The genomic region of the 3′ untranslated region (3′UTR) of PHO84, rather than the antisense RNA, promotes gene repression

Abstract PHO84 is a budding yeast gene reported to be negatively regulated by its cognate antisense transcripts both in cis and in trans. In this study, we performed Transient-transcriptome sequencing (TT-seq) to investigate the correlation of sense/antisense pairs in a dbp2Δ strain and found over 700 sense/antisense pairs, including PHO84, to be positively correlated, contrasting the prevailing model. To define what mechanism regulates the PHO84 gene and how this regulation could have been originally attributed to repression by the antisense transcript, we conducted a series of molecular biology and genetics experiments. We now report that the 3′ untranslated region (3′UTR) of PHO84 plays a repressive role in sense expression, an activity not linked to the antisense transcripts. Moreover, we provide results of a genetic screen for 3′UTR-dependent repression of PHO84 and show that the vast majority of identified factors are linked to negative regulation. Finally, we show that the PHO84 promoter and terminator form gene loops which correlate with transcriptional repression, and that the RNA-binding protein, Tho1, increases this looping and the 3′UTR-dependent repression. Our results negate the current model for antisense non-coding transcripts of PHO84 and suggest that many of these transcripts are byproducts of open chromatin.


INTRODUCTION
In both prokaryotes and eukaryotes, transcription occurs on both DNA strands (1)(2)(3)(4). The messenger RN A (mRN A) of a gene is regarded as the sense transcript while the noncoding transcript produced from the opposite DNA strand is regarded as antisense transcript (1)(2)(3)(4). In budding yeast Sacchar om y ces cer evisiae , 65% of DNA sequences on both strands are composed of non-coding sequences ( 5 , 6 ), while 98% of the human genome is composed of non-coding sequences ( 7 , 8 ).
Long non-coding RNAs (lncRNAs) are a major class of non-coding RNAs that are implicated in gene expr ession r egulation. LncRNAs ar e RN A pol ymerase II (RNAPII) products that lack an open reading frame and are longer than 200 nucleotides ( 9 , 10 ). LncRNAs undergo post-transcriptional modifications like 5 capping and 3 polyadenylation ( 11 ). They can also be spliced, giving rise to various isoforms with the potential for alternati v e functions ( 11 ). LncRNAs regulate gene expression through a variety of different mechanisms. Roles of lncRNAs range from epigenetic to transcriptional and post-transcriptional roles (12)(13)(14). For example, the lncRNA GClnc1 was found to favor specific histone modifications by binding, modulating, and coordinating the localization of WDR5 (a component of histone methyltr ansfer ase complex) and the histone acetyltr ansfer ase KAT2A complexes, thus promoting gastric carcinogenesis ( 12 , 15 ). LncRNAs can recruit transcription factors to their target promoters, as in the case of the lncRN A MALAT1 w hich recruits the Sp1 transcription factor to the promoter of LTBP3 . This leads to transcriptional activation of LTBP3 , a key gene in multiple myeloma, thus pr omoting disease pr o gression ( 13 , 16 ). LncRN As modulate alternati v e splicing of their target genes by directly interacting with splicing factors, hijacking them from binding their pre-mRNA targets ( 17 , 18 ). They can e v en act as microRNA (miRNA) sponges sequestering miRNAs in the cytoplasm and pre v enting them from binding their target mRNAs, thus affecting mRNA stability or translation ( 13 , 14 , 19 ).
LncRNAs are classified into four main categories based on their genomic locations: (i) intergenic, which are transcribed from intergenic regions and are called long intergenic non-coding RNAs (lincRNAs); (ii) intronic, which are transcribed fr om intr ons; (iii) antisense, which are transcribed from the complementary strand of the mRNAcoding strand; (iv) bidirectional, which originate from the same promoter region of the mRNA but from the opposite strand going the opposite direction ( 20 ).
In this study, we focus on antisense lncRNAs, r eferr ed to as antisense transcripts for simplicity. About 30% of human genes express antisense transcripts that are implicated in gene regulation ( 20 , 21 ). In budding yeast, the percentage of genes with overlapping transcription of sense and antisense transcripts is similar to that of the human genes (this study). Non-coding RNAs pr oduced fr om antisense transcription are generally unstable ( 22 , 23 ). Ther efor e, studies investigating non-coding RNA roles in budding yeast have been performed in mutants to stabilize those transcripts ( 3 , 22-28 ). This raises the question of whether those observed roles are innate to those transcripts or if they arise as a consequence of deleting genes encoding regulatory factors.
Dbp2 is a major RNA helicase that belongs to DEADbox protein family ( 2 , 29 , 30 ). Studies indicate that DEADbox proteins exhibit diverse biochemical activities in vitro , such as RNA duplex unwinding, RNA folding, and RNP remodeling (31)(32)(33). Dbp2 has been linked to transcriptional regulation in budding yeast through various mechanisms. It modulates RNA structure and plays a role in transcriptional termination ( 30 , 34 ). It was also reported to function in transcriptional fidelity and to r epr ess aberrant transcription initiation ( 35 ). Published studies from our group provided the first evidence that the galactose ( GAL ) gene cluster-associated lncRNAs ( GAL lncRNAs) function as transcriptional inducers via R-loop formation, an activity regulated by Dbp2 ( 2 ). This direct link between Dbp2 and lncRNAs has prompted us to investigate the global role of Dbp2 on sense / antisense correlation.
PHO84 is one example of the S. cerevisiae genes suggested to have a r eciprocal expr ession pattern of its sense and antisense transcripts. The antisense transcripts of PHO84 are 2 lncRNAs that have been previously reported to suppress the sense transcription of PHO84 ( 3 , 28 ). The sense transcript of PHO84 was reported to be repressed upon deletion of the nuclear exosome component RRP6 and subsequent stabilization of the antisense transcripts of PHO84 ( 3 , 26-28 ).
In this study, we revisited the role of the yeast PHO84 antisense transcripts and found that the model of antisensemediated r epr ession does not always hold true. Instead, we report that the 3 UTR of PHO84 constitutes a regulatory element to PHO84 sense transcript. Additionally, we provided a list of factors that promote the 3 UTR-dependent regulation of PHO84 . Our results increase our understanding of transcriptional regulatory networks, and shed light on outstanding questions in the field of RNA biology.

