CoChat
Sign Up
Cited by 529

The DOI matches its Crossref record; 12 of the 12 references it cites were found in a registry or an index.

Widespread contribution of transposable elements to the innovation of gene regulatory networks

Sundaram, Vasavi, Cheng, Yong, Ma, Zhihai, Li, Daofeng, Xing, Xiaoyun, Edge, Peter, Snyder, Michael P., Wang, Ting·2014·Genome Research
DOI 10.1101/gr.168872.113PMC PMC4248313
Open this paper in CoChat — free account

Abstract

Transposable elements (TEs) have been shown to contain functional binding sites for certain transcription factors (TFs). However, the extent to which TEs contribute to the evolution of TF binding sites is not well known. We comprehensively mapped binding sites for 26 pairs of orthologous TFs in two pairs of human and mouse cell lines (representing two cell lineages), along with epigenomic profiles, including DNA methylation and six histone modifications. Overall, we found that 20% of binding sites were embedded within TEs. This number varied across different TFs, ranging from 2% to 40%. We further identified 710 TF–TE relationships in which genomic copies of a TE subfamily contributed a significant number of binding peaks for a TF, and we found that LTR elements dominated these relationships in human. Importantly, TE-derived binding peaks were strongly associated with open and active chromatin signatures, including reduced DNA methylation and increased enhancer-associated histone marks. On average, 66% of TE-derived binding events were cell type-specific with a cell type-specific epigenetic landscape. Most of the binding sites contributed by TEs were species-specific, but we also identified binding sites conserved between human and mouse, the functional relevance of which was supported by a signature of purifying selection on DNA sequences of these TEs. Interestingly, several TFs had significantly expanded binding site landscapes only in one species, which were linked to species-specific gene functions, suggesting that TEs are an important driving force for regulatory innovation. Taken together, our data suggest that TEs have significantly and continuously shaped gene regulatory networks during mammalian evolution.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1101/gr.168872.113

