RNA polymerase II CTD interactome with 3′ processingand termination factors in fission yeast and itsimpact on phosphate homeostasisAna M. Sancheza, Stewart Shumanb,1, and Beate Schwera,1

aDepartment of Microbiology and Immunology, Weill Cornell Medical College, New York, NY 10065; and bMolecular Biology Program, Sloan-KetteringInstitute, New York, NY 10065

Edited by Fred M. Winston, Harvard Medical School, Boston, MA, and approved October 1, 2018 (received for review June 21, 2018)

The carboxy-terminal domain (CTD) code encrypted within theY1S2P3T4S5P6S7 heptad repeats of RNA polymerase II (Pol2) is deeplyrooted in eukaryal biology. Key steps to deciphering the code areidentifying the events in gene expression that are governed byindividual “letters” and then defining a vocabulary of multiletter“words” and their meaning. Thr4 and Ser7 exert opposite effectson the fission yeast pho1 gene, expression of which is repressedunder phosphate-replete conditions by transcription of an up-stream flanking long noncoding RNA (lncRNA). Here we attributethe derepression of pho1 by a CTD-S7A mutation to precocioustermination of lncRNA synthesis, an effect that is erased by muta-tions of cleavage-polyadenylation factor (CPF) subunits Ctf1,Ssu72, Ppn1, Swd22, and Dis2 and termination factor Rhn1. Bycontrast, a CTD-T4A mutation hyperrepresses pho1, as do CPF sub-unit and Rhn1 mutations, implying that T4A reduces lncRNA ter-mination. Moreover, CTD-T4A is synthetically lethal with ppn1Δand swd22Δ, signifying that Thr4 and the Ppn1•Swd22 moduleplay important, functionally redundant roles in promoting Pol2 ter-mination. We find that Ppn1 and Swd22 become essential for viabil-ity when the CTD array is curtailed and that S7A overcomes the needfor Ppn1•Swd22 in the short CTD context. Mutational synergies high-light redundant essential functions of (i) Ppn1•Swd22 and Rhn1, (ii)Ppn1•Swd22 and Ctf1, and (iii) Ssu72 and Dis2 phosphatases. CTDalleles Y1F, S2A, and T4A have overlapping synthetic lethalities withppn1Δ and swd22Δ, suggesting that Tyr1-Ser2-Thr4 form a three-letter CTD word that abets termination, with Rhn1 being a likely“reader” of this word.

Pol2 | CTD code | transcription termination | synthetic lethality

The carboxyl-terminal domain (CTD) of the Rpb1 subunit ofRNA polymerase II (Pol2) consists of tandem repeated

heptapeptides of consensus sequence Y1S2P3T4S5P6S7. The CTDis essential for cell viability because it recruits proteins thatregulate transcription, modify chromatin structure, and catalyzeor regulate RNA capping, splicing, and polyadenylation (1–4).The inherently plastic CTD structure is modulated dynamicallyby phosphorylation and dephosphorylation of the heptad Ser,Thr, and Tyr residues in rough synchrony with the steps of thetranscription cycle (e.g., preinitiation, initiation, elongation, andtermination). In turn, the primary structure of the CTD conveysinformational cues about the transcription machinery—a CTDcode—that is read by CTD-binding proteins (1–3). Insights intoCTD coding principles have been gained by (i) probing bio-chemically and structurally how individual proteins recognize theCTD and (ii) genetically manipulating the composition andstructure of the CTD and gauging effects on cell physiology.The fission yeast Schizosaccharomyces pombe is an attractive

model system for CTD structure–function analysis because thenative heptad repeat array is relatively homogeneous vis-à-visother taxa. The S. pombe CTD consists of 29 heptad repeats (SIAppendix, Fig. S1). The junction CTD segment to the body ofRpb1 consists of four repeats that deviate in size and/or se-quence from the consensus heptad; this segment is referred to as

the CTD “rump.” Distal to the rump is an array of 25 heptadrepeats that adhere perfectly to the YSPTSPS consensus, withthe single exception of an Ala in lieu of Pro3 in the fifth heptaddownstream of the rump. The in vivo requirements for all sevenamino acids of the Y1S2P3T4S5P6S7 repeat were gauged by in-troducing Ala and conservative substitutions in lieu of Tyr1,Ser2, Pro3, Thr4, Ser5, Pro6, and Ser7 of every heptad of theCTD array in the context of a fully functional Rpb1 subunit witha CTD composed of the rump plus 14 consensus heptads (5, 6).The salient findings were that (i) Tyr1, Pro3, Ser5, and Pro6 areessential for viability of fission yeast, by the criterion that Alasubstitution is lethal, whereas Ser2, Thr4, and Ser7 are not, and(ii) Y1F, Y1F+S2A, Y1F+S7A, S2A+T4A, S2A+S7A, and T4A+S7Amutants are viable, signifying that Phe is functional in lieu ofTyr1 and that Ser5 is the only strictly essential phosphorylation site.The essentiality of Ser5-PO4 in fission yeast is linked to recruitmentof the mRNA-capping apparatus to the Pol2 elongation complex(5). Indeed, the requirement for Ser5-PO4 (and for Pro6) can bebypassed by covalently fusing the capping enzyme to Pol2 (5, 7).The ability of S. pombe to grow when the Tyr1, Ser2, Thr4, and

Ser7 residues are uniformly replaced by a nonphosphorylatableside chain resonates with transcriptome analysis showing thatonly a small fraction of fission yeast mRNAs are dysregulated byCTD phospho-site mutations (8). How do we reconcile this scenariowith the strong conservation of the consensus YSPTSPS heptad

Significance

The phosphorylation pattern of the Pol2 carboxy-terminal do-main (CTD) Y1S2P3T4S5P6S7 repeats comprises an informa-tional code coordinating transcription and RNA processing. Weexploited fission yeast CTD phospho-site mutants and syntheticgenetic arraying to illuminate opposing roles for Ser7 andThr4 in transcription termination whereby: S7A elicits pre-cocious termination via cleavage-polyadenylation factor (CPF)subunits and Rhn1; and T4A reduces termination and is lethalabsent CPF subunits Ppn1 and Swd22. The findings that Y1F,S2A, and T4A are concordantly lethal with ppn1Δ and swd22Δprovide insights into a CTD vocabulary, implicating Tyr1-Ser2-Thr4 as a three-letter CTD word. This work underscores howthe effects of mutating ostensibly inessential CTD marks aregenetically buffered by other cellular factors that are func-tionally redundant to those marks.

Author contributions: A.M.S., S.S., and B.S. designed research; A.M.S. and B.S. performedresearch; A.M.S., S.S., and B.S. analyzed data; and S.S. and B.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence may be addressed. Email: s-shuman@ski.mskcc.org orbschwer@med.cornell.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1810711115/-/DCSupplemental.

Published online October 24, 2018.

E10652–E10661 | PNAS | vol. 115 | no. 45 www.pnas.org/cgi/doi/10.1073/pnas.1810711115

Dow

nloa

ded

by g

uest

on

May

19,

202

0

and the core assumption (to which we subscribe) that many es-sential steps in gene expression rely on phospho-CTD cues? Todate, the four inessential CTD phospho-sites have not been cor-related with specific events in fission yeast RNA biogenesis.Our hypothesis is that the effects of mutating these phospho-

sites are genetically buffered by other cellular factors that arefunctionally redundant to the phospho-mark or the side-chainhydroxyl. By identifying such functional redundancies and gaugingtheir specificity for a particular phospho-site mutation, we expectto discern a pattern indicating which steps in gene expression relyon particular CTD coding letters. Here we address this via syntheticenhancement genetics, entailing the screening of a select collectionof S. pombe genes for ones that, when deleted or mutated, are le-thal in an rpb1-CTD mutant strain background.In this study we focus on genetic connections between the CTD

and factors implicated in 3′-end formation and/or transcriptiontermination. We do so in light of previous findings that CTDphospho-sites Thr4, Ser5, and Ser7 govern the expression of fissionyeast genes involved in phosphate homeostasis (6, 8–11). The S.pombe phosphate regulon comprises three genes that specify, re-spectively, a cell surface acid phosphatase (Pho1), an inorganicphosphate transporter (Pho84), and a glycerophosphate transporter(Tgp1) (12). Expression of pho1, pho84, and tgp1 is actively re-pressed during growth in phosphate-rich medium by the transcrip-tion in cis of a long noncoding RNA (lncRNA) from the respective5′ flanking genes prt, prt2, and nc-tgp1 (6, 9–11, 13–15). A model forthe repressive arm of fission yeast phosphate homeostasis is thattranscription of the upstream lncRNA interferes with the expressionof the downstream mRNA genes by displacing the activatingtranscription factor Pho7 from its binding site(s) in the mRNApromoters that overlap the lncRNA transcription units (Fig. 1A)(9, 11, 16).Phosphate regulon expression under phosphate-replete condi-

tions is governed by the phosphorylation state of the Pol2 CTD. Forexample, CTD mutations that prevent installation of the Ser7-PO4or Ser5-PO4 marks derepress pho1 and pho84 in phosphate-replete

cells. By contrast, prevention of the Thr4-PO4 mark hyperrepressespho1 and pho84 under phosphate-rich conditions (6). Because suchCTD mutations do not affect the activity of the lncRNA or mRNAgene promoters per se (9, 11), it is proposed that CTD status affectsPol2 termination, and thus the propensity to displace Pho7 from thedownstream mRNA promoter, during lncRNA synthesis (Fig. 1A).Specifically, it is hypothesized that loss of the Ser7-PO4 or Ser5-PO4marks leads to precocious termination of prt lncRNA transcriptionbefore the pho1 promoter and that loss of the Thr4-PO4 mark re-duces prt termination and hence increases transcription across thepho1 promoter (Fig. 1A) (9).This scenario begets several predictions. First, mutations of

factors that normally promote cotranscriptional 3′ processing andtranscription termination—these two events being functionallycoupled (17)—might hyperrepress pho1 under phosphate-repleteconditions. Second, synergies between CTD phospho-site mu-tants and 3′ processing/termination factor mutants might illu-minate instances of overlapping or antagonistic functions. Third,genetic interactions among fission yeast proteins implicated in 3′processing/termination and between these proteins and factorsinvolved in sculpting the CTD phosphorylation array might co-here into a CTD interaction network. Here we put these ideas tothe test.

