Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

Structural basis of species-selective antagonist binding to the succinate receptor

Abstract

The tricarboxylic acid cycle intermediate succinate is involved in metabolic processes and plays a crucial role in the homeostasis of mitochondrial reactive oxygen species1. The receptor responsible for succinate signalling, SUCNR1 (also known as GPR91), is a member of the G-protein-coupled-receptor family2 and links succinate signalling to renin-induced hypertension, retinal angiogenesis and inflammation3,4,5. Because SUCNR1 senses succinate as an immunological danger signal6—which has relevance for diseases including ulcerative colitis, liver fibrosis7, diabetes and rheumatoid arthritis3,8—it is of interest as a therapeutic target. Here we report the high-resolution crystal structure of rat SUCNR1 in complex with an intracellular binding nanobody in the inactive conformation. Structure-based mutagenesis and radioligand-binding studies, in conjunction with molecular modelling, identified key residues for species-selective antagonist binding and enabled the determination of the high-resolution crystal structure of a humanized rat SUCNR1 in complex with a high-affinity, human-selective antagonist denoted NF-56-EJ40. We anticipate that these structural insights into the architecture of the succinate receptor and its antagonist selectivity will enable structure-based drug discovery and will further help to elucidate the function of SUCNR1 in vitro and in vivo.

This is a preview of subscription content, access via your institution

Access options

Rent or buy this article

Prices vary by article type

from$1.95

to$39.95

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Structure of the apo rat SUCNR1–Nanobody6 complex.
Fig. 2: Development of a human SUCNR1 antagonist and humanization of rat SUCNR1.
Fig. 3: Binding mode of the antagonist NF-56-EJ40 to humanized rat SUCNR1.
Fig. 4: Structural similarities between SUCNR1 and the P2Y1 receptor.

Similar content being viewed by others

Data availability

Structure factors and coordinates of the rat SUCNR1–Nanobody6 and the SUCNR1(K181.31E/K2697.32N)–Nanobody6–NF-56-EJ40 complex structures have been deposited in the Protein Data Bank (PDB) under accession codes 6IBB and 6RNK, respectively. All source data associated with the paper (in addition to those deposited) are provided as Supplementary Information.

References

  1. Tretter, L., Patocs, A. & Chinopoulos, C. Succinate, an intermediate in metabolism, signal transduction, ROS, hypoxia, and tumorigenesis. Biochim. Biophys. Acta 1857, 1086–1101 (2016).

    Article  CAS  Google Scholar 

  2. He, W. et al. Citric acid cycle intermediates as ligands for orphan G-protein-coupled receptors. Nature 429, 188–193 (2004).

    Article  ADS  CAS  Google Scholar 

  3. Gilissen, J., Jouret, F., Pirotte, B. & Hanson, J. Insight into SUCNR1 (GPR91) structure and function. Pharmacol. Ther. 159, 56–65 (2016).

    Article  CAS  Google Scholar 

  4. Peruzzotti-Jametti, L. et al. Macrophage-derived extracellular succinate licenses neural stem cells to suppress chronic neuroinflammation. Cell Stem Cell 22, 355–368.e13 (2018).

    Article  CAS  Google Scholar 

  5. Schneider, C. et al. A metabolite-triggered tuft cell-ILC2 circuit drives small intestinal remodeling. Cell 174, 271–284.e14 (2018).

    Article  CAS  Google Scholar 

  6. Rubic, T. et al. Triggering the succinate receptor GPR91 on dendritic cells enhances immunity. Nat. Immunol. 9, 1261–1269 (2008).

    Article  CAS  Google Scholar 

  7. Cho, E. H. Succinate as a regulator of hepatic stellate cells in liver fibrosis. Front. Endocrinol. 9, 455 (2018).

    Article  Google Scholar 

  8. Littlewood-Evans, A. et al. GPR91 senses extracellular succinate released from inflammatory macrophages and exacerbates rheumatoid arthritis. J. Exp. Med. 213, 1655–1662 (2016).

    Article  CAS  Google Scholar 

  9. Jha, A. K. et al. Network integration of parallel metabolic and transcriptional data reveals metabolic modules that regulate macrophage polarization. Immunity 42, 419–430 (2015).

