reduction in exosome function up-regulates Pap1p activity, which leads to post-transcriptional polyadenylation of dot RNA. The appearance of the newly polyadenylated RNA species correlates temporally with the loss of dots and suggests that post-transcriptional polyadenylation of dot RNA leads to its release from the site of transcription and a consequent weakening of the post-transcriptional gene-nuclear periphery tether.

DISCUSSION

Some yeast genes become localized to the nuclear periphery upon transcriptional activation. To identify factors involved in gene-pore tethering, a sub-library of deletions was screened using a visual assay to monitor the localization of a plasmid-based GAL reporter gene. Two nuclear exosome component genes, RRP6 and LRP1, specifically affect the post-transcriptional tethering of the GAL plasmid to the nuclear periphery. They have a parallel effect on post-transcriptional RNA that accumulates in foci or dots near the GAL gene. Exosome inactivation also causes the dots to disappear rapidly after transcriptional shutoff, and biochemical experiments show that the timing of this disappearance correlates with a substantial decrease in levels of unadenylated GAL1 RNA and an increase in levels of adenylated GAL1 RNA. All of these data suggest that the nuclear exosome inhibits poly-A polymerase, resulting in a pool of unadenylated GAL1 RNA that accumulates in dots. The combination of exosome inactivation and transcriptional repression leads to a decrease in dot RNA levels, by promoting post-transcriptional adenylation of dot RNA and by inhibiting the synthesis of new dot RNA. We propose that this decrease causes a loss of post-transcriptional gene-pore tethering, because dot RNP - perhaps through RNP proteins that normally contact nuclear pores - makes a substantial contribution to the physical tether.

In the last several years, multiple labs have investigated the phenomenon of gene movement to the nuclear periphery. Most of these studies examine gene localization during active transcription; they present a complicated and sometimes contradictory set of data. Although several experiments suggest that transcription is not required to establish gene-pore localization (Schmid et al., 2006), the bulk of the evidence suggests that transcriptional activation is the cause and/or the effect of gene-pore localization. For example, DNA elements as well as trans-acting protein factors are required for tethering actively transcribing genes to the nuclear periphery (Abruzzi et al., 2006; Brickner and Walter, 2004; Cabal et al., 2006; Chekanova et al., 2007; Dieppois et al., 2006; Schmid et al., 2006; Taddei et al., 2006). In addition, some experiments suggest that RNA is important for gene localization to nuclear pores (GAL1 and HXK1; Abruzzi et al., 2006; Taddei et al., 2006), although there are also studies indicating that RNA is dispensable for tethering (GAL2; Dieppois et al., 2006). Although some of the conflicting literature may reflect differences between different genes, the aggregate data are best accommodated by proposing multiple, partially redundant tethering mechanisms that work in concert to facilitate gene-pore associations.

A multiple tethering model is compatible with the different factors identified in this study (Fig. 2A, Table 1), which are involved in a variety of processes. Although it is possible that many of these deletion strains abrogate gene-pore tethering because they adversely affect GAL transcription, none of the 23 strains was noticeably slow growing in steady-state galactose conditions (data not shown). The simplest interpretation is therefore that GAL gene localization is quite sensitive to genetic perturbation. Despite the fact that no comparable tethering screen exists in the literature, plasmid tethering may also be somewhat more sensitive than the more common chromosomal assay (data not shown). Nonetheless, we suspect that this new assay is more accurate for defining tethering requirements of a single gene like GAL1 and suggests that the proteins previously proposed as direct gene-periphery tethers for GAL1 (Sus1p, Sac3p etc.; Cabal et al., 2006; Chekanova et al., 2007; Drubin et al., 2006) may function only indirectly. Most important for the main thrust of this paper is that a much smaller and more coherent set of genes affects gene maintenance at the periphery after transcriptional repression. Moreover all of the exosome-relevant results reported here have been verified with a chromosomal GAL reporter gene (Supplemental Fig 3); this includes the polyadenylation data, which were from the native GAL1 gene RNA (Fig. 6).

