CoChat

Pyruvate kinase variant of fission yeast tunes carbon metabolism, cell regulation, growth and stress resistance

Open access

Published version of a preprint

The DOI matches its Crossref record; 138 of the 139 references it cites were found in a registry or an index.

Abstract

Cells balance glycolysis with respiration to support their metabolic needs in different environmental or physiological contexts. With abundant glucose, many cells prefer to grow by aerobic glycolysis or fermentation. Using 161 natural isolates of fission yeast, we investigated the genetic basis and phenotypic effects of the fermentation–respiration balance. The laboratory and a few other strains depended more on respiration. This trait was associated with a single nucleotide polymorphism in a conserved region of Pyk1, the sole pyruvate kinase in fission yeast. This variant reduced Pyk1 activity and glycolytic flux. Replacing the “low‐activity” pyk1 allele in the laboratory strain with the “high‐activity” allele was sufficient to increase fermentation and decrease respiration. This metabolic rebalancing triggered systems‐level adjustments in the transcriptome and proteome and in cellular traits, including increased growth and chronological lifespan but decreased resistance to oxidative stress. Thus, low Pyk1 activity does not lead to a growth advantage but to stress tolerance. The genetic tuning of glycolytic flux may reflect an adaptive trade‐off in a species lacking pyruvate kinase isoforms.