    Article  CAS  Google Scholar 

  10. Chouchani, E. T. et al. Ischaemic accumulation of succinate controls reperfusion injury through mitochondrial ROS. Nature 515, 431–435 (2014).

    Article  ADS  CAS  Google Scholar 

  11. Tannahill, G. M. et al. Succinate is an inflammatory signal that induces IL-1β through HIF-1α. Nature 496, 238–242 (2013).

    Article  ADS  CAS  Google Scholar 

  12. Mills, E. L. et al. Succinate dehydrogenase supports metabolic repurposing of mitochondria to drive inflammatory macrophages. Cell 167, 457–470 (2016).

    Article  CAS  Google Scholar 

  13. de Castro Fonseca, M., Aguiar, C. J., da Rocha Franco, J. A., Gingold, R. N. & Leite, M. F. GPR91: expanding the frontiers of Krebs cycle intermediates. Cell Commun. Signal. 14, 3 (2016).

    Article  Google Scholar 

  14. Keiran, N. et al. SUCNR1 controls an anti-inflammatory program in macrophages to regulate the metabolic response to obesity. Nat. Immunol. 20, 581–592 (2019).

    Article  CAS  Google Scholar 

  15. Trauelsen, M. et al. Receptor structure-based discovery of non-metabolite agonists for the succinate receptor GPR91. Mol. Metab. 6, 1585–1596 (2017).

    Article  CAS  Google Scholar 

  16. Bhuniya, D. et al. Discovery of a potent and selective small molecule hGPR91 antagonist. Bioorg. Med. Chem. Lett. 21, 3596–3602 (2011).

    Article  CAS  Google Scholar 

  17. Chun, E. et al. Fusion partner toolchest for the stabilization and crystallization of G protein-coupled receptors. Structure 20, 967–976 (2012).

    Article  CAS  Google Scholar 

  18. Pardon, E. et al. A general protocol for the generation of Nanobodies for structural biology. Nat. Protoc. 9, 674–693 (2014).

    Article  CAS  Google Scholar 

  19. Cherezov, V. et al. High-resolution crystal structure of an engineered human β2-adrenergic G protein-coupled receptor. Science 318, 1258–1265 (2007).

    Article  ADS  CAS  Google Scholar 

  20. Rasmussen, S. G. et al. Crystal structure of the β2 adrenergic receptor–Gs protein complex. Nature 477, 549–555 (2011).

    Article  ADS  CAS  Google Scholar 

  21. Zhang, D. et al. Two disparate ligand-binding sites in the human P2Y1 receptor. Nature 520, 317–321 (2015).

    Article  ADS  CAS  Google Scholar 

  22. Che, T. et al. Structure of the nanobody-stabilized active state of the kappa opioid receptor. Cell 172, 55–67 (2018).

    Article  CAS  Google Scholar 

  23. Staus, D. P. et al. Allosteric nanobodies reveal the dynamic range and diverse mechanisms of G-protein-coupled receptor activation. Nature 535, 448–452 (2016).

    Article  ADS  CAS  Google Scholar 

  24. Burg, J. S. et al. Structural basis for chemokine recognition and activation of a viral G protein-coupled receptor. Science 347, 1113–1117 (2015).

    Article  ADS  CAS  Google Scholar 

  25. Kruse, A. C. et al. Activation and allosteric modulation of a muscarinic acetylcholine receptor. Nature 504, 101–106 (2013).

    Article  ADS  CAS  Google Scholar 

  26. Rasmussen, S. G. et al. Structure of a nanobody-stabilized active state of the β2 adrenoceptor. Nature 469, 175–180 (2011).

    Article  ADS  CAS  Google Scholar 

  27. Ballesteros, J. A. & Weinstein, H. Integrated methods for the construction of three-dimensional models and computational probing of structure-function relations in G protein-coupled receptors. Methods Neurosci. 25, 366–428 (1995).

    Article  CAS  Google Scholar 

  28. Caffrey, M. & Cherezov, V. Crystallizing membrane proteins using lipidic mesophases. Nat. Protoc. 4, 706–731 (2009).

    Article  CAS  Google Scholar 

  29. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007).

