CoChat

A novel virus-inducible enhancer of the interferon-β gene with tightly linked promoter and enhancer activities

Open access
Banerjee, A. Raja, Kim, Yoon Jung, Kim, Tae Hoon · 2014 · Nucleic Acids Research

The DOI matches its Crossref record; 72 of the 72 references it cites were found in a registry or an index.

Abstract

Long-range enhancers of transcription are a key component of the genomic regulatory architecture. Recent studies have identified bi-directionally transcribed RNAs emanating from these enhancers known as eRNAs. However, it remains unclear how tightly coupled eRNA production is with enhancer activity. Through our systematic search for long-range elements that interact with the interferon-β gene, a model system for studying inducible transcription, we have identified a novel enhancer, which we have named L2 that regulates the expression of interferon-β. We have demonstrated its virus-inducible enhancer activity by analyzing epigenomic profiles, transcription factor association, nascent RNA production and activity in reporter assays. This enhancer exhibits intimately linked virus-inducible enhancer and bidirectional promoter activity that is largely dependent on a conserved Interferon Stimulated Response Element and robustly generates virus inducible eRNAs. Notably, its enhancer and promoter activities are fully retained in reporter assays even upon a complete elimination of its associated eRNA sequences. Finally, we show that L2 regulates IFNB1 expression by siRNA knockdown of eRNAs, and the deletion of L2 in a BAC transfection assay. Thus, L2 is a novel enhancer that regulates IFNB1 and whose eRNAs exert significant activity in vivo that is distinct from those activities recapitulated in the luciferase reporter assays.

Overview

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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

