tRNA Genes Affect Chromosome Structure and Function via Local Effects

The genome is packaged and organized in an ordered, nonrandom manner, and specific chromatin segments contact nuclear substructures to mediate this organization. tRNA genes (tDNAs) are binding sites for transcription factors and architectural proteins and are thought to play an important role in the organization of the genome.

ABSTRACT

The genome is packaged and organized in an ordered, nonrandom manner, and specific chromatin segments contact nuclear substructures to mediate this organization. tRNA genes (tDNAs) are binding sites for transcription factors and architectural proteins and are thought to play an important role in the organization of the genome. In this study, we investigate the roles of tDNAs in genomic organization and chromosome function by editing a chromosome so that it lacked any tDNAs. Surprisingly our analyses of this tDNA-less chromosome show that loss of tDNAs does not grossly affect chromatin architecture or chromosome tethering and mobility. However, loss of tDNAs affects local nucleosome positioning and the binding of SMC proteins at these loci. The absence of tDNAs also leads to changes in centromere clustering and a reduction in the frequency of long-range HML-HMR heterochromatin clustering with concomitant effects on gene silencing. We propose that the tDNAs primarily affect local chromatin structure, which results in effects on long-range chromosome architecture.

INTRODUCTION

The three-dimensional (3D) organization of the Saccharomyces cerevisiae nucleus is nonrandom (reviewed in references 1 and 2). Each chromosome occupies a specific territory in the nucleus anchored to nuclear substructures via specific DNA sequences. The telomeres of each chromosome tend to associate with one another and with the nuclear envelope in small clusters, based on the lengths of the chromosome arms (3–5). The ribosomal DNA (rDNA) repeats on chromosome XII are packaged into a dense structure known as the nucleolus, which also localizes to the nuclear periphery (6). Opposite the nucleolus is the spindle pole body, which is the interphase attachment site for the centromeres (CEN) of the 16 chromosomes (7). Attachment of centromeres to the spindle pole and attachment of telomeres to the nuclear membrane depending upon chromosome arm length help organize the nucleus (8). The active genes along the chromosome arms primarily reside in the nuclear interior, though some active genes, including some tRNA genes, interact with nuclear pores and help tether the arms (1, 9, 10).

Besides DNA sequence elements, numerous proteins play a role in nuclear organization via networks of interactions between nuclear-membrane and chromatin-bound proteins. Chromatin-bound proteins involved in this organization include heterochromatin proteins (11), lamin-like proteins (12–16), specific transcription factors (17, 18), RNA polymerases (6), and DNA repair proteins (19, 20; reviewed in reference 1).

tRNA genes (tDNAs) are a class of active genes found on all chromosomes and are bound by transcription factors TFIIIB and TFIIIC and RNA polymerase (Pol) III. tDNAs are short, highly transcribed DNA sequences (21) that are usually nucleosome free with strongly positioned flanking nucleosomes (22–25). The tDNAs contain internal promoter elements called A and B boxes, which aid in the binding of the transcription factor TFIIIC (26, 27). TFIIIC helps recruit TFIIIB to AT-rich sequences upstream of the tDNA. tDNA-bound transcription factors function via interactions with cofactors. tRNA genes are sites of binding for numerous chromatin proteins, including the architectural SMC proteins, nuclear pore proteins, chromatin remodelers, and histone modifiers. Studies from several laboratories have shown that tDNAs are enriched in cohesin (Smc1/Smc3) (28) and condensin (Smc2/Smc4) complexes (29, 30), as well as the SMC-loading proteins (Scc2/Scc4) (31, 32) and some chromatin remodelers, including RSC (22, 29, 33–35).

While individual tRNA genes turn over rapidly as a result of mutational inactivation and gene loss (36–38), a subset of tDNAs are syntenic with respect to neighboring sequences (39, 40), and data suggest that these conserved tDNAs possess chromosome position-specific functions in gene regulation (reviewed in references 41 and 42). There are several position-specific effects mediated by tDNAs. First, tDNAs have been shown to function as heterochromatin barrier insulators, which stop the spread of heterochromatic domains into adjacent nonsilenced domains (35, 39, 43, 44). Second, tDNAs block communication between enhancers and promoters when located between these elements in yeast, Drosophila, mouse, and human cells by acting as enhancer blockers (39, 45–50). Third, the presence of a tDNA in close proximity to an RNA Pol II-transcribed gene promoter antagonizes transcription from the Pol II-transcribed gene in a phenomenon referred to as tRNA gene-mediated (tgm) silencing (30, 51, 52).

In many organisms, tDNAs have also been shown to cluster at sites in the nucleus (39, 42, 53–55). In S. cerevisiae, DNA fluorescent in situ hybridization (FISH) studies have shown that some tDNAs cluster together adjacent to centromeres (52, 54), while proximity ligation analysis suggests that tDNAs cluster at the outer periphery of the nucleolus, as well as near the centromeres (10), though more recent genome-wide chromosome conformation capture (HiC) studies seem unable to detect these long-range associations (56). Based on these results, it has been proposed that TFIIIC binding to discrete sites along the chromosome plays an important role in chromosome folding and organization in the yeast nucleus (54, 57, 58).

To better analyze the role of tDNAs in chromatin looping and organization, we generated a “tDNA-less” chromosome through the systematic deletion of all the tDNAs on chromosome III in S. cerevisiae. We characterized the chromatin packaging, chromosome folding, and nuclear dynamics of this chromosome. We show that tDNA loss affects nucleosome positioning and loading of SMC proteins in the vicinity of tDNAs but that this has no effect on chromatin looping. While loss of the tDNAs does not affect chromatin looping, it does affect centromere clustering and the long-range interactions of the silenced HML and HMR loci with concomitant effects on gene silencing.

RESULTS

The ∼275 tDNAs in the budding yeast genome are dispersed across all 16 chromosomes. Here, we focus on chromosome III, which is 316 kb long and has two tDNAs on the left arm and eight tDNAs on the right arm. In order to investigate the roles of tDNAs in chromatin looping and nuclear organization and function, we created a strain in which chromosome III is devoid of any functional tDNAs by deleting an internal fragment of each tDNA. The deletions eliminate the internal promoter elements (both BoxA and BoxB) and thus eliminate the binding of the transcription factors TFIIIC and TFIIIB. For simplicity, we have labeled the tDNA adjacent to the HMR locus t0 and have labeled the remaining nine tDNAs, from right to left, t1, t2, t3, etc. To delete the tDNAs, we first replaced an internal segment of the gene with a URA3 gene and then subsequently replaced URA3 with a DNA fragment containing a unique DNA bar code. This involved multiple sequential transformations. Each deletion was monitored by PCR analysis, and intermediate strains were backcrossed to the wild-type (WT) strain W-303 prior to additional rounds of transformations. All of the experiments described were performed in this strain background to avoid strain-specific effects.

Most tRNA isoacceptor families have multiple copies scattered throughout the genome, though single gene copies code for six isoacceptor families. On chromosome III, 8 of the 10 tDNAs that were deleted are members of multicopy gene families (with 10 to 16 copies in the genome) and are not essential. However, tDNA t1 [tS(CGA)C] is a single-copy gene and is essential in S. cerevisiae (59), and there are only two copies of tDNA t7 [tP(AGG)C] in the genome. Loss of t7 from chromosome III caused cells to grow more slowly. In order to remove these two genes from chromosome III and simultaneously maintain the health of the yeast, we integrated single copies of the two genes on chromosome XV at the HIS3 locus. Once the full tDNA deletion chromosome III had been constructed, the strain harboring the chromosome was backcrossed with wild-type W-303, and segregation of the deleted tDNAs was monitored by PCR using primers specific to the unique bar codes. The sequence of this modified chromosome is available.

The strain in which chromosome III lacked any tDNAs (tDNA delete) was grown in rich medium at 30°C and did not show any obvious growth defect, forming homogeneous and healthy, smooth-edged colonies. Strains bearing this tDNA-less chromosome had a doubling time of ∼90 min in liquid yeast extract-peptone-dextrose (YPD) medium, which was indistinguishable from a wild-type strain. This is consistent with data showing that loss of one copy of multicopy tDNAs in yeast cells does not lead to growth defects in rich medium (60).