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 · 10.15252/msb.20199270

Cites

138 of 138 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 A variant in conserved pyk1 region is associated with increased sensitivity to antimycin A Distribution of antimycin A resistance scores for wild isolates, compared to standard laboratory strain (red vertical line). Resistance scores are the ratio of fitness on rich glucose media with vs. without 500 μg/l antimycin A. Fitness was estimated based on colony size on solid media, corrected for spatial and plate effects (Materials and Methods). After quality control, we obtained quantitative fitness scores for 154 strains, with a signal‐to‐noise ratio of 29.8 and an unexplained variance of 0.12.Volcano plot of a genome‐wide association using mixed‐model linear regression of antimycin A resistance for 118,527 genetic variants. Variants with moderate or high impact are shown in blue. Red dot: the variant at locus I:3,845,516, which causes a T343A change in the Pyk1 amino acid sequence. This variant was among the top scoring (effect size = −0.645, rank 34; P = 0.0008, rank 70).Boxplot showing antimycin A resistance for 154 strains grouped by the two alleles at the pyk1 locus. Strains carrying a C at this genomic locus (orange box) generally have higher antimycin resistance than strains carrying the reference allele T (blue box). As is standard, this boxplot and all other boxplots in this manuscript show the median of the data as central line, the quartiles as box and the extend of the rest of the distribution as whiskers. Points which are 1.5 times the inter‐quartile range beyond the high and low quartiles are considered outliers and shown individually.Sequence alignment of the region of interest in eukaryotic PYK proteins. The threonine residue at the highlighted position 343 is unique for the reference (laboratory) S. pombe strains. For this analysis, we collected 25 Pyk1 homologues from a wide variety of eukaryotes, including animals, plants and fungi, using the HomoloGene resource (Accession 37650). We manually expanded the homologous group to include three other species of the Schizosaccharomyces genus (Rhind et al, 2011). Sequences were aligned with MAFFT (Katoh et al, 2017) and visualised with Jalview (Waterhouse et al, 2009).Phylogenetic tree based on 31 biallelic SNPs in pyk1 and in 500bp up‐ and downstream regions. Strains carrying the common A‐allele are in orange, while strains carrying the unusual T‐allele are in blue (strain names as in ref. Jeffares et al (2015)). Strains JB22 and JB50 (underlined and bold) refer to heterothallic and homothallic versions of the laboratory strain, respectively. The pie chart shows the relative allele frequencies at the genomic locus across all wild strains. Genotype calls at the locus were checked manually and were conclusive, i.e. no strain produced reads with both alleles.
Figure 1 A variant in conserved pyk1 region is associated with increased sensitivity to antimycin A Distribution of antimycin A resistance scores for wild isolates, compared to standard laboratory strain (red vertical line). Resistance scores are the ratio of fitness on rich glucose media with vs. without 500 μg/l antimycin A. Fitness was estimated based on colony size on solid media, corrected for spatial and plate effects (Materials and Methods). After quality control, we obtained quantitative fitness scores for 154 strains, with a signal‐to‐noise ratio of 29.8 and an unexplained variance of 0.12.Volcano plot of a genome‐wide association using mixed‐model linear regression of antimycin A resistance for 118,527 genetic variants. Variants with moderate or high impact are shown in blue. Red dot: the variant at locus I:3,845,516, which causes a T343A change in the Pyk1 amino acid sequence. This variant was among the top scoring (effect size = −0.645, rank 34; P = 0.0008, rank 70).Boxplot showing antimycin A resistance for 154 strains grouped by the two alleles at the pyk1 locus. Strains carrying a C at this genomic locus (orange box) generally have higher antimycin resistance than strains carrying the reference allele T (blue box). As is standard, this boxplot and all other boxplots in this manuscript show the median of the data as central line, the quartiles as box and the extend of the rest of the distribution as whiskers. Points which are 1.5 times the inter‐quartile range beyond the high and low quartiles are considered outliers and shown individually.Sequence alignment of the region of interest in eukaryotic PYK proteins. The threonine residue at the highlighted position 343 is unique for the reference (laboratory) S. pombe strains. For this analysis, we collected 25 Pyk1 homologues from a wide variety of eukaryotes, including animals, plants and fungi, using the HomoloGene resource (Accession 37650). We manually expanded the homologous group to include three other species of the Schizosaccharomyces genus (Rhind et al, 2011). Sequences were aligned with MAFFT (Katoh et al, 2017) and visualised with Jalview (Waterhouse et al, 2009).Phylogenetic tree based on 31 biallelic SNPs in pyk1 and in 500bp up‐ and downstream regions. Strains carrying the common A‐allele are in orange, while strains carrying the unusual T‐allele are in blue (strain names as in ref. Jeffares et al (2015)). Strains JB22 and JB50 (underlined and bold) refer to heterothallic and homothallic versions of the laboratory strain, respectively. The pie chart shows the relative allele frequencies at the genomic locus across all wild strains. Genotype calls at the locus were checked manually and were conclusive, i.e. no strain produced reads with both alleles.CC BY · PubMed Central
