Abstract
It is debated whether the pervasive intergenic transcription from eukaryotic genomes has functional significance or simply reflects the promiscuity of RNA polymerases. We approach this question by comparing chance promoter activities with the expression levels of intergenic regions in the model eukaryote Saccharomyces cerevisiae. We build a library of over 105 strains, each carrying a 120-nucleotide, chromosomally integrated, completely random sequence driving the potential transcription of a barcode. Quantifying the RNA concentration of each barcode in two environments reveals that 41–63% of random sequences have significant, albeit usually low, promoter activities. Therefore, even in eukaryotes, where the presence of chromatin is thought to repress transcription, chance transcription is prevalent. We find that only 1–5% of yeast intergenic transcriptions are unattributable to chance promoter activities or neighboring gene expressions, and these transcriptions exhibit higher-than-expected environment-specificity. These findings suggest that only a minute fraction of intergenic transcription is functional in yeast.
Similar content being viewed by others
Introduction
Many eukaryotes show pervasive transcriptions of intergenic regions1,2,3,4. For example, although protein-coding regions make up only ~1.5% of the human genome and genic regions constitute about one-third of the genome, RNA transcripts are detected from >75% of the genome4. The biological significance of intergenic transcription, however, remains controversial5,6. The functional hypothesis asserts that intergenic transcripts largely result from the transcriptions of uncharacterized protein-coding genes or functional noncoding genes3,7,8. Indeed, some long intergenic noncoding RNAs (lincRNAs) are functional9, although the functionality of the vast majority of annotated lincRNAs is unclear10. Furthermore, it has been suggested that, even if the transcript of a noncoding gene is functionless, the act of transcription may regulate the expressions of other genes11,12,13. By contrast, the nonfunctional hypothesis14,15 posits that most intergenic transcripts excluding those resulting from the expressions of neighboring genes16,17 are nonfunctional products of chance promoter activities of intergenic sequences18. It is notable that, in the prokaryotic model organism Escherichia coli, ~10% of random sequences of ~100 nucleotides possess promoter activities comparable to that of a functional promoter and another ~60% can become such a promoter with just one mutation19. In the eukaryotic model organism Saccharomyces cerevisiae, noncoding RNAs can arise from nucleosome-depleted genomic regions by the promiscuous binding of RNA polymerase II16,17,18,20, but neither the probability with which a random intergenic sequence drives transcription nor the resulting transcriptional level is known, especially in the presence of chromatin that substantially represses transcription21. Most lincRNAs are not evolutionarily conserved22,23, which could mean a lack of function6 or a rapid turnover of lineage/species-specific function22. Intraspecific analysis yielded mixed results on the selective constraints on lincRNAs24.
In this study, we test the nonfunctional hypothesis by characterizing the frequency distribution of promoter activities of 120-nucleotide random sequences in yeast and comparing it with the frequency distribution of yeast intergenic expressions; the functionality of intergenic expression is invoked if the expression exceeds the chance expectation. Although the median length of yeast promoters is 455 nucleotides25, a promoter as short as 69 nucleotides can be twice as strong as the CYC1 promoter26 and a 116-nucleotide synthetic promoter is similarly strong as the TDH3 promoter27. CYC1 and TDH3 are among 30% and 0.1% of the most highly expressed yeast genes, respectively. Hence, sequences of 120 nucleotides have ample opportunities to possess promoter activities yet are not too long to lower the experimental efficiency or increase the length variation of the synthesized oligonucleotides (because of the relatively high rates of insertion/deletion errors in oligonucleotide synthesis).
Our experiment differs from past studies of eukaryotic mutant promoters in both design and purpose. We investigate the promoter activities of completely random sequences, while past studies examined activities of promoters that were created by mutating a native promoter28,29 or were built on a core promoter scaffold27,30,31,32,33. We aim to estimate the probability distribution of promoter activities of random sequences, while past studies aimed to identify crucial elements of a particular functional promoter or sequence features of active promoters with a canonical scaffold.
Results
Estimating the promoter activities of random sequences
We began by synthesizing oligonucleotides each comprising a 120-nucleotide completely random sequence (the random promoter) and a 20-nucleotide completely random sequence (the barcode) interleaved with invariant sequences (primer sites) (Figs. 1a, S1a). Because the genomic location has a much smaller influence than the promoter strength on the gene expression level34, we integrated the above oligonucleotides to an intergenic position in the yeast genome that permits sensitive quantification of promoter activities (Fig. S2, Table S1), using CRISPR/Cas9 in a large-scale transformation (see Methods). A CYC1 terminator was placed upstream of the random promoter to minimize the influence of any upstream transcriptional activity (Fig. 1a). We respectively created a negative control where the random promoter was absent and a positive control where the random promoter was replaced with the promoter of the yeast PSP2 gene. We constructed eight versions of each of the negative and positive controls using different barcodes to confirm the reliability of barcode expression measurements.
a Experimental procedure. b, c DNA (b) and cDNA (c) counts of each barcode in two experimental replicates. For clarity, 1000 randomly sampled barcodes are shown. Pearson’s correlation (r) and its associated P value based on all barcodes are presented. d Estimated barcode expression levels in two replicates. Expression level is measured by the cDNA count divided by the DNA count of the barcode. The same 1000 barcodes as in (b) and (c) are shown. Pearson’s r and associated P value based on all barcodes are presented. The inset shows the expression levels of 1% of genotypes with the highest cDNA counts and associated statistics. e Expression levels measured by bulk sequencing is strongly correlated with those measured by RT-qPCR in five reconstructed genotypes, a randomly picked negative control, and a randomly picked positive control. Mean expressions and standard errors are shown by dots and error bars, respectively. Pearson’s correlation (r) and the associated P value between the two measurements are presented. All P values are from two-tailed tests. Source data are provided as a Source Data file.
The library, along with the controls, was cultured in three replicates in a rich medium (YPD) or a minimal medium (SCD) to the exponential growth phase. We extracted and amplified DNAs from barcodes (Fig. S1b) and sequenced them using 150-nucleotide paired-end Illumina sequencing. From the same samples, we extracted total RNAs, reverse-transcribed mRNAs from the barcodes, and sequenced the corresponding cDNAs using the same platform. The number of cDNA reads divided by the number of DNA reads for each barcode, upon normalization, is an estimate of the barcode expression level and the activity of the corresponding random promoter (Fig. 1a). From the library, we also Illumina-sequenced the insert to determine the sequence of the random promoter linked with each barcode (Fig. 1a).