Yeast strain construction
Strains were constructed using standard yeast genetics methods. Primers used for making PCR products used for homologous recombination are listed in Table S4. For PHO84 3 UTR Δ strain, the Delitto Perfetto technique was used as described before ( 36 ) to precisely delete the first 100 bp of the 3 UTR. Primers used for Delitto Perfetto technique are listed in Table S5. For the Synthetic Genetic Arr ay (SGA) str ains used in this study, a strain carrying ho Δ::AgSTE3pr -hygR m utation was obtained from Charles Boone lab and was further mutated to carry can1::LEU2 and either pho84::HIS3 or pho84::HIS3 3 UTR Δ mutations. Strain genotypes and oligos used are listed in Tables S2 and S4.

Plasmids and cloning
The EGFP plasmid was constructed by cloning EGFP from pYM28 into p415-ADH yeast e xpression v ector using XbaI, XmaI sites. EGFP-PHO84 3 UTR plasmid was constructed by first inserting the first 100 bp of the 3 UTR of PHO84 downstream of EGFP in pYM28 using the BamH1 site and then cloning EGFP-PHO84 3 UTR into p415-ADH yeast e xpression v ector using XbaI, XmaI sites. To insert the PHO84 promoter region into the two plasmids described above, 725 bp upstream of the PHO84 start codon were used to replace the ADH promoter upstream of EGFP using SacI, XbaI sites. Plasmid details are listed in Table S3 and primers used are listed in Table S8. For testing the trans effect of the lncRNA, site-directed mutagenesis was used to mutate three binding sites of Pho4. Plasmid details are listed in Table S3 and primers used are listed in Table S10.
Functional network analysis was performed using the ClueGo plug-in (Version 2.5.7) of the Cytoscape software (Version 3.5.0) on the 27 positi v e hits mapped to the GO term 'transcription by RN A pol ymerase II' (GO:0006366). ClueGo analysis settings include checking GO biological process, GO molecular function, and KEGG, with network specificity being set to medium and showing only pathways with pV < 0.05.

Serial dilution spot assay
Log phase cultures of pho84::HIS3 and pho84::HIS3 3 UTR Δ reporter strains were serially diluted with sterile water to make fiv e serial dilutions in a 5 × increments. 5 l aliquots were plated on YPD or on a selecti v e medium lacking histidine and having 25 mM 3-amino-1,2,4-triazole (3-AT) which is a known inhibitor of His3 function. Plates were incubated at 30 • C until growth was visible.

Northern blotting
Northern blotting was performed as described previously ( 37 ). gDNA was used as PCR templates for northern blotting pr obes. Pr obes labeled with 32 P-dCTP were generated from PCR templates using the Decaprime II kit according to manufacturer's instructions (Invitro gen). RN A samples were run on an agarose-formaldehyde gel and then transferred to a nylon membrane. After hybridization, Northern blots were quantified by densitometry using ImageQuant TL v7.0 (GE). Transcript le v els were determined relati v e to the SCR1 Control. Primers used for making Northern probes are listed in Table S6.

Yeast TT-seq:
Yeast str ains tr ansformed with an empty URA vector, pRS426, were grown at 30˚C in SD-URA + 2% glucose media to an OD 600 of 0.5. Cultur es wer e then tr eated with 4-thiouracil at a final concentration of 0.32 mg / ml for 10 min at 30˚C. Cultures were fast cooled in a dry ice / ethanol bath before pellets were harvested by centrifugation and stored a t −80˚C . RNA was extracted using a hot phenol protocol. RNAs were sonicated to an average size of 1 kb in a Covaris sonicator. 4-Thiouracil-containing RNAs were biotin-labeled using EZ-link biotin (Sigma) at a final concentration of 0.2 mg / ml for 2 h at room temperatur e. Tr eated RNAs were extracted twice with chloroform and ethanol precipita ted. Biotinyla ted RNAs were enriched using the MACS Streptavidin kit (Miltenyi) as per manufacturer's instructions. Enriched RNAs were ribodepleted using a Ribominus yeast / bacteria ribodepletion kit (Thermofisher) and size selected using an RNA Clean and Concentrator-5 kit (Zymo Research). Libraries were prepared using the NEBNext Ultra II Directional RNA Library kit (NEB) as per manufacturer's directions. Sequencing was performed on an Illumina NovaSeq (Novogene).

