To establish the zero-torsion baseline for the psoralen signal, we must establish a method to account for psoralen’s sequence preference and accessibility without torsion under conditions that preserve the native-bound protein landscape (Fig. Supplement 1). To do this, we developed a method to prepare the chromatin zero-torsion state in situ and applied psoralen photo-binding (Fig. 1a, right panel). Formaldehyde preserved the chromatin landscape and transiently bound proteins (Materials and Methods) while stopping any subsequent cellular processes. The fixed cells were permeabilized so enzymes could diffuse into the nucleus. For S. cerevisiae, the cell wall was gently digested to create spheroplasts. To release torsion in intact spheroplast nuclei, we utilized DpnII, chosen for its high cut site density but inability to processively digest DNA, thus maintaining psoralen’s crosslinking ability. Upon DNA digestion, we observed an average fragment size of ∼2 kb, sufficiently short to release torsion but long enough to provide a chromatin substrate for psoralen binding. This zero-torsion state preparation was subsequently crosslinked using 365-nm light, enabling measurement of the torsion-free TMP intercalation landscape (referred to as TMP signal without torsion). The in vivo measurement (TMP signal with torsion) was performed by directly adding TMP to the yeast media since unmodified TMP is cell-wall permeable. Cell permeability enables photo-crosslinking to be performed before cells are harvested, thus capturing native cell activity (Fig. 1a, left panel).

To locate psoralen crosslinks on the DNA, we utilized an exonuclease-based enrichment strategy followed by an Illumina-compatible library prep, similar to TMP-seq^30,33. Exonuclease-based enrichment relies on interstrand crosslink (ICL) formation, which prevents DNA denaturation before digestion by the single-strand-specific ExoI. We use the digested fraction to estimate the crosslinking density and provide quality control for ICL enrichment over the background. As controls, we found that libraries prepared without psoralen photo-binding showed a very low signal level (Fig. Supplement 2). By tuning the UV dosage, we ensured that the average densities of the in vivo and zero-torsion measurements were comparable at ∼0.1-0.25 ICLs/kb (Fig. Supplement 2). If ICL density is too low, the signal-to-noise ratio is poor; if ICL density is too high, there can be library preparation artifacts (Fig. Supplement 3). ICL densities within the range used display remarkably similar signals (Fig. Supplement 2). This allows the TMP signals with and without torsion to be compared to obtain an accurate genomic map of psoralen binding due to torsional stress. In addition, we conducted a control experiment using purified DNA to assess the extent of psoralen crosslinking due to DNA sequence preference (Fig. Supplement 2). This TMP signal was only used during an intermediate analysis step to isolate the psoralen intercalation preference not due to DNA sequence (Fig. 1b,c).

To evaluate the contributions of DNA accessibility to the TMP intercalation profile, we examined the TMP signal downstream of transcription start sites (TSSs), which are known to have positioned nucleosomes^37,38. We expect TMP to preferentially intercalate in linker DNA regions between nucleosomes in samples with and without torsion, and the ability to reveal these well-positioned nucleosomes is a measure of our TMP data quality. We found a striking similarity between the signals from the two samples (Fig. 1b). This finding demonstrated that the in vivo psoralen signal is likely dominated by accessibility so that TMP intercalation preference due to torsion may represent a smaller contribution to the overall signal than protein-mediated DNA accessibility. Therefore, using the TMP intercalation signal alone to indicate torsion may be prone to interpretation artifacts.

For both datasets, we detected a TMP peak upstream of the TSSs, with the peak amplitude being slightly larger for the condition with torsion. This peak coincides with the nucleosome-free regions (NFRs), which should be more accessible to TMP (Fig. 1c). Within the gene bodies, we also observed a clear ∼160 bp periodic signal, similar to that of the positioned nucleosomes near the promoters (Fig. 1c). However, the periodic pattern anti-correlates with H3-ChIP-MNase seq data, indicating that psoralen prefers to bind in the linker regions between nucleosomes or in NFRs (Fig. 1c)²⁷. The decay in the amplitude of the somewhat sinusoidal TMP signal away from the TSS mirrors the decay in the amplitude of the H3-ChIP-seq signal, reflecting a decrease in nucleosome phasing further from the promoters³⁸. Immediately downstream of the transcription end sites (TESs), we observed a smaller TMP peak than before the TSSs, likely due to the presence of promoter regions associated with a downstream codirectional gene. These findings show that the TMP intercalation distribution reflects many features previously observed using other chromatin accessibility assays and suggest that the TMP signal may be dominated by chromatin accessibility.