Figure 2 Multi‐omic functional investigation of Pyk1 variants PCA of metabolite data based on concentrations of 27 central carbon metabolism intermediates. To visualise this high‐dimensional data set, we divided the concentration of each metabolite by the median concentration of the T‐strain. This normalisation corrects for the large differences in concentrations observed between metabolites but maintains the relative variance within each metabolite. Biological replicates of the T‐ and A‐strains show distinct profiles, largely driven by concentrations of phosphoenolpyruvate, 2‐/3‐phosphoglyceric acid, NADH and NADPH, as indicated by top loading vectors for each principal component. The biological repeats for the three edited A‐strains (circles, squares and triangles) behave similarly, with a variance comparable to that of the biological replicates of the single T‐strain.Left boxplot: two glycolytic intermediates directly upstream of PYK were strongly depleted in the A‐strain (n T‐strain = 9, n A‐strain = 8). Right barplot: PYK activity was directly measured using a lactate dehydrogenase‐coupled colorimetric enzyme activity assay, showing the mean and standard deviation of substrate conversion rate for three biological replicates, each measured in technical duplicates.Concentrations of fructose‐1,6‐bisphosphate, which correlate with glycolytic flux, are significantly higher in A‐strain (ratio = 1.35, n T‐strain = 9, n A‐strain = 8).Boxplots for selected metabolomics data indicate differences in energy and redox status between T‐ and A‐strains (n T‐strain = 9, n A‐strain = 8).Heatmap of the 432 genes that are differentially expressed at the RNA level between the T‐ and A‐strains (FDR < 10%) and are measured in all four conditions (columns). First column: genes ordered by increasing fold‐changes for RNAs (computed as log2[T]‐log2[A]). Second column: fold‐changes for proteins (computed as log2[T]‐log2[A]). Third column: fold‐changes for RNAs in cells treated with rapamycin and caffeine (TORC1 inhibition; computed as log2[treatment]‐log2[control]) (data from ref. Rallis et al, 2013). Fourth column: fold‐changes for RNAs in cells treated with H2O2 (oxidative stress; computed as log2[treatment]‐log2[control]) (data from Chen et al, 2003). Log2 fold‐changes are capped at absolute values of 1 for all columns.Gene Ontology (GO) enrichment analysis of differentially expressed transcripts (using all measured transcripts as background). Shown are Biological Process terms only, with additional plots for Molecular Function and Cellular Component terms in Appendix Fig S5. The length of the bar represents the significance of the enrichment, while its colour reflects the direction and magnitude of differential expression (A‐ vs. T‐strain) of individual genes annotated to this term.Data information: Significance keys: *P < 0.05, **P < 0.005, ***P < 0.0005 (Welch's t‐test).
Figure 2 Multi‐omic functional investigation of Pyk1 variants PCA of metabolite data based on concentrations of 27 central carbon metabolism intermediates. To visualise this high‐dimensional data set, we divided the concentration of each metabolite by the median concentration of the T‐strain. This normalisation corrects for the large differences in concentrations observed between metabolites but maintains the relative variance within each metabolite. Biological replicates of the T‐ and A‐strains show distinct profiles, largely driven by concentrations of phosphoenolpyruvate, 2‐/3‐phosphoglyceric acid, NADH and NADPH, as indicated by top loading vectors for each principal component. The biological repeats for the three edited A‐strains (circles, squares and triangles) behave similarly, with a variance comparable to that of the biological replicates of the single T‐strain.Left boxplot: two glycolytic intermediates directly upstream of PYK were strongly depleted in the A‐strain (n T‐strain = 9, n A‐strain = 8). Right barplot: PYK activity was directly measured using a lactate dehydrogenase‐coupled colorimetric enzyme activity assay, showing the mean and standard deviation of substrate conversion rate for three biological replicates, each measured in technical duplicates.Concentrations of fructose‐1,6‐bisphosphate, which correlate with glycolytic flux, are significantly higher in A‐strain (ratio = 1.35, n T‐strain = 9, n A‐strain = 8).Boxplots for selected metabolomics