ResultsOpposite Effects of Thr4 and Ser7 Mutations on pho1 Expression. Forthe present study, we employed a series of S. pombe rpb1 mutantstrains in which the native CTD length was maintained as29 heptads (4 rump and 25 consensus repeats) and Tyr1, Ser2,Thr4, or Ser7 in every heptad was replaced by Phe, Ala, Ala, andAla, respectively (SI Appendix, Fig. S1). The full-length rpb1-CTD mutant strains thrived on YES (yeast extract, glucose,amino acid supplement) agar medium at 30 °C (Fig. 1B). Testinggrowth across a range of temperatures showed that Y1F, S2A,T4A, and S7A cells grew slowly at 20 °C and that the S2A mutantgrew slowly at 37 °C (SI Appendix, Fig. S1). Western blotting ofwhole-cell extracts from midlog cultures grown at 30 °C withantibody against the Ser5-PO4 CTD mark, the only phospho-mark essential for fission yeast growth (5), showed that the lev-els of Ser5-phosphorylated Rpb1 were similar in WT, Y1F, S2A,T4A, and S7A cells (SI Appendix, Fig. S2). We surveyed the full-length rpb1-CTD mutants for effects on pho1 expression duringexponential growth at 30 °C under phosphate-replete conditions.Acid phosphatase activity [a gauge of Pho1 enzyme level thatcorrelates with pho1mRNA level, as assayed by primer extensionas well as RT-qPCR, RNA-sequencing (RNA-seq), and North-ern blotting; refs. 6, 8–11, and 18] was quantified by incubatingsuspensions of serial dilutions of the phosphate-replete cells for5 min with p-nitrophenylphosphate and assaying colorimetricallythe formation of p-nitrophenol. The repressed basal Pho1 ac-tivity of WT rpb1-CTD cells was hyperrepressed by sevenfold inT4A cells and was derepressed by fivefold in S7A cells (Fig. 1C).By contrast, basal Pho1 expression was unaffected by the Y1Fand S2A alleles (Fig. 1C).

Effects of Mutating Termination Factors and 3′-End–Formation Factors onpho1 Expression. A model proposed for CTD control of pho1 ex-pression invokes transcription termination during prt lncRNA syn-thesis as the tunable determinant of the activity of the flanking pho1promoter controlled by the DNA-binding transcription factor Pho7(Fig. 1A). A prediction of the model is that loss-of-function muta-tions in fission yeast proteins that promote cotranscriptional 3′processing and transcription termination ought to hyperrepressPho1 under phosphate-replete conditions by increasing the proba-bility of Pol2 traversing the pho1 promoter. To test this idea, wemonitored Pho1 expression in a series of fission yeast knockoutstrains lacking Dis2, Rhn1, Ctf1, Ppn1, or Swd22 as well as in astrain with a catalytically dead (C13S) version of Ssu72. The Ctf1,

Fig. 1. CTD control of phosphate homeostasis in phosphate-replete cells.(A) Models for the pho1 repressed/hyperrepressed (Upper) and derepressed(Lower) states of the prt–pho1 locus under phosphate-replete conditions. (B)Growth of S. pombe strains with the indicated full-length rpb1-CTD alleles(see SI Appendix, Fig. S1 for CTD amino acid sequences). Cells were in-oculated in YES broth and grown at 30 °C. Exponentially growing cultureswere adjusted to A600 of 0.1 and 3-μL aliquots of serial fivefold dilutionswere spotted on YES agar and then incubated at 30 °C. (C) The indicatedrpb1-CTD strains were grown to A600 of 0.5–0.9 in liquid culture in YESmedium at 30 °C. Cells were harvested, washed with water, and assayed forPho1 acid phosphatase activity by conversion of p-nitrophenylphosphate top-nitrophenol. Activity is expressed as the ratio of A410 (p-nitrophenol pro-duction) to A600 (input cells). Each datum in the bar graph is the average ofassays using cells from at least three independent cultures ± SEM.

Sanchez et al. PNAS | vol. 115 | no. 45 | E10653

GEN

ETICS

Dow

nloa

ded

by g

uest

on

May

19,

202

0

Ssu72, Ppn1, Swd22, and Dis2 proteins are constituents of the fis-sion yeast cleavage polyadenylation factor (CPF) complex (19).Whereas Ctf1 and Ssu72 are considered CPF core subunits, Dis2,Ppn1, and Swd22 comprise a heteromeric subassembly (the DPSmodule) that associates with the core but is not necessary for coreassembly (20). Ssu72 and Dis2 are protein phosphatase enzymes ofthe cysteinyl-phosphatase and binuclear metallophosphoesterasefamilies, respectively. Dis2 is an ortholog of the budding yeastGlc7 phosphatase, which is implicated in coupling 3′ processing byCPF to transcription termination (21). A recent precision nuclearrun-on sequencing (PRO-seq) analysis of the distribution of tran-scribing Pol2 in dis2+ and dis2Δ fission yeast cells showed that lossof Dis2 caused a global ∼500-bp 3′ extension of Pol2 distributiondistal to the poly(A) site, indicative of reduced termination effi-ciency (22). Rhn1 is the fission yeast homolog of the budding yeastCTD-binding transcription termination factor Rtt103 (23). Thedis2Δ, rhn1Δ, ctf1Δ, ppn1Δ, swd22Δ, and ssu72-C13S strains allgrew well on YES agar medium at 30 °C (Fig. 2C). [The ppn1Δ andswd22Δ cells had a cold-sensitive (cs) growth defect at 20 °C, andrhn1Δ cells displayed a temperature-sensitive (ts) growth defect at37 °C (SI Appendix, Fig. S3A).]We found that Pho1 expression in exponentially growing cells

at 30 °C was uniformly hyperrepressed in dis2Δ, rhn1Δ, ctf1Δ,ppn1Δ, swd22Δ, and ssu72-C13S cells under phosphate-repleteconditions (Fig. 2A). In a previous analysis of the transcriptomeof swd22Δ cells, it was noted that only 47 genes were underex-pressed in the absence of Swd22 (20). From the perspective ofthe present study, it is noteworthy that the “swd22Δ-down” geneset included pho1 and pho84, both of which are repressed bytranscription of upstream flanking lncRNAs (11). Here we per-formed Northern analysis of pho1 and pho84 mRNA levels inWT, ctf1Δ, rhn1Δ, and ssu72-C13S cells and found that bothgenes were hyperrepressed in all three mutant strains (Fig. 2B).These results fortify the hypothesis that reducing the probabilityof Pol2 terminating during prt lncRNA synthesis leads to greaterinterference with the pho1 promoter.To address whether the CPF mutations might affect aspects of

pho1 mRNA biogenesis independently of prt lncRNA synthesis,we employed a plasmid-borne prt–pho1 reporter introduced intoCPF-WT and mutant strains in which the chromosomal pho1gene was deleted. In this experiment, we used a mutated versionof the prt–pho1 reporter construct in which the prt promoter isinactivated by nucleotide changes in the HomolD and TATA boxelements that drive prt lncRNA synthesis (SI Appendix, Fig. S3B)(9). This mutant reporter provides a readout of pho1 expressionfreed from interference by transcription of the flanking prtlncRNA. The Pho1 activity of the mutant plasmid in WT cellswas high (i.e., derepressed) and was unaffected by the ppn1Δ andswd22Δ deletions (SI Appendix, Fig. S3B). Pho1 reporter ex-pression was 17% less than the WT level in the dis2Δ and ssu72-C13S backgrounds and was 27% less than the WT level in ctf1Δcells (SI Appendix, Fig. S3B); these modest effects on Pho1 ex-pression uncoupled from prt lncRNA interference do not accountfor the fourfold to sevenfold decrements in Pho1 expression at theWT prt–pho1 locus in the CPF mutant strains (Fig. 2A). Thus, wesurmise that the hyperrepressive effects of CPF subunit mutationson pho1 expression at the native prt–pho1 locus are not caused byinhibition of pho1 mRNA biogenesis per se.