To determine psoralen intercalation due to DNA torsion, we subtracted the zero-torsion baseline from the in vivo distribution, eliminating contributions from sequence preference and accessibility (Fig. 2a,b). The resulting signal has a ∼ 5-fold reduced peak-to-valley amplitude and should reflect contributions solely due to torsional stress. Most of the nucleosome signal was abolished within the gene bodies, indicating that most accessibility bias has been removed, although a small residual remained (Fig. 2b). We refer to the corrected signal as the “torsion signal,” with a positive value indicating (-) torsion, which facilitates psoralen intercalation, and a negative value indicating (+) torsion, which restricts psoralen intercalation. Using this approach, we examined whether we could detect the torsional state more accurately during transcription. To examine the torsion distribution along a gene, we focused on regions near promoters and terminators. We found the torsion signal now reveals (-) torsion upstream of promoters and (+) torsion downstream of terminators (Fig. 2b). This may be interpreted as a statistical average of a bulk population of Pol II molecules, which have elongated to random locations along the gene in different cells. Since each Pol II molecule generates (-) torsion behind and (+) torsion ahead, according to the twin-supercoiled-domain model, the net torsion signal from different Pol II molecules averages to near zero, except at gene boundaries. We also found that torsion is confined to within 1-2 kbp of the promoters and terminators, supporting previous findings that transcription-dependent dynamic supercoiling is a short-range genomic force³².

To further examine whether the torsion signal is due to transcription, we classified genes into percentile groups based on their expression level (Fig. 2c,d). We expected that the average transcription-generated torsional stress should increase with increased gene expression. Thus, we sorted the genes based on their expression level according to previous mRNA-seq data³⁹. Consistent with this prediction, genes in the bottom 20% of the expression level do not display a significant peak near the promoters nor a significant valley near the terminators. As the gene expression level increases, the torsion signal amplitudes near the promoters and terminators monotonically increase on average (Fig. 2c,d). As expected, gene expression strongly correlated with Pol II presence along the genes (based on previous ChIP-seq data of Pol II²⁷). This shows that active transcription accumulates torsional stress near highly transcribed genes, supporting earlier findings from in vitro and in vivo studies^{2,4–7,40–42}.

The persistence of torsion near genes, even with full native topoisomerases, indicates that topoisomerases cannot fully keep up with transcription activities. To examine the correlation of torsional stress with topoisomerase activity, we analyzed previous ChIP-seq data of topoisomerase I (topo I) and topoisomerase II (topo II)²⁷, which are the primary enzymes for relaxing transcription-generated torsion. Topo I binds near promoters and terminators, and its enrichment level correlates with an increase in the torsion signal (Fig. 2c). In contrast, topo II binds near promoters but not near terminators. As the torsion signal increases, the topo I and topo II enrichment levels also increase, suggesting that torsion might facilitate topoisomerase recruitment or retention. Unlike topo I, which binds to one DNA segment, topo II relaxes DNA via a strand-passage mechanism⁴³ and has been found to prefer binding to a DNA crossover^13,44–47. Because the DNA in a nucleosome wraps around the histone octamer in a left-handed fashion, the nucleosome conformation exhibits a chiral response to DNA supercoiling^48,49. While (-) torsion readily enables DNA crossing formation via the juxtaposition of the entry and exit DNA segments in a nucleosome, (+) torsion displaces entry and exit DNA segments before bringing them together again, buffering torsional stress^14,48. Thus, there is less DNA crossing at a nucleosome available to allow topo II binding. Consistent with this interpretation, we previously found that topo II has a faster relaxation rate and is more processive on chromatin under (-) supercoiling than under (+) supercoiling⁴⁴. Thus, the absence of topo II near the terminators may not be due to any active exclusion of topo II at those regions but is a natural consequence of the interplay between the DNA chirality and nucleosome chirality.

Previous studies suggest torsion accumulates more significantly over longer genes⁵. As Pol II elongates, its rotation may be more restricted with a lengthening in the RNA transcript, which is known to be associated with large machinery, such as spliceosomes⁵⁰. Thus, torsional stress might significantly increase toward transcription termination for longer genes. We grouped data based on gene size (Fig. 2c,d), and consistent with this prediction, the (-) torsion near the promoters remained unchanged with an increase in gene size, but the (+) torsion near the terminators increased with gene size to a moderate extent. As expected, we found no significant dependence of topo I or topo II enrichment on gene size near the promoters. However, we noticed a similar lack of dependence near the terminators, likely due to the moderate dependence of torsion on gene size.

Transcription-generated torsion may dissipate over distance due to topoisomerase relaxation and DNA end rotation^6,32,51. If so, torsion generated from one gene will most likely impact the torsional state of neighboring genes. For example, (-) torsion may more likely accumulate near promoters of two divergent genes, while (+) torsion may more likely accumulate near terminators of two convergent genes. To examine this prediction, we plotted the torsion signal in three gene-pair configurations (Fig. 3). We mapped the average behavior of torsional stress in the intergenic regions after rescaling the intergenic distance between nearest neighbors. We found that between promoters of two divergent genes, the torsion signal rises throughout the intergenic region, consistent with (-) torsion accumulation between these gene pairs. Conversely, we observed that between terminators of converging genes, the torsion signal drops throughout the intergenic region, consistent with (+) torsion accumulation between these gene pairs. When two neighboring genes are oriented codirectionally, the torsion signal drops after the terminator of the trailing gene and rises before the promoter of the leading gene, consistent with generation of (+) torsion by the trailing gene followed by generation of (-) torsion by the leading gene. Thus, by grouping neighboring genes in these configurations, we demonstrate how torsion accumulates between neighboring gene pairs.

Since transcription-generated torsional stress hinders Pol II progression, topoisomerases must be recruited to active genes to resolve the torsional stress. To examine where Pol II, topo I, and topo II are located, we analyzed previous ChIP-seq data of Pol II, topo I, and topo II²⁷. We show that Pol II has a minimal presence in the intergenic region between divergent genes and is enriched in the intergenic region between convergent genes, consistent with a previous finding that after termination, Pol II tends to remain on the DNA downstream of the terminator^27,39.

We also find that topo I is enriched in all intergenic regions with significant (+) torsion or (-) torsion, indicating that topo I effectively targets regions under torsional stress. In contrast, topo II localizes to the intergenic region between divergent genes but is depleted from the intergenic region between convergent genes, indicating that topo II prefers regions with (-) supercoiling but is restricted from regions with (+) supercoiling, further supporting our interpretations above. These data further demonstrate how topo I and topo II may differentially target diverse environments in the genome.

Maintenance of three-dimensional genome structure is a crucial step in cellular function. Recent yeast genome studies have found mitotic loops near the boundaries of convergently oriented gene pairs^23,24,52. This suggests that (+) torsional stress may regulate genome folding. We examined the signal near loop boundaries (Fig. 4a) by aligning all loop boundaries and grouped signals based on the looping strength of previously published MicroC-XL Data⁵². This sorting method showed that cohesin²⁴ is enriched at the two-loop boundaries as expected. Using this sorting method, we found (+) torsion signal at the loop boundaries, which also coincide with regions of convergent gene pairs. We found that (+) torsion correlates with strengthening cohesin loop boundaries, suggesting that (+) torsion in these regions may facilitate cohesin recruitment (Fig. 4a, cartoon). These data provide evidence that the (+) supercoiling generated by transcription may facilitate genome folding in coordination with other participating proteins.

While a single (+) torsion intergenic region is present between two convergent genes, we detect two (-) torsion intergenic regions between two divergent genes (Fig. 3). The presence of the two (-) torsion peaks indicates that (-) torsion from these promoters is somewhat decoupled. However, it is unclear what limits the diffusion of (-) torsion between the two promoters. Thus, we systematically examined the promoter regions of different promoter architectures to identify factors that may play a role.

Most promoters in S. cerevisiae are not regulated by ssTFs and thus unbound by anything besides a PIC (termed “UNB”)^27,53. Interestingly, our data show a single elevated (-) torsion region between this UNB class of divergent promoters (Fig 4b, upper and middle plots), suggesting that torsion generated from one gene can impact the expression of its neighboring gene, consistent with previous findings that the expression of these genes is coupled^27,53.

Another large class of S. cerevisiae promoters (termed “TFO”) are regulated by insulator ssTFs, such as Reb1 and Abf1, which decouple interactions between neighboring genes^{27,28,53–55}. A minority of genes, such as inducible promoters (termed “STM”), are even more highly regulated by a host of additional transcription factors and usually contain an upstream activating sequence²⁷. Strikingly, these bound promoters show two distinct (-) torsion peaks, suggesting that the insulators can insulate torsion between the two genes (Fig 4b, upper and lower plots). This result indicates that bound insulators may be anchored to cellular structures in a manner that prevents rotation of the bound DNA segments and thus blocks the progression of torsion from one promoter to the next. Given that DNA unwinding facilitates transcription initiation, insulators may prevent torsional crosstalk between adjacent divergent promoters while also preventing TATA-binding protein (TBP) from sliding back and forth between promoters.

The Saccharomyces cerevisiae strain BY4741 (MAT a his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 bar1::NatMX his3Δ1::HIS3-ADH1pr-hENT-GPDpr-HSV-TK)⁵⁸ was used for all experiments. Cells were grown at 30°C in YPD media until the OD600 reached ∼0.6, and then processed for different conditions as described below.

Cells designated as zero torsion state controls were crosslinked with 1% (v/v) formaldehyde (Fisher, F79-500) for 15 minutes at 30°C. Formaldehyde was quenched with 0.4 M glycine (Sigma, 357002) for 5 minutes at room temperature. Following quenching, cells were pelleted and washed 2X with PBS at 4°C. Next, cells were spheroplasted via 0.01% (v/v) 10 mg/µl zymolyase (Sunrise Science Products, N0766391) treatment in 1 M sorbitol (Sigma, S6021), 0.5 mM 2-mercaptoethanol (Sigma, 444203) for 20 minutes. Spheroplasts were confirmed to be >90% of cells using microscopy. Formaldehyde-fixed spheroplasts were centrifuged at 1,000 g for 10 minutes at 4°C, washed with 25 ml of cold PBS (Thermo, AM9624), and resuspended in 100 µl of 1x DpnII reaction buffer (NEB, R0543) containing 100 units of DpnII (NEB, R0543) and supplemented with 0.1% (v/v) NP-40 (Thermo, 85124), and 10 µM TMP (Sigma, T6137). Samples were incubated for 2 hours at 37°C followed by TMP photobinding as described below.

The 4,5′, 8-trimethylpsoralen (TMP) was dissolved in 100% ethanol at a final stock solution of 2 mM. TMP was added to growing cells at a final concentration of 10 µM and cells were further incubated at 30°C for 1 hour. Twenty-five mL of cell solution was placed in a 10 cm petri dish and exposed to ∼0.15 W/cm² of 365-nm light for 20 seconds. Directly after TMP photobinding, the cells were fixed in 95 ml 70% −20 °C ethanol and placed at −20°C overnight.

Cells were then harvested by centrifugation at 10,000 g for 10 minutes, resuspended in 500 µl of 50 mM sodium citrate (Sigma, 567446), pH 7.4, and incubated at room temp for 10 minutes. Centrifugation and wash steps were repeated once more, followed by a 20-minute Zymolyase treatment as detailed above. Spheroplasts were harvested by centrifugation and resuspended in 100 µl of lysis buffer (300 mM NaCl (Fisher, BP358), 50 mM EDTA (Fisher, AM9260G), 2% SDS (Sigma, 71725), and 5 % (v/v) RNase A (Thermo, 12091021) for ∼2 hours and then vortexed for 2 minutes. 10 µl of Proteinase K (NEB, P8107S) was added, and cells were incubated for 2 hours at 65 °C.

Zero torsion samples followed a similar photobinding protocol but were exposed to UV light at ∼0.35 W/cm² for 10 seconds in a 1.5 ml Eppendorf tube. Samples were harvested at 10,000 g at 4°C, the supernatant was removed, and they were resuspended in 100 µl of Lysis buffer containing 10 µl Proteinase K. The zero torsion samples were incubated at 65°C overnight to reverse formaldehyde crosslinks. After lysis, genomic DNA was extracted from all samples via 2X phenol-chloroform (Sigma, 77617) extraction followed by isopropanol (Sigma, 59304) precipitation.

The TMP-seq protocol was adapted from previous work^30,33. Extracted genomic DNA was sonicated to ∼400 bp fragments using appropriate settings on a Covaris E200 sonicator. Following sonication, DNA was treated with 5 µl RNase A and purified using a Zymoclean DNA Clean and Concentrator-5 kit according to the manufacturer’s protocol using a 7:1 binding buffer:sample ratio. After quantification using a Biotum HS DNA quantification kit, 250 ng of DNA was diluted into 86 µl of 10 mM Tris-Cl pH 8.5, incubated at 98°C for 15 minutes, and then directly moved to ice for 2 minutes. DNA was then digested with 4 µl of ExoI (NEB, M0293) in 1X ExoI reaction buffer at 37°C for 1 hour. This digestion process was repeated by incubating at 98°C for 1 hour and then adding an additional 4 µl of ExoI followed by another incubation at 37°C for 1 hour. Digested samples were purified using a Zymo DNA Clean and Concentrator-5 kit using a 7:1 binding buffer: sample ratio. DNA was again quantified using the Biotum HS DNA assay. Illumina adaptors were ligated to the digested samples using the NEBNext Ultra II DNA library prep kit for Illumina (NEB, E7645) at half the standard total reaction volume of the manufacturer’s protocol. Adapter ligated DNA was purified with 2X Ampure beads (Beckman Coulter) and eluted in 40 µl of 10 mM Tris-Cl pH 8.5. DNA was stored at −20 C until further processing.

To locate psoralen interstrand crosslinks, DNA from the previous step was incubated with 5 µl of lambda exonuclease (NEB, M0262) in the provided buffer in a 50 µl total reaction volume at 37°C for 1 hour, purified with 1.8X Ampure beads, and eluted with 30 µl of 10 mM Tris pH 8.5. Lambda-digested DNA was then subjected to 10 rounds of primer extension with Q5 DNA polymerase (NEB, M0491), using 10 nM of the primer (ID): 5’-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATC*T on a thermocycler (98°C for 10s ◊ 10 rounds (98°C, 10s ◊ 65°C, 2 min)◊ 65°C, 5min ◊ 95°C 1min ◊ 4°C). Products were purified using 1.8x Ampure beads and eluted in 40 µl 10mM Tris pH 8.5. Primer extension products were ribotailed using 1.4 µl Terminal Transferase (NEB, M0315) and 0.4 mM rGTP (Roche, 11277057001) in Terminal Transferase buffer supplemented with 0.25 mM CoCl₂ by incubating at 37°C for 30 minutes followed by incubation at 75°C for 20 min. Ribotailed products were purified with 1.8X Ampure beads, eluted with 30 µl 10mM Tris pH 8.5. Purified ribotailed products were ligated with 0.33 µM annealed adapter oligos (IDT) (Top: 5’p-AGA TCG GAA GAG CGT CGT GTA GGG AAA GAG TGT-3’, Bottom: 5’ ACA CTC TTT CCC TAC ACG ACG CTC TTC CGA TCT CCC −3’) using 1.2 µl T4 DNA ligase (NEB, M0202S) in the provided buffer for 1 hour at 16 °C, and subsequently purified with 1.8X Ampure beads and eluted in 20 µl 10mM Tris pH 8.5. Finally, products were indexed and amplified using NEBNEXT Multiplex Oligos for Illumina set 1 (NEB, E7600) following the manufacturers protocol with a total of 12 PCR cycles. The PCR products were purified using 0.8X Ampure beads and eluted in TE (Invitrogen). Samples were pooled at equimolar concentrations and empty adapters were removed from the pool.

Pooled libraries were sequenced on a NextSeq 2000 (Illumina) using a NextSeq 2K P2 100 bp kit (Illumina) in paired-end mode with a 63 bp Read_1 and 59 bp Read_2. Sequencing, demultiplexing and Fastq generation were performed by the Cornell Biotechnology Resource Center. Fastq files were aligned to the SacCer3 Genome using BWA-mem2⁵⁹ with default settings for paired-end reads. Sam files were converted to BAM files followed by fixmate and duplicate marking using Samtools⁶⁰. Duplicates were again marked using picard-MarkDuplicates⁶¹, and duplicates were removed using Samtools-view. Bedgraph files were generated using deeptools⁶² bamCoverage with 10 bp bins using the following settings (--samFlagInclude 64--Offset 1) yielding the coverage of the 5’ end of Read_1 where the ICL was located.

Bedgraph files were further processed in MATLAB to ensure one bin per 10 bp interval, and normalized to the mean number of reads per bin. Replicate bedgraphs were averaged, and indicated subtractions were done using MATLAB. Normalized bedgraphs were converted to bigwig files using ucsc-bedGraphToBigWig⁶³. Heatmaps and composite plots were generated using deeptools computeMatrix, and visualized using Origin Lab. Bed files containing coordinates of verified s. cerevisiae ORFs were downloaded from the Saccharomyces Genome Database (SGD)^64,65, and a region blacklist was downloaded from the yeastepigenome.org²⁷. Bed files for each gene configuration in Figure 3 were generated by considering the boundaries of nearest neighbor verified ORFs and classifying them based on their strand configuration. Regions that had any annotations according to the list of all yeast genes from the SGD were removed to isolate true intergenic regions. The resulting Bed files were used to make heatmaps and composite plots as described above.

BAM files of ChIP-Exo, and H3-ChIP-MNase data was downloaded from yeastepigenome.org²⁷. For ChIP exo, bigwig files were calculated using the 5’ end of read 1 as in²⁷. Signals were then normalized to the mean number of reads per 10bp bin. For H3-ChIP-MNase, the center of the fragment was considered. Top1 and Top2 ChIP-Exo data were further processed by subtracting the No-TAP control (ChIP-Exo background). Previous Scc1 ChIP-seq IP and Input DNA data was processed by downloading bigwig files from GSE235560²⁴. Deeptools bigwigCompare was used to normalize Scc1 ChIP/Input DNA.

mRNA-seq data was downloaded from GSE76142³⁹ in bigwig format. The minus strand signal was subtracted from the plus strand signal. Expression level for a gene was calculated using mean absolute value of the subtracted signal within the ORF, bed files containing expression levels were used to separate data into percentile groups.

To determine positions of intrachromosomal loop boundaries and their scores, mcool files of MicroC-XL data were downloaded from GSE151553⁵². Boundaries were located and scored using cooltools dots⁶⁶ with a 500 bp bin size. Loop calls were refined by only considering the smallest loops in nested configurations, thereby only representing each region once. Boundary positions were adjusted by locating the position of the nearest (<500 bp) Scc1-ChIP-seq peak on either loop boundary.

Raw and processed sequencing data files have been deposited to GEO under accession GSE284352. The custom code used to process the raw data and generate plots is available on GitHub (https://github.com/WangLabCornell/Hall_et_al).