Distinct pathways for snoRNA and mRNA termination.

Abstract

Transcription termination at mRNA genes is linked to polyadenylation. Cleavage at the poly(A) site generates an entry point for the Rat1/Xrn2 exonuclease, which degrades the downstream transcript to promote termination. Small nucleolar RNAs (snoRNAs) are also transcribed by RNA polymerase II but are not polyadenylated. Chromatin immunoprecipitation experiments show that polyadenylation factors and Rat1 localize to snoRNA genes, but mutations that disrupt poly(A) site cleavage or Rat1 activity do not lead to termination defects at these genes. Conversely, mutations of Nrd1, Sen1, and Ssu72 affect termination at snoRNAs but not at several mRNA genes. The exosome complex was required for 3' trimming, but not termination, of snoRNAs. Both the mRNA and snoRNA pathways require Pcf11 but show differential effects of individual mutant alleles. These results suggest that in yeast the transcribing RNA polymerase II can choose between two distinct termination mechanisms but keeps both options available during elongation.

Full text

Page 1

Molecular Cell 24, 723–734, December 8, 2006 ª2006 Elsevier Inc. DOI 10.1016/j.molcel.2006.11.011
Distinct Pathways for snoRNA and mRNA Termination
Minkyu Kim,1Lidia Vasiljeva,1Oliver J. Rando,2
Alexander Zhelkovsky,3Claire Moore,3
and Stephen Buratowski1,*
1Department of Biological Chemistry
and Molecular Pharmacology
Harvard Medical School
240 Longwood Avenue
Boston, Massachusetts 02115
2Bauer Center for Genomics Research
Harvard University
7 Divinity Avenue
Cambridge, Massachusetts 02138
3Department of Molecular Microbiology
Tufts University School of Medicine
136 Harrison Avenue
Boston, Massachusetts 02111
Summary
Transcription termination at mRNA genes is linked to
polyadenylation. Cleavage at the poly(A) site gener-
ates an entry point for the Rat1/Xrn2 exonuclease,
whichdegrades thedownstream transcript topromote
termination. Small nucleolar RNAs (snoRNAs) are also
transcribed by RNA polymerase II but are not polyade-
nylated. Chromatin immunoprecipitation experiments
show that polyadenylation factors and Rat1 localize
to snoRNA genes, but mutations that disrupt poly(A)
sitecleavageorRat1activitydonotleadtotermination
defectsatthesegenes.Conversely,mutationsofNrd1,
Sen1, and Ssu72 affect termination at snoRNAs but
not at several mRNA genes. The exosome complex
was required for 30trimming, but not termination, of
snoRNAs. Both the mRNA and snoRNA pathways re-
quire Pcf11 but show differential effects of individual
mutant alleles. These results suggest that in yeast
the transcribing RNA polymerase II can choose be-
tween two distinct termination mechanisms but keeps
both options available during elongation.
Introduction
Dynamic interactions of various factors with RNA poly-
merase II (Pol II) promote initiation, elongation, or termi-
nationoftranscription.Accurateterminationisimportant
because stopping transcription too late or too early can
disrupt proper gene expression. Termination by Pol II is
tightly coupled to mRNA30end processing, and multiple
proteins, RNA/DNA sequences, and chromatin configu-
rations may be involved (Proudfoot et al., 2002; Proud-
foot, 2004; Buratowski, 2005).
Two major models have been proposed for transcrip-
tion termination of protein coding genes. The ‘‘anti-ter-