Analysis of sense versus anti-sense transcription
Coordinates for predicted non-coding transcripts (CUTs, SUT s and XUT s) ( 22 , 23 ), and coding genes ( 38 , 39 ) in Sacchar om y ces cer evisiae were collected. Genes and non-coding transcripts coordinates were converted into Granges (i.e. an efficient f ormat f or storing, manipulating, and aggrega ting genomic loca tions da ta) using the R-package Ge-nomicRanges (v1.38.0) ( 40 ). The join o ver lap inner () function from R-package Plyranges (v1.6.10) ( 41 ) was applied to determine genes and non-coding transcripts that have ov erlapping coor dina tes (a total of 2389 fea tures). Da ta wer e further filter ed to keep the featur es that ar e on opposite strands. These correspond to 2376 non-coding transcripts overlapping with 1669 yeast genes. Data were further filtered to keep the longest non-coding transcript per gene (complete table provided as supplementary table S11). Differ ential expr ession was determined (treatment versus control) using coordinates for 1669 non-coding transcripts and yeast genes in three independent datasets. For each dataset, the Pearson correlation between log (fold-change) at gene and non-coding transcripts le v el was determined using the char t.Corr elation () function from the R-package PerformanceAnalytics (v2.0.4) ( 42 ). The char t.Corr elation () function is equipped to calculate the correlation matrix based on Pearson's product-moment correlation coefficient as well as compute the statistical significance of the correlation results ( 42 ), followed by correlation visualizations using the R visualization package ggplot2 (v3.4.2).

Genetic screen
The pho84::HIS3 and the pho84::HIS3 3 UTR Δ reporters were constructed in a SGA-compatible y east str ain (Str ain genotypes are listed in Table S2). The SGA pho84::HIS3 reporter strain was crossed with an ordered array of ∼5000 strains from Yeast GST-tagged ORF collection (Horizon, YSC4423), each carrying a galactose-inducible over expr ession plasmid. After mating, diploids were selected and allowed to sporulate. Haploids carrying the pho84::HIS3 reporter and an ov ere xpression plasmid were then selected using the mating type-specific reporter ( AgSTE3pr-hygR), the plasmid marker, and the HIS3 auxotrophic marker. After se v eral pinning and selection steps, haploids were finally pinned on thr ee differ ent media: −Ura + 2% galactose, −Ura −His + 40 mM 3AT + 2% galactose and −Ura −His + 40 mM 3AT + 2% glucose. Colony sizes were compared and the ones with growth defect on the −Ura −His + 40 mM 3AT + 2% galactose media only were identified as potential hits. The same steps wer e r epeated after crossing the SGA pho84::HIS3 3 UTR Δ strain with an array of the potential positi v e hits identified from the initial screen. The hits with the galactose-specific growth defect that were completely rescued in the second screen were identified as true hits.

Chromosome conformation capture (3C)
100-ml cultures of wild-type and PHO84 3 UTR Δ strains were grown in SC + glu media to an OD 600 of ∼0.5 and then crosslinked with 1% formaldehyde and harvested. Cultures of wild-type strains harboring ov ere xpression plasmids were grown in −Ura + glu media to an OD 600 of ∼0.4, then shifted to −Ura + 2% raffinose for 2 h and then to −Ura + 2% galactose for 5 h and then crosslinked with 1% formaldehyde and harvested. Crosslinked chromatin was isolated by first cryolysing yeast cells and washing the Nucleic Acids Research, 2023, Vol. 51, No. 15 7903 resulting po w dered lysate twice with 1 × FA lysis buffer [50 mM HEPES ·KOH, pH 7.5; 140 mM NaCl; 1 mM EDTA; 1% Triton X-100] and the pellet was resuspended in 500 l 10 mM Tris ·HCl, pH 8.0. SDS was added to 50 l of the lysate at a final concentration of 0.1% and incuba ted a t 65 • C for 10 min to solubilize chromatin. Tubes were then placed immediately on ice for 5 min. SDS was quenched by adding Triton X-100 to a final concentration of 1%. Restriction digestion was performed overnight using 5 l (50 U) of NlaIII (NEB) in 1 × rCutSmart buffer (NEB) and then deactivated by heating at 65 • C for 20 min with 10 l of 10% SDS. Tubes were placed immediately on ice for 5 min and Triton X-100 was added to a final concentration of 1%. The mixture was diluted with water to a total volume of 750 l. An aliquot of 50 l was used to perform a digestion check and the ligation step was performed on the remaining 700 l. PCR using convergent primers (primer 4 & 5 listed in Table S9) was performed after RNase A treatment and phenol / chloroform extraction steps. Digestion was confirmed by the absence of a PCR product on an agarose gel. Ligation was performed by overnight incubation at 16 • C using 4.5 l (1800 U) of T4 DNA ligase (NEB) in 1 × T4 DNA ligase buffer (NEB). After incubation, 1 l (1 g) of DNase-free RNase A (Ambion) was added and incuba ted a t 37 • C for 30 min. SDS was added to a final concentration of 0.1% and 5 l of Proteinase K ( ≥800 U / ml, Sigma-Aldrich) was added and the mixture was incubated at 65 • C overnight to reverse the crosslinking. Samples were phenol extracted, ethanol precipitated and resuspended in 50 l nuclease-free wa ter. DNA concentra tion was determined using Nanodrop. PCR using di v ergent primers was then performed to detect interactions.