We focused on barcodes with at least 100 DNA reads in each replicate to allow relatively precise estimation of their expression levels; 49,169 and 146,291 barcodes passed this criterion in YPD and SCD, respectively. In YPD, the barcode DNA read number is highly correlated between replicates (Figs. 1b, S3a, b), while the cDNA read number is less well correlated (Figs. 1c, S3c, d), and the expression level correlation is even weaker (Figs. 1d, S3e, f). The reduced correlation in expression level is due to the existence of many lowly expressed barcodes; the correlation is substantially higher when the 1% of the barcodes with the most cDNA reads are examined (insets in Figs. 1d, S3ef). The same is true in SCD (Fig. S4a–i). When culturing the yeast library, we included high fractions of controls, resulting in a high expression correlation across replicates for the controls (Fig. S5). To verify the bulk sequencing-based promoter activity estimation, we selected several promoters with a wide range of activities but low across-replicate variations, reconstructed them, and measured their expressions individually by reverse transcription-quantitative polymerase chain reaction (RT-qPCR; see Methods). Expression estimates from RT-qPCR agreed well with those from bulk sequencing (Figs. 1e, S4j).
A large proportion of random sequences have promoter activities
By comparing with the negative control, we found that 63.2% and 41.4% of the random sequences have significant promoter activities in YPD (Figs. 2a, S6a, b) and SCD (Figs. 2b, S6a, b), respectively. Because our positive control—PSP2—may not have the same expression level in different environments, hereinafter we use the median expression level of all yeast native genes in the relevant medium as the reference (by comparing the PSP2 expression level with the reference in published RNA-seq data; see Methods). About 0.024% and 0.029% of the random sequences have significantly higher promoter activities than the reference in YPD and SCD, respectively (Figs. 2a, b, S6c). Similar results were obtained when different cutoffs higher than 100 DNA reads per barcode were used in analyzing barcode expressions (Fig. S6). Most (90%) random promoters have activities below the 21st (or 12th) percentile of yeast native promoter activities in YPD (or SCD), while the strongest random promoter observed is comparable in activity to the 85th percentile of the native promoters in both growth conditions (Fig. 2c, d, Table S2).
a, b Frequency distribution of the barcode expression level relative to the reference (median expression level of yeast native genes) in YPD (a) or SCD (b). The left and right vertical dashed lines indicate the mean of the negative control (no promoter) and the reference, respectively. Frequency is shown by bars whereas the cumulative frequency is indicated by the gray line. Note the difference between the left and right Y-axis scales. c, d Frequency distributions of expression levels of barcodes and yeast native genes relative to the reference in YPD (c) or SCD (d). In (c) and (d), negative infinite values come from the log10 transformation of the expression levels for barcodes with zero cDNA counts. Source data are provided as a Source Data file.
Features associated with the random promoter strength
Identifying sequence features associated with the promoter strength is important for understanding the mechanistic basis of the promoter activity and for synthetic biology27,30,31,32,33,35. In both YPD (Fig. 3a) and SCD (Fig. S7a), a positive correlation exists between random promoter strength and promoter GC content (see Methods). Interestingly, for yeast native promoters, such a positive correlation exists only for relatively strong promoters; for relatively weak promoters, the correlation is negative (Figs. S8a, b, S9a, b). About 20% of yeast native promoters contain TATA boxes36. We found that, for both random (Figs. 3b, S7b) and native (Figs. S8c, d, S9c, d) promoters, there is a positive correlation between promoter strength and TATA-box presence.
a The GC content in a random promoter increases with the promoter strength. The promoters are divided into 50 equal-size bins by their strengths. Shown are the mean GC content of each bin, with the shaded area indicating the 95% confidence intervals of the mean GC content. Spearman’s rank correlation between the promoter strength and GC content for unbinned data is ρ = 0.08 (P = 5.6 × 10−65). b Quantile-quantile plot showing the probability distributions of strengths of random promoters with and without TATA boxes. The dots show the 0th, 5th, 10th, … and 95th percentiles of the data in promoter strength. The strengths of promoters with TATA boxes are significantly higher than those without (P = 0.01, Wilcoxon rank-sum test). c Schematics illustrating the binding of a TF to a promoter in two directions (the TFBS being GGTGT). d Frequency distribution of P values from Wilcoxon rank-sum tests of equal activities between random promoters with binding sites of a particular TF (on the forward strand) and those without. The blue color indicates P values from the original data, whereas the gray color indicates P values from the data in which the promoter activity is randomly shuffled among promoters. e TF effects on the forward and reverse orientations. The effect of a TF in an orientation is measured by the median activity of promoters with the binding sites of the TF in the orientation concerned relative to that of promoters without the binding sites in the orientation concerned. Each dot represents a TF, and colors indicate results from Wilcoxon rank-sum tests. Number of TFs belonging to each of the four categories is shown in the parentheses. All P values are from two-tailed tests. Source data are provided as a Source Data file.
Due to their short and degenerate sequences, transcription factor (TF) binding sites (TFBSs) can easily arise in a random sequence31. Based on 196 known yeast TFs and their TFBSs37, we found on average 21 forward and 24 reverse TFBSs (Fig. 3c) per random promoter (see Methods). For each TF, we statistically tested if random promoters with and without its TFBSs have significantly different activities. At the false discovery rate (FDR) of 0.05, the promoter activity is significantly influenced by the forward TFBSs of 111 TFs and reverse TFBSs of 114 TFs in YPD. The corresponding numbers are 64 and 75, respectively, in SCD. For each medium and orientation, the distribution of the P values from the above tests is highly left skewed for real data but is approximately uniform upon the shuffling of promoter strengths among promoters (Figs. 3d, S10), confirming the genuine impacts of TFBSs. Previous work showed that some TFBSs have orientation-specific effects in promoters with canonical scaffolds31,33. We assessed the expression effect of each TF by the median activity of random promoters with corresponding TFBSs, relative to that without them. Overall, the effects of a TF in the forward and reverse orientations are only weakly positively correlated (Figs. 3e, S11), with many TFs showing orientation-specific effects and many showing orientation-independent effects (Figs. 3e, S11). Gene expression is often environment-dependent because of environment-dependent TF expressions or actions. When we focused on the 92 TFs with significant effects in both YPD and SCD (regardless of orientation), only three TFs showed opposite effects in the two media (Fig. S12), suggesting that opposite actions of the same TF between two environments are rare. Fig. S13 shows three examples of strong random promoters, with perfectly matching TFBSs indicated.