Integrity signals

Registry

Crossref · 10.1093/nar/gku1018

Cites

72 of 72 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. Identification of four sites interacting with the promoter of IFNB1. (A) The region of chromosome 9 queried against the promoter of IFNB1 in the 3C assay (chr9: 21 002 063–22 347 415) is displayed against the genes within the locus and the topological domains in the locus (45). (B) Results from the single-replicate preliminary 3C study are visualized as vertical bars representing the queried restriction site. Yellow bars represent sites that generated a PCR product in either the uninfected or infected state, while blue bars indicate sites with no PCR product in either condition. The black bar represents the anchor site at the promoter of IFNB1 (chr9:21068221) and is indicated along the bottom of the panel. Sites scoring positively that were >15 kb from the anchor site are labeled long the top of the panel. A site that scored negative and was used as a negative control in the validation experiment is indicated below the bars. (C) 3C results were validated with three independent biological replicates in semi-quantitative Real-time PCR. The black horizontal line indicates an RE relevance cutoff point at 0.001. Any sample that exceeded this threshold was considered to interact with the promoter of IFNB1.
Figure 1. Identification of four sites interacting with the promoter of IFNB1. (A) The region of chromosome 9 queried against the promoter of IFNB1 in the 3C assay (chr9: 21 002 063–22 347 415) is displayed against the genes within the locus and the topological domains in the locus (45). (B) Results from the single-replicate preliminary 3C study are visualized as vertical bars representing the queried restriction site. Yellow bars represent sites that generated a PCR product in either the uninfected or infected state, while blue bars indicate sites with no PCR product in either condition. The black bar represents the anchor site at the promoter of IFNB1 (chr9:21068221) and is indicated along the bottom of the panel. Sites scoring positively that were >15 kb from the anchor site are labeled long the top of the panel. A site that scored negative and was used as a negative control in the validation experiment is indicated below the bars. (C) 3C results were validated with three independent biological replicates in semi-quantitative Real-time PCR. The black horizontal line indicates an RE relevance cutoff point at 0.001. Any sample that exceeded this threshold was considered to interact with the promoter of IFNB1.CC BY · PubMed Central
Figure 2. L1 and L2 include sites with chromatin markings of enhancer activity conserved to mouse. (A) DNase Hypersensitivity, H3K4me1, H3K27ac and ChIP input enrichment plots and peaks from ENCODE were assessed for L1 both in normal human lung fibroblasts (NHLF, top half), as well as against the chromatin state in the orthologous mouse region in embryonic fibroblast (MEFs) or adult fibroblasts (bottom half) (31,44). (B) Chromatin state was assessed for L2 as in (A). (C) Chromatin state for L3 was assessed for human fibroblasts as in (A). (A and B) Black bars atop each enrichment plot indicate previously called enrichment peaks.
Figure 2. L1 and L2 include sites with chromatin markings of enhancer activity conserved to mouse. (A) DNase Hypersensitivity, H3K4me1, H3K27ac and ChIP input enrichment plots and peaks from ENCODE were assessed for L1 both in normal human lung fibroblasts (NHLF, top half), as well as against the chromatin state in the orthologous mouse region in embryonic fibroblast (MEFs) or adult fibroblasts (bottom half) (31,44). (B) Chromatin state was assessed for L2 as in (A). (C) Chromatin state for L3 was assessed for human fibroblasts as in (A). (A and B) Black bars atop each enrichment plot indicate previously called enrichment peaks.CC BY · PubMed Central
Figure 3. L1–1 and L2 include elements that are highly conserved to mouse at the sequence level. (A) Sequence conservation of L1. Top Section: The L1 genomic context was plotted against human DNA elements conserved across placental mammals, as well as against the alignment of this region to five different mammalian species, with H3K4me1 enrichment included for frame of reference to Figure 2. Second Section: L1 was aligned to its mouse ortholog and assessed for nucleotide-level sequence conservation. The height of the bar represents nucleotide identity at the centerpoint of a 5 bp window. Third Section: L1–1 was aligned to its mouse ortholog and assessed for nucleotide-level sequence conservation as in the second section. Bottom Section: Three predicted transcription factor binding sites were assessed for sequence conservation to their orthologous sequences in mouse. (B) Sequence Conservation of L2. Top and Middle Sections: Sequence conservation of L2 was assessed as in the top two sections of Figure 3A for L1. Bottom Section: The two conserved elements in L2 were assessed for their sequence conservation to their orthologous sequences in mouse, along with the sequence conservation for the TRANSFAC-predicted ISRE. (A and B) Blue brackets indicate the region that will be expanded in the following section.
Figure 3. L1–1 and L2 include elements that are highly conserved to mouse at the sequence level. (A) Sequence conservation of L1. Top Section: The L1 genomic context was plotted against human DNA elements conserved across placental mammals, as well as against the alignment of this region to five different mammalian species, with H3K4me1 enrichment included for frame of reference to Figure 2. Second Section: L1 was aligned to its mouse ortholog and assessed for nucleotide-level sequence conservation. The height of the bar represents nucleotide identity at the centerpoint of a 5 bp window. Third Section: L1–1 was aligned to its mouse ortholog and assessed for nucleotide-level sequence conservation as in the second section. Bottom Section: Three predicted transcription factor binding sites were assessed for sequence conservation to their orthologous sequences in mouse. (B) Sequence Conservation of L2. Top and Middle Sections: Sequence conservation of L2 was assessed as in the top two sections of Figure 3A for L1. Bottom Section: The two conserved elements in L2 were assessed for their sequence conservation to their orthologous sequences in mouse, along with the sequence conservation for the TRANSFAC-predicted ISRE. (A and B) Blue brackets indicate the region that will be expanded in the following section.CC BY · PubMed Central