We analyzed the wild-type and tDNA mutant strains for sensitivity to various stresses. We grew haploid cells on plates containing increasing concentrations of hydroxyurea, benomyl, and caffeine. This analysis showed that the tDNA delete strain was as resistant to these drugs as the wild-type cells (Fig. 1).

FIG 1.

Drug sensitivity of wild-type and tDNA delete strains. Tenfold serial dilutions of cells starting at 10⁷ cells were spotted on YPD plates with different concentrations of various drugs and allowed to grow for between 2 and 5 days. HU, hydroxyurea.

Changes to the local nucleosome landscape surrounding tDNAs.

The stable binding of TFIIIC and TFIIIB, as well as their interactions with chromatin remodelers, results in nucleosome eviction at the tDNA and positioning of nucleosomes adjacent to the gene (22, 61). At some tRNA genes, a single nucleosome appears to be disrupted, while at other tDNAs, multiple nucleosomes are disrupted. Since tDNAs are dispersed across the chromosome and are highly transcribed, we first asked if loss of all 10 tDNAs from the chromosome altered the nucleosome and transcription landscape of the chromosome. In order to determine if tDNAs affect nucleosome positions across chromosome III, we mapped nucleosomes in our tDNA delete strain, as well as in the wild-type strain.

Haploid yeast cells were grown to log phase and harvested, and nuclei were digested with various concentrations of micrococcal nuclease (MNase) to generate mononucleosome-protected DNAs, which were subjected to paired-end MNase-assisted sequencing (MNase-seq). Overall, the nucleosome landscape across all chromosomes except chromosome III was unaffected by the presence or absence of the chromosome III tDNAs. More focused analysis showed no change in nucleosome positioning in the proximity of the 265 tDNAs scattered on the 15 chromosomes that were not manipulated in this study (Fig. 2A).

FIG 2.

Deletion of tDNAs leads to local changes in chromatin structure. (A) Comparison of nucleosome occupancy rates at 265 tDNAs on all of the yeast chromosomes except chromosome III (ChrIII). The tDNAs were aligned with respect to their transcription start sites (TSS) (set at 0). (B) Analysis of the nucleosome occupancy at tDNAs on chromosome III in the wild-type and tDNA delete strains. (C) Comparison of global nucleosome phasing on chromosome III in wild-type and tDNA delete cells. Shown are the average nucleosome dyad positions on 106 RNA Pol II-transcribed genes on chromosome III. These genes cover most of chromosome III. The genes were aligned with respect to their TSS (set at 0). The average nucleosome dyad density was set at 1. (D) MNase-seq data for wild-type and tDNA delete strains (normalized to the genomic average [i.e., 1]). Coverage plots are shown using all DNA fragments in the 120- to 180-bp range. The reference point (0) is the nucleotide marking the 5′ end of the deletion on chromosome III. Upstream of the deletion point at 0, the DNA sequence is the same in wild-type and the tDNA delete chromosomes III. Downstream of the deletion point, the DNA sequences are different. The arrows show the locations and orientations of the tDNAs in wild-type chromosome III. Meaningful plots could not be made for two tDNAs [tP(AGG)C and tS(CGA)C] because they were moved to another chromosome. Two other tDNAs [tM(CAU)C and tK(CUU)C] are present in S288C strains but are naturally absent in W-303 strains, including the strains used here. Orange, wild-type profile; blue, tDNA delete profile.

In contrast, changes in nucleosome occupancy were observed at or immediately adjacent to the deleted tDNAs on chromosome III. Figure 2B shows the average nucleosome occupancy across 2-kb segments centered on the chromosome III tDNAs, with each tDNA in WT cells aligned at its 5′ end, while in the tDNA delete strain, the 5′ ends of the deletion points were aligned. In the wild-type strain, there is a clear nucleosome-free region centered on the tDNA flanked by positioned nucleosomes reflecting differential digestion of the TFIIIB-TFIIIC complex relative to nucleosomes (35, 62). In the tDNA delete strain, this pattern is altered, and a nucleosome is usually formed over the deletion junction (Fig. 2D). We were unable to determine the change in the chromatin landscape around t1 and t7 tDNAs, since these two genes with 100 bp of flanking sequence were transposed to the HIS3 locus. Nucleosome positions elsewhere on chromosome III that were distant from the tDNAs were not altered on the tDNA-less chromosome (Fig. 2C). These results demonstrate that tDNAs create nucleosome-free regions at the tRNA gene with positioned nucleosomes flanking the gene. The data also show that their chromatin-organizing effects are locally confined and do not extend beyond their immediate vicinity.

tDNA loss affects expression of few RNA Pol II-transcribed genes.

The presence of a tDNA in close proximity to an RNA Pol II-transcribed gene promoter antagonizes transcription from the Pol II-transcribed gene, called tRNA gene-mediated silencing (tgm silencing) (30, 51, 52). In addition, tDNAs have also been shown to function as enhancer blockers when located between an upstream activation sequence (UAS) enhancer and a promoter (47). Since the loss of the tDNAs altered nucleosomes in their vicinity, we wondered if these alterations affected the transcription landscape of genes on chromosome III. Rather than restrict the analysis to Pol II-transcribed genes adjacent to the tDNAs on chromosome III, we investigated the effects of tDNA loss on all Pol II-transcribed genes in the genome and analyzed the changes in RNA levels in the wild type and the tDNA delete strain by transcriptome sequencing (RNA-seq). Total RNA was extracted from exponentially growing yeast cultures, and RNA-seq libraries were prepared, sequenced, and analyzed as described in Materials and Methods. The RNA levels of a very small number of genes were affected upon deletion of the tDNAs. Table 1 lists the genes that were either upregulated or downregulated in the strain lacking tDNAs on chromosome III. Of the 10 tDNAs present on chromosome III, tDNAs t0, t8, and t9 were flanked by retrotransposon elements, and since these are repetitive elements, the tDNA-mediated transcription effects could not be investigated for the loci. Furthermore, tDNAs t3 and t4 are missing in W-303. The expression of only two genes on chromosome III was affected, and in both instances, a tDNA (t1 or t6) was located adjacent to the gene. In one instance, the gene was upregulated upon tDNA loss, while in the second instance, the gene was downregulated. Furthermore, we observed the upregulation of the MRM1 gene. This gene resides immediately adjacent to HIS3. The tDNAs for t1 and t7 were ectopically inserted at the HIS3 locus in the tDNA delete strain, demonstrating that the ectopic insertion of the tDNAs was the cause of the change in expression of MRM1. These data suggest that tDNA-mediated position effects are highly context dependent and affect only some Pol II-transcribed genes and not others.

TABLE 1.

Genes whose mRNA levels changed in the tDNA delete strain compared to the wild type and statistical analysis of the differences in expression levels

Gene	Q value		Beta statistic	Gene product function
	Likelihood ratio test	Wald test
Upregulated
YCR061w	0.042003857	3.22E−11	0.6580998	Protein of unknown function
YDL124w	0.042003857	6.31E−14	0.4395261	NADPH-dependent alpha-keto amide reductase
YHR214c-B	0.005672182	2.35E−210	2.5870238	Retrotransposon TYA Gag and TYB Pol genes
YNL160w	0.040272092	2.69E−20	0.5348964 YGP1	Cell wall-related secretory glycoprotein
YOR201c	0.020396347	9.18E−27	0.7685378 MRM1	Ribose methyltransferase
YOR202w	0.009117145	1.20E−76	3.7326308 HIS3	Imidazoleglycerol-phosphate dehydratase
YPL240c	0.042003857	1.20E−15	0.5188043 HSP82	Hsp90 chaperone
Downregulated
YBR068c	0.030966243	1.53E−20	−0.5156825 BAP2	High-affinity leucine permease
YBR296c	0.039894621	1.87E−19	−0.5692597 PHO89	Plasma membrane Na+/Pi cotransporter
YCR008w	0.015662463	6.48E−43	−1.1217 SAT4	Ser/Thr protein kinase involved in salt tolerance
YER073w	0.045197971	8.81E−16	−0.5691673 ALD5	Mitochondrial aldehyde dehydrogenase
YER091c	0.042003857	1.79E−20	−0.5345605 MET6	Cobalamin-independent methionine synthase
YGL009c	0.009117145	1.47E−82	−2.0202761 LEU1	Isopropylmalate isomerase
YHR208w	0.007776891	7.72E−106	−1.3235331 BAT1	Mitochondrial branched-chain amino acid aminotransferase
YJR010w	0.042003857	3.34E−13	−0.8073784 MET3	ATP sulfurylase involved in methionine metabolism
YJR016c	0.017736878	3.09E−29	−0.5823368 ILV3	Dihydroxy acid dehydratase involved in biosynthesis of branched-chain amino acids
YKL030w	0.042003857	3.46E−11	−0.6718075	Dubious open reading frame
YKL120w	0.007776891	1.47E−94	−1.4546702 OAC1	Mitochondrial inner membrane transporter
YLR355c	0.042003857	2.35E−12	−0.387429 ILV5	Acetohydroxy acid reductoisomerase involved in biosynthesis of branched-chain amino acids
YMR108w	0.042003857	3.72E−15	−0.4213145 ILV2	Acetolactate synthase involved in isoleucine and valine biosynthesis
YOR271c	0.027535084	4.63E−26	−0.5903308 FSF1	Putative protein of the sideroblastic-associated protein family