data indicate differences in energy and redox status between T‐ and A‐strains (n T‐strain = 9, n A‐strain = 8).Heatmap of the 432 genes that are differentially expressed at the RNA level between the T‐ and A‐strains (FDR < 10%) and are measured in all four conditions (columns). First column: genes ordered by increasing fold‐changes for RNAs (computed as log2[T]‐log2[A]). Second column: fold‐changes for proteins (computed as log2[T]‐log2[A]). Third column: fold‐changes for RNAs in cells treated with rapamycin and caffeine (TORC1 inhibition; computed as log2[treatment]‐log2[control]) (data from ref. Rallis et al, 2013). Fourth column: fold‐changes for RNAs in cells treated with H2O2 (oxidative stress; computed as log2[treatment]‐log2[control]) (data from Chen et al, 2003). Log2 fold‐changes are capped at absolute values of 1 for all columns.Gene Ontology (GO) enrichment analysis of differentially expressed transcripts (using all measured transcripts as background). Shown are Biological Process terms only, with additional plots for Molecular Function and Cellular Component terms in Appendix Fig S5. The length of the bar represents the significance of the enrichment, while its colour reflects the direction and magnitude of differential expression (A‐ vs. T‐strain) of individual genes annotated to this term.Data information: Significance keys: *P < 0.05, **P < 0.005, ***P < 0.0005 (Welch's t‐test).CC BY · PubMed Central
Figure 3 Metabolic network showing metabolome, transcriptome and proteome changes triggered by pyk1 allele replacement Metabolites, transcripts and proteins are coloured by their abundance ratios in A‐strain relative to T‐strain (see legend top left for details). For clarity, enzymes without protein or transcript changes are not shown (e.g. most upper glycolysis). Two metabolite pairs are indistinguishable by our LC‐MS method (marked as “+” and “#”).
Figure 3 Metabolic network showing metabolome, transcriptome and proteome changes triggered by pyk1 allele replacement Metabolites, transcripts and proteins are coloured by their abundance ratios in A‐strain relative to T‐strain (see legend top left for details). For clarity, enzymes without protein or transcript changes are not shown (e.g. most upper glycolysis). Two metabolite pairs are indistinguishable by our LC‐MS method (marked as “+” and “#”).CC BY · PubMed Central
Figure 4 Physiological characterisation of pyk1 variant Growth measured by optical density (OD) at 600 nm for 3 biological replicates each of the T‐ and A‐strains over 6 hrs of exponential growth in YES media. The x‐axis refers to the time since inoculation of the culture.The same data as in (A) plotted on a log2 scale (solid lines) with lines fitted (dotted lines). The doubling time is the inverse of the slope.Media samples were taken from the same cultures (n = 3 for both strains) at the 8 hr timepoint and the remaining glucose was quantified. The consumed glucose was calculated based on the amount of glucose measured in the same, fresh media and normalised to the OD of each culture at the time of sampling.Oxygen consumption rates in A‐ and T‐strains (ratio of means = 0.65, P = 0.0002, Welch's t‐test, three biological replicates, each measured in technical duplicates).Culture dry weight after 24 h was measured reported as a fraction of the weight of glucose put into the media for three biological replicates per strain.Data information: Vertical bars show the mean of the data. Error bars and shaded areas in all cases denote standard deviation. Significance keys: *P < 0.05, ***P < 0.0005 (Welch's t‐test).
Figure 4 Physiological characterisation of pyk1 variant Growth measured by optical density (OD) at 600 nm for 3 biological replicates each of the T‐ and A‐strains over 6 hrs of exponential growth in YES media. The x‐axis refers to the time since inoculation of the culture.The same data as in (A) plotted on a log2 scale (solid lines) with lines fitted (dotted lines). The doubling time is the inverse of the slope.Media samples were taken from the same cultures (n = 3 for both strains) at the 8 hr timepoint and the remaining glucose was quantified. The consumed glucose was calculated based on the amount of glucose measured in the same, fresh media and normalised to the OD of each culture at the time of sampling.Oxygen consumption rates in A‐ and T‐strains (ratio of means = 0.65, P = 0.0002, Welch's t‐test, three biological replicates, each measured in technical duplicates).Culture dry weight after 24 h was measured reported as a fraction of the weight of glucose put into the media for three biological replicates per strain.Data information: Vertical bars show the mean of the data. Error bars and shaded areas in all cases denote standard deviation. Significance keys: *P < 0.05, ***P < 0.0005 (Welch's t‐test).CC BY · PubMed Central