Figure 4. Virus-inducible transcription factor recruitment and eRNA production at L1 and L2. (A) ChIP-PCR primers were designed to detect transcription factor enrichment within the DHS of L1–1 (left) and L2 (right). (B–E) Enrichment of TF binding binding enrichment over input chromatin at L1, L2 and a negative control site were assessed in uninfected (dark gray) and virus-infected (light gray) IMR90 cells for pIRF3 (B), RNAPII (C), SMC1A (D) and Med1 (E). Each column represents at least three independent experiments. Error bars represent standard error. (F and G) GRO-seq was performed both in uninfected as well as infected IMR90 cells. The EcoRI sites represent the outer bounds of the restriction fragment that defines L1 (F) and L2 (G). GRO-seq output was visualized with normalized RPM (RNA tags per million sequenced) multiplied by 100 plotted against the L2 region at single-nucleotide resolution. Top two tracks show the GRO-seq data from uninfected cells. The next two tracks show that for the SeV-treated state.
Figure 4. Virus-inducible transcription factor recruitment and eRNA production at L1 and L2. (A) ChIP-PCR primers were designed to detect transcription factor enrichment within the DHS of L1–1 (left) and L2 (right). (B–E) Enrichment of TF binding binding enrichment over input chromatin at L1, L2 and a negative control site were assessed in uninfected (dark gray) and virus-infected (light gray) IMR90 cells for pIRF3 (B), RNAPII (C), SMC1A (D) and Med1 (E). Each column represents at least three independent experiments. Error bars represent standard error. (F and G) GRO-seq was performed both in uninfected as well as infected IMR90 cells. The EcoRI sites represent the outer bounds of the restriction fragment that defines L1 (F) and L2 (G). GRO-seq output was visualized with normalized RPM (RNA tags per million sequenced) multiplied by 100 plotted against the L2 region at single-nucleotide resolution. Top two tracks show the GRO-seq data from uninfected cells. The next two tracks show that for the SeV-treated state.CC BY · PubMed Central
Figure 5. L1 is inert, but L2 includes both promoter and enhancer activity in a virus-inducible fashion. (A) The location of the L1F and L2F fragments amplified for the luciferase assays are show at the bottom of the panel, plotted against H3K4me1 enrichment in (Figure 2) for reference. L1 is shown on the left, L2 is shown on the right. (B) For luciferase promoter activity assays, the fragment of interest is placed immediately upstream of a promoterless luciferase gene (pGL3-basic). To calculate promoter activity, the luciferase activity of the fragment-driven reporter is divided by the activity of the promoter-less reporter. (C) For Luciferase enhancer activity assays, the fragment of interest is immediately downstream of an SV40-driven luciferase reporter (pGL3-promoter). To calculate enhancer activity, the luciferase activity of the fragment-enhancer SV40-driven reporter is divided by the activity of the SV40-driven reporter. (D) Promoter activity of L1F in both orientations, before and after virus infection. All constructs were analyzed in at least three independent experiments (n ≥ 3). (E) Promoter activity of L1F in both orientations, before and after virus infection. All constructs were analyzed in at least seven independent experiments (n ≥ 7). (F and G) Luciferase promoter and enhancers assays were performed as in (D and E) respectively, but with L2F instead of L1F. All constructs were analyzed in at least seven independent experiments (n ≥ 7). (D–G) Error bars indicate Standard Error. We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.
Figure 5. L1 is inert, but L2 includes both promoter and enhancer activity in a virus-inducible fashion. (A) The location of the L1F and L2F fragments amplified for the luciferase assays are show at the bottom of the panel, plotted against H3K4me1 enrichment in (Figure 2) for reference. L1 is shown on the left, L2 is shown on the right. (B) For luciferase promoter activity assays, the fragment of interest is placed immediately upstream of a promoterless luciferase gene (pGL3-basic). To calculate promoter activity, the luciferase activity of the fragment-driven reporter is divided by the activity of the promoter-less reporter. (C) For Luciferase enhancer activity assays, the fragment of interest is immediately downstream of an SV40-driven luciferase reporter (pGL3-promoter). To calculate enhancer activity, the luciferase activity of the fragment-enhancer SV40-driven reporter is divided by the activity of the SV40-driven reporter. (D) Promoter activity of L1F in both orientations, before and after virus infection. All constructs were analyzed in at least three independent experiments (n ≥ 3). (E) Promoter activity of L1F in both orientations, before and after virus infection. All constructs were analyzed in at least seven independent experiments (n ≥ 7). (F and G) Luciferase promoter and enhancers assays were performed as in (D and E) respectively, but with L2F instead of L1F. All constructs were analyzed in at least seven independent experiments (n ≥ 7). (D–G) Error bars indicate Standard Error. We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.CC BY · PubMed Central
Figure 6. A 2-bp ISRE mutation impairs, but does not abolish, L2F's virus-induced transcriptional activities. (A) Top half: The specific nucleotides of the L2 ISRE sequence were assessed for their enrichment score in the primary motifs for ISGF3 and IRF3 from the Uniprobe database (68). Bottom half: The specific nucleotides of the mutated L2 ISRE sequence were assessed for their enrichment score in the primary motifs for ISGF3 and IRF3 from the Uniprobe database (68). (B) ISRE mutation effect on the promoter activity was determined for Orientation 1. All constructs were analyzed in 10 independent experiments (n = 10). (C) ISRE mutation effect on the promoter activity was determined for Orientation 2. All constructs were analyzed in seven independent experiments (n = 7). (D) The impact of the ISRE mutation on enhancer activity was determined for Orientation 1. All constructs were analyzed in at least seven independent experiments (n ≥ 7) E. Impact of ISRE mutation on enhancer activity was determined for Orientation 2. All constructs were analyzed in 13 independent experiments (n = 13). (B–E) Error bars indicate Standard Error. We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.
Figure 6. A 2-bp ISRE mutation impairs, but does not abolish, L2F's virus-induced transcriptional activities. (A) Top half: The specific nucleotides of the L2 ISRE sequence were assessed for their enrichment score in the primary motifs for ISGF3 and IRF3 from the Uniprobe database (68). Bottom half: The specific nucleotides of the mutated L2 ISRE sequence were assessed for their enrichment score in the primary motifs for ISGF3 and IRF3 from the Uniprobe database (68). (B) ISRE mutation effect on the promoter activity was determined for Orientation 1. All constructs were analyzed in 10 independent experiments (n = 10). (C) ISRE mutation effect on the promoter activity was determined for Orientation 2. All constructs were analyzed in seven independent experiments (n = 7). (D) The impact of the ISRE mutation on enhancer activity was determined for Orientation 1. All constructs were analyzed in at least seven independent experiments (n ≥ 7) E. Impact of ISRE mutation on enhancer activity was determined for Orientation 2. All constructs were analyzed in 13 independent experiments (n = 13). (B–E) Error bars indicate Standard Error. We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.CC BY · PubMed Central