Cites

12 of 12 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. Different human and mouse TFs had different numbers and fractions of their binding derived from TEs. (A) Percentage of TF binding peaks that occurred in TEs, in human (left panel, red) and mouse (right panel, blue). (B) Correlation between the number of TF binding peaks in the genome and the percentage of the TF binding peaks in TEs in human (left panel) and mouse (right panel). Correlation between these two variables was measured with Pearson’s correlation coefficient (r).
Figure 1. Different human and mouse TFs had different numbers and fractions of their binding derived from TEs. (A) Percentage of TF binding peaks that occurred in TEs, in human (left panel, red) and mouse (right panel, blue). (B) Correlation between the number of TF binding peaks in the genome and the percentage of the TF binding peaks in TEs in human (left panel) and mouse (right panel). Correlation between these two variables was measured with Pearson’s correlation coefficient (r).CC BY-NC · PubMed Central
Figure 2. Specific TE classes and subfamilies enriched for TF binding peaks. (A) Proportion of the TF binding peaks that occurred in each TE class in human (left panel) and mouse (right panel). For comparison, “Background” represents the proportion of the genome (in bp) that each TE class constitutes. (B) Enrichment of TF binding peaks in TE subfamilies. We used a log-odds ratio (see Methods) for the definition of enrichment and estimated this for each TE subfamily in human (left panel), and mouse (right panel). Dots with the same color and shape represent TE subfamilies that belong to the same TE class. Several TFs lack data points, which represents no enrichment of binding peaks in TE subfamilies because they had ≤10 binding peaks occurring within the subfamily (see Methods).
Figure 2. Specific TE classes and subfamilies enriched for TF binding peaks. (A) Proportion of the TF binding peaks that occurred in each TE class in human (left panel) and mouse (right panel). For comparison, “Background” represents the proportion of the genome (in bp) that each TE class constitutes. (B) Enrichment of TF binding peaks in TE subfamilies. We used a log-odds ratio (see Methods) for the definition of enrichment and estimated this for each TE subfamily in human (left panel), and mouse (right panel). Dots with the same color and shape represent TE subfamilies that belong to the same TE class. Several TFs lack data points, which represents no enrichment of binding peaks in TE subfamilies because they had ≤10 binding peaks occurring within the subfamily (see Methods).CC BY-NC · PubMed Central
Figure 3. Including nonuniquely mapped reads (multireads) captured additional binding events on TEs. (A) Comparison between the enrichment of reads (see Methods) for all TF–TE pairs and TF–TE pairs that were enriched for interactions defined by peaks (see Methods). (B) Distribution of mappability scores (for 36-mer and 75-mer, respectively) of TE subfamilies. (Peaks) TE subfamilies enriched for TF binding peaks; (reads) TE subfamilies enriched for ChIP-seq reads (including both unique and nonunique reads). We calculated the mappability score (level of sequence uniqueness) for each genomic copy of a particular TE subfamily using mappability tracks downloaded from the UCSC Genome Browser, where 1.0 = 100% mapped uniquely, and 0 = 0% mapped uniquely. (C) Distribution of the mappability scores for the genomic copies of the RLTR4 subfamily in mouse by using 36-mer and 75-mer sequence reads. (D) Comparison of the ChIP-seq signal using unique and all reads in RLTR4. The first panel shows genomic coverage of RLTR4 copies on the RLTR4 consensus sequence. The second panel shows the average mappability (75-mer) score (mappability file was downloaded from the UCSC Genome Browser). The third panel shows the accumulation of POL2 ChIP-seq reads over RLTR4 consensus, with purple representing unique reads, and green representing both unique and non-unique reads. The fourth panel shows the accumulation of ChIP input reads over the RLTR4 consensus sequence.
Figure 3. Including nonuniquely mapped reads (multireads) captured additional binding events on TEs. (A) Comparison between the enrichment of reads (see Methods) for all TF–TE pairs and TF–TE pairs that were enriched for interactions defined by peaks (see Methods). (B) Distribution of mappability scores (for 36-mer and 75-mer, respectively) of TE subfamilies. (Peaks) TE subfamilies enriched for TF binding peaks; (reads) TE subfamilies enriched for ChIP-seq reads (including both unique and nonunique reads). We calculated the mappability score (level of sequence uniqueness) for each genomic copy of a particular TE subfamily using mappability tracks downloaded from the UCSC Genome Browser, where 1.0 = 100% mapped uniquely, and 0 = 0% mapped uniquely. (C) Distribution of the mappability scores for the genomic copies of the RLTR4 subfamily in mouse by using 36-mer and 75-mer sequence reads. (D) Comparison of the ChIP-seq signal using unique and all reads in RLTR4. The first panel shows genomic coverage of RLTR4 copies on the RLTR4 consensus sequence. The second panel shows the average mappability (75-mer) score (mappability file was downloaded from the UCSC Genome Browser). The third panel shows the accumulation of POL2 ChIP-seq reads over RLTR4 consensus, with purple representing unique reads, and green representing both unique and non-unique reads. The fourth panel shows the accumulation of ChIP input reads over the RLTR4 consensus sequence.CC BY-NC · PubMed Central
Figure 4. The majority of TE-derived binding peaks were species-specific, whereas the shared ones exhibited sequence conservation. (A) Bar plot showing the numbers of binding events in various categories of conservation. We classified TE-derived binding events in each species into four categories: (1) no orthologous region identifiable for the TF binding peak; (2) orthologous region lacks TF binding; (3) conserved TF binding, not annotated as the same TE subfamily in both species; and (4) conserved TF binding, annotated as the same TE subfamily in both species. (B) Distribution of phastCons scores (see Methods) in 6-kb windows centered on the TE sequences in the four categories defined above. The signal was averaged across 25-bp bins and smoothened for this plot. TE-derived binding peaks that could be mapped to binding peaks derived from the same TE subfamily in the other species exhibited increased sequence constraint over the background. (C,D) UCSC Genome Browser images of TE-derived TF binding peaks, whose occupancy was conserved between human (upper panel) and mouse (lower panel). The browser images correspond to (C) CTCF binding encoded on MER91B fragments in human and mouse, and (D) BHLHE40 binding encoded on L3 fragments in human and mouse.