Figure 5 Cellular phenotypes mediated by pyk1 allele replacement Spot assays on solid media from a threefold dilution series of exponential cultures at the same cell density in 96‐well plates (3 biological replicates of each strain) and spotted in 16 technical replicates (each dilution in 4 × 4 square). The A‐strain is more resistant to antimycin A but less resistant to oxidative stress triggered by H2O2 or diamide. A control without toxin (left) was included in each batch of spot assays performed and a representative image is shown here.Fitness (approximated by maximum slope of smoothed growth curves) of A‐ and T‐strains on 12 carbon sources, with either yeast extract (YE) or ammonium (NH4), with or without 0.1% priming glucose to support initial growth. For all 48 conditions, two biological replicates of A‐ and T‐strains were grown in technical quadruplicates each. Dotted lines in panels B‐D mark a fitness ratio of 1 (i.e. same fitness).Fitness of A‐ and T‐strains on 95 nitrogen sources on Biolog Phenotype MicroArrays. Conditions with no substantial growth were excluded (black circles, maximum slope < 0.015). Red lines show arbitrary significance cut‐off, put at |log2(A‐strain/T‐strain)| > 0.75.Fitness of A‐ and T‐strains on 72 different drugs and toxins, at 4 concentrations each, on Biolog Phenotype MicroArrays. Graph details and cut‐off as in (C). (Results for benzamidine were inconclusive, with both strains comparatively resistant in one concentration each.)Spot assays as in (A) to validate selected results from (C) and (D). Several assays were performed with both h‐ and h90 strains, and no mating‐type‐specific differences were evident between the two sets of allele swap strains.Chronological lifespan of A‐ and T‐strains, i.e. the proportion of non‐dividing cells in stationary phase that maintain proliferative potential after refeeding. The data show colony forming units (CFUs) per ml of culture over 7 days of stationary phase in glucose‐depleted rich media. Three biological repeats were carried out for both strains, with each repeat measured as technical triplicates. Error bars represent standard error of the biological replicates.Boxplot showing resistance to 3 mM H2O2 grouped by pyk1 allele for 156 strains from our collection. The T‐strains had a higher mean fitness in H2O2 than the A‐strains (1.01 ± 0.06 vs. 0.95 ± 0.11; P = 0.0021, Welch's t‐test). The resistance score was obtained as for antimycin A.Boxplot showing growth fitness on rich media, grouped by pyk1 allele for 158 S. pombe strains (0.88 ± 0.16 vs. 0.94 ± 0.24; P = 0.18, Welch's t‐test).
Figure 5 Cellular phenotypes mediated by pyk1 allele replacement Spot assays on solid media from a threefold dilution series of exponential cultures at the same cell density in 96‐well plates (3 biological replicates of each strain) and spotted in 16 technical replicates (each dilution in 4 × 4 square). The A‐strain is more resistant to antimycin A but less resistant to oxidative stress triggered by H2O2 or diamide. A control without toxin (left) was included in each batch of spot assays performed and a representative image is shown here.Fitness (approximated by maximum slope of smoothed growth curves) of A‐ and T‐strains on 12 carbon sources, with either yeast extract (YE) or ammonium (NH4), with or without 0.1% priming glucose to support initial growth. For all 48 conditions, two biological replicates of A‐ and T‐strains were grown in technical quadruplicates each. Dotted lines in panels B‐D mark a fitness ratio of 1 (i.e. same fitness).Fitness of A‐ and T‐strains on 95 nitrogen sources on Biolog Phenotype MicroArrays. Conditions with no substantial growth were excluded (black circles, maximum slope < 0.015). Red lines show arbitrary significance cut‐off, put at |log2(A‐strain/T‐strain)| > 0.75.Fitness of A‐ and T‐strains on 72 different drugs and toxins, at 4 concentrations each, on Biolog Phenotype MicroArrays. Graph details and cut‐off as in (C). (Results for benzamidine were inconclusive, with both strains comparatively resistant in one concentration each.)Spot assays as in (A) to validate selected results from (C) and (D). Several assays were performed with both h‐ and h90 strains, and no mating‐type‐specific differences were evident between the two sets of allele swap strains.Chronological lifespan of A‐ and T‐strains, i.e. the proportion of non‐dividing cells in stationary phase that maintain proliferative potential after refeeding. The data show colony forming units (CFUs) per ml of culture over 7 days of stationary phase in glucose‐depleted rich media. Three biological repeats were carried out for both strains, with each repeat measured as technical triplicates. Error bars represent standard error of the biological replicates.Boxplot showing resistance to 3 mM H2O2 grouped by pyk1 allele for 156 strains from our collection. The T‐strains had a higher mean fitness in H2O2 than the A‐strains (1.01 ± 0.06 vs. 0.95 ± 0.11; P = 0.0021, Welch's t‐test). The resistance score was obtained as for antimycin A.Boxplot showing growth fitness on rich media, grouped by pyk1 allele for 158 S. pombe strains (0.88 ± 0.16 vs. 0.94 ± 0.24; P = 0.18, Welch's t‐test).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