Figure 7. L2F enhances the activity of the previously known IFNB1 proximal regulatory elements. (A) Four constructs were used for the luciferase assay. The endogenous IFNB1 promoter and proximal enhancer together (−110) were placed upstream of the Luciferase reporter gene. L2F was inserted downstream of the reporter gene in one of two orientations. (B) The Luciferase activity of each test construct was examined in uninfected cells. The vertical line across all three bars represents baseline activity. Significance markings adjacent to this line indicates that the construct indicate differences between the marked bar and baseline activity. Each construct was tested in at least three independent experiments (n ≥ 3). (C) The Luciferase activity of each test construct was examined in SeV-infected cells. Each construct was tested in at least three independent experiments (n ≥ 3). (B and C) Error bars indicate Standard Error. We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.
Figure 7. L2F enhances the activity of the previously known IFNB1 proximal regulatory elements. (A) Four constructs were used for the luciferase assay. The endogenous IFNB1 promoter and proximal enhancer together (−110) were placed upstream of the Luciferase reporter gene. L2F was inserted downstream of the reporter gene in one of two orientations. (B) The Luciferase activity of each test construct was examined in uninfected cells. The vertical line across all three bars represents baseline activity. Significance markings adjacent to this line indicates that the construct indicate differences between the marked bar and baseline activity. Each construct was tested in at least three independent experiments (n ≥ 3). (C) The Luciferase activity of each test construct was examined in SeV-infected cells. Each construct was tested in at least three independent experiments (n ≥ 3). (B and C) Error bars indicate Standard Error. We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.CC BY · PubMed Central
Figure 8. The L2 fragment lacking eRNA sequences retains transcriptional activities, even though the eRNA is necessary for endogenous IFNB1 expression. (A) The genomic fragments used for the L2 fragment (L2F) and the eRNA-excluding L2 fragment (0.3L2F) and the larger eRNA-inclusive fragment (4L2F) are shown against the GRO-seq reads at L2 for reference (from Figure 4). For the GRO-seq data, data from both strands are displayed in a single track. (B) The impact of the exclusion of remaining eRNA sequence was assessed on L2 promoter activity in Orientation 1. (C) The impact of the exclusion of remaining eRNA sequence was assessed on L2 promoter activity in Orientation 2. (D) The impact of the presence or absence of eRNA sequence was assessed on L2 enhancer activity in Orientation 1. (E) The impact of the presence or absence of eRNA sequence was assessed on L2 enhancer activity in Orientation 2. (B–E) All constructs were analyzed in five independent experiments (n = 5). Error bars indicate Standard Error. (F) The effectiveness of both siRNA pools on knockdown of SeV-induced eRNA expression levels from each strand was assessed. (G) The effectiveness of treatment with each siRNA pool on IFNB1 induction in response to SeV was assessed. (F and G) Expression levels were normalized to GAPDH. Each treatment was tested in three independent experiments (n = 3). Error bars indicate Standard Deviation. (B–G) We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.
Figure 8. The L2 fragment lacking eRNA sequences retains transcriptional activities, even though the eRNA is necessary for endogenous IFNB1 expression. (A) The genomic fragments used for the L2 fragment (L2F) and the eRNA-excluding L2 fragment (0.3L2F) and the larger eRNA-inclusive fragment (4L2F) are shown against the GRO-seq reads at L2 for reference (from Figure 4). For the GRO-seq data, data from both strands are displayed in a single track. (B) The impact of the exclusion of remaining eRNA sequence was assessed on L2 promoter activity in Orientation 1. (C) The impact of the exclusion of remaining eRNA sequence was assessed on L2 promoter activity in Orientation 2. (D) The impact of the presence or absence of eRNA sequence was assessed on L2 enhancer activity in Orientation 1. (E) The impact of the presence or absence of eRNA sequence was assessed on L2 enhancer activity in Orientation 2. (B–E) All constructs were analyzed in five independent experiments (n = 5). Error bars indicate Standard Error. (F) The effectiveness of both siRNA pools on knockdown of SeV-induced eRNA expression levels from each strand was assessed. (G) The effectiveness of treatment with each siRNA pool on IFNB1 induction in response to SeV was assessed. (F and G) Expression levels were normalized to GAPDH. Each treatment was tested in three independent experiments (n = 3). Error bars indicate Standard Deviation. (B–G) We used unpaired t-tests to assess statistical significance. A single asterisk, ‘*’, indicates P < 0.05. A double asterisk, ‘**’, indicates P < 0.01. The letters ‘ns’ mean ‘not significant and indicate P > 0.05.CC BY · PubMed Central
Figure 9. The 0.3L2 fragment is necessary for proper IFNB1 expression. (A) The BAC CTD-2104N16 is mapped against the human genome. (B) The recombineering process for the deletion of 0.3L2F out of CTD-2104N16 is illustrated. (C) Recomineering success was assessed by PCR performed for 0.3L2F on wild-type, GalK-inserted and 0.3KO deleted BAC clones. (D) The impact of the 0.3L2F deletion on IFNB1 inducibility was assessed by transient transfection into NIH3T3 cells with or without SeV treatment. Each bar represents two independent experiments (n = 2). Error bars indicate one standard deviation.
Figure 9. The 0.3L2 fragment is necessary for proper IFNB1 expression. (A) The BAC CTD-2104N16 is mapped against the human genome. (B) The recombineering process for the deletion of 0.3L2F out of CTD-2104N16 is illustrated. (C) Recomineering success was assessed by PCR performed for 0.3L2F on wild-type, GalK-inserted and 0.3KO deleted BAC clones. (D) The impact of the 0.3L2F deletion on IFNB1 inducibility was assessed by transient transfection into NIH3T3 cells with or without SeV treatment. Each bar represents two independent experiments (n = 2). Error bars indicate one standard deviation.CC BY · PubMed Central