    Article  CAS  Google Scholar 

  30. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D 66, 213–221 (2010).

    Article  CAS  Google Scholar 

  31. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D 60, 2126–2132 (2004).

    Article  Google Scholar 

  32. Jacobson, M. P. et al. A hierarchical approach to all-atom protein loop prediction. Proteins 55, 351–367 (2004).

    Article  CAS  Google Scholar 

  33. Milletti, F., Storchi, L., Sforna, G. & Cruciani, G. New and original pK a prediction method using grid molecular interaction fields. J. Chem. Inf. Model. 47, 2172–2181 (2007).

    Article  CAS  Google Scholar 

  34. Halgren, T. A. et al. Glide: a new approach for rapid, accurate docking and scoring. 2. Enrichment factors in database screening. J. Med. Chem. 47, 1750–1759 (2004).

    Article  CAS  Google Scholar 

  35. Friesner, R. A. et al. Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. J. Med. Chem. 47, 1739–1749 (2004).

    Article  CAS  Google Scholar 

  36. McWilliam, H. et al. Analysis tool web services from the EMBL-EBI. Nucleic Acids Res. 41, W597–W600 (2013).

    Article  Google Scholar 

  37. Robert, X. & Gouet, P. Deciphering key features in protein structures with the new ENDscript server. Nucleic Acids Res. 42, W320–W324 (2014).

    Article  CAS  Google Scholar 

  38. Ashkenazy, H. et al. ConSurf 2016: an improved methodology to estimate and visualize evolutionary conservation in macromolecules. Nucleic Acids Res. 44, W344–W350 (2016).

    Article  CAS  Google Scholar 

Download references

Acknowledgements

We thank C. Schleberger for help with implementing crystallographic data processing tools; T. Huber, R. Link and A. Winterhalter for help with nanobody generation; P. Loesle for biochemical assay support; S. Haenni and S. Holzinger for molecular biology support; and E. Loetscher for providing the rat SUCNR1-CHO cell line.

Author information

Authors and Affiliations

Authors

Contributions

M.H., A.L.-E., K.K., M.D. and V.P.J. designed the experiments. G.R., J.B. and V.P.J. performed initial protein expression and purification screening and established receptor purification protocols. M.H. expressed and purified all proteins, established complex purification and crystallization, solved and analysed structures, performed thermostability assays and proposed structure-based mutagenesis. D.F. and K.K. established and performed radioligand-binding and GTPγS assays. M.D. initiated and supervised nanobody generation. S.C. designed compounds and built human SUCNR1 homology models. F.S. built structural models of the humanized rat and human SUCNR1 and performed docking analysis. N.G. designed and synthesized compounds. F.F. designed expression constructs. M.H., K.K., A.L.-E. and V.P.J. wrote the manuscript with input from all authors.

Corresponding authors

Correspondence to Matthias Haffke, Klemens Kaupmann or Veli-Pekka Jaakola.

Ethics declarations

Competing interests

All authors are employees of Novartis Pharma AG.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Peer review information Nature thanks Edward Chouchani, Mike Murphy, Janos Peti-Peterdi and Gebhard Schertler for their contribution to the peer review of this work.

Extended data figures and tables

Extended Data Fig. 1 The succinate–SUCNR1 signalling axis.

Levels of the Krebs cycle intermediate succinate are increased under certain conditions such as hypoxia, necrosis, ischaemia reperfusion and inflammation. The ways in which succinate concentrations increase in the mitochondrion are shown in green. Mitochondrial reactive oxygen species result from the reversed electron transport (RET) chain driven by an increase in succinate. Succinate is transported into the cytoplasm, where it can stabilize HIF1α and increase the expression of genes that have HIF-responsive elements, such as IL1Β. Further succinate is exported into the local extracellular environment, where it accumulates and binds and activates SUCNR1. Several of the consequences of this are shown in red. GABA, γ-aminobutyric acid; HIF, hypoxic inducible factor; IL-1, interleukin-1; mROS, mitochondrial reactive oxygen species; SDH, succinate dehydrogenase.