Figure 4. The majority of TE-derived binding peaks were species-specific, whereas the shared ones exhibited sequence conservation. (A) Bar plot showing the numbers of binding events in various categories of conservation. We classified TE-derived binding events in each species into four categories: (1) no orthologous region identifiable for the TF binding peak; (2) orthologous region lacks TF binding; (3) conserved TF binding, not annotated as the same TE subfamily in both species; and (4) conserved TF binding, annotated as the same TE subfamily in both species. (B) Distribution of phastCons scores (see Methods) in 6-kb windows centered on the TE sequences in the four categories defined above. The signal was averaged across 25-bp bins and smoothened for this plot. TE-derived binding peaks that could be mapped to binding peaks derived from the same TE subfamily in the other species exhibited increased sequence constraint over the background. (C,D) UCSC Genome Browser images of TE-derived TF binding peaks, whose occupancy was conserved between human (upper panel) and mouse (lower panel). The browser images correspond to (C) CTCF binding encoded on MER91B fragments in human and mouse, and (D) BHLHE40 binding encoded on L3 fragments in human and mouse.CC BY-NC · PubMed Central
Figure 5. Binding motifs of TFs were enriched in TE-derived binding peaks. (A) Distribution of motif enrichment scores (log-odds ratio) (see Methods) in non-TE, TF binding peaks (training set for de novo motif prediction), TE-derived TF binding peaks (test set), and TEs without peaks (control). (B) Distribution of motif enrichment scores in TE subfamilies that were enriched in TF binding peaks compared with all other TE subfamilies. (C) Multiple sequence alignment of MER91B genomic copies (n = 215) from the mouse genomes and the MER91B consensus sequence (bottom row of the alignment). Beside the alignment are indications of genomic copies that had CTCF binding peaks (purple) and CTCF motifs (green). (D) Similarly, multiple sequence alignment of LTR18A genomic copies (n = 259) from the human genome along with the LTR18A consensus sequence (bottom row of the multiple-sequence alignment). Annotated on the right are indications of genomic copies that had MAFK binding peaks and MAFK motifs. Nucleotides in alignments are color-coded: (A) green; (C) blue; (G) yellow; (T) red.
Figure 5. Binding motifs of TFs were enriched in TE-derived binding peaks. (A) Distribution of motif enrichment scores (log-odds ratio) (see Methods) in non-TE, TF binding peaks (training set for de novo motif prediction), TE-derived TF binding peaks (test set), and TEs without peaks (control). (B) Distribution of motif enrichment scores in TE subfamilies that were enriched in TF binding peaks compared with all other TE subfamilies. (C) Multiple sequence alignment of MER91B genomic copies (n = 215) from the mouse genomes and the MER91B consensus sequence (bottom row of the alignment). Beside the alignment are indications of genomic copies that had CTCF binding peaks (purple) and CTCF motifs (green). (D) Similarly, multiple sequence alignment of LTR18A genomic copies (n = 259) from the human genome along with the LTR18A consensus sequence (bottom row of the multiple-sequence alignment). Annotated on the right are indications of genomic copies that had MAFK binding peaks and MAFK motifs. Nucleotides in alignments are color-coded: (A) green; (C) blue; (G) yellow; (T) red.CC BY-NC · PubMed Central
Figure 6. Epigenetic profile of TE-derived TF binding peaks in lymphoblast and leukemia cell lines in human and mouse. All figures represent the average signal density at 50-bp resolution over a 10-kb window centered on the genomic regions of interest (i.e., TEs or peaks). Left panels display profiles of histone modification marks, and the right panels display DNA methylation data. (A) Epigenetic profile of TE-derived TF binding peaks, which represents TEs that contained binding peaks for any one TF in human (upper panels) and mouse (lower panels). (B) Epigenetic profile of TF binding peaks that do not overlap any TE. (C) Epigenetic profile of TEs that do not overlap any TF binding peak.
Figure 6. Epigenetic profile of TE-derived TF binding peaks in lymphoblast and leukemia cell lines in human and mouse. All figures represent the average signal density at 50-bp resolution over a 10-kb window centered on the genomic regions of interest (i.e., TEs or peaks). Left panels display profiles of histone modification marks, and the right panels display DNA methylation data. (A) Epigenetic profile of TE-derived TF binding peaks, which represents TEs that contained binding peaks for any one TF in human (upper panels) and mouse (lower panels). (B) Epigenetic profile of TF binding peaks that do not overlap any TE. (C) Epigenetic profile of TEs that do not overlap any TF binding peak.CC BY-NC · PubMed Central
Figure 7. TE-derived, cell type-specific TF binding peaks. (A) Venn diagrams represent the numbers of cell type-specific TE-derived binding events in human (top panel) and mouse (bottom panel). The numbers represent the total number of instances in which a TE-derived TF binding peak for a particular TF was found in one cell type but not in the other. (B) Comparison of the epigenetic profiles of TEs that contained TF binding peaks. These represent TEs that contained a TF binding peak for any TF in one cell line but not in the other. For comparison, we plotted the average epigenetic profile of TEs that contributed cell type-specific peaks in both cell types. (C) UCSC Genome Browser view of a lymphoblast-specific TE-derived (L1MB5) BHLHE40 binding peak that was conserved between human and mouse.
Figure 7. TE-derived, cell type-specific TF binding peaks. (A) Venn diagrams represent the numbers of cell type-specific TE-derived binding events in human (top panel) and mouse (bottom panel). The numbers represent the total number of instances in which a TE-derived TF binding peak for a particular TF was found in one cell type but not in the other. (B) Comparison of the epigenetic profiles of TEs that contained TF binding peaks. These represent TEs that contained a TF binding peak for any TF in one cell line but not in the other. For comparison, we plotted the average epigenetic profile of TEs that contributed cell type-specific peaks in both cell types. (C) UCSC Genome Browser view of a lymphoblast-specific TE-derived (L1MB5) BHLHE40 binding peak that was conserved between human and mouse.CC BY-NC · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 12 references