Derepression of Pho1 Expression by CTD-S7A Depends on CPFSubunits and Rhn1. The property of precocious termination dur-ing prt lncRNA synthesis proposed for the Pol2-S7A polymerasecould reflect either (i) an inherent change in the Pol2-S7Aelongation/termination balance in favor of termination (i.e., in-dependent of the cleavage/polyadenylation and termination fac-tors) or (ii) a change in the responsiveness of elongating Pol2-S7Ato the action of cleavage/polyadenylation and termination factors.These two models beget distinct predictions concerning epistatic

relationships between CTD-S7A and the CPF and Rhn1 mutants.To address this issue, we introduced the rpb1-CTD WT and rpb1-CTD-S7A alleles into the CPF subunit and Rhn1-mutant strains.The CTD-S7A allele did not affect their growth at 30 °C (Fig. 2C).We then assessed Pho1 expression under phosphate-replete con-ditions. The instructive findings were that dis2Δ, rhn1Δ, ssu72-C13S, ppn1Δ, and swd22Δ eliminated the derepression of Pho1elicited by CTD-S7A, and the effect of CTD-S7A was severely at-tenuated in ctf1Δ cells (Fig. 2A). Thus, the increase in Pho1 ex-pression in S7A cells requires CPF subunits and Rhn1, consistentwith model (ii) above.In contrast to the CPF and Rhn1 mutants, the derepression of

Pho1 expression by CTD-S7A was not attenuated in nab3Δ orsen1Δ cells (SI Appendix, Fig. S4), which lack the fission yeasthomologs (24) of the Nab3 and Sen1 subunits of the budding yeastNNS complex implicated in noncoding RNA transcription termi-nation (17), or in din1Δ cells that lack Din1 (SI Appendix, Fig. S4),an RNA pyrophosphohydrolase whose budding yeast homologRai1 interacts with the “torpedo” 5′–3′ exoribonuclease Rat1/Xrn1 (Dhp1 in fission yeast) that drives postcleavage transcription

Fig. 2. Role of termination factor Rhn1 and 3′-end–formation factors inpho1 expression in rpb1-CTD WT and CTD-S7A cells. (A) S. pombe cellsbearing the indicated rpb1-CTD alleles were grown in liquid culture at 30 °Cand assayed for acid phosphatase activity as described in Fig. 1C. Each datumin the bar graph is the average of assays using cells from at least three in-dependent cultures ± SEM. (B) Northern blot analysis of total RNA fromexponentially growing phosphate-replete WT, ctf1Δ, rhn1Δ, and ssu72-C13Scells. The RNA was resolved by agarose gel electrophoresis and stained withethidium bromide to visualize 28S and 18S rRNA (Right) before transfer tomembrane and hybridization with 32P-labeled probes specific for pho1 (Left)or pho84 (Middle). Annealed probes were visualized by autoradiography.The positions and sizes (in kilobases) of RNA size standards are indicated. TheNorthern blots shown are representative of three biological replicates of theexperiment, entailing isolation and analysis of RNA from three separatecultures of each of the fission yeast strains specified. (C) Growth of dis2Δ,rhn1Δ, ssu72-C13S, ctf1Δ, ppn1Δ, and swd22Δ strains with rpb1-CTD-WT orrpb1-CTD-S7A alleles. Serial fivefold dilutions of exponentially growing cellswere spotted on YES agar and then incubated at 30 °C. The pairs of WT andS7A strains were spotted on the same agar plate in every case. In the threeinstances shown in which the WT and S7A spottings are separated by a whitespace, other rows of cell spottings on the plate separating WT and S7Astrains were cropped out of the image.

E10654 | www.pnas.org/cgi/doi/10.1073/pnas.1810711115 Sanchez et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

0

termination and with termination factor Rtt103 (25). Moreover,unlike the CPF and Rhn1 mutants, nab3Δ and sen1Δ mutationsdid not cause hyperrepression of Pho1 expression in CTD WT cells(SI Appendix, Fig. S4). The din1Δ mutation resulted in a modest40% decrement in Pho1 activity in CTDWT cells. These resultssuggest that the proposed precocious prt lncRNA termination isnot dependent on Nab3, Sen1, or Din1.

Distance Between prt and pho1 Start Sites Is a Tunable Governor ofPho1 Expression. We hypothesize that the basal level of Pho1expression in phosphate-replete cells is affected by whetherPol2 terminates prt transcription before reaching the pho1 pro-moter. The efficiency of this termination event is, in principle,governed by two parameters: (i) the probability that Pol2 will ter-minate over any interval segment of template DNA of the prt gene;and (ii) the total DNA length that Pol2 must travel to displacePho7 from the pho1 promoter. The altered Pho1 expression causedby erasing Pol2 CTD phospho-sites or by mutations of CPF subunitsand Rhn1 (Figs. 1 and 2) presumably reflects a reset in the first ofthese two parameters. To address the role of the second parameter,we altered the length of the prt transcription unit and assessed ex-pression of the downstream pho1 gene. The experiments wereperformed using a plasmid-borne prt–pho1 reporter cassette in-troduced into otherwise WT fission yeast cells in which the chro-mosomal pho1 gene was deleted (11). This reporter faithfullyreflects the homeostatic controls on the native pho1 locus (9). Asdepicted in SI Appendix, Fig. S5A, a 392-nt internal segment of theprt gene (from nucleotides +472 to +864 relative to the prt start site)was inverted (to maintain gene length but alter the sequence of thetemplate and nascent RNA), deleted (to shorten the distance to thepho1 promoter), or tandemly duplicated and triplicated (to increasethe distance to the pho1 promoter). Whereas segmental inversionhad a modest effect on Pho1 activity (a 35% reduction) underphosphate-replete conditions, deletion of the segment resulted in12-fold hyperrepression, and duplication and triplication of thesegment elicited incremental threefold and fivefold derepression ofPho1 activity, respectively, compared with theWT prt–pho1 reporter(SI Appendix, Fig. S5B). These effects accord with the predictionthat a longer transcription interval leads to a higher fraction oftermination within that interval (and hence less interference withthe pho1 promoter), and vice versa.

Mapping 3′ Polyadenylated Ends of prt lncRNA. Previous charac-terization of the prt lncRNA included mapping of the tran-scription initiation site and detection in rrp6Δ cells (lacking asubunit of the nuclear exosome) of a prt–pho1 read-throughtranscript extending from the prt initiation site to the pho1 poly(A)site (6, 13, 14). A key prediction of our prt termination-centricmodel of the control of pho1 expression by CTD and CPF/Rhn1is that the nascent prt transcript is polyadenylated and termi-nated at sites upstream of the pho1 promoter. However, to ourknowledge, there has been no prior demonstration of a specificpolyadenylated prt RNA other than the prt–pho1 read-through[notwithstanding that a ChIP-seq analysis found that the essen-tial fission yeast termination factor Seb1 is present across the prttranscription unit and photoactivatable ribonucleoside-enhancedcross-linking and immunoprecipitation (PAR-CLIP) methodsdocumented association of Seb1 with the 5′ segment of the prtRNA (26)].Here, to query the existence of terminated prt transcripts, we

mapped poly(A) sites in prt RNA by 3′-RACE using as templatetotal RNA isolated from (i) rpb1-CTD-S7A cells expressing prt–pho1 at its natural chromosomal locus (four independent cDNAclones sequenced); (ii) rpb1-CTD-S7A pho1Δ cells bearing theprt–pho1 reporter plasmid (six independent cDNA clones se-quenced); and (iii) rpb1-CTD-WT pho1Δ cells bearing the prt–pho1 reporter plasmid (eight independent cDNA clones se-quenced). Fourteen of eighteen cDNAs (78%) had an identical

junction to a poly(A) tail 351 nt downstream of the prt tran-scription start site (SI Appendix, Fig. S6A). The predominant prtpoly(A) site is located 10 nt downstream of the first nucleotide ofa fission yeast ATTTTT polyadenylation signal (PAS) (27). TheATTTTT hexanucleotide is distinctive among fission yeast PASmotifs in that it is situated closer to the cleavage site than thecanonical AATAAA element (27). Four additional prt poly(A)junctions were each recovered once, at positions +282, +325,+333, and +338 of the prt transcript (SI Appendix, Fig. S6A).To better interrogate the existence of the polyadenylated prt

transcript suggested by 3′-RACE, we performed a Northern blotanalysis of RNAs isolated from three independent cultures ofrpb1-CTD-WT pho1Δ cells bearing the prt–pho1 reporter plas-mid (SI Appendix, Fig. S7, WT lanes). Probing the blot witha 32P-labeled ssDNA complementary to the segment of the prtRNA from nucleotides +159 to +198 [i.e., upstream of the +351poly(A) site identified by 3′-RACE] highlighted two classes of prttranscripts: an ∼2.5-kb RNA corresponding to the prt–pho1 read-through transcript and an ∼0.4-kb species (labeled “prt PAS” inSI Appendix, Fig. S7, Left) that we surmise corresponds to prtRNA that was cleaved and polyadenylated at the +351 site.Probing with a 32P-labeled ssDNA complementary to the seg-ment of the pho1 mRNA from nucleotides +83 to +115 alsohighlighted two distinct RNA classes: a predominant ∼1.5-kbspecies corresponding to the pho1 mRNA and a minority speciescorresponding to the longer prt–pho1 read-through transcript (SIAppendix, Fig. S7, Right, WT lanes). No discrete pho1 RNAssmaller than the mature mRNA were detected by Northernblotting. Taken together, the 3′-RACE and Northern blot anal-yses affirm that the prt lncRNA is terminated at a site well up-stream of the pho1 promoter, albeit clearly not with 100%efficiency, as indicated by the prt–pho1 read-through transcript.