Extended Data Fig. 2 Interaction of Nanobody6 with rat and human SUCNR1 and characterization of the effects of glycerol and 2,5-hexanediol on rat SUCNR1.

a, b, Nanobody6 increases the thermal stability of rat SUCNR1 in nano-DSF thermal shift assays. The average melting curves in the absence (a) and presence (b) of Nanobody6 are shown (n = 3; technical replicates). The experiment was repeated independently 3 times with similar results. c, Analytical size-exclusion chromatography shows a clear shift in the peak of the rat SUCNR1–Nanobody6 complex compared to the peak of the receptor alone. The complex samples contain a 1.2 molar excess of Nanobody6 over receptor. One of n = 2 independent experiments is shown. d, [35S]GTPγS assay on wild-type rat SUCNR1 in the absence or the presence of increasing concentrations of Nanobody6. The average curves of n = 3 independent experiments are shown; data are mean ± s.d. Average half-maximum effective concentration (EC50) values from n = 3 independent experiments are listed; data are mean ± s.d. e, Analytical size-exclusion chromatography shows a clear shift in the peak of the N-terminal BRIL-fused human SUCNR1–Nanobody6 complex compared to the peak of the receptor alone. One of n = 2 independent experiments is shown. f, [35S]GTPγS assay on wild-type human SUCNR1 in the absence or the presence of increasing concentrations of Nanobody6. The average curves of n = 3 independent experiments are shown and average EC50 values from n = 3 independent experiments are listed; data are mean ± s.d. g, Analytical size-exclusion chromatography of rat SUCNR1 purified in the absence or the presence of 10% glycerol. One of n = 2 independent experiments is shown. h, Glycerol increases the thermal stability of rat SUCNR1 as evidenced from nano-DSF assays (n = 3 technical replicates; bars represent mean values; individual data points are indicated by circles). The control sample was purified in the presence of 10% glycerol. All other samples contain rat SUCNR1 purified without glycerol, to which the respective final glycerol concentration was added. The experiment was repeated independently twice with similar results. i, 2,5-Hexanediol decreases the thermal stability of rat SUCNR1 in nano-DSF assays (n = 3 technical replicates; bars represent mean values; individual data points are indicated by circles). The experiment was repeated independently twice with similar results. j, k, [35S]GTPγS assay of rat SUCNR1 (j) and human S1P1R (k) in the absence or the presence of 1% (109 mM) glycerol. The average curve of n = 3 independent experiments and the individual data points of each experiment are shown. Data are mean ± s.d.

Source Data

Extended Data Fig. 3 Purification, crystallization and electron-density map quality of the rat SUCNR1–Nanobody6 complex, with detailed binding modes of glycerol and 2,5-hexanediol.

a, Analytical size-exclusion chromatography and SDS–PAGE analysis of crystallization samples of the rat SUCNR1–Nanobody6 complex. Shown is a representative experiment of n = 5 independent experiments with similar results. For gel source data, see Supplementary Fig. 1a. b, Initial crystallization hits for the rat SUCNR1–Nanobody6 complex (top) and optimized crystals used for data collection (bottom). Shown are representative experiments of n = 20 independent experiments. c, The 2FoFc electron density map contoured at 1.5σ for a part of Nanobody6. d, The 2FoFc electron-density map contoured at 1.5σ for helix VII in rat SUCNR1. e, f, FoFc composite omit map for glycerol (e) and corresponding 2FoFc electron-density map after refinement (f). Both maps are contoured at 1.5σ. g, h, FoFc composite omit map for 2,5-hexanediol (g) and corresponding 2FoFc electron-density map after refinement (h). Both maps are contoured at 1.5σ. i, j, Detailed views of the side-chain environment around 2,5-hexanediol (i) and glycerol (j). Hydrogen bonds are indicated by dashed lines. For clarity, only residues within a distance of 4 Å are shown.

