tRNA Gene Identity Affects Nuclear Positioning

The three-dimensional organization of genomes is dynamic and plays a critical role in the regulation of cellular development and phenotypes. Here we use proximity-based ligation methods (i.e. chromosome conformation capture [3C] and circularized chromosome confrmation capture [4C]) to explore the spatial organization of tRNA genes and their locus-specific interactions with the ribosomal DNA. Directed replacement of one lysine and two leucine tRNA loci shows that tRNA spatial organization depends on both tRNA coding sequence identity and the surrounding chromosomal loci. These observations support a model whereby the three-dimensional, spatial organization of tRNA loci within the nucleus utilizes tRNA gene-specific signals to affect local interactions, though broader organization of chromosomal regions are determined by factors outside the tRNA genes themselves.


Introduction
Structural genome organization is manifested on different levels, such as linear arrays of genes and spatial arrangement of chromosome territories [1]. Recent studies have implicated interactions that form between genomic loci in the regulation of genes [2][3][4] and of cellular processes such as development [5].
Examination of the spatial organization of gene families can provide insight into how position relates to evolutionary or functional imperatives. The largest family of co-regulated genes in the eukaryotic genome is the RNA polymerase III (Pol III)transcribed tRNA gene family. The budding yeast Saccharomyces cerevisiae has 274 tRNA genes that are dispersed throughout the linear maps of the 16 chromosomes. Fluorescence in situ hybridization (FISH) microscopy has shown that these tRNA genes are clustered throughout the cell cycle, with the assistance of condensin complexes bound at each gene, and that clusters localize to the boundary of the nucleolus in a microtubule-dependent manner [6][7][8]. Condensin has also been localized to the nucleolar ribosomal DNA (rDNA) repeats, and mutants of condensin affect proper compaction of the rDNA repeats [8][9][10][11][12]. Clustering of tRNA genes has also been observed in fission yeast [13,14], although their subnuclear localization is different from that seen in S. cerevisiae. Proximity-based ligation methodologies, which crosslink spatially adjacent loci, now permit investigation of direct physical interactions among genes in greater detail. Two of these techniques, Genome Conformation Capture (GCC) and a variant of HiC, have previously been used to produce a yeast genome contact map [15,16] and confirm microscopy results by showing preferential interactions between tRNA genes [15], consistent with the physical clustering observed using fluorescent microscopy.
Since the localization of a large number of dispersed genes to a single subnuclear region necessarily requires a vast rearrangement of the genome, it is of interest to investigate whether individual tRNA gene associations are a controlling influence on the overall organization of the genome, or merely serve as non-specific ''fasteners,'' providing some level of local condensation, while global organization is determined by other factors. Here we use three methods that rely on proximity-GCC, chromosome conformation capture (3C), and circularized chromosome conformation capture (4C)-to examine the contributions that tRNA genes make to the positioning of specific loci within the S. cerevisiae nucleus.

Results and Discussion
Genomic fragments that contain tRNA genes are spatially associated with the nucleolus GCC using unsynchronized S. cerevisiae revealed that many tRNA genes formed multiple interactions with the ribosomal DNA locus (RDN) on chromosome XII [15], which contains multiple tandem copies of the ribosomal RNA genes and forms the nucleolus. While numerous of these interactions were well above background, there was an extremely frequent interaction between one particular DNA fragment containing a lysine tRNA gene on chromosome XVI, tK(CUU)P (Chr XVI: 581,025-583,522, Figure 1A), and the non-transcribed spacer sequence (NTS1) in the RDN locus, adjacent to the Pol III-transcribed 5S rRNA gene (Chr XII: 460,025-460,609). None of the MspI restriction fragments adjacent to the tK(CUU)P lysine tRNA gene fragment interacted with NTS1. In fact only two of the nearby fragments (Chr XVI: 585884-589137 and 549477-580469, respectively) interacted with the rDNA at levels even slightly above the background ( Figure S1, Table S1, and Supplementary statistics in Methods S1). It is theoretically possible that some of the tK(CUU)P-NTS1 interactions involve the extra-chromosomal rDNA circles that are present within the yeast nucleus [17,18]. However, preliminary data indicate that high ERC copy number does not correlate with increased interaction frequencies (data not shown). We conclude that the interaction is driven by signals within the fragment and, given that tRNA genes are known to cluster with the nucleolus [6][7][8], we hypothesized that the tRNA gene was responsible for this interaction.