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 138 references

Research works printed but not linked: 1

  1. Ephrussi B, Hottinguer H, Tavlitzki J (1949) Action de l'acriflavine sur les levures. I. La mutation ‘petite colonie’. Ann Inst Pasteur 76: 351–367

Cited by

See all 40

Metadata

Authors
Kamrad, S., Grossbach, J., Rodríguez‐López, M., Mülleder, M., Townsend, S., Cappelletti, V., Stojanovski, G., Correia‐Melo, C., Picotti, P., Beyer, A., Ralser, M., & Bähler, J.
Affiliations
  • The Francis Crick Institute (GB)
  • University of Cologne (DE)
  • University College London (GB)
  • ETH Zurich (CH)
Corresponding author
  • Beyer, Andreas
  • Ralser, Markus
  • Bähler, Jürg
Venue
Molecular Systems Biology
Published
2020-04-01
Volume
16
Issue
4
Pages
e9270-e9270
ISSN
1744-4292 (ISSN-L)
OpenAlex
W3017345718
Cited by
40
References
139
Citations by year
2026: 3 · 2025: 8 · 2024: 4 · 2023: 5 · 2022: 6 · 2021: 7 · 2020: 4
FWCI
1.78
Type
journal article
Language
en
Open access
Licence
Locations
  • Molecular Systems Biology (publishedVersion, other-oa)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Kölner Universitäts PublikationsServer (Universität zu Köln) (publishedVersion)
  • Repository for Publications and Research Data (ETH Zurich) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • UCL Discovery (University College London) (submittedVersion)
  • Figshare (submittedVersion, cc-by)
  • ETH Zürich Research Collection (cc-by)
Funders
  • Wellcome Trust (095598/Z/11/Z, FC001134)
  • Cancer Research UK (FC001134)
  • UK Research and Innovation|Medical Research Council (FC001134)
  • Deutsche Forschungsgemeinschaft (CRC 680, CRC 1310)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Fungal and yeast genetics research
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Cancer Research › Cancer, Hypoxia, and Metabolism
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Metabolism, Diabetes, and Cancer
Keywords
pyruvate kinase, Schizosaccharomyces pombe, glycolysis, respiration, fermentation, single nucleotide polymorphisms, glycolytic flux, oxidative stress resistance, chronological lifespan, transcriptome, proteome, wild isolate, metabolic adaptation
MeSH terms
Carbon, Fermentation, Glycolysis, Pyruvate Kinase, Schizosaccharomyces, Gene Expression Regulation, Fungal, Gene Expression Regulation, Enzymologic, Oxidative Stress, Mutation, Missense, Polymorphism, Single Nucleotide, Gene Expression Profiling, Schizosaccharomyces pombe Proteins, Proteomics

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: Pyruvate kinase variant of fission yeast tunes carbon metabolism, cell regulation, growth and stress resistance. https://app.cochat.ai/papers/10.15252/msb.20199270
The paper
Kamrad, S., Grossbach, J., Rodríguez‐López, M., Mülleder, M., Townsend, S., Cappelletti, V., Stojanovski, G., Correia‐Melo, C., Picotti, P., Beyer, A., Ralser, M., & Bähler, J. (2020). Pyruvate kinase variant of fission yeast tunes carbon metabolism, cell regulation, growth and stress resistance. Molecular Systems Biology. https://doi.org/10.15252/msb.20199270

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.