Most yeast intergenic expressions are explainable by chance promoter activities or neighboring gene expressions
To understand yeast intergenic expressions in the light of chance promoter activities, we first examined two existing RNA sequencing datasets38,39 of the same yeast strain and similar growth conditions as in our random promoter experiments. To allow fair comparisons with the expressions of 20-nucleotide barcodes, we divided each genic or intergenic region into 20-nucleotide contiguous windows, estimated the expression level of each window (Fig. S14a), and validated this measurement by benchmarking with the canonical estimates of genic expressions (Fig. S14b; see Methods). Genic as well as intergenic expressions are highly correlated across replicates (Fig. S14c, d). We subsequently generated the expression distribution of intergenic windows (Fig. 4a, b). Notably, only 0.8% and 1.3% of intergenic windows are significantly more highly expressed than the reference in YPD and SCD, respectively.
a, b Frequency distribution of relative expression levels (RELs) of intergenic windows in YPD (a) or SCD (b). REL equals the expression level divided by the reference (median expression level of yeast native genes) under the same environment. Frequency is shown by bars whereas the cumulative frequency is indicated by the gray line. The vertical dashed line indicates REL = 1. Note the difference between the left and right Y-axis scales. c Fractions of barcodes or intergenic windows with RELs in YPD significantly higher than various cutoffs. For instance, the bin of 0.6 includes all barcodes or intergenic windows with RELs significantly exceeding 0.6. d Fractions of intergenic windows whose YPD expressions are attributable (blue) or unattributable (gray) to chance promoter activities. For example, the bin of 0.5–0.6 includes intergenic windows with RELs significantly higher than 0.5 but not significantly higher than 0.6. Only bins containing gray areas are shown. e Fraction of barcodes or intergenic windows with RELs significantly exceeding a cutoff in YPD that have RELs significantly exceeding the same cutoff in SCD. Symbols follow (c). Source data are provided as a Source Data file.
To directly compare the expression levels of intergenic windows with those of the barcodes, we computed relative expression levels (RELs) by dividing the raw expression levels by the reference. We then compared the fraction of intergenic windows whose RELs significantly exceed various cutoffs with the corresponding fraction of barcodes. While both fractions decrease with the cutoff, the latter drops more quickly than the former and becomes smaller than the former when the REL cutoff is 0.3 in YPD (Fig. 4c) and 0.1 in SCD (Fig. S15a). We then computed for each expression bin the proportion of intergenic windows whose expressions can or cannot be explained by random promoter activities. In YPD, all intergenic expressions with RELs not significantly higher than 0.5 are explainable by chance promoter activities. Starting from the REL bin of 0.5–0.6 (i.e., REL is significantly higher than 0.5 but not significantly higher than 0.6), we observed larger and larger proportions of intergenic windows whose expressions cannot be explained by chance promoter activities (Fig. 4d). Overall, 1.7% of intergenic windows could not have their YPD expressions explained by chance promoter activities. The corresponding value is 7.6% in SCD (Fig. S15b).
Although we have minimized the potential influence of neighboring gene expressions on intergenic expressions by extending the 5′ and 3′ untranslated regions (UTRs) of neighboring genes (see Methods), it remains possible that some intergenic expressions reflect the bi-directional promoter activities or transcriptional readthroughs of neighboring genes. Indeed, we observed a significant positive correlation between the expression level of an intergenic region (i.e., the mean expression level of all windows in the intergenic region) and the mean expression level of its two neighboring genes (Fig. S16; see Methods). We progressively excluded intergenic windows with the highest neighboring gene expressions till the potential influence of neighboring gene expressions was no longer significant (Fig. S17; see Methods). Afterwards, only 1% and 5% of intergenic windows have expressions unexplained by chance promoter activities in YPD and SCD, respectively.
Another confounding factor is that we studied random promoter sequences of 120 nucleotides while an intergenic region can be longer than 120 nucleotides. More importantly, the expressions of different windows in an intergenic region are likely interdependent. To circumvent these problems, instead of using each intergenic window as a unit, we used each intergenic region as a unit and estimated that 2.9% and 4.9% of intergenic regions have expressions unexplainable by chance promoter activities or neighboring gene expressions (Tables S3, S4; see Methods).
Higher-than-expected environment-specificity of a minority of intergenic expressions
The activities of a promoter in different environments could be correlated. For example, 36.3% of barcodes have RELs significantly exceeding 0.1 in YPD (Fig. S18a). Among barcodes with RELs significantly exceeding 0.1 in SCD, however, 64.8% have RELs significantly exceeding 0.1 in YPD (Fig. S18c), indicating nonindependent (or shared) barcode expressions in the two environments. Such nonindependence also exists for intergenic expressions (Fig. S19). If higher intergenic expressions are more likely to be functional (Fig. 4d), we might also expect them to show less sharing between environments as a result of environment-specific demand of function. Indeed, the shared fraction of intergenic windows decreases with the REL cutoff, while no such trend exists for barcodes, which have no functional expressions (Figs. 4e, S15c).
Variation among replicates
Throughout the analysis, we merged the data from the three biological replicates when measuring random promoter activities (i.e., combined analysis). To assess the variation among the replicates, we individually analyzed the data from each replicate (i.e., individual analysis). The results from the individual analysis are overall similar to those from the combined analysis (Table S5). Furthermore, the active promoters (i.e., those driving significantly higher expressions than the negative control) discovered from each replicate largely overlap with those discovered from the combined analysis (Fig. S20).
Discussion
In summary, we found that 41–63% of 120-nucleotide random sequences have significant promoter activities in S. cerevisiae, demonstrating the easiness for a random sequence to be transcribed by chance even in eukaryotes. However, the probability is ~0.025% for a random promoter to be significantly stronger than the median promoter activity of yeast native genes, contrasting the observation in E. coli where 2.5% of random promoters are as strong as the induced lac promoter19,40, which ranks in the top 3% of all E. coli native promoters in strength. This disparity could be due to the chromatin structure in eukaryotes21 and/or the lack of consensus sequence in yeast that is analogous to the short motifs bound by the canonical σ70-RNAP in E. coli40. Indeed, although random promoters with TATA boxes tend to be stronger than those lacking TATA boxes, a sizable fraction of the former (22.8% in YPD and 39.4% in SCD) do not have detectable activities (Fig. S21).