Mutating prt poly(A) Signals Attenuates pho1 Derepression by CTD-S7A. If S7A leads to precocious termination of prt lncRNA syn-thesis via a CPF-mediated pathway of prt polyadenylation at the+351 cleavage/polyadenylation site, then we might expect thatmutation of the ATTTTT PAS immediately upstream of the+351 site would have an impact on the derepression of pho1expression in phosphate-replete CTD-S7A cells and perhaps onbasal pho1 expression in phosphate-replete CTD WT cells. Totest this idea, we changed the PAS hexanucleotide to CTCGAGin the prt–pho1 reporter resident in rpb1-CTD-WT pho1Δ andrpb1-CTD-S7A pho1Δ cells and gauged acid phosphatase activity(SI Appendix, Fig. S6B). In CTD WT cells, the PAS mutationresulted in a 69% decrement in Pho1 expression. In CTD-S7Acells, the PAS mutation attenuated the derepression of Pho1, sothat acid phosphatase activity was reduced by 56% comparedwith the WT PAS. These results implicate utilization of the+351 cleavage/polyadenylation site as a modulator of lncRNArepression of the flanking pho1 gene.The finding that the PAS mutation did not entirely eliminate

the derepressive effect of S7A (i.e., unlike the CPF subunit andRhn1 mutations) suggested that there might be additional cleavage/polyadenylation sites at distal positions in the prt transcription unit.However, because such putative polyadenylated prt transcriptswould be subject to rapid intranuclear decay triggered by a clusterof cis-acting determinants of selective removal (DSR) elementslocated at +399 to +438 of the prt transcript (9, 13, 28), they wouldnot be readily detected by 3′-RACE. To address this point, weperformed 3′-RACE using RNA isolated from a fission yeaststrain that lacks Mmi1 (29), the RNA-binding protein that rec-ognizes the DSR motifs and elicits RNA decay. Analysis of thePCR-amplified poly(A)+ prt cDNAs from mmi1Δ cells revealed afragment longer than that detected in mmi1+ cells. Sequencing of13 independent cDNA clones showed that 12 had a junction to apoly(A) tail 589 nt downstream of the prt transcription start site,whereas one clone had a junction 1 nt 5′ of the other 12 cDNAs

Sanchez et al. PNAS | vol. 115 | no. 45 | E10655

GEN

ETICS

Dow

nloa

ded

by g

uest

on

May

19,

202

0

(SI Appendix, Fig. S6C). These prt poly(A) sites are located 22 and21 nt downstream of the first nucleotide of a canonical fissionyeast AATAAA PAS (27), hereafter referred to as “prt PAS-2,”and 151 nt downstream of the last DSR motif of the prt DSRcluster (SI Appendix, Fig. S6C).Additional evidence for 3′ processing and termination of prt at

the +589 poly(A) site was obtained by Northern analysis ofRNAs isolated from three independent cultures of rpb1-CTD-WT pho1Δ cells bearing a previously characterized prt–pho1 re-porter plasmid in which the clustered DSR motifs in the prt genewere mutated to eliminate Mmi1 binding (SI Appendix, Fig. S7)(9). The DSR mutations result in 10-fold lower Pho1 acidphosphatase expression and a fivefold increase in prt RNA (asgauged by RT-qPCR) under phosphate-replete conditions, andthey strongly attenuate the pho1 response to phosphate starva-tion (9). Here we found that the DSR mutations resulted in theappearance of an ∼0.6-kb prt transcript (referred to as “prtPAS2” in SI Appendix, Fig. S7) that we surmise correspondsto prt RNA that was cleaved and polyadenylated at the +589poly(A) site. The DSR mutations also elicited an approximatelyfourfold increase in the abundance of the prt–pho1 read-throughtranscript (detected with either prt or pho1 probes) and a dec-rement in the level of the pho1 mRNA (SI Appendix, Fig. S7).Mutating the prt PAS-2 AATAAA hexamer to GGCGGG

resulted in a 50% reduction in Pho1 expression in CTD WT cellsand a 29% decrement in Pho1 levels in CTD-S7A cells (SI Ap-pendix, Fig. S6B). Combining the ATTTTT (PAS) and AATAAA(PAS-2) prt mutations additively reduced Pho1 expression by 84%in CTD WT cells and 78% in CTD-S7A cells (SI Appendix, Fig.S6B). Taken together, these genetic experiments and RNA analyseshighlight 3′ processing and termination of prt lncRNA synthesisupstream of pho1 via the canonical poly(A) pathway as a tunableinfluence on pho1 expression.

Pho1 Derepression by Csk1 and Cdk9 Kinase Mutants and Epistasiswith CPF Subunits and Rhn1. The protein kinase Csk1 is a negativeregulator of the phosphate response, as judged by the fact thatpho1 expression is constitutively turned on in csk1Δ cells underphosphate-replete conditions (Fig. 3A) (6, 12, 18). Csk1 is acyclin-dependent kinase (CDK)-activating kinase with severalphysiological targets, including Cdk9 (30, 31). The essentialCdk9 kinase, in a complex with its cyclin partner Pch1, catalyzesSer phosphorylation of the Pol2 CTD (31, 32). Csk1 stimulatesthe CTD kinase activity of Cdk9•Pch1 by phosphorylatingCdk9 on residue Thr212 of the activating “T-loop” segment. Theactivated Cdk9•Pch1 enzyme phosphorylates the Pol2 CTD atpositions Ser2 and Ser5 of the CTD heptad (31). The non-phosphorylatable Cdk9-T212A mutant protein is refractory to ac-tivation by Csk1 in vitro. The phosphomimetic Cdk9 mutationT212E enhances the kinase activity of recombinant Cdk9•Pch1 byabout threefold compared with the WT enzyme that had not beenactivated by Csk1 and also compared with the T212A mutant en-zyme. Nonetheless, the kinase activity of Cdk9-T212E•Pch1 isthreefold lower than that of WT Cdk9•Pch1 that had been activatedby Csk1 (31). Thus, Cdk9-T212E is a hypomorphic mutation.The fission yeast csk1Δ, cdk9-T212A, and cdk9-T212E strains

phenocopy each other with respect to derepression of Pho1 underphosphate-replete conditions (Fig. 3, black bars) (6). A key questionis whether derepression by Csk1 kinase deletion or hypomorphicCdk9 mutations entails effects on 3′ processing or termination.Epistatic relationships were illuminated by measuring the effectof csk1Δ, cdk9-T212A, and cdk9-T212E on Pho1 expression inthe various 3′ processing and termination factor-mutant back-grounds (Fig. 3).The derepression by csk1Δ was erased in the rhn1Δ and ssu72-

C13S strains, signifying that the protein phosphatase activity ofSsu72 and the presence of the CTD-binding termination fac-tor Rhn1 are critical for the effects of Csk1 loss on phosphate

homeostasis. By contrast, there was no decrement in csk1Δ de-repression of Pho1 in the dis2Δ strain (Fig. 3A), implying that theprotein phosphatase activity of Dis2 does not come into play forcsk1Δ dysregulation of Pho1. [The derepression by csk1Δ isthereby distinguished from derepression by CTD-S7A, which isdependent on Dis2.] The derepressive effect of csk1Δ wasmaintained in the ctf1Δ, ppn1Δ and swd22Δ strains with respectto the fold increase in Pho1 activity compared with low back-ground in the equivalent csk1+ strains (13-fold for ctf1Δ, 20-foldfor ppn1Δ, 13-fold for swd22Δ) but was diminished by two- to2.5-fold with respect to the absolute level of derepressed Pho1 inthe csk1Δ single mutant (Fig. 3A).The hierarchy of effects of processing/termination factor mu-

tants on the derepression of Pho1 by cdk9-T212A was akin tothat seen in csk1Δ cells, insofar as the cdk9-T212A phenotypewas similarly unaffected by dis2Δ and the impacts of the othermutations clustered into two groups comprising (i) ctf1Δ, ppn1Δ,and swd22Δ and (ii) rhn1Δ and ssu72-C13S (Fig. 3B). The dis-tinctions were that ctf1Δ, ppn1Δ, and swd22Δ did not antagonizederepression by cdk9-T212A, although they partially blunted thecsk1Δ phenotype, and rhn1Δ and ssu72-C13S cells maintainedderepression by cdk9-T212A, albeit to an intermediate level,whereas they virtually abolished derepression by csk1Δ. Theseresults suggest that the events that lead to Pho1 derepression incdk9-T212A cells are less acutely dependent on the ensemble ofprocessing/termination factors than the underlying events incsk1Δ cells.A different set of epistatic effects was seen for cdk9-T212E

than for cdk9-T212A. The derepression of Pho1 by cdk9-T212Ewas erased in rhn1Δ, ssu72-C13S, ppn1Δ, and swd22Δ cells andwas partially blunted in dis2Δ and ctf1Δ cells (Fig. 3B). Of thethree protein kinase alleles analyzed in Fig. 3, cdk9-T212E wasmost similar to CTD-S7A with respect to the effects of CPFsubunit and Rhn1 mutations on dysregulation of Pho1 expression.

Fig. 3. Pho1 derepression by Csk1 and Cdk9 kinase mutants and epistasiswith CPF subunits and Rhn1. S. pombe strains bearing the indicated csk1alleles (A) or cdk9 alleles (B) in combination with CPF subunit or Rhn1 mu-tations as specified were grown in liquid culture at 30 °C and assayed for acidphosphatase activity. Each datum in the bar graph is the average of assaysusing cells from at least three independent cultures ± SEM.