Does the lysine tRNA sequence direct the interaction with NTS1?
Clustering of the tRNA genes is mediated by one or more condensin complexes bound to the tRNA genes [8,19]. Since condensin is also bound to the rDNA repeats and concentrated adjacent to the 5S rRNA gene [19], we hypothesized that the tK(CUU)P-NTS1 association might be determined by a condensincondensin interaction between the tRNA gene complex and the NTS1 region. However, we remained open to the possibility that there might be other contributors that direct this particular tRNA gene to this specific region of the RDN locus. Therefore, to examine to what extent the tRNA gene was responsible for this tight association, we performed 3C on both a wild type (WT) strain and the same strain from which only the tRNA gene coding region, including its intragenic transcription promoter, had been precisely deleted (yDP97, tK(CUU)P::kanMX6 [20]; Figure 1). Quantitative analysis of locus proximities by 3C [15,21] showed that precise deletion of the tK(CUU)P tRNA coding sequence did not significantly alter the frequency of interaction between the general locus (fragments F1 or F2' on Chromosome XVI) and a HindIII fragment spanning the 5S rRNA gene and including the NTS1 region ( Figure 1B). Furthermore, the deletion did not significantly affect growth rate, determined by co-culturing the two strains for 100 generations (not shown). This lack of growth defect suggests an absence of serious disruption to nuclear organization, in contrast to the strong growth defects that were previously observed in mutants that disrupted general tRNA gene clustering or nucleolar organization [6,8,22]. Together, these results indicate that the tK(CUU)P lysine tRNA gene itself does not provide the major driving force that determines the proximity of this chromosomal segment to the rDNA repeats. Rather, the general spatial arrangement of these chromosomal loci is driven by factors outside the tRNA gene. Despite this, this result does not preclude the possibility that the tRNA gene might determine the positions of local contacts.
There are internal and external drivers for tRNA spatial positioning The coding regions of yeast tRNA genes contain the major transcriptional promoter elements and are bound by identical sets of transcription components (i.e. TFIIIC, TFIIIB, and Pol III) [23][24][25], consistent with the finding that all tested tRNA gene loci can be expressed. Therefore, it was predicted that the nature of the tRNA coding sequences would not alter the interaction behaviour of a locus. To test this prediction, we precisely replaced the mature tRNA coding regions of two different leucine tRNA loci-tL(UAA)B2 on Chromosome II, and tL(CAA)G3 on Chromosome VII. The coding regions of these leucine tRNA genes interacted with the RDN locus in the parental strain, although with different patterns across the rDNA repeats ( Figure 2B). The tL(UAA)B2 and tL(CAA)G3 tRNA coding regions were precisely replaced with the coding region from a tyrosine tRNA gene-variant, the SUP4-1 ochre suppressor [26], to allow selection for the insertion (Figure 2A). In each case the 59 and 39 flanking regions, including upstream transcription initiation and downstream termination sites, and the primary transcript processed leader and trailer sequences were retained from the original leucine tRNA locus. As expected, the SUP4-1 tyrosine tRNA gene replacements of the leucine tRNA coding regions continued to allow association of both loci with the RDN locus, yet unexpectedly the pattern of preferred positions of the associations along the RDN locus were altered. As a control for general disruption of tRNA gene contacts, quantitative 3C analyses of the S. cerevisiae yPH499, yDP77, and yDP84 strains identified no significant differences in the interactions between the wild-type SUP4 tyrosine tRNA gene (tY(GUA)J2) (Chr X: 542960-543119) and the 25S rDNA (Chr XII: 451928-452600) locus ( Figure S2). These results are consistent with the tK(CUU)P lysine tRNA gene replacement data, above, suggesting that external sequences specify general positioning within the nuclear space but that internal factors affect precise local positioning.
By contrast, other interactions identified within the dataset demonstrate that the identity of the internal tRNA gene sequences does influence partner selection. For example, some preferred partners of the SUP4-1 sequence were gained at the new locus following replacement of the original leucine tRNA gene sequence (Table S2). The most pronounced of these involved interactions between SUP4-1 at the new loci and MED6 or MED20 (Table S2). This ability of the internal tRNA gene sequences to contribute to binding partner selection is intriguing, though puzzling. However, the overall effect can be explained by the hypothesis that the flanking sequences act to direct a locus to a particular region of the nucleus. Once within this region, partner selection is influenced by the tRNA gene sequence itself.
How does internal sequence affect tRNA positioning and partner selection?
The effect of the tRNA gene internal sequences on the positioning of the tRNA gene was completely unexpected, given that all tRNA gene transcription complexes bind the same required components to the internal promoters, as far as is known [23][24][25]. However, this phenomenon could have several causes. Firstly, TFIIIC, the transcription factor that initially recognizes the gene, has a bipartite binding site within the tRNA coding region [27]. The spacing between these sites in the original leucine tRNA genes and in the replacement tyrosine tRNA gene (SUP4) is slightly different, and the resulting change in the topology or strength of the TFIIIC-DNA interaction might subtly alter its interaction with other binding partners such as condensin. Moreover, since the degree of occupation of all the genomic tRNA genes by TFIIIC, TFIIIB, and Pol III is variable across the genome [24,25,28,29], the nature of the complexes could be changed by having different geometries or stability of occupation by one or more components. A more speculative yet possible explanation might also be that tRNA gene complexes have tRNA-specific protein components due to differences in their transcript sequences. For example, although both leucine tRNAs and tyrosine tRNAs are cleaved at their 59 and 39 mature ends early in biosynthesis, the exact order and location of nucleotide modification events is not clear. Because processing reactions are tRNA specific [30], it therefore remains possible that some enzymes associate with tRNA gene complexes through the nascent RNAs in a sequence-dependent manner.

Conclusion
These findings support a model whereby the locus-specific tRNA transcription complexes serve as ''fasteners'' to determine local interactions and promote clustering, but that this occurs in combination with other determinants that dominate global nuclear positioning.

Strain construction
The coding sequence of tK(CUU)P in S. cerevisiae BY4741 (MATa his3D1 leu2D0 met15D0 ura3D0), including mature coding sequence and intron, was precisely replaced with the kanMX6 expression cassette by recombination using PCR fragments generated from plasmid pFA6a-kanMX6 [20] to create yDP97. Transformants Figure 2. tRNA identity dictates interaction patterns with the ribosomal DNA locus. A) Two leucine tRNA genes (tL(CCA)G3 and tL(UAA)B2), which are located on separate chromosomes, were replaced with the tyrosine tRNA SUP4-1 suppressor gene to generate strains yDP84 and yDP77. Grey bars denote MspI restriction sites, arrows denote 4C primer positions. B) Interactions with the rDNA locus were identified by 4C coupled to high through-put sequencing (Methods S1). Raw interaction frequencies have been plotted across a tandem rDNA repeat for clarity. Note, only patterns of interaction along the rDNA locus should be compared as inter-experiment variation has not been corrected for. doi:10.1371/journal.pone.0029267.g002 were selected on medium containing G418, and exact gene replacement was verified by PCR and sequencing.

Genome Conformation Capture [15]
Briefly, chromatin was prepared from 15 sets of 10 8 (i.e. a total of 1. Chromosome conformation capture (3C) [31] S. cerevisiae strains (i.e. WT and yDP97) were grown (30uC, 160 rpm) to an OD 600 = 0.6 in synthetic complete media containing amino acid supplements and glucose (2% w/v). Chromatin was prepared according to [15] using HindIII or MspI restriction enzyme. HindIII cleaves the kanMX6 expression cassette and thus results in three restriction fragments in yDP97, as opposed to two fragments in the WT. Interactions between F1 and F2 of the WT strain (F1 and F29 of the yDP97 strain) on Chr XVI, and the rDNA HindIII fragment (Chr XII: 457,910-460,634 bp) were measured for three biological replicates.