Strand-specific RT-qPCR
As described previously ( 2 ). Primers are listed in Table S7.

Sense / antisense pairs show positive correlation
The antisense transcripts of PHO84 have been suggested to silence sense transcription (Figure 1 A) ( 3 , 26-28 ). Dbp2, a major DEAD-box RNA helicase in S. cerevisiae , was reported to be involved in antisense-dependent gene regulation at the GAL gene cluster ( 2 ). In this study, we wanted to investigate the effect of DBP2 on the sense / antisense r egulation. Mor e specifically, we wanted to see how the loss of DBP2 would influence the sense and antisense transcript le v els of the genes with over lapping tr anscription. Tr ansient-tr anscriptome sequencing (TT-seq), a technique that allows for detection of nascent transcripts ( 43 ), was performed in wild-type and dbp2 Δ strains. This technique has not been previously employed to study the transcriptional impact of antisense RNA. For TT-seq, cultures were treated with 4-thiouracil for a short period of time and nascent transcripts tha t incorpora te the 4-thiouracil were then enriched and sequenced. We identified 1669 gene loci showing overlapping sense and antisense transcripts (Figure 1 B-D). We then performed differential gene expression analysis (DE) between wild-type and dbp2 Δ strains for sense and antisense transcripts. Interestingly, we found a strong positi v e correlation between sense and antisense transcript le v els with a correla tion coef ficient of 0.92 (Figure 1 B). Surprisingly, PHO84 also showed a sense / antisense positi v e corr elation (Figur e 1 B), which dir ectly contradicts the current model for antisense-mediated repression of PHO84 ( 3 , 26-28 ). To determine whether this observation is also observed at steady-state RNA le v els, we reanalyzed our prior published dbp2 Δ RNA-seq dataset ( 30 ). Consistent with the TT-seq, we saw a positi v e correlation in sense / antisense stead y-sta te RNA le v els (Figure 1 C). This was again true for PHO84 (Figure 1 C).
The PHO84 antisense transcripts have been proposed to function in trans , so we asked if deleting the cytoplasmic decay factor, XRN1 would influence sense / antisense correlation. Xrn1 is a 5 -3 RNA exonuclease that is known to regulate a class of non-coding RNAs termed XUTs, and loss of XRN1 has been suggested to enhance antisense-mediated r epr ession ( 23 , 44 ). Thus, we r e-anal yzed publicl y available RNA-seq data of xrn1 Δ strains ( 23 ). Interestingly, this data showed no global correlation between sense and antisense RNA le v els with a correla tion coef ficient of 0.00 (Figure 1 D). Since we see different correla tion pa tterns in dif ferent strains, this suggests that it is not the antisense transcripts tha t regula te the sense transcription, but rather the protein factors that play roles in the regulation of this locus. It is also possible that Xrn1-degraded RNAs that become stable after Xrn1 deletion play a role opposite to that of non-Xrn1dependent ones.

UTR deletion of the PHO84 sense transcript leads to increased sense expression at the PHO84 locus
The observation of sense / antisense positi v e correla tion a t the PHO84 gene locus in the dbp2 Δ strain prompted us to study the regulation of PHO84 more closely. PHO84 has been cited as a well-studied example of a gene that is r epr essed by its cognate antisense transcripts ( 3 , 26-28 ). Howe v er, our data showing positi v e correlation of sense / antisense transcripts suggests that the model of antisense-mediated r epr ession of PHO84 is not correct. To pinpoint the effect of the antisense transcripts of PHO84 on sense expression, we deleted 100 bp downstream of the PHO84 stop codon using the Delitto Perfetto technique to remove the previously published PHO84 antisense promoter site ( 3 ) (Figure 2 A). We then performed strand-specific re v erse transcription qPCR (ssRT-qPCR) for sense and antisense transcripts of PHO84 in wild-type and PHO84 3 UTR Δ strains. Surprisingly, we found that the deletion of the PHO84 3 UTR led to a significant 2fold increase in the sense transcript le v el without affecting the antisense transcript le v els (Figure 2 B, C). This suggests that elements within the 3 UTR of PHO84 are responsible for r epr ession, r ather than the antisense tr anscripts. It also suggests that the promoter of the antisense lncRNA is misannotated in the previous study ( 3 ).