Source Data

Extended Data Fig. 4 Rat SUCNR1 adopts an inactive conformation in complex with Nanobody6, which binds to the intracellular side via an extended CDR3 loop.

a, Structural alignment of helix VI in rat SUCNR1 and the active and inactive states of β2-AR, with the positions of key residues as hallmarks for inactive and active receptor states. SUCNR1 is shown in blue, inactive β2-AR in red and active β2-AR in pink. Alignment of helix VI (left) and side-chain positions of key residues (R3.50 and Y7.35) (right) indicate an inactive state for rat SUCNR1. b, Structural alignment of helix VI and key residues (R3.50 and Y7.35) of rat SUCNR1 (blue) and the P2Y1 receptor (orange) in the inactive conformation. c, Superposition of the rat SUCNR1–Nanobody6 complex with the GαS subunit from the β2-AR GS-protein trimer structure (PDB ID: 3SN6). Rat SUCNR1 is shown in blue, Nanobody6 in orange and the GαS subunit in red. The G-protein and the Nanobody6-binding site partially overlap. d, Magnified view of the overlap between the G-protein and the Nanobody6-binding site of rat SUCNR1. e, Structural alignment of nanobodies used to crystallize GPCRs. The extended CDR3 of Nanobody6 forms a helical secondary structure. f, Sequence alignment of the GPCR-stabilizing nanobodies shown in e. The PDB IDs of the respective GPCR–nanobody complex structures are listed and the CDR3 region is highlighted by a black bar.

Extended Data Fig. 5 Sequence alignment of rat, human and mouse SUCNR1 and ConSurf analysis.

a, The helical structure elements as observed in the crystal structure of apo rat SUCNR1 are indicated. Sequences corresponding to ECL1 and ECL2 are boxed in blue and the non-conserved residues K181.31 and K2697.32 are marked by blue arrows. Yellow arrowheads indicate residues that were previously reported to be involved in succinate-induced receptor activation. Green dots indicate residues that are involved in NF-56-EJ40 binding in the humanized rat SUCNR1 structure. b, The sequence alignment of SUCNR1 from various species is colour-coded from turquoise (variable) to dark pink (conserved), similar to the colours used in Extended Data Fig. 6a, b, on the basis of analysis with ConSurf. Residues involved in the binding of Nanobody6 are indicated by blue triangles.

Extended Data Fig. 6 Potential ligand-binding site in apo SUCNR1 is partially occluded by ECL2.

a, Side view (top) and top view (bottom) of the hydrophobic pocket located below the glycerol molecule. The sequence conservation within the SUCNR1 receptor family is indicated by colour, ranging from turquoise for highly variable residues to dark pink for highly conserved residues. The hydrophobic pocket is shown as a surface, colour-coded by charge. The glycerol molecule is shown as yellow sticks. b, The same orientations as in a are shown. Residues forming the deep hydrophobic pocket are shown as sticks, colour-coded by sequence conservation as in a. Residues that were previously reported to be involved in succinate-induced receptor activation are coloured yellow. c, Residues in the environment of ECL2 are shown as sticks and hydrogen bonds are shown as dashed black lines. Residues that have previously been reported to have an effect on succinate binding by the receptor are shown in yellow2; R2516.58, which was identified in a second study15, is shown in green.

Extended Data Fig. 7 Identification of critical residues that impart species selectivity for antagonist binding.

a, Side view (left) and top view (right) of the potential binding mode of the antagonist NF-56-EJ40 (shown in pink), obtained by molecular modelling based on the crystal structure of apo rat SUCNR1. Residues within 4 Å of NF-56-EJ40 are shown as sticks. Red arrows point towards two sites in which potential steric clashes may occur with rat-SUCNR1-specific residues K181.31 and K2697.32, which are shown in blue. For clarity, ECL2, helix IV and helix V are omitted in the side view. b, Radioligand competition binding experiments with unlabelled NF-56-EJ40 on human SUCNR1 mutant proteins. Curves were calculated from n = 3 independent experiments; data are mean ± s.d. Individual data points are shown.