We investigated the relationship between various sequence features (nucleotide composition, TATA box, and TFBSs) and the promoter strength in the absence of a core promoter scaffold. Contrasting the observation in native promoters35,41, we found the GC content to be positively correlated with the promoter activity in random sequences, suggesting that GC content and the core promoter scaffold might interact in influencing the promoter activity. About 82% of all types of TFBSs examined had a significant effect on the random promoter activity in at least one condition, suggesting that the random promoter activity may simply require the opening of the chromatin, which can be accomplished by the binding of TFBSs by TFs42. However, we also observed pervasive orientation-specific effects of TFBSs on the promoter activity (Figs. 3e, S11). Therefore, regulatory sequences by themselves can influence expression independently of the core promoters in both orientation-dependent and orientation-independent manners.
We found that 1–5% of yeast intergenic transcriptions, especially those exhibiting relatively high expressions, are attributable to neither chance promoter activities nor neighboring gene expressions, so are putatively functional. Consistent with this finding is the observation of a higher-than-expected environment-specificity of the relatively high intergenic expressions. Due to the drastic genome size variation across eukaryotes, it is unclear whether our findings in yeast on intergenic expression are directly applicable to other eukaryotes. But our approach is likely adaptable for studying the functional significance of intergenic transcriptions in a wide variety of eukaryotes.
Methods
Media used
YPD medium: 10 g/l of yeast extract, 20 g/l of peptone, and 20 g/l of glucose. YPAD medium: YPD medium plus 80 mg/l of adenine hemisulfate. SCD medium: 1.7 g/l of yeast nitrogen base (YNB), 5 g/l of ammonium sulfate, 0.79 g/l of complete supplement mixture (CSM), and 20 g/l of glucose. SCD − Ura medium: 1.7 g/l of YNB, 5 g/l of ammonium sulfate, 0.77 g/l of CSM without uracil, and 20 g/l of glucose.
Random promoter library construction
The 120-nucleotide random promoter and 20-nucleotide random barcode were synthesized by IDT as part of 200-nucleotide oligos. In the synthesis, equal amounts of A, T, G, and C were used for the promoter and barcode regions. We used 20-nucleotide barcodes because the large barcode space (420) relative to the number of barcodes in the library means that sequencing or PCR errors are extremely unlikely to convert one barcode in the library to another one in the library. The random promoter and barcode were flanked by constant regions as primer binding sites (Fig. S1a). The primer binding site downstream of the barcode has no homologous sequence in the yeast genome, ensuring that the cDNA generated is exclusively from the mRNA of the barcode.
Genomic integration location
We compiled several RNA-seq datasets (PRJNA392312, PRJNA315924, PRJNA238899, and PRJNA239408 from NCBI) to screen for an intergenic region with a potentially high expression dynamic range. We found several candidates based on the following criteria: (1) at least 50 nucleotides long, (2) not overlapping with genes, (3) far from telomeres and centromeres, and (4) with a total of about 50–200 reads mapped to the region in 20 RNA-seq datasets in the above NCBI BioProjects. Using RT-qPCR, we quantified the expressions of the positive and negative controls (see the following section) integrated into five candidate locations (Table S1). To quantify the amount of genomic DNA contamination in the RNA sample, we performed a control experiment with no reverse transcriptase; signals in this control would arise from genomic DNA instead of RNA. Due to extremely low expressions that we are attempting to measure in this study, the above control is critical. We picked the intergenic region between HSP31 and FIT1 on Chr. IV as the site of integration of our library (Table S1). This site showed a relatively large difference in expression level between the positive and negative controls. Furthermore, virtually no signal was detected in the control experiment without reverse transcriptase (Fig. S2).
Strain construction
To improve the efficiency of CRISPR/Cas9-based integration of our library of random promoters, we first used CRISPR/Cas9 to integrate a synthetic landing pad (SLP) into the aforementioned genomic integration site of the laboratory yeast strain BY4741. The SLP contains a CYC1 terminator, used to prevent the transcription from the upstream of the integration site, and three de novo CRISPR/Cas9 targeting sites each with a 20-nucleotide Cas9 target sequence plus a three-nucleotide protospacer adjacent motif (PAM) site. Next, we integrated the random promoter library at the SLP by CRISPR/Cas9 in a large-scale liquid transformation modified from an existing protocol43. Specifically, we followed the same procedure of a 100× transformation until the plating step. Instead of plating cells onto the selective plate, we grew all the transformants in liquid culture for selection. We expected to obtain about 1000 transformants per 1× transformation, equivalent to 100,000 transformants in a 100× transformation ideally. In the end, with five parallel transformations, we acquired about 200,000 transformants (estimated from sequencing). Besides random promoters, we also created a positive control (PSP2 promoter) and a negative control (no promoter) by CRISPR/Cas9. The PSP2 promoter is one of the weakest constitutive promoters used in yeast synthetic biology44 so is suitable for comparison with random promoters, which are expected to be weak.
Five random promoters covering a large dynamic range (~50 fold) were chosen based on their relatively high consistency in expression level across replicates. Their promoter-barcode pairs were synthesized, amplified, and integrated into SLP to create these genotypes independently.
Large-scale liquid transformation
YPAD medium (250 ml) was inoculated by a single colony of the yeast strain BY4741 picked from a fresh YPAD plate; the yeast culture was incubated for 24 h at 200 rpm and 30 °C. About 1.25 × 1010 cells were added to 2250 ml pre-warmed 2× YPAD medium (initial density = 5 × 106 cells/ml). The cells were allowed to grow for 4.5 h at 30 °C and 200 rpm until the density reached at least 2 × 107 cells/ml. Standard transformation steps were then performed. 0.1% of the final transformants were plated onto SCD − URA plates to estimate the transformation efficiency, and the rest of the transformants were resuspended in 2000 ml of 2× SCD − URA liquid culture and redistributed to 250 14 ml falcon tubes each containing 8 ml of culture. The falcon tubes were put on a large roller and incubated at 30 °C for 48 h. Afterwards, 1 × 109 cells were collected, washed, and resuspended into 200 ml YPD medium and cultured for 24 hrs. This step was to remove the Cas9 plasmid and to revive cells. About 5×109 cells were then collected and diluted in 15% glycerol to a density of 5 × 107 cell/ml and were stored at −80 °C.
Library preparation and Illumina sequencing
The glycerol stocks of the random promoter library and controls were mixed in a 100:5:1 ratio of the random promoter library to the negative control to the positive control. The mixture was precultured in a 250 ml flask with 50 ml YPD at an initial density of 5 × 106 cells/ml for 24 h. The resulting culture was diluted in 20 ml YPD or SCD medium to an initial density of 5 × 106 cells/ml in a 100 ml flask, with three replicates per medium type. The cell cultures were always incubated at 30 °C with 250 rpm. After 8 hrs of culturing, genomic DNA was extracted from 3 × 108 cells per replicate using MasterpureTM Yeast DNA Purification Kit, whereas mRNA was extracted from 1.5 × 108 cells per replicate using RNeasy Mini Kit.
To retrieve the linkage information between promoters and barcodes, we used pairs of primers containing Illumina sequencing adapters to amplify the promoter-barcode cassette in the construction of the sequencing library. Using 200 ng of genomic DNA as templates, we conducted two parallel 18-cycle PCR reactions for each biological replicate. The resulting amplicons were combined, purified, and sequenced by 150-nucleotide paired-end Illumina sequencing (HiSeq 4000).
To generate DNA reads for the barcode region, we used a two-step PCR strategy. In the first step, primers with unique molecular identifiers (UMIs) were used to amplify the barcode region in a PCR reaction with only three cycles. The UMIs used were 6-nucleotide random sequences to mark individual DNA molecules. We conducted four parallel PCR reactions per biological replicate. The resulting PCR products for each biological replicate were purified and concentrated. In the second step, we used pairs of primers containing both the sample index and Illumina sequencing adapters to amply the previous products in an 18-cycle PCR reaction. The resulting amplicons were combined, purified, and sequenced by 150-nucleotide paired-end Illumina sequencing (HiSeq 4000).
To generate cDNA reads for the barcode region, we first reverse-transcribed the mRNA into cDNA from 2.4 µg mRNA per reaction (SuperScript® III First-Strand Synthesis System for RT-PCR). The cDNA was then amplified using the aforementioned two-step PCR strategy with UMIs. The resulting amplicons were combined, purified, and sequenced by 150-nucleotide paired-end Illumina sequencing (HiSeq 4000).
Notably, the number of genotypes obtained in YPD is only about one-third of that in SCD. This difference may be due to the higher growth rate of yeast in YPD than in SCD. Specifically, some cells may have a longer lag time by chance during the preculture stage, rendering their frequencies lower than those with a shorter lag time. This effect is intensified in YPD because of the higher growth rate in YPD than in SCD, reducing the number of genotypes obtained in YPD.
The total sequence space for a 120-nucleotide random promoter is as large as 4120. Whether our random library could accurately capture the distribution of chance promoter activities depends on whether our library is randomly distributed in the large sequence space. In theory, the expected Hamming distance between two random 120-nucleotide sequences is \(120[\left(1\,-\,{f}_{A}\right){f}_{A}\,+\,\left(1\,-\,{f}_{T}\right){f}_{T}\,+\,\left(1\,-\,{f}_{G}\right){f}_{G}\,+\left(1\,-\,{f}_{C}\right){f}_{C}]\), where fX is the frequency of nucleotide X. Under equal frequencies for the four nucleotides, two random promoters should differ at 90 sites on average. However, we found that our random promoter library does not have equal frequencies for the four nucleotides, probably because of the variation introduced in the oligonucleotide synthesis. Instead, we found fA = 24.3%, fT = 32.4%, fG = 25.7%, and fC = 17.6%. So, the GC content of the random promoters in our library is 43.3%, which incidentally is closer than the GC content in our design (50%) to yeast’s genomic GC content (38%). The mean Hamming distance expected from the above nucleotide frequencies is 88.78. We observed an average Hamming distance of 88.71 in our library, suggesting that the promoters in our library are randomly distributed in the large sequence space given the nucleotide frequencies.
Influences of potential errors in library construction and sequencing
Potential PCR and sequencing errors have no impact on identifying random promoters or random barcodes because the expected sequence difference between two different barcodes or promoters (see the next section) is much greater than typical PCR/sequencing errors. One drawback of the HiSeq 4000 platform is a relatively high probability of index hopping. However, index hopping is unlikely in our experiments for the following reasons. First, we used two PCRs (Fig. S1b): the first PCR added sample indices while the second PCR added Illumina adapters. The mixing of amplicons from different samples took place after two steps of PCR right before sequencing. The free adapters in the sequencing pool would not have both adapters and index sequences. Second, we performed PCR purification after each PCR, which should have removed the remaining primers.
Sequencing-based expression level estimation
For the promoter-barcode pair sequencing, we first filtered the sequencing reads to ensure that the barcodes were 20-nucleotide long and promoters were 120-nucleotide long. We clustered promoters and barcodes separately. We used Bartender45 to cluster the barcodes with a tolerance of two mismatches, and used CD-HIT46 to cluster the promoters with a tolerance of six mismatches. We allowed relatively high numbers of mismatches because the promoters (or barcodes) are completely random such that two distinct promoters (or barcodes) have an exceedingly low probability to be different by fewer than 7 (or 3) nucleotides. We discarded any barcode cluster connected with different promoter clusters. For any promoter cluster connected with multiple barcode clusters, all of these barcodes were counted toward the promoter cluster.
For cDNA and DNA sequencing of the barcodes, we counted only one of the reads when these reads shared the same UMI. Afterwards, we divided the read count of each barcode by the total read count in each sample to obtain the normalized read count (NRC) for the barcode. The expression level of barcode i is measured by NRCicDNA/NRCiDNA.
Promoter GC content and activity
Given that the GC content is positively correlated with nucleosome occupancy41 and that nucleosomes are depleted in yeast promoters47, random promoters with higher GC contents are expected to have lower activities. However, a trend opposite to this expectation was found. This trend cannot arise from the potential impact of the GC content on the efficiency of expression measurement because it was the barcode not the promoter sequence that was amplified in expression quantification.
TFBS identification and analysis
Based on the position weight matrices (PWMs) of 196 yeast TFs and the suggested cutoffs in the ScerTF database37, we identified the TFBSs of all these TFs on the forward and reverse strands of the random promoters, respectively. The forward strand is the strand with the same orientation as the barcode transcription that we intended to measure. For each TF and each strand, we separated promoters into two groups based on whether there is at least one binding site in the promoter sequence. We then tested whether the promoter activity differs between the two groups by a Wilcoxon rank-sum test.
Promoter sequences of S. cerevisiae native genes
The promoter sequences of native genes were defined by from −500 nucleotides to either the translation start site (Fig. S8) or the transcription start site (Fig. S9).
Expressions of yeast intergenic regions
We analyzed the RNA-seq data from the YPD medium generated by Chou et al. 38. Specifically, the data from 10 wild-type samples were used. The RNA-seq data from the SCD medium were generated by Berg et al. 39, and only the data from three wild-type samples were used.
The reference genome of strain S288C was dissected into genic and intergenic regions. Our genic regions are defined conventonally48 and include both protein-coding genes and RNA genes (rRNAs, tRNAs, snRNAs, snoRNAs, and ncRNAs) based on the annotations from Ensembl (http://useast.ensembl.org/Saccharomyces_cerevisiae/Info/Index) and SGD (http://sgd-archive.yeastgenome.org/sequence/S288C_reference/rna/). Intergenic regions are the entire genome subtracting protein-coding genes with their UTRs, RNA genes, centromeres, telomeres, long terminal repeats (LTRs), and LTR retrotransposons from the reference genome. We combined the annotated UTRs from two previous studies49,50 and made additional 200-nucleotide outward extensions of both 5′ and 3′ UTRs. For protein-coding genes without annotated UTRs, we allocated 200 nucleotides outward from each end of the coding sequence as 5′ and 3′ UTRs, respectively. For RNA genes, we also added 50 nucleotides to each end of the gene. By doing the above, we aimed to minimize the influence of transcription from nearby genes or special sequence features on intergenic expression measures. We also varied the UTR extension length but found the results similar (Fig. S14e).
To measure intergenic expressions, we used a sliding window approach. For each intergenic region, we divided it into sliding windows of 20 nucleotides with a step size of 20 nucleotides, starting from the site of the intergenic region with the smaller genomic coordinate. We discarded the window at the end of an intergenic region if it is shorter than 20 nucleotides. The expression level of a window is measured by the number of reads mapped to the window in a strand-specific manner normalized by the total number of reads (in millions) of the RNA-seq data. For genic regions, a similar approach was used. For all windows from the same gene, their expressions were averaged to obtain the expression estimate for the gene. To estimate genic expressions by a canonical method, we employed the software StringTie51.
Comparing the expressions of intergenic windows or barcodes with the median expression level of yeast native genes
We divided the expression level of each intergenic window by the median expression level of all annotated genes (including RNA genes) in S. cerevisiae to obtain the relative expression level (REL) in each replicate. We then used a one-sample t-test to test if the REL is significantly different from 1 at a false discovery rate (FDR) of 0.05 by Benjamini–Hochberg’s procedure52. We similarly tested if the REL differs significantly from 0.1, 0.2, and so on. This allows computing the fraction of intergenic windows with RELs significantly higher than any REL cutoff (Fig. 4c), as well as identifying intergenic windows whose RELs are significantly higher than an REL cutoff but not significantly higher than the next (higher) cutoff (Fig. 4d).
For barcode i, we first merged its NRC from the three replicates to obtain the total NRC (TNRC) and estimated its expression level Ei = TNRCicDNA/TNRCiDNA. Because eight of the barcodes are associated with the positive control, we also had eight Ei values for the positive control. For each random promoter-associated barcode i, we obtained eight estimates of its expression relative to the positive control by Ei/Ej, where j = 1 to 8 refers to the eight barcodes of the positive control. Based on the expression level of PSP2 relative to the median expression level of all yeast genes in RNA-seq data, we obtained the eight estimates of REL of each barcode, which is the expression level of the barcode relative to the median expression level of yeast genes. Finally, we used a one-sample t-test to test if the REL of a barcode is significantly different from 1 at FDR = 0.05 by Benjamini–Hochberg’s procedure. We similarly tested if the REL of a barcode differs significantly from 0.1, 0.2, and so on. This allows computing the fraction of barcodes with RELs significantly higher than any REL cutoff (Fig. 4c), as well as identifying barcodes whose RELs are significantly higher than an REL cutoff but not significantly higher than the next (higher) cutoff (Fig. 4d). We similarly used a one-sample t-test to test if the expression level of barcode i (Ei) is significantly different from the negative control (using the eight expression estimates of the negative control) at FDR = 0.05 by Benjamini–Hochberg’s procedure.
Environment-specific expressions
First, expression levels of barcodes or intergenic windows are measured relative to the reference, which is the median expression level of yeast native genes. Second, at each expression level cutoff (0.1, 0.2, 0.3, …), we calculated the fraction of barcodes or intergenic windows with significantly higher expressions than the cutoff in YPD and SCD, respectively. Third, for barcodes or intergenic windows with significantly higher expressions than the cutoff in one environment (YPD or SCD), we calculated the fraction of them with significantly higher expressions than the same cutoff in the other environment; the fraction is referred to as the shared fraction.
Correlation between the expression of an intergenic region and that of their neighboring genes
For each intergenic region, we considered the transcriptions of the two directions separately. For a given transcriptional direction, the mean expression of all windows in an intergenic region was used as an estimate of the expression level of the intergenic region for that direction. To consider its neighboring gene expression, we used (1) the mean expression level of both of its neighboring genes (one on each side), (2) the expression level of the upstream neighboring gene given the transcriptional direction under consideration, or (3) the readthrough level of the upstream neighbor, which is the expression level of the upstream neighboring gene only when it has the same transcriptional direction as the direction under consideration; otherwise, the expression of the neighbor is set at 0. We then correlated the expression level of an intergenic region with the expression of its neighboring genes in each of the above three ways.
Intergenic transcriptions unattributable to neighboring gene expressions
The highest REL bin where expressions of intergenic windows are fully attributable to chance promoter activities is 0.3–0.4 in YPD (Fig. 4d) and 0.0–0.1 in SCD (Fig. S15b), respectively. For intergenic windows with REL > 0.4 in YPD (or >0.1 in SCD), we ranked them based on their neighboring gene expression level (following the first definition of neighboring gene expression in the preceding section) and then separated them into 50 groups with equal numbers of intergenic windows per group according to the ranking. We iteratively removed the group with the highest neighboring gene expression. For the remaining groups, we calculated the median expression level of their neighboring genes (ME1). From all intergenic windows (regardless of REL), we sampled the same number of windows as that in the remaining groups and calculated the median expression level of their neighboring genes (ME2); this sampling was repeated 1000 times to allow the estimation of the fraction of times when ME2 ≥ ME1. When this fraction exceeds the cutoff of 0.05, the remaining intergenic windows should be minimally influenced by the expressions of neighboring genes. We varied the cutoff and found the result similar (Fig. S17).
Fraction of intergenic regions whose expressions are explainable by chance promoter activities
For each intergenic region, we used the maximal (or the 95th percentile) expression level of its windows as a proxy for its expression level Eobs. Given the length (L) of the intergenic region, we calculated the number of nonoverlapping 120-nucleotide segments within the intergenic region by \(N\,=\,\left[\frac{L}{120}\right]\), where [x] is the smallest integer equal to or greater than x. We then randomly chose N promoters from the random promoter library and compared the maximal expression of them (Esampled) with Eobs. We repeated the random sampling 100,000 times and used the fraction of times when Esampled > Eobs as the nominal P value for the null hypothesis that the expression of an intergenic region can be explained by chance promoter activities. The number of intergenic regions with expressions that cannot be explained by chance promoter activities is calculated by counting the number of intergenic regions with a P value <0.05 upon a Benjamini–Hochberg multiple-testing correction. The expressions of neighboring genes of the intergenic regions whose expressions cannot be explained by chance promoter activities are not significantly higher than the corresponding values for the rest of the intergenic regions (Table S4). Hence, there is no need to correct the potential influence of neighboring gene expressions on intergenic expressions in this analysis.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
Sequencing data have been submitted to NCBI under accession code PRJNA876017. The RNA-seq datasets used for intergenic expression analysis are available under the accession number PRJNA728585 and PRJNA392312. Additionally, data files for active promoters are available at https://github.com/JasperXuEvolution/Random_promoter/tree/main/Data. Intermediate data files are available at https://figshare.com/articles/dataset/Intermediate_data_for_Chance_promoter_activities_illuminate_the_origins_of_eukaryotic_intergenic_transcriptions_/22231603. Source data are provided with this paper.
Code availability
Computer code is available at https://github.com/JasperXuEvolution/Random_promoter.
References
Johnson, J. M., Edwards, S., Shoemaker, D. & Schadt, E. E. Dark matter in the genome: evidence of widespread transcription detected by microarray tiling experiments. Trends Genet. 21, 93–102 (2005).
Hangauer, M. J., Vaughn, I. W. & McManus, M. T. Pervasive transcription of the human genome produces thousands of previously unidentified long intergenic noncoding RNAs. PLOS Genet. 9, e1003569 (2013).
Berretta, J. & Morillon, A. Pervasive transcription constitutes a new level of eukaryotic genome regulation. EMBO Rep. 10, 973–982 (2009).
Kellis, M. et al. Defining functional DNA elements in the human genome. Proc. Natl Acad. Sci. U.S.A. 111, 6131–6138 (2014).
Kapranov, P. & St Laurent, G. Dark matter RNA: existence, function, and controversy. Front. Genet. 3, 60 (2012).
Ponting, C. P. & Haerty, W. Genome-wide analysis of human long noncoding RNAs: a provocative review. Annu Rev. Genomics Hum. Genet 23, 153–172 (2022).
Mattick, J. S. The genetic signatures of noncoding RNAs. PLOS Genet. 5, e1000459 (2009).
Willingham, A. T. & Gingeras, T. R. TUF love for “junk” DNA. Cell 125, 1215–1220 (2006).
Ulitsky, I. & Bartel, D. P. lincRNAs: genomics, evolution, and mechanisms. Cell 154, 26–46 (2013).
Ransohoff, J. D., Wei, Y. N. & Khavari, P. A. The functions and unique features of long intergenic non-coding RNA. Nat. Rev. Mol. Cell Bio 19, 143–157 (2018).
Martens, J. A., Laprade, L. & Winston, F. Intergenic transcription is required to repress the Saccharomyces cerevisiae SER3 gene. Nature 429, 571–574 (2004).
Pande, A., Brosius, J., Makalowska, I., Makalowski, W. & Raabe, C. A. Transcriptional interference by small transcripts in proximal promoter regions. Nucleic Acids Res. 46, 1069–1088 (2018).
Pande, A., Makalowski, W., Brosius, J. & Raabe, C. A. Enhancer occlusion transcripts regulate the activity of human enhancer domains via transcriptional interference: a computational perspective. Nucleic Acids Res. 48, 3435–3454 (2020).
Brosius, J. Waste not, want not-transcript excess in multicellular eukaryotes. Trends Genet. 21, 287–288 (2005).
Palazzo, A. F. & Lee, E. S. Non-coding RNA: what is functional and what is junk? Front. Genet. 6, 2 (2015).
Neil, H. et al. Widespread bidirectional promoters are the major source of cryptic transcripts in yeast. Nature 457, 1038–1042 (2009).
Xu, Z. et al. Bidirectional promoters generate pervasive transcription in yeast. Nature 457, 1033–1037 (2009).
Struhl, K. Transcriptional noise and the fidelity of initiation by RNA polymerase II. Nat. Struct. Mol. Biol. 14, 103–105 (2007).
Yona, A. H., Alm, E. J. & Gore, J. Random sequences rapidly evolve into de novo promoters. Nat. Commun. 9, 1530 (2018).
Porrua, O. & Libri, D. Transcription termination and the control of the transcriptome: why, where and how to stop. Nat. Rev. Mol. Cell Biol. 16, 190–202 (2015).
Rando, O. J. & Winston, F. Chromatin and transcription in yeast. Genetics 190, 351–387 (2012).
Pang, K. C., Frith, M. C. & Mattick, J. S. Rapid evolution of noncoding RNAs: lack of conservation does not mean lack of function. Trends Genet. 22, 1–5 (2006).
Babak, T., Blencowe, B. J. & Hughes, T. R. A systematic search for new mammalian noncoding RNAs indicates little conserved intergenic transcription. BMC Genomics 6, 104 (2005).
Haerty, W. & Ponting, C. P. Mutations within lncRNAs are effectively selected against in fruitfly but not in human. Genome Biol. 14, R49 (2013).
Kristiansson, E., Thorsen, M., Tamas, M. J. & Nerman, O. Evolutionary forces act on promoter length: identification of enriched cis-regulatory elements. Mol. Biol. Evol. 26, 1299–1307 (2009).
Decoene, T., De Maeseneire, S. L. & De Mey, M. Modulating transcription through development of semi-synthetic yeast core promoters. PLOS ONE 14, e0224476 (2019).
Redden, H. & Alper, H. S. The development and characterization of synthetic minimal yeast promoters. Nat. Commun. 6, 7810 (2015).
Alper, H., Fischer, C., Nevoigt, E. & Stephanopoulos, G. Tuning genetic control through promoter engineering. Proc. Natl Acad. Sci. U.S.A. 102, 12678–12683 (2005).
McIsaac, R. S., Gibney, P. A., Chandran, S. S., Benjamin, K. R. & Botstein, D. Synthetic biology tools for programming gene expression without nutritional perturbations in Saccharomyces cerevisiae. Nucleic Acids Res. 42, e48 (2014).
Kotopka, B. J. & Smolke, C. D. Model-driven generation of artificial yeast promoters. Nat. Commun. 11, 2113 (2020).
de Boer, C. G. et al. Deciphering eukaryotic gene-regulatory logic with 100 million random promoters. Nat. Biotechnol. 38, 56–65 (2020).
Gertz, J., Siggia, E. D. & Cohen, B. A. Analysis of combinatorial cis-regulation in synthetic and genomic promoters. Nature 457, 215–218 (2009).
Sharon, E. et al. Inferring gene regulatory logic from high-throughput measurements of thousands of systematically designed promoters. Nat. Biotechnol. 30, 521–530 (2012).
Chen, X. & Zhang, J. The genomic landscape of position effects on protein expression level and noise in yeast. Cell Syst. 2, 347–354 (2016).
Lubliner, S., Keren, L. & Segal, E. Sequence features of yeast and human core promoters that are predictive of maximal promoter activity. Nucleic Acids Res. 41, 5569–5581 (2013).
Basehoar, A. D., Zanton, S. J. & Pugh, B. F. Identification and distinct regulation of yeast TATA box-containing genes. Cell 116, 699–709 (2004).
Spivak, A. T. & Stormo, G. D. ScerTF: a comprehensive database of benchmarked position weight matrices for Saccharomyces species. Nucleic Acids Res. 40, D162–D168 (2012).
Chou, H. J., Donnard, E., Gustafsson, H. T., Garber, M. & Rando, O. J. Transcriptome-wide analysis of roles for tRNA modifications in translational regulation. Mol. Cell 68, 978–992.e974 (2017).
Berg, M. D. et al. The amino acid substitution affects cellular response to mistranslation. G3 11, jkab218 (2021).
Lagator, M. et al. Predicting bacterial promoter function and evolution from random sequences. Elife 11, e64543 (2022).
Tillo, D. & Hughes, T. R. G+C content dominates intrinsic nucleosome occupancy. BMC Bioinform. 10, 442 (2009).
Guertin, M. J. & Lis, J. T. Mechanisms by which transcription factors gain access to target sequence elements in chromatin. Curr. Opin. Genet Dev. 23, 116–123 (2013).
Gietz, R. D. & Schiestl, R. H. Large-scale high-efficiency yeast transformation using the LiAc/SS carrier DNA/PEG method. Nat. Protoc. 2, 38–41 (2007).
Lee, M. E., DeLoache, W. C., Cervantes, B. & Dueber, J. E. A Highly characterized yeast toolkit for modular, multipart assembly. Acs Synth. Biol. 4, 975–986 (2015).
Zhao, L., Liu, Z., Levy, S. F. & Wu, S. Bartender: a fast and accurate clustering algorithm to count barcode reads. Bioinformatics 34, 739–747 (2018).
Li, W. & Godzik, A. Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics 22, 1658–1659 (2006).
Sekinger, E. A., Moqtaderi, Z. & Struhl, K. Intrinsic histone-DNA interactions and low nucleosome density are important for preferential accessibility of promoter regions in yeast. Mol. Cell 18, 735–748 (2005).
Tsai, Z. T., Lloyd, J. P. & Shiu, S. H. Defining functional genic regions in the human genome through integration of biochemical, evolutionary, and genetic evidence. Mol. Biol. Evol. 34, 1788–1798 (2017).
Park, D., Morris, A. R., Battenhouse, A. & Iyer, V. R. Simultaneous mapping of transcript ends at single-nucleotide resolution and identification of widespread promoter-associated non-coding RNA governed by TATA elements. Nucleic Acids Res. 42, 3736–3749 (2014).
Nagalakshmi, U. et al. The transcriptional landscape of the yeast genome defined by RNA sequencing. Science 320, 1344–1349 (2008).
Pertea, M. et al. StringTie enables improved reconstruction of a transcriptome from RNA-seq reads. Nat. Biotechnol. 33, 290–295 (2015).
Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate—a practical and powerful approach to multiple testing. J. R. Stat. Soc. B 57, 289–300 (1995).
Acknowledgements
We thank D. Jiang, A. Mahilkar, and X. Shen for valuable comments. This work was supported by U.S. National Institutes of Health research grant R35GM139484 to J.Z.
Author information
Authors and Affiliations
Contributions
J.Z. conceived of the project and acquired funding; H.X., C.L., and J.Z. designed the study; C.L. performed some pilot experiments; H.X. performed the experiments; H.X. and C.X. analyzed the data; H.X. and J.Z. wrote the paper, with input from all authors.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Xu, H., Li, C., Xu, C. et al. Chance promoter activities illuminate the origins of eukaryotic intergenic transcriptions. Nat Commun 14, 1826 (2023). https://doi.org/10.1038/s41467-023-37610-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-023-37610-w
- Springer Nature Limited