minator’’ or ‘‘allosteric’’ model proposes that transcrip-
tion through the poly(A) signal changes the properties
of the elongating Pol II complex, perhaps by dissocia-
tion of positive elongation factors or recruitment of ter-
mination factors (Logan et al., 1987). Indeed, this model
is supported by evidence for exchange of factors asso-
ciated with polymerase at the 30end of genes (Kim et al.,
2004a) and genes where termination apparently occurs
without cleavage (Osheim et al., 1999; Tran et al., 2001;
Sadowski et al., 2003). The ‘‘torpedo’’ model postulates
that cleavage of the nascent RNA transcript at the
poly(A) site transmits a signal to Pol II, leading to the de-
stabilization of the elongation complex (Connelly and
Manley, 1988). Recently, the 50to 30exoribonuclease
Rat1/Xrn2 was shown to degrade the RNA transcript
downstream of a poly(A) cleavage or cotranscriptional
cleavage (CoTC) site, and this degradation somehow
promotes transcription termination (Kim et al., 2004b;
West et al., 2004). These findings strongly support the
torpedo model. Since both models have experimental
support, it seems likely that termination can occur by
more than one mechanism. One pathway or another
may predominate in particular gene contexts.
In addition to mRNAs, Pol II transcribes noncoding
RNAs such as the small nuclear RNAs (snRNAs) and
small nucleolar RNAs (snoRNAs). snoRNAs are required
for maturation of pre-rRNAs (cleavage and base modifi-
cation), and they are divided into two classes (C/D and
H/ACA) based on conserved secondary structures
(Kiss, 2002). Each class of snoRNAs is associated with
aspecificsetofproteinstoformstablesnoRNPs.Recent
studies indicate that snoRNP assembly occurs cotran-
scriptionally (Morlando et al., 2002; Ballarino et al.,
2005; Yang et al., 2005). So far, 66 snoRNAs have been
identified in Saccharomyces cerevisiae, and 42 of them
are independently transcribed (Samarsky and Fournier,
1999).
The30ends ofsnoRNAs arenot polyadenylated bythe
mRNA poly(A) polymerase (Pap1), but the basis for this
difference with mRNAs is still unclear. One proposal is
that snoRNA 30end formation is a variation of the mech-
anism used for mRNAs. Cleavage factor IA, which is
required for mRNA 30end formation, could promote
cleavageofsnoRNAtranscriptswithoutsubsequentpol-
yadenylation of the 30ends (Fatica et al., 2000; Morlando
et al., 2002). Additionally, 30extended forms of snoRNAs
accumulate in strains with mutations in the exosome,
acomplexof30to50exonucleases,indicatingthatlonger
premature snoRNA transcripts are processed by 30trim-
ming (Allmang et al., 1999; van Hoof et al., 2000). Several
proteins have been implicated in snoRNA 30end forma-
tion. Mutations in either of two RNA-binding proteins
(Nrd1 and Nab3) or a helicase (Sen1) cause defective
termination at multiple snoRNA genes (Steinmetz et al.,
2001; Ursic et al., 2004). Subunits of the APT complex
(Pti1, Ref2, Ssu72, and Swd2), which associate with the
mRNA cleavage/polyadenylation factor Pta1, are also
implicatedinsnoRNAtermination(Dheuretal.,2003;Ga-
nemetal.,2003;SteinmetzandBrow,2003;Chengetal.,
2004; Dichtl et al., 2004). Recently, a role for the Paf1
complexinsnoRNA30endformationwasreported(Shel-
don et al., 2005). However, the mechanism or mecha-
nisms by which these proteins promote snoRNA termi-
nation are not understood.
*Correspondence: steveb@hms.harvard.edu

Page 2

Here, we compare factor requirements for mRNA and
snoRNA 30end formation. Chromatin immunoprecipita-
tion (ChIP) experiments show that Rat1 and the known
mRNA cleavage factors are present at snoRNA genes,
as are Sen1, Nrd1, and Nab3. However, we find that
termination at SNR13 or SNR33 is normal in strains de-
fective for Rat1 activity or mRNA 30end cleavage. Con-
versely, mutations in the Nrd1/Nab3/Sen1 complex
leadtodefectsinsnoRNA30endformationbutdonotaf-
fect30endsatseveral mRNAs tested.These results sug-
gestthatsnoRNAterminationcanoccurbyamechanism
distinct from the mRNA torpedo model. The predomi-
nant snoRNA pathway may involve cleavage-indepen-
dent termination and subsequent 30trimming by the
exosome/Nrd1complex.However,theChIPresultssug-
gest that elongating Pol II interacts with both machiner-
ies and that the choice between pathways is made while
transcribing.
Results
Rat1-Independent Transcription Termination
of snoRNA Genes
We previously showed that the exonuclease Rat1/Xrn2
promotes transcription termination by yeast Pol II in
vivo (Kim et al., 2004b). In this study, crosslinked chro-
matin was immunoprecipitated for the Pol II subunit
Rpb3 and used to probe a microarray covering Saccha-
romyces cerevisiae chromosome III. At most mRNA
genes with a clear Rpb3 signal, crosslinking of Pol II ex-
tendedfurtheratthe30endinthemutantascomparedto
wild-type cells. For example, the highly transcribed
PGK1 gene (transcription rate 110.5 mRNA/hr; see Hol-
stege et al., 1998) shows 30extended Pol II crosslinking
into the POL4 locus (0.3 mRNA/hr, which is below the
level of detection by ChIP for Rpb3) in rat1-1 cells at
the nonpermissive temperature (Figure 1).
Even though it is robustly transcribed by Pol II, tran-
scriptional termination of the nearby H/ACA class
snoRNA gene SNR33 was not affected by the rat1-1
mutation (Figure 1). Indeed, in the rat1-1 mutant no
termination defects were observed at any of the other
snoRNAs on chromosome III (SNR43, SNR65, and
SNR189; data not shown). However, Pol II density on
all of these snoRNA genes was slightly reduced in the
mutant. Since Rat1 also participates in the 50processing
or degradation of snoRNA precursors (Lee et al., 2003),
we speculate there may be a regulatory mechanism to
reduce snoRNA transcription when snoRNA precursors
accumulate.
To investigate the mechanism used during transcrip-
tion termination ofsnoRNAs, we monitored two snoRNA
genes: the C/D class SNR13 and the H/ACA type SNR33
(Figure 2A). The recruitment of several known RNA pro-
cessing factors to these snoRNA genes was monitored
using ChIP assays with TAP-tagged factors. Because
of the small size of these genes, it was not possible to
easily resolve 50and 30crosslinking. Proteins involved
in cleavage and polyadenylation of mRNAs, including
subunits of CFIA (Rna14, Rna15, and Pcf11), CFIB
(Hrp1), and CPF (Cft2), were strongly recruited to
SNR33 (Figure 2B) and SNR13 (data not shown). Similar
results were seen with subunits of the APT complex, a
set of proteins associated with CPF (Nedea et al., 2003).
Crosslinking of Ssu72, a C-terminal domain (CTD) phos-
phatase important for mRNA and snoRNA 30end forma-
tion, to snoRNA genes was previously reported (Nedea
et al., 2003).
Surprisingly, despite the fact that snoRNA termination
is unaffected in the rat1-1 mutant strain, Rat1 is clearly
recruitedtotheSNR33gene(Figure2B).Asimilarrecruit-
ment pattern was observed on SNR13 (data not shown).
To confirm Rat1-independent termination of snoRNAs
as seen on the microarray, conventional ChIP assays
with anti-Rpb3 antibody were performed using rat1-1
cells. As a control to show that we can detect snoRNA
termination defects in the ChIP assay, temperature-sen-
sitive strains mutated in SSU72 were used to generate
readthrough transcription (Ganem et al., 2003). When
shifted to 37?C for 30 min, rat1-1 cells showed a 30ex-
tended Pol II signal at the ADH1 gene, but ssu72-2 has
no effect (Figure 3A). Conversely, a 30extension of Pol II
crosslinking at SNR33 was seen in ssu72-2 and ssu72-3
cells but not in the rat1-1 mutant (Figure 3B). As seen in
the ChIP/microarray experiment, Pol II crosslinking to
snoRNAs was slightly reduced in the rat1-1 strain. In
agreement with the ChIP results, SNR13 and SNR33
readthrough transcripts were detected by northern
Figure
SNR33 by Pol II Is Not Dependent on Rat1
1. TranscriptionalTerminationof
Chromatin immunoprecipitated with anti-
Rpb3 was amplified and analyzed using tiled
microarrays of chromosome III as described
(Kim et al., 2004b). The PGK1-SNR33 region
is shown. The y axis indicates the log2value
of the immunoprecipitation (IP) to input ratio,
and the x axis represents position along
the chromosome. Dark blue arrows denote
open reading frames (ORFs). Note that
PGK1 transcription is from left to right, while
SNR33 is from right to left. The numbers in
parentheses show the transcription rates
(mRNA/hr), with N/D being below the detec-
tion limit (Holstege et al., 1998).
Molecular Cell
724

Page 3

blottinginthessu72mutantsbutnotintherat1-1mutant
(Figure 3C). Therefore, termination of transcription at
these snoRNA genes is not dependent upon Rat1, and
thisfindingsuggeststhatsnoRNAterminationmayoccur
byamechanismdifferentfromthatusedatmRNAgenes.
Cleavage-Independent Termination
of snoRNA Genes
Rat1-independent termination at snoRNA genes raises
the question of whether there is cleavage of nascent
RNA transcripts at the 30end of these RNAs. For mRNA
genes, endonucleolytic cleavage of the transcript by
the polyadenylation machinery generates an uncapped
50phosphorylated RNA end that is thought to provide
an entry site for Rat1-dependent degradation. These
events contribute to termination by a mechanism that
is still unclear. To see whether a similar mechanism ap-
plies to snoRNAs, termination was assayed in mutant
strains where mRNA cleavage is inhibited.
Pcf11 is a component of cleavage factor IA (Rna14,
Rna15, Pcf11, and Clp1). Pcf11 binds the CTD of the
Pol II largest subunit and is required for pre-mRNA 30
end processing (Amrani et al., 1997; Licatalosi et al.,
2002). Keller and colleagues characterized a useful col-
lection of conditional alleles of PCF11 (Sadowski et al.,
2003)as follows: pcf11-2leads todefectiveCFIAactivity
but has normal CTD binding; pcf11-13 lacks CTD-bind-
ing ability but still supports mRNA cleavage in vitro;
pcf11-9 is defective for both CFIA activity and CTD-
binding ability. At mRNA genes like ADH1 and PMA1,
the pcf11-2 and pcf11-9 mutants, but not pcf11-13,
have termination defects (Figure 4A and see Figure S1
in the Supplemental Data available with this article on-
line; note that the effect of pcf11-2 is much stronger on
PMA1 than ADH1). This argues that at mRNAs, cleavage
of the nascent RNA transcripts is necessary for termina-
tion. In contrast, transcription readthrough at snoRNAs
was seen in pcf11-9 and pcf11-13 but not pcf11-2 cells,
indicating that CTD interaction but not cleavage was
important at these genes. Rpb3 crosslinking levels to
snoRNA genes and PMA1 were greatly reduced in the
pcf11-9 strain at the nonpermissive temperature. The
basis for this is unclear, but it may be an indirect effect
because the cells are dying. In rna14-1 strains, which
also have mRNA cleavage defects, snoRNA termination
was normal (data not shown).
Cleavage/polyadenylation factor Brr5/Ysh1 (the yeast
homologof CPSF-73) isthought tobethe30
Figure 2. Crosslinking of RNA 30Processing Factors to SNR33
(A) Schematic diagrams of the SNR33, SNR13, and ADH1 gene-flanking regions. The coding region (ADH1) or mature transcript (SNR33 and
SNR13) is represented by an open box. Black arrows indicate promoters and direction of transcription. Open arrow in ADH1 indicates the po-
sition ofthe majorpolyadenylation site. Barsabove genes show the positions ofPCR products usedinthe ChIP assays. Numbersinparentheses
show the transcription rates (mRNA/hr).
(B) ChIP analysis was carried out on the SNR33 gene in strains carrying the indicated TAP-tagged proteins. Precipitated chromatin was PCR
amplified with primers shown in (A). The top band is the SNR33-specific band, while the common lower band (marked by asterisk) is an internal
background control from a nontranscribed region on chromosome V. The input panel shows control used to normalize the PCR amplification
efficiency of each primer pair. Similar results were seen for SNR13 (data not shown).
snoRNA Termination Pathway
725

Page 4

endonuclease responsible for cleavage at the poly(A)
site (Ryan et al., 2004). More recently, CPSF-73 has
also been implicated in the endonucleolytic 30end pro-
cessing of the nonpolyadenylated metazoan histone
pre-mRNAs (Dominski et al., 2005; Kolev and Steitz,
2005). To investigate whether Brr5 has a role in termina-
tion, we depleted Brr5 in vivo using a temperature-
inducible degron strain (Zhelkovsky et al., 2006). Upon
Brr5 depletion, termination was defective at all mRNA
genes tested (Figure 5 and Figure S2). In contrast, termi-
nation atSNR13 and SNR33 was normal (Figures 5B and
5C), further indicating that cleavage by the polyadenyla-
tion machinery is not required for snoRNA termination.
We considered whether another protein might func-
tion as an endonuclease to generate an exonuclease
entry site. Rnt1 is the yeast RNase III and is involved in
processing and maturation of multiple RNAs, including
ribosomal and snoRNAs. Endonucleolytic cleavage by
Rnt1 generates entry points for Rat1 and the related
Xrn1 nuclease to carry out subsequent 50trimming of
unprocessed snoRNA precursors (Lee et al., 2003). Al-
though Rnt1 cleaves nascent snoRNA transcripts (Lee
et al., 2003), its deletion does not lead to any defects
in termination of snoRNA or mRNA genes (Figure S3).
We also tested for a role of the nuclear exosome in ter-
mination, since exosome and Rnt1 participate in RNA
30end trimming and degradation of snoRNAs (Allmang
et al., 1999). Deletion of nuclear exosome subunit Rrp6
did not affect termination in either mRNAs or snoRNAs
(Figure S3).
Requirement of Distinct Protein Factors
for snoRNA Termination
TheNrd1complexincludesNrd1,Nab3,andSen1(Stein-
metz et al., 2001; Dheur et al., 2003; Ganem et al., 2003;
Ursic et al., 2004; Vasiljeva and Buratowski, 2006). This
complex is required for termination at snoRNA genes,
perhaps in combination with other factors (Steinmetz
et al., 2001; Dheur et al., 2003; Ganem et al., 2003; Ursic
etal.,2004).However,theNrd1complexisalsodetected
by ChIP at mRNA genes (Nedea et al., 2003), suggesting
itmayfunctionatbothclassesofgenes.Accordingly,we
testedwhether mRNAtermination isaffectedin nrd1-51,
nrd1-102, and sen1-1 conditional mutants. Termination
at the ADH1 and PMA1 genes was not affected at the
nonpermissive temperature (Figure 6A and Figure S4),
Figure 3. Transcription Termination of SNR13 and SNR33 Is Not Dependent upon Rat1
(A) ChIP analysis with anti-Rpb3 antibody on ADH1 in isogenic wild-type, ssu72-2, and rat1-1 strains. Cells were shifted to nonpermissive tem-
perature(37?C)for30min.TheleftpanelshowsPCRproducts,andnumbersbeloweachlanerefertoPCRprimersasinFigure2A.Therightpanel
shows quantification of results. The y axis is the ratio of ADH1 signal relative to the internal negative control after normalizing to the input con-
trols. A value of 1.0 therefore indicates no signal above background. Crosslinking of Rpb3 (i.e., Pol II) extended beyond position 4 in the rat1-1
strain, confirming its role in mRNA termination. In contrast, the ssu72-2 mutation has no effect on termination of ADH1.
(B) ChIP wasperformedon SNR33.Two ssu72 mutants, but not rat1-1,show termination defects. Similar results (data not shown) wereseen with
SNR13.
(C) Northern blot analysis of total RNAs isolated at 23?C or 37?C (30 min). Mature form and readthrough transcripts of SNR13 or SNR33 are
indicated by arrows.
Molecular Cell
726

Page 5

suggestingthatalthoughthesefactorsmaybepresentat
mRNA genes they are not necessary for normal termina-
tion. In contrast, SNR13 showed a strong termination
defect in nrd1 and sen1 mutants (Figure 6B). Somewhat
surprisingly, sen1-1 and two ssu72 mutants exhibited
transcriptional readthrough at SNR33, but termination
defects at SNR33 were not observed in nrd1-51 or
nrd1-102 mutants (Figures 6C, 3B, and 3C). Considering
that Nrd1 is a sequence-specific RNA-binding protein,
this variation in Nrd1 dependency within snoRNA genes
is probably due to different sequence contexts at the 30
end. Indeed, while SNR13 scores very highly for pre-
dicted Nrd1 sites, SNR33 does not (Carroll et al., 2004).
Thecap-bindingcomplexassociateswiththeNrd1com-
plex (Vasiljeva and Buratowski, 2006) and crosslinks to
both mRNA and snoRNA genes (Figure 2 and data not
shown). However, deletion of the Cbc1 subunit gene
did not result in termination defects at either class of
genes (Figure S3).
Detection of Precursor and Readthrough
snoRNA Transcripts
To confirm the ChIP results and to analyze how defec-
tive termination affects RNA synthesis, northern blot as-
says were used to monitor ADH1, SNR13, and SNR33
transcripts (Figure 7). Transcripts were separated by
agarose (Figure 7A) or acrylamide (Figures 7B and 7C)
gel electrophoresis. While a single major transcript
wasobservedforeachgeneinwild-type cells,twotypes
of new snoRNA transcript species were observed in mu-
tant strains. One class was 200–300 nucleotides (nt) lon-
ger than the processed transcript. Based on the Pol II
ChIP data (Figures 1 and 3–6), this RNA is the size ex-
pected for the full-length transcription product before
30trimming by exosome (pre-snRNA). Indeed, this tran-
scriptisseeninanexosomemutant(rrp6D)wheretermi-
nation is normal but processing is partially defective
(Figure 7A, lanes 12–15; Figure 7B, lanes 3 and 4; Fig-
ure 7C, lanes 3 and 4). Interestingly, a small amount of
pre-snR33 is also seen in nrd1 mutants at nonpermis-
sive temperature (Figure 7A, lanes 7 and 9; Figure 7C,
lanes 17 and 19). We confirmed that these RNAs were
not due to 50extension by RT-PCR (data not shown).
Since nrd1 mutants have no termination defect on this
gene (Figure 6C), we believe the appearance of pre-
snR33 reflects the other function of Nrd1 in recruiting
and activating exosome at snoRNA 30ends (Vasiljeva
and Buratowski, 2006).
The second new class of transcript seen in mutant
strains represents transcriptional readthrough of the
Figure 4. Allele-Specific Effects of PCF11 Mutations on mRNA and snoRNA Termination
Wild-type, pcf11-2, pcf11-9, and pcf11-13 cells were shifted to nonpermissive temperature (37?C) for 30 min, and a-Rpb3 ChIP assays were
carried out. Quantification data are shown in the right panels. (A) ADH1 gene. (B) SNR13 gene. (C) SNR33 gene. See also Figure S1 for the
PMA1 gene.
snoRNA Termination Pathway
727