Of the genes that were downregulated in the tDNA delete strain, several are involved in amino acid biosynthesis, though these genes are scattered throughout the genome and do not localize near tDNAs. The reason expression of these genes was reduced is unclear, given that the two yeast strains used are isogenic with respect to nutritional markers and there are between 10 and 16 copies of each of the six deleted tDNAs in the genome (t0, 11 copies; t2, 10 copies; t5, 16 copies; t6, 11 copies; t8, 10 copies; and t9, 15 copies). It is possible that there is a reduction in transcript levels of these genes due to the small reduction in the tDNA copy number without any other cell phenotype. This is consistent with a recent study where single tDNAs in yeast were deleted, and these single deletions in multicopy tDNA families also led to changes in the expression of a small set of genes involved in translation (60).

Scc2 binding at tDNAs is dependent upon a functional tDNA, but other binding sites are tDNA independent.

The SMC proteins play an important role in nuclear organization (63), and tDNAs are major binding sites for SMC proteins and the SMC loaders Scc2/Scc4 and Rsc. Our nucleosome-mapping data indicated that loss of the tDNAs altered nucleosome positions at tDNAs. Since nucleosome-free tDNAs are sites for the recruitment of RSC and Scc2/Scc4 proteins (31, 34, 64, 65), we asked if loss of all the tDNAs on chromosome III reduced recruitment of Scc2 proteins at these loci and whether it also affected loading of Scc2 at other sites along the chromosome.

We performed chromatin immunoprecipitation sequencing (ChIP-seq) of Myc-tagged Scc2 to compare the distributions of the protein genome-wide in the WT and tDNA delete strains (Fig. 3). This analysis showed Scc2 binding at multiple sites along the chromosome, including tDNAs. At some tDNAs, the Scc2 binding was focused, forming a sharp peak, while at other tDNAs, the binding was spread over a wider region. Comparison between the wild type and the tDNA delete strain showed that Scc2 levels did not decrease at any of the sites on the 15 chromosomes. Upon tDNA loss, Scc2 binding decreased at the tDNA loci on chromosome III or at sites in the immediate vicinity of tDNAs, such as LEU2 (adjacent to tDNA t8) (Fig. 3) and HMR (adjacent to t0). On chromosome III, the analysis also showed that there was no significant change in Scc2 binding at other non-tDNA sites on the chromosome. For example, we saw a large peak of Scc2 binding at Tel3L. This peak at Tel3L was unchanged upon tDNA deletion, and similarly, we did not record any change in Scc2 levels at centromere III (CEN3), confirming that tDNAs are not the sole determinants for the recruitment of Scc2 to chromosomes.

FIG 3.

Scc2 binding along chromosome III in the wild-type and tDNA delete strains. Shown is ChIP-seq mapping of Myc-Scc2. (Top) Distribution of Scc2 at tDNAs on chromosome III in wild-type cells (left) and tDNA delete strains (right). (Bottom) Distribution of Scc2 at 265 tDNAs on all chromosomes except chromosome III in the wild-type (left) and tDNA delete (right) strains.

We confirmed this result by ChIP-quantitative PCR (qPCR) against Scc2. A site at the OCA4 gene was used as an internal control, since the site does not bind Scc2 in wild-type cells. We were unable to design unique primers at t6 due to the presence of repetitive sequences in the immediate vicinity of the gene and therefore could not map the localization of these proteins at the tDNA. Some primer pairs flank the tDNAs, while others are adjacent to the tDNAs. Consistent with the ChIP-seq data, in wild-type cells, Scc2 was enriched at several of the tDNAs present on chromosome III (Fig. 4A). We observed ∼3.5-fold enrichment at t8 and ∼2.5-fold enrichment at t0, t2, and t5. When the same protein was mapped in the tDNA delete strain, we observed a significant reduction in Scc2 binding at these tDNAs. The levels dropped to those observed for the negative control OCA4 except for the t8 tDNA, where the level dropped 2-fold but there was some residual Scc2 still present (Fig. 4A). The amount of Scc2 did not change at CEN3 when the tDNAs were absent from the chromosome, indicating that the binding of Scc2 to the centromere was independent of the tDNAs.

FIG 4.

Scc2 and Brn1 binding at tDNAs on chromosome III. (A) ChIP-qPCR mapping of Myc-Scc2. The data show the distribution of Scc2 at specific sites along chromosome III in the wild-type and tDNA delete strains. The data are the results of two independent cross-links on which four IPs were performed. For each amplicon, the fold enrichment compared to input was first calculated, and the data were then normalized to the OCA4 locus. An unpaired t test assuming unequal standard deviations (SD) was used to test for significance of differences between the wild-type and tDNA delete strains. (B) ChIP-qPCR mapping of HA-Brn1 and condensin. Fold enrichment and statistical significance were calculated in the same way as for the Scc2 ChIP and normalized to the OCA4 locus. The error bars indicate SD.

Scc2, in association with Scc4, helps recruit the SMC proteins to chromatin (29, 32). Condensins localize to tDNAs and are necessary for the clustering of tDNAs in the nucleus (29, 30). We therefore mapped the binding of condensins at tDNAs on chromosome III using the hemagglutinin (HA)-tagged Brn1 subunit. In wild-type cells, the Brn1 profile was very similar to that previously observed for Scc2, with significant binding of Brn1 at specific tDNAs. Correspondingly, the binding of the condensins was significantly reduced at these sites upon deletion of the tDNA promoters (Fig. 4B).

Chromosome mobility on the tDNA-less chromosome.

TFIIIC binding sites and tDNAs are described as chromosome-organizing clamps because of their consistent association with specific landmarks within the nucleus (54). The localization of tDNAs with the kinetochore is dependent upon condensins, while the interactions of tDNAs with nuclear pores in the G₂ phase of the cell cycle are dependent upon cohesins. These associations likely help tether the chromosome. Since loss of tDNAs from chromosome III led to a decrease in SMC proteins at these sites, we wondered if this loss would affect chromosome tethering and the mobility of the chromosome. To assess mobility, we fluorescently labeled specific sites on chromosome III and used them to monitor chromosome mobility in the wild-type and tDNA deletion chromosomes. The location of a point on the chromosome was mapped in three-dimensional space over a defined period in relation to another point within the nucleus—the spindle pole body (marked with the Spc29-red fluorescent protein [RFP] fusion protein)—and mobility was characterized by mean square distance (MSD) analysis as described previously (66–68). Six chromosomal loci across chromosome III were assayed (Fig. 5). These loci were tagged by inserting LacO arrays at the sites and monitored using LacI-green fluorescent protein (GFP) fusion protein-mediated fluorescence. Time-lapse movies of >35 individual unbudded cells (in the G₁ phase of the cell cycle) were imaged over the course of 10 min. Z-stack images of the cells were taken every 30 s during the time-lapse movie, and the MSD was calculated at each time point using the following equation: MSD = (X_{t + T} − X_t)² + (Y_{t + T} − Y_t)², where X and Y are the coordinates of the fluorescent dot, T is the time lag, and t is the time. Using this information, MSD curves were generated for each locus in both the WT and tDNA delete strains (data not shown). For the wild-type chromosome III, CEN3 was the most constrained locus (radius of constraint [Rc] = 415 nm), with loci located further from the centromere exhibiting greater mobility. For example, LEU2, which is approximately 30 kb from the centromere, had an Rc of 522 nm, while HMR, which is approximately 180 kb from the centromere, had an Rc value of 688 nm. This is consistent with previous data showing that the location of a locus in relation to the centromere is critical in determining its mobility, with loci closer to the centromere displaying decreased mobility compared to loci farther from the centromere (9, 68, 69). Comparison of the mobilities of segments in the wild-type and tDNA delete strains showed a small decrease in mobility in the tDNA delete strain at two sites (SRO9 and CEN3). However, these differences were not statistically significant (P = 0.15 and 0.19). The data indicate that tDNAs are not major determinants in constraining chromosome arm motion or that they are a subset of the factors involved, and the redundancy precludes observation of their contribution.