Research works printed but not linked: 54

  1. 10.1126/science.287.5461.2185 · DOI
  2. 10.1073/pnas.90.17.7915 · DOI · Paper page
  3. 10.1038/nature04696 · DOI
  4. 10.1016/j.gde.2009.10.013 · DOI
  5. 10.1101/gr.080663.108 · DOI
  6. 10.1126/science.165.3891.349 · DOI
  7. Cheng Y , Ma Z , Kim BH , Cayting P , Boyle AP , Wu W , Sundaram V , Xing X , Li J , Euskirchen G , . 2014. Principles of regulatory information conservation between mouse and human. Nature doi: 10.1038/nature13985. · DOI
  8. 10.1038/ng.2553 · DOI · Paper page
  9. 10.1371/journal.pgen.1002384 · DOI
  10. 10.1093/molbev/mst045 · DOI
  11. 10.1038/sj.embor.7400314 · DOI
  12. 10.1038/284601a0 · DOI
  13. 10.1073/pnas.2134464100 · DOI
  14. 10.4161/epi.19462 · DOI
  15. 10.1038/nature11247 · DOI · Paper page
  16. 10.1038/nrg2337 · DOI
  17. 10.1038/nrg3199 · DOI
  18. 10.1093/bioinformatics/btr064 · DOI · Paper page
  19. 10.1016/j.molcel.2010.05.004 · DOI · Paper page
  20. 10.1038/35057062 · DOI · Paper page
  21. 10.1371/journal.pgen.1003504 · DOI
  22. Johnson, 2007, Genome-wide mapping of in vivo protein-DNA interactions, Science, 80, 1497, 10.1126/science.1141319 · DOI · Paper page
  23. 10.1016/S0168-9525(02)00006-9 · DOI · Paper page
  24. 10.1159/000084979 · DOI
  25. 10.1126/science.1183621 · DOI
  26. 10.1101/gr.229102. Article published online before print in May 2002 · DOI · Paper page
  27. 10.1038/nbt.1508 · DOI
  28. 10.1038/ng.600 · DOI
  29. 10.1101/gr.136184.111 · DOI · Paper page
  30. 10.1038/ng.917 · DOI
  31. 10.1038/nature09165 · DOI · Paper page
  32. 10.1073/pnas.36.6.344 · DOI
  33. 10.1101/SQB.1956.021.01.017 · DOI
  34. 10.1074/jbc.M006557200 · DOI
  35. 10.1016/j.cell.2013.02.029 · DOI · Paper page
  36. 10.4049/jimmunol.172.5.2731 · DOI · Paper page
  37. 10.1371/journal.pone.0019470 · DOI
  38. 10.1186/1759-8753-2-8 · DOI
  39. 10.1016/j.cell.2009.06.001 · DOI
  40. 10.1073/pnas.1004139107 · DOI
  41. 10.1186/1471-2164-9-226 · DOI
  42. 10.1186/gb-2012-13-10-r89 · DOI
  43. 10.1146/annurev-genet-110711-155621 · DOI
  44. 10.1073/pnas.0708366105 · DOI
  45. Samuelson, 1990, Retroviral and pseudogene insertion sites reveal the lineage of human salivary and pancreatic amylase genes from a single gene during primate evolution, Mol Cell Biol, 10, 2513
  46. 10.1073/pnas.0709398105 · DOI
  47. 10.1016/j.cell.2011.11.058 · DOI
  48. 10.1038/msb.2011.75 · DOI
  49. 10.1101/gr.152231.112 · DOI
  50. 10.1038/nature07730 · DOI · Paper page
  51. 10.1073/pnas.0703637104 · DOI
  52. 10.1101/gr.100594.109 · DOI
  53. 10.1038/ng.2649 · DOI
  54. 10.1101/gr.156539.113 · DOI

Cited by

Sorted by citations

See all 529

Metadata

Authors
Sundaram, V., Cheng, Y., Ma, Z., Li, D., Xing, X., Edge, P., Snyder, M. P., & Wang, T.
Venue
Genome Research
Publisher
Cold Spring Harbor Laboratory
Published
2014-10-15
Volume
24
Issue
12
Pages
1963-1976
ISSN
  • 1088-9051
  • 1549-5469
DOI
10.1101/gr.168872.113
PMC
PMC4248313
Cited by
529
References
12
Language
en
Funders
  • Mouse ENCODE Consortium (3RC2HG005602)
  • March of Dimes Foundation (5-FY10-491)
  • American Cancer Society (RSG-14-049-01-DMC)
  • NIH (5U01ES017154, R01HG007354, R01HG007175, R01ES024992)

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Widespread contribution of transposable elements to the innovation of gene regulatory networks. https://app.cochat.ai/papers/10.1101/gr.168872.113
The paper
Sundaram, V., Cheng, Y., Ma, Z., Li, D., Xing, X., Edge, P., Snyder, M. P., & Wang, T. (2014). Widespread contribution of transposable elements to the innovation of gene regulatory networks. Genome Research. https://doi.org/10.1101/gr.168872.113