Our previous studies showed that dot formation and the ability of a gene to associate with the nuclear periphery were impacted in parallel by 3′-UTRs. Our preferred interpretation was that inefficient 3′-end formation promotes dot RNA formation, which then helps tether genes to the periphery in the absence of transcription (Abruzzi et al., 2006). The experiments in this paper add substantially to this view. They include the abrogation of post-transcriptional tethering in the GAL1 ORFLESS gene (Fig. 3A), the parallel effects of the exosome deletions and rrp4-1 on dot RNA and post-transcriptional tethering (Figs. 2 and 4), and the tethering insensitivity of an RZ-terminated gene to the exosome mutations (Fig. 5) Although all of these phenomena could still occur indirectly, through effects on chromatin or gene looping for example, a causal role for RNP in post-transcriptional gene-pore tethering is parsimonious.

Moreover, our experiments indicate that the exosome functions to inhibit efficient 3′ end formation of GAL transcripts. This is because the absence of the exosome allows GAL1 3′ ends to undergo enhanced polyadenylation, which occurs well after Pol II has left the GAL gene (Fig. 6). It is consistent with reported genetic and physical interactions between the exosome and poly-A polymerase mutants (Burkard and Butler, 2000; Milligan et al., 2005), and direct inhibition of polyadenylation by the exosome has been recently observed in biochemical experiments (Cyril Saguez et al., 2008). This suggests that polyadenylation neither needs to be co-transcriptional nor tightly coupled to 3′ cleavage.

LM-3′ RACE and sequencing confirmed the presence of a pool of unadenylated, stable RNA terminated exactly at the proximal GAL1 cleavage and polyadenylation sites (Fig 7). Note that the initial oligo selection indicates that it is full-length or nearly so (Fig. 7A, red). Although this RNA could be generated by core nuclear exosome degradation of RNA that terminates at the distal GAL1 3′ cleavage site or even further upstream (Torchet et al., 2002; Vasiljeva and Buratowski, 2006), its sequence suggests that most if not all transcripts are generated by proper 3′ end formation without polyadenylation (Fig. 7C). Given the post-transcriptional increase in both pA+ species in the core exosome mutant strain (Supplemental Fig. 4), the data suggest that unadenylated RNA terminated at the distal site is normally efficiently degraded by the core exosome. Degradation from the proximal site is probably less efficient. As a result, there remains a pool of properly 3′ cleaved but unadenylated pA-RNA, which we suggest is dot RNA. Because the observed post-transcriptional adenylation events apparently require transcriptional inhibition as well as an exosome mutation, they might benefit from the lack of competing substrate sites generated by ongoing transcription.

The proposed role of the nuclear exosome in post-transcriptional dot RNA formation/stability is reminiscent of previous experiments implicating the nuclear exosome in mRNA surveillance near transcription sites (Hilleren et al., 2001; Libri et al., 2002; Rougemaille et al., 2007). In these studies, heat shock mRNA export was blocked in different RNA processing mutant strains (pap1-1, rna15-2, hpr1Δ, etc), and in situ hybridization revealed that some non-exported mRNA is retained in foci or dots like what we describe here for a GAL gene under wild-type conditions. However, the mutant strain heat shock dots require the exosome even during active transcription (Hilleren et al., 2001; Rougemaille et al., 2007). This might reflect a difference in dot-gene tethers between heat shock genes and GAL genes, or some feature of polyadenylation at the proximal GAL1 site may be unusually inefficient.

The inferred less efficient exosome degradation of pA-RNA terminated at the proximal site may indicate that it remains decorated with 3′ end formation machinery. This may be important for dot formation, by linking the transcript to chromatin or to a chromatin-bound pol II CTD. This relationship to polyadenylation also predicts that other RNA processing mutants that give rise to dot RNA will/may display polyadenylation defects. Indeed, the dot-forming mutant strain Δ mft1 generates improperly polyadenylated mRNAs, and the RRP6 deletion rescues both dot formation and the polyadenylation defects of the Δmft1 mutant strain (Cyril Saguez et al., 2008). Other dot-forming mutants may function similarly, because they affect cleavage and polyadenylation or because they present an inappropriate mRNP substrate to this machinery. A prediction of the connection between dot RNA and post-transcriptional gene-periphery tethering is that export mutant strains like mft1Δ will manifest gene-periphery tethering as well as dot formation.

MATERIALS AND METHODS

Yeast strains and plasmids

All yeast strains and plasmids used in this study are described in Supplementary Materials and Methods and Supplemental Table 1.

Microscopy

Cells for intranuclear localization experiments were prepared as described previously in Abruzzi et al., (2006). Temperature sensitive (ts) strains were grown at 25°C to an OD600 ~ 0.8 to 1 and then divided equally into two halves. Equal volumes of 25°C or 50°C media (containing either glucose or galactose) were added and then grown at 25°C or 37°C respectively. For each sample or time point between 150 to 200 cells were scored for the position of the reporter loci and categorized as either peripheral (reporter locus touching or overlapping the nuclear rim) or internal (reporter locus away from the nuclear rim). Fluorescence microscopy was performed using an IX70 inverted microscope (Olympus) and where required the cells were counted using the Openlab software (Improvision). Deconvolution microscopy was performed using the Volocity software (Improvision). 51 Z-stack images with a step size of 0.2 μm were obtained using Volocity acquisition software and iterative deconvolution was performed on the images. The significance for iterative restoration was determined using 99.5 % confidence limit or 25 iterations.

Fluorescence in-situ hybridization (FISH) was performed as described in Abruzzi et al., 2006 except that the sensitivity of our experiments was increased by purifying the probes using QIAquick Nucleotide removal kit (Qiagen). Slides were visualized using a100X objective and the exposure time was kept constant throughout each experiment. Images were captured using Openlab software (Improvision). Where required for each sample or time points, between 150 to 200 cells were scored for the presence or absence of dots in two independent experiments.

RNA Analysis

Total cellular RNA from intact yeast cells was extracted as described by (Ausubel et al., 1994). 3′-RACE was performed as described in GeneRacer kit (Invitrogen) with minor modifications. RNaseH cleavage assays were performed as described in (Decker and Parker, 1993) with few minor changes. 10 μg of total RNA was first treated with RQ1 DNase (Promega) and then 5 μg of the total RNA was cleaved using oligo (VS26) and RNaseH (Ambion). Cleaved RNA was used for the reverse transcription reaction and the resulting cDNA was diluted 1:250 and 3 μl was used as template for the second strand synthesis PCR (VS24 and VS25). The PCR samples were analyzed on a 2% agarose gel. For quantitation, a 1% agarose gel stained with SYBR Green I (Molecular probes) was analyzed using a STORM 860 imager (GE) and analyzed using the ImageQuant TL software (GE).

Total GAL1 mRNA pull down was performed using two oligonucleotides tagged with biotin at the 5′ end (VS32 and VS33). 4μg of each oligonucleotide was resuspended in 500μls of 0.5 X SSC and conjugated to 400μls of prewashed Streptavidin sepharose (GE Healthcare) on ice for 30 min. The beads were washed with 2 mls of 0.5 X SSC and resuspended in 300 μls of 0.5 X SSC. 100μg of total RNA was resuspended in 300 μls of 0.5 X SSC and incubated at 80 °C for 5 min. The denatured RNA was added to the biotinylated oligonucleotide conjugated sepharose and incubated at 50 °C for 1h vortexing every 10 minutes. The resin was washed with 0.5 X SSC for 5 min at 50°C and eluted with 500μls of TE at 80 °C. GAL1 mRNA was precipitated and passed over an oligo dT resin (Ambion) to generate the poly A- pool of GAL1 mRNA (as described by the supplier).

The oligonucleotide (VS27) used for LM-3′RACE was phosphorylated at the 5′ end and the 3′ end was amino modified to prevent concatamerization. 50ng of VS27 was ligated to equal amounts of polyA-GAL1 mRNA using T4 RNA ligase (NEB) and RT-PCR was performed as described above using the primer VS28. cDNA was amplified using primers VS25 and VS28 and the PCR products were analyzed using 2% agarose gels. Discrete bands were purified and the DNA was cloned into pGEM-T vectors (Promega) for sequencing using primers VS25.