SGA screen identified factors promoting 3 UTR-dependent r epr ession in pho84::HIS3 reporter
To understand how 3 UTR-dependent PHO84 gene regulation occurs, we constructed a reporter strain for PHO84 Genes and non-coding transcripts with overlapping coordinates and located on opposite strands were determined using R-packages. Differential expression analysis was performed on 1669 genes identified as having overlapping sense / antisense transcription. Sense / antisense correlation for the regions that are significant at FDR < 0.25 at gene le v el is shown. The correlation coefficient and statistical significance denoted by asterisks (*** equivalent to Pvalue < 0.001) are shown in the bottom right quadrant of the plot. ( C ) Correlation plot showing positi v e correlation between sense and antisense RNA stead y-sta te le v els in dbp2 Δ compared to wild-type. RNA-seq da ta was obtained from a prior stud y ( 30 ). Dif fer ential expr ession analysis was performed as above. Sense / antisense correlation for the regions that are significant at FDR < 0.25 at gene le v els is shown. The correlation coefficient and statistical significance denoted by asterisks (*** equivalent to p-value < 0.001) are shown in the bottom right quadrant of the plot. ( D ) Correlation plot showing no correlation between sense and antisense differential expression in xrn1 Δ compared to wild-type. RNA-seq data was obtained from a prior study ( 23 ). Differ ential expr ession analysis was performed as above. Sense / antisense correlation for the r egions that ar e significant at FDR < 0.25 at gene le v el are shown. The correlation coefficient is shown in the bottom right quadrant of the plot (no asterisks means non-significant). similar to that used in a prior study ( 28 ). To this end, we replaced the open reading frame (ORF) of PHO84 with the open reading frame of the auxotrophic marker HIS3 (Figure 3 A). We also constructed another reporter strain with the first 100 bp of the 3 UTR deleted. When HIS3 is expressed, cells can synthesize histidine and, ther efor e, can grow on a media lacking histidine. To confirm the functionality of the reporter, a serial dilution spot assay was performed using the pho84::HIS3 and the pho84::HIS3 3 UTR Δ strains (Figure 3 B). This resulted in a weak growth of the pho84::HIS3 strain on a media lacking histidine and having 3-amino-1,2,4-triazole (3-AT), which is a known inhibitor of His3 function ( 45 ). The deletion of the 3 UTR rescued the growth defect seen in the pho84::HIS3 strain, which confirms the observation that the 3 UTR is r epr es-si v e to PHO84 . We then performed ssRT-qPCR and found that the deletion of the PHO84 3 UTR leads to a 2-fold increase in the sense transcript le v el of the pho84::HIS3 reporter with no significant effect on the antisense transcript le v els (Figure 3 C, D). To further confirm that HIS3 expression in the pho84::HIS3 3 UTR Δ strain is higher than that of the pho84::HIS3 strain, a Northern blot was performed. This showed a ∼2-fold increase in pho84::HIS3 transcript le v els upon the deletion of the 3 UTR (Figure 3 E, F).
Next, we constructed a plasmid that allows for ectopic expression of the antisense transcripts. A plasmid carrying the pho84::HIS3 reporter was constructed and the three binding sites of the transcription factor Pho4 were mutated using site-directed mutagenesis to knockout sense transcription from the plasmid (primers listed in Table S10) showing the effect of deletion of the PHO84 3 UTR on the sense ( B ) and the antisense transcript le v els ( C ) . RNA was pr epar ed from 3-8 biological replicates of wild-type PHO84 and PHO84 3 UTR Δ strains grown in SC + glu medium. SsRT-qPCR was conducted to determine the RNA le v els of sense and antisense PHO84 . ACT1 was used as an internal control. A two-tailed P value was calculated using an unpaired t -test using GraphPad Prism 9. Error bars r epr esent the standar d de viation (s.d.), while ns denotes no significance ( P value > 0.05). dilution spot assay ( Figure S1) showed that the deletion of the 3 UTR rescues the growth of the pho84::HIS3 reporter strain on a media lacking histidine, as we showed earlier.
The deletion of the nuclear exosome component RRP6 led to suppression of the sense transcript in the pho84::HIS3 strain, which was also rescued by the deletion of the 3 UTR ( Figure S1). The ectopically expressed antisense transcripts did not have any effect on the growth of the reporter strain e v en when the RRP6 was deleted, which also questions the in trans model of these antisense transcripts ( 26 ) ( Figure S1).
Next, we employed a synthetic genetic array (SGA) screen to identify factors promoting the 3 UTR-dependent r epr ession (Figur e 4 ). The rationale of the screen was to use the yeast GST-tagged ORF collection (Horizon, YSC4423), w hich has a pproximatel y 5000 strains carrying each ORF under a galactose-inducible promoter in individual strains. These strains were mated with the two reporter strains abov e, and standar d y east genetics methods were employ ed to select for haploids encoding the reporter and the overexpression plasmid. We then grew the strains on galactose media (-His + 3AT) to induce ov ere xpression and on non-inducing glucose media for comparison. Strains were also grown on a galactose media that is not selecti v e for HIS3 to control f or an y indirect toxic effects of galactose. Plates were then scored for galactose-specific growth defects on the HIS3 -selecti v e medium. 211 hits were identified to be lethal in the pho84::HIS3 strain, while having no effect in the pho84::HIS3 3 UTR Δ strain (Table S1).
Gene Ontology (GO) term analysis was then performed using the Sacchar om y ces Genome Da tabase (SGD) Gene Ontolo gy Ma pper tool ( https://www.yeastgenome. org/goSlimMa pper ). Interestingl y, the GO term having the highest number of hits mapped to it (27 out of 211) was 'transcription by RN A pol ymerase II (GO:0006366)' (Figure 5 A). Functional network analysis was performed for those 27 hits using the ClueGo plug-in (Version 2.5.7) of the Cytoscape software (Version 3.5.0). This showed that hits are implicated in a number of di v erse transcriptional regulatory roles as well as DNA binding (Figure 5 B), suggesting a link between the 3 UTR and transcriptional regulation. For instance, Glc7 was one of the hits identified. Glc7 is a phospha tase tha t is involved in RNA PolII C-termainal domain (CTD) Tyr1 dephosphorylation which is essential for tran- SsRT-qPCR was conducted to determine the RNA le v els of sense and antisense pho84::HIS3 . ACT1 was used as an internal control. A two-tailed P value was calculated using an unpaired t-test using Graph-Pad Prism 9. Error bars represent the s.d. and ns denotes no significance ( P value > 0.05). (E, F) Northern blot of HIS3 expressed in pho84::HIS3 and pho84::HIS3 3 UTR Δ carrying an empty vector (EV), pRS426 ( E ) . Three biological replicates of cultures were grown in a -Ura + glu medium, shifted to -Ura + 2% raffinose for 2 hours, then shifted to -Ura + 2% galactose for 5 hours befor e harvesting. RNA was pr epar ed and Northern blot was performed. HIS3 and SCR1 bands were quantified by densitometry using ImageQuant TL v7.0 (GE) and the intensity of HIS3 bands relati v e to those of the loading control, SCR1, was plotted ( F ) . A two-tailed P value was calculated using an unpaired t -test using GraphPad Prism 9. Error bars represent the s.d. . Schematic r epr esentation of genetic screen to identify factors that promote 3 UTR-dependent r epr ession in the pho84::HIS3 reporter. The galactose-inducible ORF collection (Horizon, YSC4423) was crossed with the SGA pho84::HIS3 reporter strain. After selection for diploids, strains were pinned in liquid sporulation medium and allowed to sporulate. Se v eral pinning steps were then performed to select for haploids harboring pho84::HIS3 reporter and the ORF ov ere xpression plasmids. Strains were then pinned on galactose to induce expression and on glucose for comparison. Strains were scored for galactose-specific growth defects and those were identified as initial hits. A counter screen was then performed by crossing the SGA pho84::HIS3 3 UTR Δ strain with an array of the initial positi v e hits. All the pinning steps described above were repeated and strains were scored for galactose-specific growth defects. The strains whose growth was rescued in the counter screen were identified as true hits. scription termination ( 46 ). In addition, it associates with the cleav age / poly adenylation factor (CPF) ( 46 , 47 ). Tho1, an RNA-binding protein that functions in transcription elongation ( 48 ), was also identified from our screen. It binds transcribed chromatin in an RNA-dependent manner ( 48 ). Spt6, a well-known histone chaperone, was also identified from the screen. Spt6 was reported to associate with the 3 ends of transcribed genes and to interact with elongating RNA Pol II ( 49 , 50 ). Mor e r ecently, Spt6 was found to also play a critical role in controlling the fidelity and specificity of transcription initiation at thousands of genic and intragenic promoters ( 51 ).
To further confirm the screen results, we conducted ssRT-qPCR on 4 randomly selected hits from those mapped to the 'transcription by RN A pol ymerase II' GO term.
SsRT-qPCR of the sense transcript of the pho84::HIS3 and pho84::HIS3 3 UTR Δ reporter strains harboring some selected plasmids showed that the selected hits r epr ess the sense transcript onl y w hen the 3 UTR is present (Figure 6 ), which is consistent with the SGA screen results.