Verification

Claims the paper makes

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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 72 references

Research works printed but not linked: 2

  1. Collins J.R.M. Friedman RM Finter NB Interferon 1984 3 NY Elsevier 33 83
  2. Smit A. Hubley R. Green P. RepeatMasker Open-3.0 1996–2010 http://www.repeatmasker.org (21 October 2014, datelast accessed)

Cited by

Metadata

Authors
Banerjee, A. R., Kim, Y. J., & Kim, T. H.
Affiliations
  • Yale University (US)
  • The University of Texas at Dallas (US)
Corresponding author
Kim, Tae Hoon
Venue
Nucleic Acids Research
Published
2014-10-27
Volume
42
Issue
20
Pages
12537-12554
ISSN
  • 0305-1048 (ISSN-L)
  • 1362-4962
OpenAlex
W2160339448
Cited by
39
References
72
Citations by year
2026: 3 · 2025: 5 · 2024: 2 · 2023: 5 · 2022: 5 · 2021: 1 · 2020: 2 · 2019: 3 · 2018: 4 · 2017: 2 · 2016: 2 · 2015: 5
FWCI
1.23
Type
journal article
Language
en
Open access
Locations
  • Nucleic Acids Research (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • National Institutes of Health
  • U.S. Public Health Service (T32M007499)
  • National Cancer Institute (R01CA140485)
  • National Institute of Allergy and Infectious Diseases (R21AI107067)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Chromatin Dynamics
  • Life Sciences › Immunology and Microbiology › Immunology › interferon and immune responses
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Cancer Research › Cancer-related molecular mechanisms research
Keywords
interferon-beta, IFNB1, long-range enhancers, bidirectional promoter, enhancer RNA, eRNA, interferon-stimulated response element, transcriptional regulation, nascent RNA, transcription factor binding, siRNA knockdown
MeSH terms
Animals, Cell Line, Chromatin, Enhancer Elements, Genetic, Humans, Promoter Regions, Genetic, RNA, Transcription Factors, Transcription, Genetic, Interferon-beta, Sendai virus, Mice, HEK293 Cells

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

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.