FIG 5.

Effect of tDNA deletion on chromosome III mobility. Mean square displacement analysis of seven loci along chromosome III in wild-type and tDNA delete strains is shown. The box plots represent the data obtained from the MSD experiments. The components of the box plots are as follows: red lines, means; pink bars, 95% confidence intervals; purple bars, standard deviations; gray dots, individual values obtained from each cell analyzed. The arrowheads below the chromosome III schematic show the locations of the loci assayed. The Rc measurement was calculated from MSD graphs that were generated over the course of a 10-minute time-lapse movie. The experiments were the result of time-lapse images taken from at least 35 cells per locus assayed. A t test was used to determine the significance of differences observed between the wild-type and tDNA delete strains for each locus.

tDNAs are not required for proper chromatin folding.

tRNA genes have been proposed to affect chromatin fiber folding via the clustering of dispersed tRNA genes. The promoters in tDNAs are the binding sites for the transcription factor TFIIIC, and foci comprised of multiple TFIIIC-bound sites have been proposed to function in chromatin looping and folding (10, 39, 52–54, 57, 58, 70). If tDNAs are major drivers of chromatin folding and looping, then elimination of these loci from an entire chromosome should lead to changes in the folding of the chromatin fiber or result in changes in chromosome packaging in the nucleus. We set out to determine the detailed three-dimensional organization of chromosome III lacking functional tDNAs. We used a modified chromosome conformation capture technique called Micro-C XL (71, 72). We chose Micro-C XL over HiC because it can capture both short 3D interactions and some long interactions, and the method is not dependent upon the distribution of restriction sites along the DNA. In brief, yeast cells were first cross-linked with formaldehyde and disuccinimidyl glutarate (DSG), and chromatin was then fragmented into nucleosomes via micrococcal nuclease digestion. The cross-linked, digested chromatin was ligated to capture chromosomal interactions. Size-selected ligation products were then purified and subjected to paired-end high-throughput sequencing. The sequencing reads were mapped back to the reference genome to determine the interacting regions of the chromosome, as previously described. For each strain, two independent cross-linking experiments and ligations were performed, and QuASAR (quality assessment of spatial arrangement reproducibility) and Genome DISCO (differences between smoothed contact maps) were used to assess the reproducibility of the data. The two independent measurements both gave a score of >95%, confirming the reproducibility between the two biological replicates for the wild-type and tDNA delete strains. The reproducibility of the data was also analyzed by measuring contact probability over genomic distance, and the decay curves between replicates overlapped almost completely, which is consistent with our reproducibility measurements.

Overall, Micro-C maps for the wild-type and tDNA mutant strains both exhibited previously described features of yeast chromosome folding, with no difference in chromatin folding between the tDNA delete and wild-type strains. The chromatin interaction maps of chromosome IX showed that most Micro-C interactions occurred close to the diagonal in both strains, though there was significant variation in the density along the diagonal. The data clearly show ∼2- to 10-kb contact domains (chromosomally interacting domain [CIDs]/topologically associating domains [TADs]) encompassing ∼1 to 5 genes in both strains (Fig. 6A). We calculated the insulation score across bins, and a scatter plot of the insulation scores for wild-type and mutant strains was consistent with the conclusions that the overall architecture of chromosomes in the two strains was not altered (Fig. 6B).

FIG 6.

Micro-C interaction plots of chromosome III. (A) Snapshot of chromosome IX at 2.5-kb resolution and 1-kb resolution. The wild type and tDNA delete mutant are shown in the contact matrices. The bottom matrix is a view of a segment at 100-bp resolution, and the insulation scores are plotted above the contact matrix. The insulation score is the value obtained by calculating the number of contacts within a 10-kb-by-10-kb sliding window using the 1-kb-resolution contact matrix. The local minima identify boundaries that help demarcate CIDs. (B) Scatter plot of insulation scores showing that there was no change in insulation strength between the wild-type and tDNA delete strains. (C) Interaction decay curve for chromosome III showing that there were no significant changes between the wild-type and tDNA delete strains. (D) Circular chromosome conformation capture (4C)-type contact graph using a region immediately adjacent to the deleted tRNA gene. The 1D contact matrix at 5-kb resolution was plotted. The y axis is the log₂ ratio of differential contacts between the wild-type and tDNA delete strains.

Inspection of chromosome III in the wild-type and tDNA delete cells showed that these interaction domains persisted on the chromosome even upon loss of the tDNAs from chromosome III. The interaction decay curves for chromosome III were very similar in the wild-type and tDNA delete strains, indicating that the overall folding of the chromosome had not altered (Fig. 6C). There was no significant change in the contact frequency versus genomic distance in the two strains, indicating no local chromatin decondensation or change in chromatin-looping interactions.

We analyzed the contact frequencies of sites immediately adjacent to the 8 tDNAs on chromosome III in the wild-type and tDNA delete strains. At some sites, the loss of the adjacent tDNA did not alter long-range interactions at all, while at other sites there were small changes, though the significance of these remains to be elucidated (Fig. 6D).

Thus, tDNAs do not appear to be responsible for the general folding of the chromatin fiber and the CID/TAD architecture.

tDNAs affect CEN-CEN interaction frequency.

While the overall folding of the chromatin fiber of chromosome III was not altered, Micro-C analysis identified changes in contact frequency at specific sites along chromosome III. In order to identify sites where contact frequency had changed, the contact maps were normalized by distance for the wild-type and tDNA delete strains (Fig. 7A). These matrices were used to identify differential contact sites (generated by dividing the tDNA delete matrix by the wild-type matrix). This analysis identified increased sites of contacts in the tDNA delete strain (shown in red) around the centromere and near the telomeres of chromosome III (Fig. 7B).

FIG 7.

Micro-C analysis of the centromeres. (A) Contact map of chromosome III for the wild-type and tDNA delete strains normalized by distance (obs/exp, observed/expected). (B) Differential contact maps were generated by dividing the tDNA matrix by the wild-type matrix. Increased contacts in the tDNA delete strain are shown in red, and reduced contacts in the tDNA delete strain compared to the wild-type strain are shown in blue. Tel, telomere; Cen, centromere; HM, HML and HMR. (C and D) Graphs quantifying CEN-CEN interactions. (C) Graph examining the interaction of CEN16 with all other centromeres. The x axis shows the interaction counts of a 50-kb segment centered on each centromere. (D) Graph examining the interaction of CEN3 with all other centromeres. The x axis shows the interaction counts of a 50-kb segment centered on each centromere. The increase in the CEN3-CEN interactions in the tDNA delete strain was significantly greater (P = 1.22 × 10⁻¹⁴) than the values of all CEN16-CEN interactions (excluding CEN16-CEN3).

The 16 centromeres in yeast are in close physical proximity to one another and cluster adjacent to the spindle pole body (7, 73, 74). These CEN-CEN interactions are readily captured by chromosome conformation capture (3C) methods, including HiC (10, 71, 75), and are recapitulated in this study in the W-303 strain background. Interestingly, compared to the wild-type strain, the centromere of chromosome III in the tDNA delete strain showed an increased frequency of interactions with the other centromeres. Focusing on the 50-kb pericentric region of each chromosome, we found that most CEN-CEN interactions were minimally affected by the loss of chromosome III tDNAs. For instance, interactions between the chromosome XVI centromere and the remaining centromeres showed that interactions between CEN16 and the majority of centromeres remained unchanged but that there was an ∼20% increase in interaction strength between CEN16 and CEN3 when chromosome III lacked tDNAs (Fig. 7C).

This increase in CEN3 interaction was not confined to CEN16. When the same analysis was performed using CEN3 as an anchor, we observed increased frequency of interactions between CEN3 and all of the other chromosomal centromeres in the tDNA delete strain (Fig. 7D). Most of the interaction counts increased approximately 20% compared to the WT, with the greatest increase seen at CEN3-CEN9. The increase in the CEN3-CEN interactions in the tDNA deletion strain was significantly greater (P = 1.22 × 10⁻¹⁴) than the values of all CEN16-CEN interactions (excluding CEN16-CEN3). These results show that upon deletion of all tDNAs across chromosome III, interchromosomal interactions increase between CEN3 and the other centromeres, suggesting that functional tDNAs likely antagonize CEN-CEN associations during interphase.

We next measured chromosome loss rates of wild-type and tDNA delete diploid cells. We first constructed homozygous diploid cells containing URA3, TRP1, and the MAT locus located on chromosome III (in both the wild-type and tDNA delete strains). Single diploid colonies were grown for 20 doublings in yeast extract-peptone-dextrose (YPD) medium, and approximately 10⁷ cells were plated onto 5-fluoroorotic acid (5-FOA) plates to measure the number of cells that had become URA3 negative. Cells could have become uracil auxotrophs by gene mutations or via loss of the chromosome. These two types of mutants could be distinguished by replica plating the 5-FOA-resistant colonies onto plates lacking tryptophan or by checking for the appearance of MATα pseudohaploid cells. The cells unable to grow on medium lacking tryptophan and also able to mate were counted, and chromosome loss rates were calculated by dividing the number of these colonies by the total number of cells plated. In wild-type cells, the loss rate for chromosome III was 4.08 × 10⁻⁵, which is in agreement with previous reports (76). The loss rates for the tDNA delete chromosome decreased slightly to 3.26 × 10⁻⁵, suggesting that loss of tDNA slightly helped stabilize the chromosome (P = 0.02).

tDNAs play a role in HML-HMR long-range association.

The silent loci HML and HMR reside on chromosome III separated by approximately 300 kb along the linear chromosome. However, the HML locus, located 11 kb from TEL3L, is in close three-dimensional proximity to the HMR locus, located 23 kb from TEL3R. This long-range interaction has previously been detected using both live-cell microscopy and HiC analysis (20, 75, 77), and we recapitulated this finding in the Micro-C experiment with the wild-type strain (Fig. 8A). Comparing wild-type cells to the tDNA delete strain, we noticed that the interaction of HML with HMR was slightly altered in the tDNA delete strain. In wild-type cells, there was an interaction between HML and HMR; this interaction zone became less defined and more diffuse upon deletion of the tDNAs, and a slightly increased interaction frequency was observed across a broader region of the end of chromosome III. While HMR still interacted with HML in the deletion strain, it appeared to also display interactions with other loci (including TEL3L). Similarly, the segment containing HML/TEL3L showed increased interactions with TEL3R rather than being restricted to interacting with sequences at HMR. These results suggest that deletion of chromosome III tDNAs subtly perturbed HML-HMR long-range interactions.

FIG 8.

Long-range HML-HMR association. (A) Deletion of tDNAs on chromosome III leads to a change in HML-HMR interaction as measured by Micro-C. The heat maps display the interaction profiles between segments on chromosome III that include HML and HMR (obtained from the Micro-C data). Increased interactions are denoted by red, and decreased interactions are denoted by blue. The data are displayed in a log₂ format. The x and y axes show the region of the chromosome displayed on each axis of the heat map. (B) Deletion of tDNA t0 leads to perturbation of HML-HMR long-range association. The violin plots show data on the distances between HML::TetR-YFP and HMR::CFP-LacI foci in asynchronously growing strains. Mann-Whitney U tests were performed to determine statistical significance between the wild type and the various mutants. Wild type, n = 305; sir4Δ strain, n = 134 (P = 6.7e−16); tDNA t0Δ strain, n = 317 (P = 3.1e−14); t0Δ:NL1 strain, n = 330 (P = 2.2e−16). The dark horizontal lines represent the median distances. The data for sir4Δ are shown as a control and are the same as those in reference 20. (C) Replication repair proteins are necessary for HML-HMR interactions. Violin plots of the distance between TetR-YFP and CFP-LacI foci in a given wild-type or mutant strain are shown. rrm3Δ, n = 208 (P = 4.8e−12); dpb3Δ, n = 134 (P = 1.7e−10); top1Δ, n = 139 (P = 4.2e−12); scc2D730V, n = 188 (P = 1.9e−14). The data for the scc2D730V allele are shown as a control and are the same as those in reference 20.

Given that Micro-C measures population averages of stable long-range interactions, we decided to measure HML-HMR interactions in live cells using fluorescence microscopy. We wished to determine if the tDNAs influenced HML-HMR interactions. We generated a strain with multiple Lac operator sequences inserted adjacent to HMR (at the GIT1 gene) and multiple copies of the Tet operator sequences inserted adjacent to HML. Expression of the fusion proteins cyan fluorescent protein (CFP)-LacI and yellow fluorescent protein (YFP)-TetR in this strain enabled us to visualize these loci in living yeast by fluorescence imaging. The distance between HML and HMR was then measured in the wild type and a strain lacking the t0 tDNA (Fig. 8B). We found that in wild-type cells, HML was in close proximity to HMR. Consistent with our expectations, deletion of Sir proteins resulted in separation of these loci, validating the assay. Importantly, when we eliminated the t0 tDNA, it led to a change in the distance between HML and HMR compared to wild-type cells, with the median distance between HML and HMR increasing upon deletion of the tDNA. Given the presence of outliers in the data, we used a Mann-Whitney U test to determine statistical significance between the wild type and the mutant. With approximately 300 cells for the wild-type and tDNA delete strains, we observed a P value of 3.1e−14, showing that the differences observed between the two strains were significant. Closer analysis of the plot indicated that upon deletion of the tDNA, there was heterogeneity in the distances between the two loci and a continuum of values. Thus, there were cells where the two loci were in very close proximity, as well as cells where the two loci were further apart. It should be noted that the HiC approach and the microscopy method to measure proximity are inherently different and distinct; each method provides a different type of information, and each approach has specific limitations. Use of long arrays bound tightly by repressor proteins that dimerize may influence chromatin architecture and fluorescence measurements (78). Similarly, the HiC-based methods are influenced by indirect cross-linking, especially in regions of condensed heterochromatin and nuclear substructures, like sites near centromeres and the nuclear envelope (79–81).

tDNA t0 is necessary for the recruitment of cohesins to the silenced loci, and the SMC proteins are necessary for long-range HML-HMR interactions (20). However, not all tDNAs are equivalent in their ability to recruit cohesins to the silenced loci (43, 82, 83). We therefore inquired if tDNAs that are unable to recruit cohesins are able to restore long-range association between HML and HMR. We replaced HMR tDNA^THR (t0) with tDNA^THR (NL1) from chromosome XIV. This tDNA has sequence identical to that of the t0 tDNA in the body of the gene and therefore has BoxA and BoxB promoter sequence and spacing identical to that of tDNA t0. However, sequences flanking this tDNA are distinct, and the NL1 tDNA is unable to recruit/bind cohesins (82, 83). The NL1 tDNA has reduced binding of TFIIIC and a narrower nucleosome-free region, as well (62, 84). This tDNA also has reduced nucleoporin binding and reduced histone turnover frequency (61, 85). When we replaced a 300-bp t0 tDNA-containing fragment with a 300-bp NL1 tDNA-containing fragment, we found that the NL1 tDNA was not able to robustly restore long-range HML-HMR interactions, suggesting that tDNA-mediated local chromatin organization might be necessary for these long-range interactions.

Replication fork pausing at tDNA mediates long-range chromosomal interactions.

We had previously shown that DNA double-strand repair proteins help deposit cohesins at the silenced loci, which then leads to homology-dependent long-range interactions between HML and HMR (19, 20). Since we had discovered that a specific tDNA helped in the clustering of HML and HMR, we wished to know the mechanism by which this phenomenon occurred. Replication fork pausing/stalling is observed at many tDNAs. It results in the deposition of γH2A at the tDNA, which is necessary for fork recovery from the pause/stall (20, 29, 86–90). Rrm3 and topoisomerases play a role in the recovery of stalled replication forks at protein-bound sites in the genome, such as tDNAs (88, 91, 92). Therefore, we analyzed the effects of these mutants on HML-HMR long-range association. The data show that deletion of Rrm3, as well as mutants in the DNA polymerase ε subunit Dpb3 and the topoisomerase Top1, lead to a statistically significant decrease in HML-HMR long-range association (Fig. 8C). Thus, the presence of a tDNA, as well as normal Rrm3, Top1, and Dpb3 function, is necessary for the establishment or maintenance of the long-range association between HML and HMR.

tDNA enhances epigenetic gene silencing at clustered HML-HMR.

Since the tDNA is necessary for the long-range clustering of the silenced domain, we wondered if reduction in clustering had any effect on gene silencing. We asked whether tDNA-mediated loss of HML-HMR interactions affected gene silencing at HML and HMR. Silencing can be assayed by insertion of reporter genes within or immediately adjacent to the silenced domains. In wild-type yeast, when a reporter gene is inserted immediately adjacent to these loci, the gene is metastably silenced. A cassette containing an H2B (HTB1) promoter driving HTB1-EYFP was integrated to the right of HML, while a cassette containing the HTB1 promoter driving HTB1-ECFP was integrated to the left of HMR. In addition, on chromosome XV, a cassette containing an HTB1 promoter driving HTB1-mCherry was integrated as a control euchromatic marker (93). The HTB1-mCherry gene was active in all the cells in the population. The HML::YFP and HMR::CFP reporter genes are present immediately outside HML and HMR but reside in a region bound by Sir proteins (4, 22). These genes adopt one of two expression states, either active or silent. For visualization, single cells were placed on microfluidic plates and monitored continuously by fluorescence microscopy. Fluorescent signal from each individual cell was recorded every 40 min over a period of ∼24 h. This allowed us to trace the lineage of each daughter from the founder cell and score the cells according to the expression of the reporter genes at HML and HMR. Cell lineage trees were traced, and each cell in the lineage was assigned a positive or negative value for expressing each reporter as it underwent cell division (Fig. 9A).

FIG 9.

Silencing of reporter genes at HML and HMR. (A) tDNAs on chromosome III modulate silencing of reporter genes at HML and HMR. Representative lineage trees of the different strains that were analyzed are shown. Wild type refers to a strain containing all the tDNAs on chromosome III. tDNA delete refers to a strain lacking all of the tDNAs on chromosome III; +T0 refers to chromosome III containing only the T0 tDNA; −T0 refers to chromosome III lacking only the T0 tDNA. The expression of HML::EYFP or HMR::ECFP in each generation of cells was monitored and quantitated and is indicated by the presence of their respective colors in the cells of the tree. (B) Deletion of tDNAs on chromosome III leads to a change in the maintenance of silencing at HML and HMR. The graphs quantify the changes in expression state of HML::EYFP and HMR::ECFP between generations in the different genotypes studied. Expressed-to-repressed transitions identify reporters that were expressed in one generation but not expressed in the next. Repressed-to-expressed transitions represent reporter genes that were not expressed in one generation but were expressed in the next. The error bars indicate SD.

We initially analyzed the silencing of the reporter genes in the wild-type strain. Consistent with previous data (93), reporters at HML and HMR were regulated so that the reporters maintained their activity state over many generations and occasionally switched to the opposite expression state. Once they switched, they maintained the new state for several generations. Furthermore, when one reporter was active, the other was also more likely to be active, suggesting long-range coordination between HML and HMR, though this coordination is not absolute.

We next investigated silencing of the reporters in a strain where chromosome III lacked all the tDNAs. In this strain, the reporter at HMR was active more often than in the wild-type strain. While the effect was not as pronounced, the same effect was also observed at HML (Fig. 9A, tDNA delete). Furthermore, the silenced state was less stable and switched to the active state more often. This suggests that silencing at these loci is influenced by the tDNAs.

While expression states at both HML and HMR were stably inherited, the transcriptional state did flip in daughter cells (Fig. 9B). An expressed-to-repressed transition was a less frequent event than the repressed-to-expressed transitions regardless of genotype. This is not entirely surprising, since the reporter genes were inserted immediately outside the two silencers in a zone where the silent state is metastable (94, 95). However, when analyzing the repressed-to-expressed transitions, we saw a discernible difference in the frequency of the expression of the reporter genes at HMR. The full tDNA delete strain showed an increased frequency of cells undergoing the transitions at HMR compared to wild-type cells, and the inverse was seen for the expressed-to-repressed transition (Table 2).

TABLE 2.

Statistical analysis of differences in expression of HML::EYFP and HMR::ECFP in the wild-type and tDNA delete strains

Strain	Frequency ofa:
	HMR		HML
	Expressed to repressed	Repressed to expressed	Expressed to repressed	Repressed to expressed
WT	0.2 (0.019)	0.25 (0.031)	0.06 (0.005)	0.3 (0.075)
tDNA delete	0.13 (0.057)	0.56 (0.145)	0.02 (0.006)	0.41 (0.041)
t0+	0.22 (0.085)	0.25 (0.070)	0.08 (0.014)	0.19 (0.049)
t0−	0.18 (0.049)	0.48 (0.041)	0.11 (0.072)	0.46 (0.12)

^aNumbers in parentheses are standard deviations.

Given that the transcription states of the reporters were affected at both HML and HMR in a strain containing full tDNA deletions on chromosome III, even though only HMR has a tDNA adjacent to it, we decided to focus on the tDNA (t0) that resides immediately adjacent to HMR and functions as an insulator at HMR. Importantly, t0 lies to the right of HMR while the reporter gene lies to the left of HMR, so any effect of t0 on the transcriptional state of the reporter is not due to the barrier function of the tDNA.

To test whether t0 is necessary for regulating silencing states at HML and HMR, we built a strain where only t0 was deleted on chromosome III (t0−). The lineage tree showed that the strain behaved similarly to the strain lacking all tDNAs in that both reporters were active most of the time and rarely switched to the repressed state. Like the full tDNA delete strain, the t0− strain showed an increased frequency of cells undergoing the transitions at both HML and HMR compared to the wild-type strain, where a reporter gene that was not expressed in one generation was more likely to be expressed in the next generation.

Alternatively, to determine if the t0 tDNA was sufficient to mediate silencing effects at HML and HMR, we constructed a strain lacking 9 of the 10 tDNAs on chromosome III and where t0 was the only tDNA still present at its normal location adjacent to HMR (t0+). Again, we monitored expression of the reporters at HML and HMR in this background. The lineage tree showed that the strain behaved similarly to the wild-type strain in that both reporters were silent more often than for the full tDNA delete and t0− strains and inherited the silent state with greater fidelity.

Taken together, the data suggest that deletion of tDNAs on chromosome III had an effect on the ability of HMR to interact with HML and that diminution of this clustering led to an alteration in the stability of the silenced state at these loci.

DISCUSSION

tDNAs are middle repetitive DNA sequences (i.e., moderately repetitive DNA [usually rRNA and tRNA genes] [140]) scattered across all 16 chromosomes, and their primary function is the synthesis of tRNAs. Here, we show that tDNAs affect local chromatin structure, which then impinges on chromosome architecture. tDNAs (i) affect chromatin structure by maintaining local nucleosome-free regions along the fiber and precisely positioned nucleosomes immediately outside the tDNAs, (ii) recruit cohesins and condensins, (iii) affect nuclear architecture by influencing centromere clustering, and (iv) alter heterochromatin clustering, leading to changes in the fidelity of inheritance of gene silencing.

The binding of specific proteins, such as CTCF, to a site on the DNA can affect nucleosome positions over long distances (96). Nucleosome depletion at the gene and positioned nucleosomes flanking the gene are hallmarks of tDNAs (22, 24, 25, 35, 61, 97–100). Our data show that loss of the tDNA promoters affects only nucleosome positions in the immediate vicinity of the tDNA. The nucleosome-positioning effects mediated by the tDNA-bound transcription factors TFIIIC and TFIIIB are not transmitted over long distances.

tDNAs, SMC proteins, and chromatin folding.

The SMC proteins are involved in higher-order chromosome organization in all eukaryotes and have been extensively mapped. tDNAs are binding sites for all three classes of SMC proteins (cohesins, condensins, and SMC5 and SMC6 proteins involved in DNA repair [repairsins]), the SMC protein loaders Scc2 and Scc4, and the meiotic Rec8 SMC protein (28, 29, 32, 101–105). Given these intimate connections between tDNAs and the SMC proteins, our data indicate that loss of the tDNA promoters does lead to loss of SMC proteins from tDNAs, but this effect is tDNA specific, since we do not see a loss of SMC proteins from centromeres or of Scc2 from other sites in the genome. Surprisingly, the loss of Scc2 and Brn1 from tDNAs does not affect chromatin folding. While clustering of tDNAs in the nucleus (as measured by fluorescence microscopy) is dependent upon the SMC proteins (54, 70), the precise contribution of tDNAs in this process remained unclear. Our Micro-C analysis suggests that tDNAs play a minor role in chromatin folding and tethering to nuclear substructures, since we observed only subtle changes in contact frequency across the chromosome and small effects on chromosome loss rates. It is likely that tDNA-independent SMC protein binding sites mask the tDNA-mediated effects. SMC proteins bind only half of the tDNAs in the nucleus, and only a third of the SMC protein binding sites localize at or near tDNAs (29). The lack of phenotype is also consistent with previous data that showed that a reduction in the levels of the SMC proteins does not affect the properties of the chromosome arm (106). Recently a synthetic yeast chromosome III was generated and characterized (107, 108). The synthetic chromosome lacks repetitive sequences, such as TY elements, long terminal repeats (LTRs), and tRNA genes. The 3D structure of the chromosome was determined using HiC, and the data show that there were no major differences between this chromosome and the wild-type chromosome except for a decrease in the length. While the chromosome lacks multiple elements, the three-dimensional folding data are consistent with our conclusions from the Micro-C analysis that the same chromosome lacks only tDNAs.

While it is possible that redundancy of structural elements masks tDNA-mediated effects on chromatin folding, it is also possible that chromatin folding is driven by underlying DNA sequence-mediated nucleosome organization and not tDNA-mediated long-range interactions. The yeast chromosomes have isochores with G-C-rich, gene-rich R band segments alternating with AT-rich G band segments (109, 110), which exhibit different functional properties and chromosome conformation (111, 112). Chromosome III has a G-C segment from kb 20 to 100 on the left arm, followed by an A-T-rich central segment from kb 100 to 200 on the right arm and then a second G-C-rich segment from kb 200 to 290 on the same arm. In this scenario, the underlying A-T-rich DNA sequence likely plays a dominant role in the three-dimensional folding of chromatin. tDNAs are often syntenic along chromosomes (39, 113), and it is possible that these positions have been selected for optimal gene activity rather than being involved in long-range chromatin loop formation (114). Thus, while the A-T-rich isochore is structurally and functionally distinct (75, 115, 116) and is the region rich in tDNAs (Fig. 1), our results suggest that the tDNAs do not play a significant role in either tethering of this isochore or the overall folding of the segment. The tDNA clustering observed by microscopy could simply be a function of conservation of tDNA positions along the chromatin fiber.

tDNAs and centromere clustering.

Chromosome tethering to nuclear substructures enables nuclear organization (1, 114), and centromeres and telomeres, along with their associated proteins, play a key role in this process (7, 9, 10, 12, 15, 16, 73, 117–120). All sixteen centromeres cluster together in a ring around the membrane-embedded spindle pole body. The centromeres are tethered to the spindle pole body via direct interactions between kinetochore-associated proteins and the spindle pole body-associated microtubules in interphase (7, 73, 75, 117). Other factors are likely to influence this phenomenon but remain unknown. tDNA density is almost 2-fold higher in the pericentric region of S. cerevisiae chromosomes, including chromosome III (121). While tDNAs have been shown to help tether centromeres to the spindle axis during mitosis (121), in interphase nuclei, the loss of tDNAs results in increased interactions between the clustered centromeres. The physical presence of tDNAs in the pericentric region could interfere with the close packaging of centromeres during interphase. This could be due to transcription-mediated effects, since tRNA genes are highly active. In Schizosaccharomyces pombe, mutations that reduce tDNA transcription result in increased tDNA association with the kinetochore and increased chromosome condensation during mitosis. Furthermore, tDNA association with kinetochores increases when these genes become inactive (55). Thus, tDNA clustering at sites of active tRNA transcription near centromeres could hinder centromere-centromere interactions during interphase, while a decrease in tDNA transcription during mitosis could help tether centromeres to the spindle axis during mitosis (121). This would also explain the observation that the chromosome with tDNA deleted had a slightly lower chromosome loss rate.

An alternative, though not mutually exclusive, possibility is based on the observation that transcriptionally active tDNAs interact with nuclear pores in the G₂/M phase of the cell cycle (4, 122, 123). It is thus possible that there is competition between pericentric tDNA–nuclear-pore interactions in opposition to centromere-centromere interactions. In this scenario, the loss of tDNA tethering to the nuclear pore would provide the centromere greater freedom of movement, thus enabling closer centromere-centromere interactions.

tDNA effects on HML-HMR interactions and the inheritance of gene silencing.

Gene silencing is primarily a function of the Sir proteins, though numerous other factors influence the process (124). Protosilencers are sequence elements that on their own are unable to silence a gene but when located near a silencer increase the efficiency of silencing (125, 126). Our demonstration that tDNA affects silencing of a reporter adjacent to the silent HMR domain suggests that tDNAs function as protosilencers. Our data suggest that tDNA-mediated clustering of silent loci might be important in the silencing of these loci and that the loss of long-range association might reduce the efficient inheritance of the silent state. This is analogous to the observations that gene clustering at active chromatin hubs and transcription factories increases the efficiency of transcription, as well as the data showing that telomere clustering increases the efficiency of silencing at subtelomeric sequences (127).

This unexpected observation also raises the question of how tDNAs might influence long-range HML-HMR interactions. tDNAs, including the tDNA next to HMR, are sites of replication slowing/pausing (86, 87, 91, 128–130). The tDNA adjacent to HMR is a site of replication fork pausing (89, 131). We recently showed that long-range HML-HMR interactions require homologous sequences to be present at these loci (19, 20), and we now show that mutations in replication-coupled homologous recombination repair proteins, including the SMC proteins, Rrm3, Top1, and Dpb3, lead to a reduction in HML-HMR interactions. Based on the accumulated data, we posit that replication fork slowing/pausing results in the deposition of γH2A and SMC proteins at tDNAs, followed by a homology search leading to HML-HMR interactions. The re-formation of silenced chromatin following replication precludes the eviction of γH2A (132), thereby stabilizing SMC protein binding, which then maintains the long-range HML-HMR association. The tDNAs thus help initiate a network of interactions mediated by the SMC proteins and the Sir proteins, leading to HML-HMR association and chromosome folding. We posit that a series of transient interactions during replication aid in setting up the final optimal nuclear architecture found in the interphase nucleus.

In conclusion, tDNAs primarily affect local chromatin structure. Each tDNA affects nucleosome positions and protein binding in its immediate vicinity. These local perturbations functionally and structurally interact with neighboring regulatory regions, resulting in tDNA-mediated pleiotropic effects. In some instances, tDNAs affect the expression of neighboring Pol II-transcribed genes by the phenomenon of local tgm silencing. In another context, tDNA-mediated replication pausing results in the establishment of long-range heterochromatin interactions, which then influence the inheritance of silencing states at these loci.

MATERIALS AND METHODS

Yeast strains and primers.

Tables 3 and 4 list the yeast strains and the primer sequences that were used in this study.

TABLE 3.

Strains and genotypes

^at8Δ+T9Δ indicates that the two genes were deleted simultaneously. T1+T7 indicates that the two genes were inserted adjacent to one another.

TABLE 4.

Sequences of PCR primers used in this study

Primer name	Sequence (5′–3′)	Amplicon
yOH58	TACTACAAGAGAAAGGCCATCTCC	t1
yOH59	AATGCAGCGCAGACAGCACAGTT	t1
QJK61	TTGAGATACAAAATATTACAAGAAGTCCTG	t2
QJK62	GCGTTCTTCTGTATCTGAAGATAGTG	t2
QJK63	TCATGTATCAAGATTACTAGCGCAAGTG	t5
QJK64	TTCTATTCTTATGTACCGTTCCGCC	t5
yOH62	GCAAGCGAAGTTGTTCCCGTTAT	t7
yOH63	GTTCGGTCACTTAGAGGATATAATTG	t7
QJK69	CTCTATTTCTCAACAAGTAATTGGTTGTTT	t8
QJK70	GCCCCTGTGTGTTCTCGTTATGT	t8
yOH64	GACAAGAAAGATAACGACACAGTGA	t9
yOH65	GGCCCTCGTATAGTCTCTTTTC	t9
R197	GAGACCAGGTTTATTCAACCGGTAAC	t0
LOU120	GGGTGTCACCGAATAACGTGAT	t0
GRO39	TAAGACAATTGTGGACAACAAAGCAAA	OCA4
GRO40	ATTTATTAATGTCAAAAGCCGCTGAGG	OCA4
yOH66	TCACTCATATAAACCGAACCCTTCC	CEN3
yOH67	GGATTTTCCATATTGTTTGGCGCTG	CEN3

MNase-seq.

MNase-seq experiments were carried out as previously described (25). In brief, isolated nuclei were digested with MNase to mononucleosomes. Paired-end sequencing libraries were prepared (Illumina). Paired reads (50 nucleotides [nt]) were mapped to the reference genome (SacCer2) using Bowtie-2 (133–135). For analysis of nucleosome occupancy (coverage) at tDNAs, both across the genome and on chromosome III, tDNAs were aligned on their start sites or at the deletion points. Data sets were normalized to their genomic average (i.e., 1) using only DNA fragments in the 120- to 180-bp range. In one experiment, mononucleosomal DNA was gel purified, but not in the replicate, in which short fragments (<120 bp) derived from digestion of the TFIIIB-TFIIIC complex at tDNAs (97) were observed.

ChIP-seq and RNA-seq.

Chromatin immunoprecipitation reactions were performed essentially as described above, but elution of the precipitated DNA from protein A/G beads was carried out with two successive washes in 175 μl of 0.1 M NaHCO₃-1% SDS; 50 μl of each input sample was diluted to 350 μl with the elution buffer. NaCl was added to a final concentration of 0.2 M, and cross-links were reversed with an overnight incubation at 65°C in a Thermomixer (Eppendorf). All samples were treated with 60 μg of RNase A (Sigma) at 37°C for 60 min, followed by proteinase K (Roche) treatment at 50°C for 60 min. DNA was purified with successive phenol chloroform and chloroform extractions, followed by precipitation with 2 volumes of ethanol and 50 μg of glycogen (Roche).

The ChIP and input DNAs were spun, washed with 70% ethanol, and resuspended in deionized water. DNA quantitation was performed using a Qubit double-stranded DNA (dsDNA) HS assay kit prior to confirmation by qPCR.

Libraries for ChIP-seq were prepared at the Functional Genomics Laboratory, University of California—Berkeley (UC Berkeley), and sequenced on an Illumina HiSeq 4000 at the Vincent J. Coates Laboratory at UC Berkeley.

For RNA-seq, yeast strains JRY2334 and JKY690 were grown in duplicate in 50 ml YPD to a cell density of 6 × 10⁶ to 7 × 10⁶ cells/ml, spun, washed in 25 ml phosphate-buffered saline (PBS), divided into 4 aliquots per culture, and transferred to 1.5-ml microcentrifuge tubes. The cell pellets were flash frozen in liquid N₂ and transferred to −70°C. RNA library preparation and sequencing for RNA-seq were performed by ACGT, Inc., Wheeling, IL.

Transcript abundances were estimated using Kallisto (136). Differential analysis of gene expression data was performed using the R package Sleuth (137). The likelihood ratio test and Wald test were used to identify the differentially expressed genes (the false-discovery rate-adjusted P value [Q value] was <0.05 in both tests). Since the likelihood ratio test does not produce any metric equivalent to the fold change, we used the Wald test to generate the beta statistic, which approximates to the log₂ fold change in expression between the two conditions.

ChIP.

ChIP-qPCR experiments on Brn1 and Scc2/Scc4 were performed as previously described (20, 35). In brief, yeast cells of a strain of interest were inoculated and grown overnight in 300 ml of YPD medium to an optical density (OD) of 1 to 2. These cells were then fixed in 1% formaldehyde for a duration of 2 h at room temperature. The reaction was then quenched with glycine, and the cells were spun down and washed in 1× PBS. The cross-linked cells were then flash frozen in dry ice and stored at −70°C. In preparation for immunoprecipitation (IP), the cells were thawed on ice, broken apart by bead beating, and sonicated to achieve a desired chromatin size of ∼300 bp. Once the size of the chromatin was checked, cell debris was cleared from the sample by high-speed centrifugation. The cross-linked, sized chromatin was split into 2 samples, and IPs were done overnight in the presence of both an antibody to the protein of interest and preblocked A/G-Sepharose beads at 4°C; 50 μl of input chromatin was also taken from each IP sample prior to addition of the antibody. Chromatin elution was done using 10% Chelex 100 (Bio-Rad), along with proteinase K treatment. After elution, both input and IP DNAs were quantitated via a PicoGreen fluorescent quantification assay (Invitrogen). For each qPCR, input DNA was run in triplicate and IP DNA was run in duplicate. Equal amounts of input and IP DNAs were used in each individual reaction. The enrichment for a given probe was then calculated as IP divided by input and was further normalized to the OCA4 locus. The results of each ChIP-qPCR were comprised of two independent cross-links for each strain assayed, and for each cross-link, two independent IPs were done.

Mean squared distance analysis.

Mean squared distance (MSD) analysis was carried out as previously described (68, 138, 139). In brief, we built strains that contained a 64×LacO array at specific points along chromosome III. We then integrated a cassette containing an spc29-RFP fusion protein elsewhere in the genome. This protein is an essential kinetochore protein and therefore serves as a marker for the spindle pole body. The spindle pole body served as a fixed point at which we could measure the movement of our GFP-tagged loci in 3D space over a period of 10 min. Z-stack images of the cells were taken every 30 s during the time-lapse movie, and the data were used to calculate the radius of constraint using the following formula: MSD = (X_{t + T} − X_t)² + (Y_{t + T} − Y_t)², where X and Y are the coordinates of the fluorescent dot, T is the time lag, and t is the time. MSD curves were generated for each locus in both the WT and tDNA delete strains. The MSD curves were used to calculate the Rc for each locus. The analysis was performed in no less than 35 cells per genotype assayed. The data were plotted in NotBoxPlots (the source code was obtained from https://github.com/raacampbell/notBoxPlot).

HML-HMR colocalization analysis.

Distance assays between HML and HMR were performed as previously described (20). Fluorescence microscopy was performed on live yeast cells after growing the cells in yeast minimal dextrose (YMD) with leucine, uracil, tryptophan, lysine, adenine, and histidine. The cells were grown to an A₆₀₀ of approximately 0.6. The cells were washed in YMD, placed on YMD-agar patches on slides, and imaged. Microscopy was performed with an Olympus xi70 inverted wide-field microscope with a DeltaVision precision stage using a Coolsnap HQ2 camera and a 100×/1.4-numerical-aperture oil objective. The 20 image stacks for each image were acquired with a step size of 200 nm using the appropriate wavelength for CFP, YFP, GFP, or mCherry. The acquisition software used was softWoRx3.7.1. The images were cropped using Adobe Photoshop. For the distance analysis between HML and HMR, the distances between the yellow and cyan dots were calculated in nanometers using the “measure” tool in three dimensions. The measured distances were loaded into R software (www.r-project.org), and the data were plotted as a box plot. The box included the middle 50% of the data, with the line in the box indicating the median value. The data presented are the sums of at least two independent strains.

Single-cell expression analysis.

Single-cell expression analysis was performed as previously described (93). Briefly, cells were grown in YPD at 30°C and placed in a microfluidics device. Time-lapse photographs of growing cells were recorded using an Axio Observer Z1 microscope with a 40× objective. The ECYP and EYFP fluorescence intensities were normalized to the highest level of fluorescence observed and to the euchromatic mCherry signal.

Micro-C.

Micro-C was performed as previously described (71). The detailed method has been described previously (72). In brief, the technique provides nucleosome level resolution of all of the interactions occurring across the genome by using MNase digestion in lieu of a restriction enzyme, as in traditional Hi-C techniques.

Antibodies.

The antibodies used in ChIP were as follows: Scc2-Myc, anti-myc 9E10 (Abcam; 5 μl); Brn1-HA, anti-HA HA.11 (Covance; 5 μl).

Accession number(s).

The interactome data were deposited in the GEO database under accession no. GSE98543. The MNase-seq data are available in the GEO database under accession no. GSE98304. CHIP-seq and RNA-seq data have been deposited in the GEO database under accession no. GSE106250.