Sorting involved some pre-processing of sequence tags. Each of the individual samples was isolated from the sequence files according to its 6 bp tag and primer sequence and trimmed to 34 bp (with the MspI recognition sequence in the centre). The sequences for fragment 1 and 2 for each mutant (yDP77 and yDP84) were pooled. Since SUP4-1 shares considerable identity with other tyrosine tRNAs, particularly around the MspI restriction site, the primer sequences which hybridised adjacent to the novel restriction site within this locus could not be uniquely positioned to the bait fragment on the reference genome. Therefore, the unique primer sequences from the opposite ends of the bait fragments were substituted for the 17 bp repetitive sequences that abut the novel restriction site within SUP4-1, prior to analysis. Thus sequences that crossed the SUP4-1 restriction site within the bait fragments were accurately mapped to either tL(UAA)B2 or tL(CAA)G3, depending on the interaction under investigation. For analysis, adjacent interaction frequencies were used to correct for between sample comparisons. Manipulated datasets are provided as Data S1, S2, S3. Figure S1 The high frequency interactions between Chr XVI: 581,025-583,522 (fragment 13476) and the NTS1 sequence adjacent to the 5S rDNA are isolated and not mirrored at adjacent sites within Chr XVI. Critically, of the six fragments which immediately flank fragment 13476, only the -2 and +2 fragments interact with the rDNA. However, neither interact with the same NTS1 fragment (Chr XII:460,025-460,609) and the maximum number of interactions we observed was 3 orders of magnitude lower than for fragment 13476. Global chromosome capture was performed on unsynchronized exponentially growing S. cerevisiae cells (Materials and Methods, Methods S1). Interactions that occurred above the experimental false detection rate (Methods S1) were counted and mapped between restriction fragments surrounding fragment 13476 on Chr XVI (illustrated to the left), and the restriction fragments present across the rDNA (illustrated below the graphs for reference). A, map of interaction frequencies between fragment +2 (Chr XVI:585884-589137) and the rDNA locus; B, map of interaction frequencies between fragment 13476 and the rDNA locus; and C, map of interaction frequencies between fragment 22 (Chr XVI:549477-580469) and the rDNA locus. (DOCX) Figure S2 Genetic background does not affect the interaction frequency between the tyrosine tRNA tY(-GUA)J2 (Chr X: 543044-542956) locus and the 25S rDNA (Chr XII: 451928-452600). 3C was performed using MspI on cross-linked chromatin isolated from yPH499, yDP77 and yDP84 cells grown in SC-glucose to an OD 600 = 0.660.14. Interaction frequencies were determined by quantitative 3C analyses using a fluorescent probe and primers that are specific for the tY(GUA)J2 -25S rDNA interaction (Table S3) and have been corrected for  nuclear genome copy number to facilitate inter-strain comparisons  (see Methods). Interaction values are expressed as percentages of the yPH499 sample (set at 100%) +/2 standard error of the mean (n = 3). (DOCX)  (22), 13475 (21), 13476, 13477 (+1), 13478 (+2), 13479 (+3)). These interactions were identified from interaction networks produced by analysis of Genome Conformation Capture (GCC) libraries constructed using the MspI restriction enzyme and chromatin from glucose grown S. cerevisiae cells (Methods S1). Only data which occurs above the experimental noise threshold has been included (see Supplementary Statistics in Methods S1). These tables do not contain adjacent interactions as they are not included in the statistical calculations. (XLSX) Table S2 SUP4-1 directed interactions detected by circularized chromosome conformation capture (4C). Captured "prey" fragments whose coordinates are outlined in columns F, G and H are listed alongside the 4C bait fragments (which contain either the wildtype tL(UAA)B2, tL(CAA)G3 loci, or these loci replaced with a SUP4-1 sequence) with which they interact. Only interactions which are conserved by the presence of the SUP4-1 sequence are present in this file. These interactions are conserved regardless of the genomic position of the SUP4-1 sequence. (XLSX) Data S1 Comparison of interactions for wild-type and yDP77 strains. Topography v1.19 program output for the high throughput sequencing of the 4C PCR product. The complete data set is presented along with tabs in which interactions have been classified according to the feature present on the partner sequence (i.e. tRNA, autonomous replicating sequences [ARS], YSCPLASM [2 micron plasmid], ribosomal DNA, Telomeres). The read-me tab explains file specific variations and presents a diagram of the fragments that were analysed to determine the interaction patterns. (XLSX) Data S2 Comparison of interactions for wild-type and yDP84 strains. Topography v1.19 program output for the high throughput sequencing of the 4C PCR product. The complete data set is presented along with tabs in which interactions have been classified according to the feature present on the partner sequence (i.e. tRNA, autonomous replicating sequences [ARS], YSCPLASM [2 micron plasmid], ribosomal DNA, Telomeres). The read-me tab explains file specific variations and presents a diagram of the fragments that were analysed to determine the interaction patterns.