The 3 UTR is necessary but not sufficient to affect mRNA levels
To determine how the 3 UTR impacts PHO84 le v els, we constructed a reporter with the 3 UTR of PHO84 inserted downstream of an EGFP ORF on a yeast expression plasmid (Figure 7 A). We started by testing the EGFP-PHO84 3 UTR construct against the EGFP construct that has the wild-type 3 UTR of EGFP (Figure 7 B). RT-qPCR showed  RNA was pr epar ed from 3 biological r eplicates of each strain harboring an empty vector or a plasmid of each ORF under a galactose-inducible promoter. Cultur es wer e grown in -Ura + glu medium and shifted to -Ura + galactose for 17 h. SsRT-qPCR was conducted to determine the RNA le v els of sense pho84::HIS3 . ACT1 was used as an internal control. A two-tailed P value was calculated using an unpaired t-test using GraphPad Prism 9. Asterisk(s) denote significance ( P value < 0.05), and ns denotes no significance ( P value > 0.05). Error bars r epr esent the s.d. no significant difference in the transcript le v els between the two constructs, meaning that the PHO84 3 UTR is not sufficient to affect the mRNA le v els of PHO84 (Figure 7 B). Next, we replaced the ADH1 promoter with the PHO84 promoter region. This led to a significant, 2-fold difference in the mRNA le v els between the 3 UTR (+) and the 3 UTR (-) constructs (Figure 7 A, B). Thus, the PHO84 promoter also plays a role in 3 UTR-dependent regulation. This finding also confirms our initial observa tion tha t PHO84 sense transcript regulation is not linked to its antisense transcripts.
A recent stud y investiga ting the model of antisensemediated r epr ession of PHO84 suggested that the HIR histone cha perone complex, w hich is involved in de novo histone deposition, is involved in this process ( 28 ). This study showed that the loss of HIR2 , one of the four subunits of the HIR histone chaperone complex, leads to an increase in the sense transcript of PHO84 , an effect that is attenuated upon knocking down the antisense transcripts. In the study, it was suggested that Hir2 and the antisense transcript act together to mediate antisense-mediated r epr ession ( 28 ).
We wondered whether HIR2 deletion would influence the sense transcription in a way that is completely independent of the antisense transcripts. To test this, we conducted an RT-qPCR experiment using wild-type and hir2 Δ strains carrying the PHO84pr-EGFP-PHO84 3 UTR plasmid (Figure 7 C). Interestingly, despite the lack of antisense transcription in this construct, we saw an 8-fold increase in the EGFP expression in the hir2 Δ strain, showing that Hir2 plays a role in regulating the sense expression regardless of the presence of the antisense transcripts. This suggests that prior observations may have been the result of impacts on sense transcription rather than a role for the antisense transcripts. Similarly, we also tested the effect of deleting HIR2 on the EGFP expression from the PHO84pr-EGFP construct, and we also saw a significant increase in the sense transcript le v el (Figure 7 D). This shows that the role Hir2 plays in regulating the sense expression of PHO84 is not dependent on the 3 UTR and is solely dependent on the promoter region.

The PHO84 gene forms 3 UTR-dependent gene loops
Since the promoter region of PHO84 is a key player in 3 UTR-dependent gene r epr ession, we wonder ed whether a physical interaction between the promoter and terminator regions exists. Gene looping has been found to influence transcription both positi v ely and negati v ely (52)(53)(54). The most frequent effect of looping is transcriptional activation ( 52 , 53 ). In these cases, promoter-terminator regions were found to juxtapose each other during acti v e transcription to facilitate RNAPII re-initiation ( 52 , 53 ). Gene looping has also been found to correlate with transcriptional repression at BRCA1 gene, in which looping is thought to reduce promoter accessibility ( 55 ).
To test if the promoter and terminator regions of PHO84 interact, we conducted a chromosome conformation capture (3C) assay using the endogenous PHO84 gene in wildtype and PHO84 3 UTR Δ strains ( Figure 8 A). PCR using di v ergent primers was done using PHO84 and PHO84 3 UTR Δ strains. It showed gene looping between the promoter and terminator regions of PHO84 , as evidenced by obtaining a PCR product using di v ergent primers (Figure  8 B, C). It also shows that this interaction is less frequent when the 3 UTR is deleted (Figure 8 B-C); band intensity from the 3 UTR Δ strain was 58% of that of the wild-type (Figure 8 C). This indicates that the promoter and terminator regions of PHO84 interact and this interaction might play a role in the 3 UTR-dependent r epr ession seen in the PHO84 gene ( Figure 9 ).
We then sought to test the effect of the positi v e hits identified from the SGA screen on the PHO84 gene looping. We used the 4 selected hits previously confirmed to produce 3 UTR-dependent (Figur e 6 ). Inter estingly, we found that the ov ere xpression of THO1 lead to a significant increase in the PHO84 gene looping ( Figure S2), which again links r epr ession to increased looping frequency. Howe v er, no significant change in looping frequency was seen with the other three hits tested which suggests that their effect on PHO84 expression is achieved via alternative mechanisms.

DISCUSSION
Since 2007, the PHO84 gene has been cited as a model for antisense-mediated r epr ession ( 3 , 28 ). In line with this, the deletion of the nuclear exosome component, RRP6 , was found to stabilize the antisense transcripts of PHO84 and r epr ess the sense transcription ( 3 ). It was shown that the antisense transcripts silence sense transcription via recruiting the histone deacetylase Hda1, resulting in local chromatin compaction ( 3 ). This activity of the antisense transcripts is only observed when they are stabilized by the deletion of RRP6 . A mor e r ecent stud y suggested tha t HIR histone RT-qPCR performed using a wild-type yeast strain transformed with each of the four constructs . 3-4 biological replicates of each strain were grown in -Leu + glu medium, and shifted to -Leu + 2% raffinose and then shifted to -Leu + galactose then harvested. RNA was pr epar ed and RT-qPCR was performed for EGFP using ACT1 as an internal control. A two-tailed P value was calculated using an unpaired t-test using GraphPad Prism 9. Error bars r epr esent the s.d. while ns denotes no significance ( P value > 0.05). ( C , D ) RT-qPCR performed using wild-type and hir2 Δ strains carrying either the PHO84pr-EGFP-PHO84 3 UTR or the PHO84pr-EGFP plasmid. 3-4 biological replicates of each strain were grown and RT-qPCR was performed for EGFP as described above. A two-tailed P value was calculated using an unpaired t-test using GraphPad Prism 9. Error bars represent the standard deviation. chaperone complex is r equir ed for the antisense transcripts to silence sense transcription ( 28 ).
Most studies of antisense transcripts in S. cerevisiae have been conducted in mutant strains where antisense transcripts can be stabilized and / or extended ( 3 , 22 , 23 ). This is a caveat for characterizing the influence of those non-coding transcripts on the overlapping mRNAs, because the loss of those factors that regulate antisense transcripts might have either direct or indirect effect on the sense transcription itself.
Recent studies have also employed loss-of-function approaches to study the influence of antisense transcripts. RN A interference (RN Ai) using oligonucleotides that target antisense transcripts and strand-specific CRISPR interference (CRISPRi) have been among the main approaches for knocking down antisense transcripts ( 56 , 57 ). Howe v er, these techniques suffer from limitations such as off-target effects and incomplete strand specificity ( 56 , 57 ). Off-target effects have been a common issue with CRISPR / Cas9 techniques, and lack of strand specificity could be because the physical presence of a catal yticall y-dead Cas9 that would interfere with the transcriptional machinery and / or chroma tin structure a t the opposite strand ( 56 ). In addition, almost all studies to da te investiga ting the roles of antisense transcripts in S. cerevisiae have employed techniques that look into stead y-sta te le v els of RNA ( 22 , 23 ). This may not gi v e an accura te estima tion of transcriptional activity, as stead y-sta te le v els depend both on transcription and decay.
Taking PHO84 as an example, the fact that sense / antisense pairs were found to be positi v ely  . Proposed model for the 3 UTR-dependent r epr ession at PHO84 locus. Promoter-terminator region interaction at PHO84 locus is thought to be promoted by transcription factors, resulting in r epr ession. Those factors can be any known factor that promotes gene looping, which could be any of the identified hits from the genetic screen performed in this study. Tho1 was identified in this study to promote gene looping and to cause 3 UTR-dependent r epr ession. When the 3 UTR is deleted, some of those factors can no longer bind, resulting in less frequent interaction between the promoter and terminator regions leading to transcriptionally acti v e state. Other factors may still bind any region downstream of the 3 UTR resulting in looping that is not dependent on the 3 UTR. correla ted suggests tha t the antisense-media ted r epr ession model is not valid, and that other factors might be contributing, either directly or indirectl y. Additionall y, the previously proposed activity of the PHO84 antisense transcripts was only seen when cells were grown in minimal growth conditions after chronological aging or when RRP6 is deleted ( 3 ), raising concerns of whether these antisense transcripts have physiological roles in normal conditions. Here, we provide the first evidence that the 3 UTR of PHO84 , together with its promoter r egion, r egulates the sense transcript le v els of PHO84 , and that this regulation is independent on the antisense transcripts in healthy, wildtype cells. Additionally, we provide the first evidence that the PHO84 locus forms gene loops, and that interaction between the promoter and terminator regions is modulated by the 3 UTR, suggesting that the looping at this locus is r epr essi v e. Our data shows that the 3 UTR is not r equir ed for the looping to occur, howe v er its deletion significantly lowers the interaction frequency. With the data in hand that shows that Tho1, the RNA-binding protein that is known to bind transcribed chromatin ( 48 ), increases the looping and 3 UTR-dependent r epr ession, it's probable that some protein factors bind either the promoter region of PHO84 or the terminator region including the 3 UTR. It is possible that these factors interact with each other to help bring the promoter and terminator regions in close proximity and this in turn reduces promoter accessibility. This may explain why the loss of the 3 UTR does not completely abolish the looping. Further mechanistic studies are needed to decipher the link between the PHO84 gene looping and its r epr ession, to investigate whether the looping is r epr essi v e per se, or it's the trans -acting factors, such as Tho1, that bind the 3 UTR interfering with transcription efficiency.
Antisense RNAs are ubiquitous in nature ( 1 ). We suggest that r esear ch focused on transcriptional roles for antisense transcripts use methods that measure transcription directly rather than the steady-state RNA le v els. Moreov er, we suggest that caution should be used when determining a role for antisense transcripts solely from data using mutant strains. Approaches investigating the transcriptional roles of antisense RNAs in wild-type cells rather than mutants are immensely needed to avoid experimental bias arising from indirect effects.
Many factors known to be involved in transcriptional r egulation wer e identified in our SGA scr een and shown to regulate transcript le v els in a 3 UTR-dependent manner. Mor e studies ar e needed to better characterize the mechanism of these factors which can lead to a better understanding of the role of 3 UTRs in transcriptional regulation. Future studies using a combination of TT-seq and comparison with ChIP-seq may provide clarification of the mechanisms in play. Regardless, our studies provide evidence that the lncRNA-mediated r epr ession model at the PHO84 locus, and likely many more gene loci, is incorr ect. Futur e studies are necessary to understand the role(s) between transcriptional regulation and promoter-terminator looping.

DA T A A V AILABILITY
The data reported in this study is available under the GEO accession number: GSE227252.