E10656 | www.pnas.org/cgi/doi/10.1073/pnas.1810711115 Sanchez et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

0

Synthetic Genetic Interactions of CTD Phospho-Site Mutants. Toquery genetic interactions of CTD phospho-site mutations withcleavage/termination factors and kinases, haploid fission yeaststrains with the full-length rpb1-CTD alleles Y1F, S2A, T4A, andS7A (marked with a 3′ flanking natMX gene) were mated tohaploid strains with null or missense mutations in CPF subunits,Rhn1, and protein kinases (marked with 3′ flanking kanMX,hygMX, or ura4 genes). The resulting heterozygous diploids weresporulated and, for each allelic pair, a random collection of 500–1,000 viable haploid progeny was screened by serial replica-plating for the presence of the flanking markers. A failure torecover any viable haploids with both markers while recoveringthe three other haploid progeny (the unmarked and the twosingly marked haploids) with the expected frequencies was takenas evidence of synthetic lethality between the CTD allele and thetest allele. (The lethal allelic pairs are indicated by red boxes inthe matrix shown in Fig. 4A.) The double-mutant haploids thatpassed selection were spotted on YES agar at 20–37 °C in par-allel with the component single mutants. Two allelic pairs thatgrew very poorly at 30 and 34 °C and that failed to grow at 37 or25 °C (classified as “very sick”) are denoted by yellow boxes inFig. 4A. The double mutants that thrived at 30 °C are denoted bylight green boxes in Fig. 4A. Double mutants that displayed anenhanced ts or cs growth defect vis-à-vis one of the single mu-tants are annotated as such in Fig. 4A.The results of the synthetic genetic array highlight allele-

specific interactions of CTD phospho-site mutations that provideinsights into the function of the CTD-PO4 marks (or the aminoacid side-chain hydroxyl groups) in fission yeast. To wit, we seethat S7A displays few or no synthetic growth defects with theCPF subunit or Rhn1 mutants (Fig. 2C), consistent with ourinferences from the phosphate homeostasis experiments thatS7A elicits precocious prt transcription termination and that theCPF subunit and Rhn1 mutants diminish prt termination. Bycontrast, T4A, which is imputed to antagonize termination inthe prt–pho1 system, is synthetically lethal in the absence ofPpn1 or Swd22 (Fig. 4A). Previous tiling-array studies of theswd22Δ transcriptome showed that the absence of Swd22 leadsto a defect in termination of a limited subset of S. pombe genes(20). We surmise from the synthetic lethality observed here thatThr4-PO4 (or the Thr-OH) and the Ppn1/Swd22 subcomplex ofCPF each play important but genetically redundant roles indirecting essential transcription termination events in fissionyeast. A salient outcome of the genetic array was that Y1F andS2A cluster with T4A with respect to their synthetic lethalitywith ppn1Δ and swd22Δ (and not with ctf1Δ, dis2Δ, or rhn1Δ)(Fig. 4A), implying that Tyr1-PO4 (or the Tyr-OH) and Ser2-PO4 (or the Ser-OH) also function in cleavage/polyadenylationand/or termination in a manner that is redundant to the Ppn1/Swd22 component of CPF. The strong genetic interactions ofCTD phospho-sites with CPF subunits contrasted with the lackof synergy between CTD mutations Y1F, S2A, T4A, or S7A andnab3Δ, sen1Δ, and din1Δ.In keeping with their opposing effects on phosphate homeo-

stasis, S7A and T4A diverged in their genetic interactions withthe Csk1 and Cdk9 kinases. A key finding was that S7A and thethree kinase mutants, each of which elicited derepression of prt-regulated pho1 expression in a manner that relied wholly or inpart on the subunits of the CPF complex and Rhn1, were syn-thetically lethal for the S7A cdk9-T212A and S7A cdk9-T212Eallelic pairs or were synthetically very sick in the case of the S7Acsk1Δ pair (Fig. 4A). Taken together, these results suggest thatSer7-PO4 (or the Ser-OH) and the Csk1 and Cdk9 kinasesfunction redundantly in preventing precocious termination. Bycontrast, the csk1Δ and two cdk9 mutants were viable in com-bination with T4A (Fig. 4A). We took advantage of these viabledouble mutants to probe epistatic relations between the kinase allelesthat derepress Pho1 and the CTD-T4A allele that hyperrepresses

Pho1 (Fig. 4B). The salient findings were that T4A “wins out” againstcsk1Δ and cdk9-T212E, i.e., the derepression elicited in the csk1Δand cdk9-T212E single mutants was erased in the csk1Δ T4A andcdk9-T212E T4A double mutants (Fig. 4B). T4A attenuated but didnot eliminate the derepression of Pho1 expression by cdk9-T212A(Fig. 4B). Whereas csk1Δ and cdk9-T212E were viable when pairedwith Y1F and S2A (albeit with varying conditional or synthetic sickgrowth defects), the cdk9-T212A allele was synthetically lethal withY1F and S2A (Fig. 4A). These results underscore the inferencesfrom Fig. 3B that the Cdk9 T212A and T212E mutations are notequivalent in their impact on the various cellular events in whichCdk9 participates.

Ppn1 and Swd22 Are Essential for Viability When the CTD Array IsShortened. In light of previous findings that the phosphataseactivity of Ssu72 becomes essential for fission yeast growth whenCTD length is shortened to 16 repeats (rump plus 12 consensusheptads) (6), we tested for mutational synergies between theseveral CPF subunit and Rhn1 mutants and our collection ofrpb1 alleles with serially truncated CTDs (33). In an otherwiseWT background, a 16-repeat rpb1-CTD allele has no effect onfission yeast growth (33). Here we found that (i) shortening theCTD array to 20 or 18 repeats in the ppn1Δ and swd22Δ strainselicited a slow-growth defect at 25 and 30 °C and a failure togrow at 34 and 37 °C and (ii) ppn1Δ and swd22Δ were lethal

Fig. 4. Synthetic genetic interactions of CTD phospho-site mutants. (A)Synthetically lethal pairs of alleles are highlighted in red boxes. Viabledouble mutants without a synthetic defect are indicated by plain greenboxes. Viable double mutants that displayed ts or cs defects are annotated assuch in their respective green boxes. The yellow boxes indicate severe syn-thetic growth defect (very sick). (B) S. pombe strains bearing the indicatedcsk1 or cdk9 alleles in combination with CTD WT or CTD-T4A alleles asspecified were grown in liquid culture at 30 °C and assayed for acid phos-phatase activity. Each datum in the bar graph is the average of assays usingcells from at least three independent cultures ± SEM.

Sanchez et al. PNAS | vol. 115 | no. 45 | E10657

GEN

ETICS

Dow

nloa

ded

by g

uest

on

May

19,

202

0

when the CTD was truncated to 16 repeats (Fig. 5). By contrast,growth of rhn1Δ cells at 20–34 °C was unaffected by shorteningthe CTD to 16 repeats. The 16-heptad CTD allele also had noeffect on the growth of dis2Δ cells at 20–37 °C. Growth of thectf1Δ strain at 20–37 °C was unaffected by CTD shortening to20 repeats; further shortening to 16 repeats had no effect ongrowth of ctf1Δ at 30–37 °C but did elicit a cs defect seen asfailure to thrive at 20 °C. Thus, like Ssu72, the Ppn1 andSwd22 subunits of CPF are distinctive in that they become es-sential for vegetative growth when the CTD array is curtailed.

S7A Overcomes the Requirement for Ppn1 and Swd22 When the CTD IsShort. A likely consequence of CTD truncation is that it pro-motes competition among the many cellular CTD-binding pro-teins and protein complexes for occupancy of the residual CTDheptads of the Pol2 elongation complex. Accordingly, the le-thality or impaired growth of ppn1Δ and swd22Δ when theheptad array is reduced to 16 or 18 repeats indicates a com-petitive disadvantage in CPF recruitment to the shorter CTDwhen Ppn1 or Swd22 is absent. Given the results of our phos-phate homeostasis experiments indicating that CTD-S7A exertseffects on prt termination opposite those of ppn1Δ and swd22Δ,we queried whether replacing Ser7 with Ala might amelioratethe requirement for Ppn1 and Swd22 for growth when the CTDis truncated. The salient findings were that the slow-growth andts defects of ppn1Δ and swd22Δ in the context of CTD shortenedto 18 repeats (rump plus 14 consensus heptads) were fully sup-pressed by mutating the 14 Ser7 positions to Ala (Fig. 5).However, a shortened 18-heptad CTD in which the 14 Ser7 and14 Thr4 positions were mutated to Ala (T4A+S7A; ref. 8) did notameliorate the slow-growth and ts defects of ppn1Δ and swd22Δ.Collectively, our results point to S7A as a gain-of-function mu-tation and to T4A as a loss-of-function mutation with respect toCTD interactions with components of the cleavage/polyadenylation/termination machinery.

Mutational Synergies Among 3′ Processing Factors and TerminationFactors. Haploids with mutations in the inessential CPF subunitsCtf1, Ssu72, Dis2, Ppn1, and Swd22 and the termination factorRhn1 were crossed in all pairwise combinations to test for syn-

thetic genetic interactions. The results, compiled in Fig. 6, pro-vide further insights into the organization of the fission yeast 3′processing/termination machinery. Previous studies indicatedthat CPF comprises a core complex (of which Ctf1 and Ssu72 aresubunits) plus a DPS module composed of Dis2, Ppn1, andSwd22 (20). A series of affinity-tag purification experiments insubunit-null strains showed that (i) the absence of Ppn1 resultedin failure of Dis2 and Swd22 to associate with one another andwith the CPF core; (ii) the absence of Swd22 did not affect theassociation of Ppn1 with Dis2 but resulted in failure of Ppn1•Dis2 toassociate with the CPF core; and (iii) the absence of Dis2 did notaffect the assimilation of Ppn1 and Swd22 with the CPF core (20).Here, we observed no synthetic growth defects in ppn1Δ swd22Δ,ppn1Δ dis2Δ, and swd22Δ dis2Δ double mutants (Fig. 6). Takentogether, the physical and genetic interactions indicate that the DPSmodule is dispensable en bloc for vegetative growth.The key findings in regard to the DPS module were that

ppn1Δ, swd22Δ, and dis2Δ were each synthetically lethal with thessu72-C13S allele that specifies a catalytically dead version of theSsu72 protein phosphatase (Fig. 6). Because the three DPSmutants all result in a failure to assimilate the Dis2 proteinphosphatase into the CPF complex, we surmise that the syntheticlethality of DPS mutants and ssu72-C13S signifies that (i) a CPF-associated protein phosphatase activity is essential for fissionyeast viability; and (ii) either Dis2 or Ssu72 can fulfill this re-quirement. This conclusion resonates with an earlier observationthat an ssu72Δ-null allele is synthetically lethal with an allele ofppn1 that is defective for interaction with Dis2 (20).We observed additional instructive synergies whereby ppn1Δ

and swd22Δ were each synthetically lethal with ctf1Δ and rhn1Δ(Fig. 6). By contrast, dis2Δ was viable (albeit ts and cs) whenpaired with rhn1Δ and was viable (although cs) when paired withctf1Δ (Fig. 6). Thus, the Ppn1•Swd22 portion of the modulebecomes essential for CPF function when Ctf1 is missing fromthe core, in a manner that does not rely on the protein phos-phatase component of DPS. Genetic redundancy of Ppn1 andSwd22 with Rhn1 underscores a role for Ppn1•Swd22 in tran-scription termination. Note that there was no synthetic growthdefect when ctf1Δ and rhn1Δ were paired (Fig. 6). The ctf1Δ andrhn1Δ alleles were each viable but were synthetically very sickwhen paired with ssu72-C13S (Fig. 6), a phenotype that wasclearly more severe than the effect of dis2Δ in the ctf1Δ andrhn1Δ backgrounds. These results indicate that the Ssu72 andDis2 phosphatases are not functionally equivalent, with Ssu72 beingthe more important for growth in the absence of Ctf1 or Rhn1.Finally, we found no synergies between CPF and Rhn1 mutants andnab3Δ, sen1Δ, or din1Δ.

csk1Δ Suppresses the Synthetic Lethality of dis2Δ ssu72-C13S. Thelethality accompanying simultaneous inactivation of any of sevenpairwise combinations of CPF subunits and/or Rhn1 (Fig. 6) is

Fig. 5. Ppn1 and Swd22 are essential for viability when the CTD array isshortened. Exponentially growing cultures of S. pombe ppn1Δ (Upper) andswd22Δ (Lower) strains bearing chromosomal rpb1-CTD alleles with 29, 26,20, or 18 CTD heptads were adjusted to A600 of 0.1, and aliquots of serialfivefold dilutions were spotted to YES agar and incubated at the indicatedtemperatures. As noted below the panels, ppn1Δ and swd22Δ were lethal inthe context of a 16-heptad CTD array. The slow-growth and ts defects ofppn1Δ and swd22Δ in the context of CTD shortened to 18 repeats were fullysuppressed by mutating the Ser7 positions of all consensus heptads to Ala.

Fig. 6. Mutational synergies among 3′ processing factors and terminationfactors. Synthetically lethal pairs of alleles are highlighted in red boxes. Vi-able double mutants without a synthetic defect are indicated by plain cyanboxes. Viable double mutants that displayed ts or cs defects are annotated assuch in their respective cyan boxes. The yellow boxes indicate a severe syn-thetic growth defect (very sick).

E10658 | www.pnas.org/cgi/doi/10.1073/pnas.1810711115 Sanchez et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

0

presumably the consequence of a severe termination defectimpacting the expression of essential S. pombe genes. We con-sidered the prospect that such a lethal termination defect in adouble mutant might be ameliorated by a third mutation thatexerts an opposite effect, e.g., the csk1Δ allele that we hypoth-esize promotes precocious termination in the prt transcriptionunit. To test this idea, we crossed viable double mutants of csk1Δplus ctf1Δ, rhn1Δ, or ssu72-C13S (Fig. 3A) with differentiallymarked single mutants ppn1Δ, swd22Δ, and dis2Δ and thenscreened by random spore analysis for viable triple mutants.Only in the cross of csk1Δ ssu72-C13S with dis2Δ did we recovera viable csk1Δ dis2Δ ssu72-C13S triple mutant. The csk1Δ dis2Δssu72-C13S strain grew as well as the csk1Δ ssu72-C13S doublemutant on YES agar at 30 °C and 34 °C (said growth beingslower than that of WT and of the dis2Δ single mutant) but had amore severe ts and cs defect than did csk1Δ ssu72-C13S at 37 °Cand 25 °C (SI Appendix, Fig. S8A). The finding that loss of theCsk1 kinase rescued the lethality incurred by simultaneous in-activation of the two protein phosphatase subunits of CPF for-tifies the model that Csk1 exerts an effect on terminationopposite that of Dis2 and Ssu72 and hints that the Dis2 andSsu72 phosphatase might act on an essential phosphoproteinsubstrate that is generated (directly or indirectly) by Csk1. Withrespect to Pho1 expression, we saw that the csk1Δ dis2Δ ssu72-C13S triple mutant phenocopied the csk1Δ ssu72-C13S doublemutant in erasing the derepression of Pho1 caused by csk1Δrather than maintaining the derepression seen in the csk1Δ dis2Δdouble mutant (SI Appendix, Fig. S8B).

Effect of Mutating the CTD-Binding Site of Rhn1. Fission yeast Rhn1is a homolog of the budding yeast phospho-CTD–binding tran-scription termination factor Rtt103 (23). Rtt103 is a “reader” ofthe Thr4-PO4 and Ser2-PO4 letters of the CTD code. Structures ofthe Rtt103 CTD interaction domain in complex with CTD-Thr4-PO4and CTD-Ser2-PO4 peptides have delineated the determinants ofCTD recognition (34–37), as depicted in Fig. 7A for the Rtt103complex with the CTD peptide PSY1S2P3(pT4)S5P6S7YSPTSPScontaining a single Thr4-PO4 heptad and a downstream unmodifiedheptad (36). Alignment of the amino acid sequence of the seg-ment of Rtt103 that comprises the CTD interface to the corre-sponding sequence of Rhn1 highlights the conservation of functionalgroups that engage the CTD (Fig. 7A). The CTD contacts in Rtt103involve Asn65 (H-bond to the Tyr1-OH of the phosphorylatedheptad), Gln70 (H-bond to the Ser7-OH of the phosphorylatedheptad), Arg116 (bidentate H-bonds to the pThr4 and Ser5 main-chain carbonyls), Arg108 (bidentate contacts to the pThr4 phosphateoxygens), Lys72 (H-bond to the Thr4-OH of the downstreamheptad), and Lys75 (H-bond to Ser7 of the downstream heptad)(Fig. 7A). The equivalent amino acids in Rhn1 are Asn63, Gln68,Arg114, Tyr106, Arg70, and Lys73. To assess the contributions ofCTD binding to Rhn1 function in fission yeast, we replacedthe chromosomal rhn1+ gene with a series of three rhn1 alleles,N63A-Q68A, R70A-K73A, and Y106A-R114A, in which pairs ofCTD-interacting side chains were mutated to Ala. The mutatedRhn1 proteins evidently had biological activity, insofar as theyrestored WT growth at 37 °C vis-à-vis the ts growth defect of therhn1Δ strain (Fig. 7B). However, in the ppn1Δ and swd22Δ ge-netic backgrounds, in which Rhn1 is essential for growth, therhn1-N63A-Q68A and rhn1-Y106A-R114A alleles were both lethal(Fig. 7C), signifying that Rhn1 contacts to the phosphorylatedCTD heptad are crucial for Rhn1 in vivo activity when the DPSmodule of CPF is compromised. The rhn1-R70A-K73A allele con-ferred a severe growth defect when paired with ppn1Δ or swd22Δ,i.e., the double mutants formed pinpoint colonies at 25 and 30 °Cand failed to grow at higher or lower temperatures. We surmise thatRhn1 contacts to the unmodified flanking CTD heptad are alsoimportant for Rhn1 function in vivo. The rhn1 N63A-Q68A, R70A-K73A, and Y106A-R114A mutations phenocopied rhn1Δ with

respect to elimination of the derepression of Pho1 expressionby CTD-S7A (Fig. 7D).

DiscussionThe present study highlights 3′ processing and termination as acontrol point in the lncRNA-mediated repression of 3′-flankinggene expression that underlies fission yeast phosphate homeo-stasis. In particular, the prt–pho1 locus is established as a sensi-tive readout of cellular influences on 3′ processing/terminationvia the degree of interference with Pho1 expression by prt lncRNAtranscription. By exploiting this system to interrogate the effects ofCTD phospho-site mutations and the genetic interactions of CTDmutations with 3′ processing and termination factors (summarizedin SI Appendix, Fig. S9), we provide insights into the functions ofSer7, Thr4, and Tyr1, the “orphan” letters (4) of the CTD code.

S7A Elicits Precocious Termination. The findings that derepressionof pho1 expression at the prt-pho1 locus by S7A is expunged byloss-of-function mutations of CPF subunits Ctf1, Ssu72, Ppn1,Swd22, and Dis2 and termination factor Rhn1 implicate Ser7status as a determinant of Pol2 termination; to wit, Pol2-S7A isprone to precocious termination during prt lncRNA synthesis.Our inference is that (i) Ser7-PO4 (or the Ser7-OH) normallyexerts a negative influence on the interaction of the 3′ process-ing/termination machinery with the Pol2 elongation complex (SIAppendix, Fig. S9) and (ii) said interaction is enhanced when allSer7 marks are changed to Ala. This idea is underscored by ourdemonstration that S7A rescues the severe growth defects caused byabsence of Ppn1 or Swd22 (present study) or a catalytically inacti-vated Ssu72-C13S (6) in the context of a shortened CTD heptadarray, i.e., S7A affords a gain of function for the core CPF complexthat lacks the Ppn1•Swd22 module.

T4A Reduces Termination. The CTD-T4A allele exerts a hyper-repressive effect on prt-pho1 opposite that of S7A. Indeed,hyperrepression of pho1 by T4A wins out in a CTD-(T4A+S7A)double mutant (8), consistent with the model that T4A reducestermination during prt lncRNA transcription either by dimin-ishing interaction of the 3′ processing/termination machinerywith the Pol2 elongation complex or rendering the elongationcomplex less responsive to the action of those factors. Theconcordant effects of T4A and processing/termination factormutations in negating the derepressive impact of S7A (Fig. 2Aand ref. 8), csk1Δ, and cdk9-T212E (Figs. 3 and 4B) furthersupport the notion that Thr4-PO4 (or the threonine-OH) acts asa positive effector of Pol2 termination (SI Appendix, Fig. S9).However, it is clear that the Thr4-PO4 mark is not essential perse for termination in fission yeast, based on the fairly benigneffect of T4A on vegetative growth and the finding that T4Aaffects the expression of only seven protein-coding genes (8). Ahighly instructive outcome in the present study is that T4A issynthetically lethal in the absence of the otherwise inessentialCPF subunits Ppn1 and Swd22. Thus, we propose that the Thr4-PO4 mark and Ppn1•Swd22 play critical but genetically re-dundant roles in termination in fission yeast.Rhn1, the homolog of budding yeast Rtt103, is the likely

reader of the Thr4-PO4 CTD coding letter with respect to ter-mination. Rhn1 is not a subunit of the fission yeast CPF complex(20), and it is inessential for vegetative growth. We find here thatrhn1Δ (like T4A) is synthetically lethal in the absence ofPpn1•Swd22 and that the activity of Rhn1 in sustaining cellgrowth in the ppn1Δ or swd22Δ backgrounds is abolished bymutations in its phospho-CTD–binding site. In the WT CPFcomplex, the CTD-T4A rhn1Δ double mutation (which simplylowers the restrictive temperature of the rhn1Δ allele to 34 °C)has little impact. These results would place Thr4-PO4 andRhn1 in the same subpathway of termination in fission yeast.Our data are consistent with recent studies in budding yeast that

Sanchez et al. PNAS | vol. 115 | no. 45 | E10659

GEN

ETICS

Dow

nloa

ded

by g

uest

on

May

19,

202

0

establish a physical association of Thr4-phosphorylated CTDwith Rtt103 (35, 36, 38), although in budding yeast the T4Amutation impacted termination specifically at small nucleolarRNA genes (35). A budding yeast T4V mutant strain exhibited amore general increase in Pol2 density distal to poly(A) sites ofprotein-coding genes (36).

Tyr1 and Termination. The fission yeast CTD-Y1F mutation hadno overt impact on pho1 expression under phosphate-repleteconditions, and hence there was no sign that the absence ofthe Tyr1-PO4 mark impacts lncRNA-mediated transcriptionalinterference. Instead, a persuasive genetic connection between theTyr1-PO4 mark or the Tyr hydroxyl and 3′ processing/terminationwas made via the findings that Y1F is synthetically lethal in theabsence of Ppn1•Swd22 or when the Ssu72 phosphatase is cata-lytically dead. We would attribute these synthetic phenotypes di-rectly to the Y1Fmutation and not to a possible indirect impact onthe adjacent Ser2-coding letter because (i) the spectrum of syntheticlethality of Y1F with three different CPF mutations is broader thanthat of the S2A allele and (ii) we showed previously that a Y1F allelehad no effect on the overall level of the Ser2-PO4 mark, as probedwith a phospho-specific antibody (7).Our results point to either Tyr1-PO4 or the Tyr hydroxyl acting

as a positive effector of Pol2 termination in fission yeast. Thisscenario is distinct from the negative role assigned to the Tyr1-

PO4 mark in budding yeast Pol2 termination, whereby CTD Tyrphosphorylation impairs recruitment of the termination factorsRtt103 and Pcf11 (39) and the removal of the Tyr1-PO4 mark bythe Glc7 protein phosphatase subunit of CPF as Pol2 traversesthe poly(A) site leads to engagement of Rtt103 and Pcf11 that thenleads to CPF-coupled termination (21). Whereas Glc7 is essentialfor viability in Saccharomyces cerevisiae, the orthologous CPFphosphatase subunit Dis2 is inessential in S. pombe. We show herethat loss of Dis2 is genetically buffered by the Ssu72 phosphatasesubunit of fission yeast CPF.Our invocation of a positive termination function for the Tyr1-

PO4 or the Tyr hydroxyl in fission yeast based on the Y1F geneticsis in tune with a recent report that replacement of three-fourths ofthe Tyr residues in the mammalian CTD with Phe resulted in ageneralized termination defect manifest as widespread read-throughtranscription over long distances 3′ of mammalian genes (40).

Tyr1-Ser2-Thr4 as a Three-Letter Word in the CTD Code. Our besthope was that a synthetic genetic array of CTD phospho-sitemutations in fission yeast would (i) identify or affirm the cellulartransactions to which individual CTD coding letters contributeand (ii) shed light on the vocabulary of the code, i.e., if twodifferent phospho-site mutations elicit highly overlapping muta-tional synergies, then the two CTD letters comprise a “word” thatregulates certain cellular transactions. The array experiments

Fig. 7. Effect of mutating the CTD-binding site ofRhn1. (A, Upper) Stereoview of the structure of theRtt103 CTD-interaction domain in complex with aCTD peptide containing Thr4-PO4 (Protein Data BankID code 5LVF). The CTD peptide is depicted as a stickmodel with gray carbons; the CTD amino acids arelabeled in plain font. Rtt103 side chains that contactthe CTD are shown as stick models with cyan carbonsand are labeled in italics. Hydrogen-bond interac-tions between Rtt103 and the CTD are indicated byblack dashed lines; an intramolecular H-bond fromSer2-OH to the Thr4 phosphate is indicated by amagenta dashed line. (A, Lower) Alignment of theamino acid sequence of the CTD-interacting segmentof Rtt103 to the equivalent segment of Rhn1. Aminoacids targeted for Ala scanning in Rhn1 are shadedand denoted by j. Other positions of side-chainidentity/similarity are indicated by dots above theRhn1 sequence. (B) Serial dilutions of the indicatedrhn1 strains were spot tested for growth on YES agarat the indicated temperatures. All rhn1 strains werespotted on the same agar plate in every case. Thewhite space between the WT and N63A-Q68A spot-tings indicates that a single intervening row of cellspottings (of a strain not relevant to the experiment)was cropped out of the image. (C) Mutational syn-ergies. Synthetically lethal pairs of alleles are high-lighted in red boxes. The yellow boxes indicate asevere synthetic growth defect (very sick). (D) S.pombe strains bearing the rpb1-CTD-S7A allele incombination with Rhn1 mutants as specified weregrown in liquid culture at 30 °C and were assayed foracid phosphatase activity. Each datum in the bargraph is the average of assays using cells from atleast three independent cultures ± SEM.

E10660 | www.pnas.org/cgi/doi/10.1073/pnas.1810711115 Sanchez et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

0

conducted here show that Y1F, S2A, and T4A have overlappingsynthetic lethalities with deletions of CPF subunits Ppn1 andSwd22 (SI Appendix, Fig. S9), suggesting that Tyr1-Ser2-Thr4 forma three-letter word in the CTD code that promotes Pol2 termi-nation (and mutations of which impair termination in certain ge-netic backgrounds). There was already a strong suggestion thatTyr1-Ser2 comprise a two-letter CTD word in fission yeast, basedon the observation that the majority of the protein-coding RNAsthat were affected in Y1F cells were coordinately affected inS2A cells (8). With respect to termination, we hypothesize thatthe “spelling” of the three-letter word is likely to be Tyr1-Ser2-Thr4(PO4), based on how the Rtt103/Rhn1 termination factorrecognizes the Thr4-phospho-CTD (Fig. 7A). The terminationfactor makes a hydrogen bond from a conserved Asn side chainto the CTD Tyr1 hydroxyl, and mutation of the Asn to Ala is det-rimental to Rhn1 function. We envision that changing the CTDTyr1 letter to Phe, thereby subtracting the hydroxyl group, hasthe same negative effect on Rhn1 activity. The Ser2 hydroxyl of

the phospho-heptad makes no contacts with Rtt103, but it doesdonate a hydrogen bond to the phosphate group of Thr4-PO4

(Fig. 7A), and it is likely that this interaction aids in establishingor stabilizing a conformation of the phospho-CTD that is suitablefor binding to the termination factor.

MethodsDetailed methods are provided in SI Appendix for (i) the construction offission yeast strains with Rpb1 CTD mutations; (ii) deletions of the ctf1, rhn1,dis2, ppn1, swd22, nab3, sen1, and din1 genes; and (iii) allelic exchanges atthe rhn1 locus. The strains used in this study and their relevant genotypesare listed in SI Appendix, Table S1. The methods for testing mutationalsynergies, assaying cell-associated acid phosphatase activity, prt-pho1 re-porter plasmids and assays, and RNA analyses via primer extension, 3′-RACE,and Northern blotting are described in detail in SI Appendix.

ACKNOWLEDGMENTS. We thank Robert Fisher for the dis2Δ strain. Thiswork was supported by NIH Grants R01-GM52470 and R35-GM126945.

1. Eick D, Geyer M (2013) The RNA polymerase II carboxy-terminal domain (CTD) code.Chem Rev 113:8456–8490.

2. Corden JL (2013) RNA polymerase II C-terminal domain: Tethering transcription totranscript and template. Chem Rev 113:8423–8455.

3. Jeronimo C, Bataille AR, Robert F (2013) The writers, readers, and functions of theRNA polymerase II C-terminal domain code. Chem Rev 113:8491–8522.

4. Yurko NM, Manley JL (2018) The RNA polymerase II CTD “orphan” residues: Emerginginsights into the functions of Tyr-1, Thr-4, and Ser-7. Transcription 9:30–40.

5. Schwer B, Shuman S (2011) Deciphering the RNA polymerase II CTD code in fissionyeast. Mol Cell 43:311–318.

6. Schwer B, Sanchez AM, Shuman S (2015) RNA polymerase II CTD phospho-sitesSer5 and Ser7 govern phosphate homeostasis in fission yeast. RNA 21:1770–1780.

7. Schwer B, Sanchez AM, Shuman S (2012) Punctuation and syntax of the RNA poly-merase II CTD code in fission yeast. Proc Natl Acad Sci USA 109:18024–18029.

8. Schwer B, Bitton DA, Sanchez AM, Bähler J, Shuman S (2014) Individual letters of theRNA polymerase II CTD code govern distinct gene expression programs in fissionyeast. Proc Natl Acad Sci USA 111:4185–4190.

9. Chatterjee D, Sanchez AM, Goldgur Y, Shuman S, Schwer B (2016) Transcription oflncRNA prt, clustered prt RNA sites for Mmi1 binding, and RNA polymerase II CTDphospho-sites govern the repression of pho1 gene expression under phosphate-replete conditions in fission yeast. RNA 22:1011–1025.

10. Sanchez AM, Shuman S, Schwer B (2018) Poly(A) site choice and Pol2 CTD Serine-5 status govern lncRNA control of phosphate-responsive tgp1 gene expression infission yeast. RNA 24:237–250.

11. Garg A, Sanchez AM, Shuman S, Schwer B (2018) A long noncoding (lnc)RNA governsexpression of the phosphate transporter Pho84 in fission yeast and has cascadingeffects on the flanking prt lncRNA and pho1 genes. J Biol Chem 293:4456–4467.

12. Carter-O’Connell I, Peel MT, Wykoff DD, O’Shea EK (2012) Genome-wide character-ization of the phosphate starvation response in Schizosaccharomyces pombe. BMCGenomics 13:697.

13. Shah S, Wittmann S, Kilchert C, Vasiljeva L (2014) lncRNA recruits RNAi and theexosome to dynamically regulate pho1 expression in response to phosphate levels infission yeast. Genes Dev 28:231–244.

14. Lee NN, et al. (2013) Mtr4-like protein coordinates nuclear RNA processing for het-erochromatin assembly and for telomere maintenance. Cell 155:1061–1074.

15. Ard R, Tong P, Allshire RC (2014) Long non-coding RNA-mediated transcriptional in-terference of a permease gene confers drug tolerance in fission yeast. Nat Commun 5:5576.

16. Schwer B, Sanchez AM, Garg A, Chatterjee D, Shuman S (2017) Defining the DNAbinding site recognized by the fission yeast Zn2Cys6 transcription factor Pho7 and itsrole in phosphate homeostasis. MBio 8:e01218–17.

17. Porrua O, Libri D (2015) Transcription termination and the control of the tran-scriptome: Why, where and how to stop. Nat Rev Mol Cell Biol 16:190–202.

18. Henry TC, et al. (2011) Systematic screen of Schizosaccharomyces pombe deletioncollection uncovers parallel evolution of the phosphate signal transduction pathwayin yeasts. Eukaryot Cell 10:198–206.

19. Roguev A, et al. (2004) A comparative analysis of an orthologous proteomic envi-ronment in the yeasts Saccharomyces cerevisiae and Schizosaccharomyces pombe.Mol Cell Proteomics 3:125–132.

20. Vanoosthuyse V, et al. (2014) CPF-associated phosphatase activity opposes condensin-mediated chromosome condensation. PLoS Genet 10:e1004415.

21. Schreieck A, et al. (2014) RNA polymerase II termination involves C-terminal-domaintyrosine dephosphorylation by CPF subunit Glc7. Nat Struct Mol Biol 21:175–179.

22. Parua PK, et al. (2018) A Cdk9-PP1 switch regulates the elongation-terminationtransition of RNA polymerase II. Nature 558:460–464.

23. Sugiyama T, Sugioka-Sugiyama R, Hada K, Niwa R (2012) Rhn1, a nuclear protein, isrequired for suppression of meiotic mRNAs in mitotically dividing fission yeast. PLoSOne 7:e42962.

24. Lemay JF, et al. (2016) The Nrd1-like protein Seb1 coordinates cotranscriptional 3′ endprocessing and polyadenylation site selection. Genes Dev 30:1558–1572.

25. Kim M, et al. (2004) The yeast Rat1 exonuclease promotes transcription terminationby RNA polymerase II. Nature 432:517–522.

26. Wittmann S, et al. (2017) The conserved protein Seb1 drives transcription terminationby binding RNA polymerase II and nascent RNA. Nat Commun 8:14861.

27. Mata J (2013) Genome-wide mapping of polyadenylation sites in fission yeast revealswidespread alternative polyadenylation. RNA Biol 10:1407–1414.

28. Kilchert C, et al. (2015) Regulation of mRNA levels by decay-promoting introns thatrecruit the exosome specificity factor Mmi1. Cell Rep 13:2504–2515.

29. Sugiyama T, Sugioka-Sugiyama R (2011) Red1 promotes the elimination of meiosis-specific mRNAs in vegetatively growing fission yeast. EMBO J 30:1027–1039.

30. Saiz JE, Fisher RP (2002) A CDK-activating kinase network is required in cell cyclecontrol and transcription in fission yeast. Curr Biol 12:1100–1105.

31. Pei Y, et al. (2006) Cyclin-dependent kinase 9 (Cdk9) of fission yeast is activated by theCDK-activating kinase Csk1, overlaps functionally with the TFIIH-associated kinaseMcs6, and associates with the mRNA cap methyltransferase Pcm1 in vivo.Mol Cell Biol26:777–788.

32. Pei Y, Shuman S (2003) Characterization of the Schizosaccharomyces pombe Cdk9/Pch1 protein kinase: Spt5 phosphorylation, autophosphorylation, and mutationalanalysis. J Biol Chem 278:43346–43356.

33. Schneider S, Pei Y, Shuman S, Schwer B (2010) Separable functions of the fission yeastSpt5 carboxyl-terminal domain (CTD) in capping enzyme binding and transcriptionelongation overlap with those of the RNA polymerase II CTD. Mol Cell Biol 30:2353–2364.

34. Lunde BM, et al. (2010) Cooperative interaction of transcription termination factorswith the RNA polymerase II C-terminal domain. Nat Struct Mol Biol 17:1195–1201.

35. Nemec CM, et al. (2017) Different phosphoisoforms of RNA polymerase II engage theRtt103 termination factor in a structurally analogous manner. Proc Natl Acad Sci USA114:E3944–E3953.

36. Jasnovidova O, Krejcikova M, Kubicek K, Stefl R (2017) Structural insight into recog-nition of phosphorylated threonine-4 of RNA polymerase II C-terminal domain byRtt103p. EMBO Rep 18:906–913.

37. Jasnovidova O, et al. (2017) Structure and dynamics of the RNAPII CTDsome withRtt103. Proc Natl Acad Sci USA 114:11133–11138.

38. Harlen KM, et al. (2016) Comprehensive RNA polymerase II interactomes reveal dis-tinct and varied roles for each phospho-CTD residue. Cell Rep 15:2147–2158.

39. Mayer A, et al. (2012) CTD tyrosine phosphorylation impairs termination factor re-cruitment to RNA polymerase II. Science 336:1723–1725.

40. Shah N, et al. (2018) Tyrosine-1 of RNA polymerase II CTD controls global terminationof gene transcription in mammals. Mol Cell 69:48–61.e6.

Sanchez et al. PNAS | vol. 115 | no. 45 | E10661

GEN

ETICS

Dow

nloa

ded

by g

uest

on

May

19,

202

0