Source Data

Extended Data Fig. 8 Purification, crystallization and electron-density map quality of the rat SUCNR1–Nanobody6–NF-56-EJ40 complex and structural changes in SUCNR1 after antagonist binding.

a, Analytical size-exclusion chromatography and SDS–PAGE analysis of crystallization samples of the humanized rat SUCNR1–Nanobody6–NF-56-EJ40 complex. Shown is a typical result from n = 2 independent experiments. Although partial complex aggregation was observed, this did not interfere with crystallization. For gel source data, see Supplementary Fig. 1b. b, Top, initial crystallization hits for the humanized rat SUCNR1–Nanobody6–NF-56-EJ40 complex, shown in normal (left) and cross-polarization (right) imaging modes. Bottom, optimized crystals used for data collection shown in normal (left) or cross-polarization (right) imaging modes. A typical result from n = 3 independent experiments is shown. c, The 2FoFc electron density map contoured at 1.5σ for a part of Nanobody6. d, The 2FoFc electron density map contoured at 1.5σ for helix VII in humanized rat SUCNR1. e, FoFc composite omit map (top) for NF-56-EJ40 and glycerol contoured at 1.5σ is shown in orange. The 2FoFc map (bottom) for NF-56-EJ40 and glycerol after refinement contoured at 1.5σ is shown in blue. f, Top view of apo rat SUCNR1 (shown in orange) and humanized rat SUCNR1 (shown in blue) in complex with NF-56-EJ40 (shown as green sticks). Large structural rearrangements are indicated by red arrows. Note that ECL2 is completely structured in the humanized rat SUCNR1 structure. g, Top view of the NF-56-EJ40-binding site in humanized rat SUCNR1 overlaid with apo rat SUCNR1. Important side chains around NF-56-EJ40 (shown in green) are shown as sticks and are coloured blue for humanized rat SUCNR1 or orange for apo wild-type rat SUCNR1. For clarity, only the backbone of the humanized rat SUCNR1 is shown in cartoon representation. h, Side chains that directly interact with NF-56-EJ40 via hydrogen bonding, π–π stacking and cation–π stacking are listed in black. The hydrogen-bonding interactions are shown by black dashed lines and the π–π and cation–π interactions are shown by green dashed lines. Additional residues with van der Waals interactions are listed in green, and their interaction surfaces are indicated by solid green lines. i, Top view of the humanized rat SUCNR1 (blue) in complex with NF-56-EJ40 (green sticks) and of the apo rat SUCNR1-derived model of human SUCNR1 (red) with the binding mode of NF-56-EJ40 (pink) from molecular-docking studies, for which details are shown in Extended Data Fig. 7a. Note how both NF-56-EJ40 poses differ considerably, as indicated by red arrows. j, Detailed views of the NF-56-EJ40-binding site. For clarity, only side chains are shown. Humanized rat SUCNR1 is shown in blue; wild-type rat SUCNR1 is shown in orange. Note the side-chain flips for R953.29, L983.32, H993.33 and Y171 between both structures. k, Top view of wild-type rat SUCNR1 structure (left), the apo rat SUCNR1-based homology model of human SUCNR1 in complex with NF-56-EJ40 (middle) and the humanized rat SUCNR1 structure in complex with NF-56-EJ40 (right). The surface is shown coloured by electrostatic charge. The two key positions (K/E1.31 and K/N7.32) are highlighted by arrows. NF-56-EJ40 is shown in yellow as a ball-and-stick model. Note the differences between the NF-56-EJ40-binding mode determined in the modelled structure and in the crystal structure, and between the surface charge distributions in rat, human and humanized rat SUCNR1.

Source Data

Extended Data Table 1 Data collection and refinement statistics
Extended Data Table 2 Data for the binding of radioligand [3H]NF-56-EJ40 with human and rat SUCNR1 mutants

Supplementary information

Supplementary Information

This file contains: Supplementary Figure 1 - uncropped SDS-PAGE gels for items displayed in Extended Data Figure 3a and Extended Data Figure 8a; Supplementary Methods - Chemical synthesis of SUCNR1 antagonists JC-59-GF68, PB-20-OV24 and NF-58-EJ40.

Reporting Summary

Source data

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Haffke, M., Fehlmann, D., Rummel, G. et al. Structural basis of species-selective antagonist binding to the succinate receptor. Nature 574, 581–585 (2019). https://doi.org/10.1038/s41586-019-1663-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41586-019-1663-8

This article is cited by

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing