Defining fundamental steps in the assembly of the Drosophila RNAi enzyme complex

Abstract

Small RNAs such as small interfering RNAs (siRNAs) and microRNAs (miRNAs) silence the expression of their complementary target messenger RNAs1,2 via the formation of effector RNA-induced silencing complexes (RISCs), which contain Argonaute (Ago) family proteins at their core. Although loading of siRNA duplexes into Drosophila Ago2 requires the Dicer-2–R2D2 heterodimer3,4,5 and the Hsc70/Hsp90 (Hsp90 also known as Hsp83) chaperone machinery6,7,8, the details of RISC assembly remain unclear. Here we reconstitute RISC assembly using only Ago2, Dicer-2, R2D2, Hsc70, Hsp90, Hop, Droj2 (an Hsp40 homologue) and p23. By following the assembly of single RISC molecules, we find that, in the absence of the chaperone machinery, an siRNA bound to Dicer-2–R2D2 associates with Ago2 only transiently. The chaperone machinery extends the dwell time of the Dicer-2–R2D2–siRNA complex on Ago2, in a manner dependent on recognition of the 5′-phosphate on the siRNA guide strand. We propose that the chaperone machinery supports a productive state of Ago2, allowing it to load siRNA duplexes from Dicer-2–R2D2 and thereby assemble RISC.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Figure 1: In vitro reconstitution of Ago2–RISC assembly.
Figure 2: Single-molecule observation of pre- and mature RISC formation.
Figure 3: The chaperone machinery extends the dwell time of Dicer-2–R2D2–siRNA on Ago2.
Figure 4: Recognition of the guide 5′-phosphate is required for dwell-time extension.

Change history

  • 27 May 2015

    Minor formatting changes were made to Figs 2 and 4.

References

  1. 1

    Carthew, R. W. & Sontheimer, E. J. Origins and mechanisms of miRNAs and siRNAs. Cell 136, 642–655 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  2. 2

    Kawamata, T. & Tomari, Y. Making RISC. Trends Biochem. Sci. 35, 368–376 (2010).

    CAS  PubMed  Google Scholar 

  3. 3

    Liu, Q. et al. R2D2, a bridge between the initiation and effector steps of the Drosophila RNAi pathway. Science 301, 1921–1925 (2003).

    ADS  CAS  PubMed  Google Scholar 

  4. 4

    Pham, J. W., Pellino, J. L., Lee, Y. S., Carthew, R. W. & Sontheimer, E. J. A. Dicer-2-dependent 80S complex cleaves targeted mRNAs during RNAi in Drosophila. Cell 117, 83–94 (2004).

    CAS  PubMed  Google Scholar 

  5. 5

    Tomari, Y. et al. RISC assembly defects in the Drosophila RNAi mutant armitage. Cell 116, 831–841 (2004).

    CAS  PubMed  Google Scholar 

  6. 6

    Iki, T. et al. In vitro assembly of plant RNA-induced silencing complexes facilitated by molecular chaperone HSP90. Mol. Cell 39, 282–291 (2010).

    CAS  PubMed  Google Scholar 

  7. 7

    Iwasaki, S. et al. Hsc70/Hsp90 chaperone machinery mediates ATP-dependent RISC loading of small RNA duplexes. Mol. Cell 39, 292–299 (2010).

    CAS  PubMed  Google Scholar 

  8. 8

    Miyoshi, T., Takeuchi, A., Siomi, H. & Siomi, M. C. A direct role for Hsp90 in pre-RISC formation in Drosophila. Nature Struct. Mol. Biol. 17, 1024–1026 (2010).

    CAS  Google Scholar 

  9. 9

    Kawamata, T., Seitz, H. & Tomari, Y. Structural determinants of miRNAs for RISC loading and slicer-independent unwinding. Nature Struct. Mol. Biol. 16, 953–960 (2009).

    CAS  Google Scholar 

  10. 10

    Kanellopoulou, C. et al. Dicer-deficient mouse embryonic stem cells are defective in differentiation and centromeric silencing. Genes Dev. 19, 489–501 (2005).

    CAS  PubMed  PubMed Central  Google Scholar 

  11. 11

    Murchison, E. P., Partridge, J. F., Tam, O. H., Cheloufi, S. & Hannon, G. J. Characterization of Dicer-deficient murine embryonic stem cells. Proc. Natl Acad. Sci. USA 102, 12135–12140 (2005).

    ADS  CAS  PubMed  Google Scholar 

  12. 12

    Ye, X. et al. Structure of C3PO and mechanism of human RISC activation. Nature Struct. Mol. Biol. 18, 650–657 (2011).

    CAS  Google Scholar 

  13. 13

    Betancur, J. G. & Tomari, Y. Dicer is dispensable for asymmetric RISC loading in mammals. RNA 18, 24–30 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  14. 14

    Kim, Y. et al. Deletion of human tarbp2 reveals cellular microRNA targets and cell-cycle function of TRBP. Cell Rep. 9, 1061–1074 (2014).

    CAS  PubMed  Google Scholar 

  15. 15

    Yoda, M. et al. ATP-dependent human RISC assembly pathways. Nature Struct. Mol. Biol. 17, 17–23 (2010).

    CAS  Google Scholar 

  16. 16

    Johnston, M., Geoffroy, M. C., Sobala, A., Hay, R. & Hutvagner, G. HSP90 protein stabilizes unloaded argonaute complexes and microscopic P-bodies in human cells. Mol. Biol. Cell 21, 1462–1469 (2010).

    CAS  PubMed  PubMed Central  Google Scholar 

  17. 17

    Matranga, C., Tomari, Y., Shin, C., Bartel, D. P. & Zamore, P. D. Passenger-strand cleavage facilitates assembly of siRNA into Ago2-containing RNAi enzyme complexes. Cell 123, 607–620 (2005).

    CAS  PubMed  Google Scholar 

  18. 18

    Miyoshi, K., Tsukumo, H., Nagami, T., Siomi, H. & Siomi, M. C. Slicer function of Drosophila Argonautes and its involvement in RISC formation. Genes Dev. 19, 2837–2848 (2005).

    CAS  PubMed  PubMed Central  Google Scholar 

  19. 19

    Rand, T. A., Petersen, S., Du, F. & Wang, X. Argonaute2 cleaves the anti-guide strand of siRNA during RISC activation. Cell 123, 621–629 (2005).

    CAS  PubMed  Google Scholar 

  20. 20

    Liu, Y. et al. C3PO, an endoribonuclease that promotes RNAi by facilitating RISC activation. Science 325, 750–753 (2009).

    ADS  CAS  PubMed  PubMed Central  Google Scholar 

  21. 21

    Smith, D. F. & Toft, D. O. The intersection of steroid receptors with molecular chaperones: observations and questions. Mol. Endocrinol. 22, 2229–2240 (2008).

    CAS  PubMed  PubMed Central  Google Scholar 

  22. 22

    Rüdel, S. et al. Phosphorylation of human Argonaute proteins affects small RNA binding. Nucleic Acids Res. 39, 2330–2343 (2011).

    PubMed  Google Scholar 

  23. 23

    Dittmar, K. D. & Pratt, W. B. Folding of the glucocorticoid receptor by the reconstituted Hsp90-based chaperone machinery. The initial hsp90.p60.hsp70-dependent step is sufficient for creating the steroid binding conformation. J. Biol. Chem. 272, 13047–13054 (1997).

    CAS  PubMed  Google Scholar 

  24. 24

    De, N. et al. Highly complementary target RNAs promote release of guide RNAs from human Argonaute2. Mol. Cell 50, 344–355 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  25. 25

    Nakanishi, K., Weinberg, D. E., Bartel, D. P. & Patel, D. J. Structure of yeast Argonaute with guide RNA. Nature 486, 368–374 (2012).

    ADS  CAS  PubMed  PubMed Central  Google Scholar 

  26. 26

    Schirle, N. T. & MacRae, I. J. The crystal structure of human Argonaute2. Science 336, 1037–1040 (2012).

    ADS  CAS  PubMed  PubMed Central  Google Scholar 

  27. 27

    Tomari, Y., Matranga, C., Haley, B., Martinez, N. & Zamore, P. D. A protein sensor for siRNA asymmetry. Science 306, 1377–1380 (2004).

    ADS  CAS  PubMed  Google Scholar 

  28. 28

    Flaherty, K. M., DeLuca-Flaherty, C. & McKay, D. B. Three-dimensional structure of the ATPase fragment of a 70K heat-shock cognate protein. Nature 346, 623–628 (1990).

    ADS  CAS  PubMed  Google Scholar 

  29. 29

    Prodromou, C. et al. Identification and structural characterization of the ATP/ADP-binding site in the Hsp90 molecular chaperone. Cell 90, 65–75 (1997).

    CAS  Google Scholar 

  30. 30

    Miyazono, Y., Hayashi, M., Karagiannis, P., Harada, Y. & Tadakuma, H. Strain through the neck linker ensures processive runs: a DNA-kinesin hybrid nanomachine study. EMBO J. 29, 93–106 (2010).

    CAS  PubMed  Google Scholar 

  31. 31

    Nykänen, A., Haley, B. & Zamore, P. D. ATP requirements and small interfering RNA structure in the RNA interference pathway. Cell 107, 309–321 (2001).

    PubMed  Google Scholar 

  32. 32

    Haley, B., Tang, G. & Zamore, P. D. In vitro analysis of RNA interference in Drosophila melanogaster. Methods 30, 330–336 (2003).

    CAS  PubMed  Google Scholar 

  33. 33

    Förstemann, K. et al. Normal microRNA maturation and germ-line stem cell maintenance requires Loquacious, a double-stranded RNA-binding domain protein. PLoS Biol. 3, e236 (2005).

    PubMed  PubMed Central  Google Scholar 

  34. 34

    Tomari, Y., Du, T. & Zamore, P. D. Sorting of Drosophila small silencing RNAs. Cell 130, 299–308 (2007).

    CAS  PubMed  PubMed Central  Google Scholar 

  35. 35

    Kawamata, T. & Tomari, Y. Native gel analysis for RISC assembly. Methods Mol. Biol. 725, 91–105 (2011).

    CAS  PubMed  Google Scholar 

  36. 36

    Izumi, N. et al. AAA+ proteins RUVBL1 and RUVBL2 coordinate PIKK activity and function in nonsense-mediated mRNA decay. Sci. Signal. 3, ra27 (2010).

    PubMed  Google Scholar 

  37. 37

    Tsutsumi, A., Kawamata, T., Izumi, N., Seitz, H. & Tomari, Y. Recognition of the pre-miRNA structure by Drosophila Dicer-1. Nature Struct. Mol. Biol. 18, 1153–1158 (2011).

    CAS  Google Scholar 

  38. 38

    Haley, B. & Zamore, P. D. Kinetic analysis of the RNAi enzyme complex. Nature Struct. Mol. Biol. 11, 599–606 (2004).

    CAS  Google Scholar 

  39. 39

    Pall, G. S. & Hamilton, A. J. Improved northern blot method for enhanced detection of small RNA. Nature Protocols 3, 1077–1084 (2008).

    CAS  PubMed  Google Scholar 

  40. 40

    Saito, K. et al. Roles for the Yb body components Armitage and Yb in primary piRNA biogenesis in Drosophila. Genes Dev. 24, 2493–2498 (2010).

    CAS  PubMed  PubMed Central  Google Scholar 

  41. 41

    Zhou, Z. P. et al. Single molecule imaging of the trans-translation entry process via anchoring of the tagged ribosome. J. Biochem. 149, 609–618 (2011).

    CAS  PubMed  Google Scholar 

  42. 42

    Tseng, Q. et al. Spatial organization of the extracellular matrix regulates cell–cell junction positioning. Proc. Natl Acad. Sci. USA 109, 1506–1511 (2012).

    ADS  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We thank Q. Liu for providing vectors for Dicer-2 and R2D2 expression, M. Horwich and P. D. Zamore for Ago2 cDNA, and H. Siomi and M. C. Siomi for anti-Ago2 antibody and ago2414 flies. We are also grateful to P. B. Kwak and A. Tsutsumi for assistance with plasmid constructions, S. Katsuma for support for Dicer-2–R2D2 preparation, and A. Yamashita for advice on protein purification. We thank H. Taguchi and T. Ueda for fruitful discussions and P. D. Zamore, H. Seitz and the members of the Tomari laboratory for critical comments on the manuscript. This work was supported in part by Grants-in-Aid for Scientific Research on Innovative Areas ('Functional machinery for non-coding RNAs' and 'Non-coding RNA neo-taxonomy'), a Grant-in-Aid for Young Scientists (A) (to H.T. and Y.T.), a Grant-in-Aid for Research Activity start-up (to S.I.), and Grants-in-Aid for challenging Exploratory Research (to H.M.S.) from The Ministry of Education, Culture, Sports, Science and Technology in Japan.

Author information

Affiliations

Authors

Contributions

S.I. performed biochemical experiments and H.M.S. performed single-molecule experiments. Y.S. and T.S. performed mass spectrometry analysis. S.I., H.M.S., H.T. and Y.T. designed experiments, analysed data and wrote the manuscript.

Corresponding authors

Correspondence to Hisashi Tadakuma or Yukihide Tomari.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Extended data figures and tables

Extended Data Figure 1 Validation of small RNA pull-down assay.

a, CBB staining of recombinant proteins used in this study. b, Small RNA pull-down assay using immunopurified wild type (WT), catalytic mutant (Cmut; D965A) and 5′ nucleotide-binding pocket mutant (Pmut; Y897E) of Ago2. Ago2 alone (Buffer) could not load siRNA, but supplementation of lysate from ago2414 mutant flies directed efficient formation of pre- and mature Ago2–RISC. The catalytic mutant Ago2 predominantly formed pre-Ago2–RISC, while the 5′ nucleotide-binding pocket mutant failed to load siRNA. c, Western blotting of Ago2 and tubulin in lysate from S2 cells, wild-type embryo and ago2414 mutant embryo. d, Anti-Flag western blotting of wild-type and mutant Ago2 proteins in lysate. All the constructs are expressed at similar levels. e, Hsp70 inhibitor (PES) and p23 inhibitor (Celastrol) block duplex loading into Ago2.

Extended Data Figure 2 In vitro reconstitution of Ago2–RISC assembly by seven recombinant proteins.

a, Quantification of Fig. 1a. b, Quantification of Fig. 1b. c, Time course of target cleavage in the reconstituted system. Figure 1c represents the data at 30 min. d, Activities of the reconstituted system and ago2414 mutant lysate. e, Cleavage assay in a multiple turnover condition by Ago2–RISCs assembled in the reconstituted system or ago2414 mutant embryo lysate (d). ago2414 lysate was diluted to normalize the amount of the assembled Ago2–RISC. f, Estimated concentrations of mature single-strand-containing Ago2–RISCs by the pull-down assay (d) and the pre-steady-state 'burst' kinetics of target cleavage (e). Owing to a technical limitation in the quantification of small amounts of radiolabelled RNAs, the concentrations of RISCs estimated by the burst kinetics were higher than those estimated by the pull-down assay. However, the ratios between them were essentially identical between RISCs assembled in the reconstituted system and ago2414 mutant lysate, suggesting that the reconstituted Ago2–RISC is as fully active as the one assembled in lysate. g, Dicer-2–R2D2 and the chaperone machinery are dispensable for loading of single-stranded RNAs whereas they are essential for duplex loading. h, Quantification of g. Data represent mean ± s.d. (n = 3).

Extended Data Figure 3 Unloading is not accompanied by loading.

a, Quantification of Fig. 1d. b, Experimental strategy for c. c, The amount of pre-loaded CXCR4 siRNA did not change during new loading of let-7 siRNA. d, Quantification of c. Red, reconstituted system; black, GST. Data represent mean ± s.d. (n = 3).

Extended Data Figure 4 The reconstituted system faithfully recapitulates siRNA loading asymmetry and ATP dependency.

a, Ago2–RISC was assembled with siRNAs with either the guide strand or the passenger strand radiolabelled. Left, asymmetry-flipped let-7 siRNA. In contrast to the normal let-7 siRNA (Fig. 1e), the let-7 sequence in this siRNA duplex serves as the passenger strand due to the introduction of a mismatch at the opposite end. Middle, highly asymmetric CXCR4 siRNA. Right, relatively symmetric luc siRNA, in which the two strands can serve as the guide strand with similar chances. b, Small RNA pull-down assay was performed in the reconstituted system with 0.5 mM ATP, ATP-γS, AMP-PNP or ADP. c, Quantification of b. d, Quantification of Fig. 1f.

Extended Data Figure 5 Experimental scheme for single-molecule analysis.

a–c, Halo-TEV-Flag–Ago2 was tethered to the surface as illustrated. For snapshot observation (a), the reaction mixture was infused into the observation chamber and incubated for 1 h. After washing of the surface, the images of the wash-resistant spots were acquired for 1 s. For snapshot observation after protease treatment (b), the protease mixture was infused into the chamber, which had been imaged and incubated for 30 min. For continuous monitoring (c), images were acquired for 20 min without washing immediately after the reaction was started on the surface.

Extended Data Figure 6 Halo-tagged Ago proteins and fluorescently labelled siRNAs used in single-molecule analysis.

a, Visualization of Halo-TEV-Flag-tagged wild type, catalytic mutant (Cmut; D965A) and 5′ nucleotide-binding pocket mutant (Pmut; Y897E) of Ago2 by a Halo-TMR ligand. b, RNA pull-down assay with siRNAs with or without 3′ fluorescent modifications. Red, guide strand; black, passenger strand. The 5′-phosphate of the guide strand was radiolabelled as denoted by yellow circles. Fluorescent labelling did not compromise the efficiency of RISC assembly in vitro. c, Time course of the guide-alone (red) or double-stranded (yellow) spot formation for wild-type Ago2 with the reconstituted system. d, TEV-protease-mediated cleavage of Halo-TEV-Flag–Ago2. After S2 lysate containing Halo-TEV-Flag–Ago2 was incubated with proteases (HRV3C or TEV) or lysis buffer for 30 min, western blotting was performed using anti-Halo and anti-Flag antibodies. e, f, Representative single-molecule images (e) and quantification (f) after protease treatment for 30 min (Extended Data Fig. 5b). Wash-resistant spots observed in the WT, +D/R and +Cpn conditions (Fig. 2b, c) largely disappeared after TEV-protease-mediated cleavage but not after the control treatment of HRV3C protease. The colour scheme is the same as in Fig. 2b, c. g, h, Validation of pre- and mature RISC formation with inversion of the strand colours. Representative single-molecule images (g) and quantification (h) of wash-resistant spots after 60 min incubation with the reconstituted system together with an siRNA with a 3′ Alexa555 (green)-labelled guide strand and a 3′ Alexa647 (red)-labelled passenger strand. Data represent mean ± s.d. (WT in e, n = 4; others, n = 3).

Extended Data Figure 7 Determination of the time constants by cumulative curve fitting.

a, The data in Fig. 3c for short binding were accumulated and fitted to the following equation derived from the one-step reaction model: C˙[1 − exp(−t/τ)], where C is the normalized parameter, τ is the time constant, and t is the reaction time. b, The data in Fig. 3e for the yellow-to-red transition in wild type were accumulated and fitted to the following equation derived from the two-step reaction model: [C0/(C1C2)]˙[C1˙exp(−C2˙t) − C2˙exp(−C1˙t)], where C0 is the normalized parameter, C1 and C2 are the inverse of the time constants τ1 and τ2, respectively, and t is the reaction time. ce, The data in Fig. 3f–h were accumulated and fitted to the following equation: C˙exp(−t/τ), where C is the normalized parameter, τ is the time constant, and t is the reaction time.

Extended Data Figure 8 The 5′-phosphate of the guide strand is dispensable for binding to Dicer-2–R2D2 but is essential for duplex loading into Ago2.

a, Gel shift assay with increasing concentrations of recombinant Dicer-2–R2D2 and passenger-strand-radiolabelled siRNA duplex with the 5′-phosphate (left) or 5′-hydroxyl group (right) on the guide strand. The 5′-phosphate of the guide strand is dispensable for the interaction with Dicer-2–R2D2. Red, guide strand; black, passenger strand. b, c, Small RNA pull-down assay with siRNA duplexes containing 5′-phosphate (left) or 5′-hydroxyl group (right) on the guide strand with wild-type or the catalytic mutant Ago2. The 3′ end of the guide strand (b) or the 5′ end of the passenger strand (c) of the siRNA duplex was radiolabelled, as denoted by yellow circles. The 5′-phosphate on the guide strand is essential for duplex loading into Ago2; in the absence of the guide 5′-phosphate, the siRNA duplex is hardly loaded while maintaining the functional asymmetry. d, Determination of the time constants of short binding in the absence of 5′-phosphate recognition. The data in Fig. 4d were accumulated and fitted to the one-step reaction model as in Extended Data Fig. 7a.

Extended Data Figure 9 Multiple fundamental steps in the assembly of Drosophila Ago2–RISC.

In the absence of the chaperone machinery, an siRNA bound to Dicer-2–R2D2 binds Ago2 only transiently, dissociating rapidly. The Hsc70/Hsp90 chaperone machinery extends their dwell time, presumably by inducing a productive state of Ago2, in a manner dependent on the recognition of the 5′-phosphate on the guide strand by Ago2. Successful anchoring of the 5′-phosphate would trigger loading of the rest of the siRNA duplex from Dicer-2–R2D2 into Ago2, leading to passenger ejection and mature RISC formation.

Extended Data Figure 10 Single-molecule properties of the observed fluorescent spots.

a, Representative time courses of one-step (left) and two-step (right) photobleaching for the guide (red) or passenger (green) strand. Spots for wild-type Ago2 with the reconstituted system were analysed. Arrows show the timings of photobleaching. b, 95% of the fluorescent spots showed one-step photobleaching for each colour, indicating that each spot contained a single RNA molecule. c, Photobleaching dependence on power density. Left, Alexa647. Right, Alexa555. The photobleaching time constants of Alexa647 and 555 under continuous monitoring condition were estimated as 22,800 and 5,900 s, respectively. Data represent mean ± s.d. (n = 3). d, Representative observed spots in snapshot observation without deconvolution (raw data). The distribution of fluorescence intensity was fitted to Gaussian function (bottom). The diameter of the spots is comparable to the diameter of the Airy disk (533 nm), indicating that the spots are diffraction limited. Scale bar, 500 nm. e, The signal intensity distributions of observed spots and single-molecule benchmarks for Alexa647 (red, left) and Alexa555 (green, right) in snapshot observation. Observed, fluorescent spots observed in snapshot observation (Fig. 2c; WT, +D/R, +Cpn, yellow). Benchmark, the catalytic mutant Ago2 programmed with siRNA duplex, of which the guide and passenger strands were labelled with Alexa647 and 555, respectively (1×, middle) or with the same dye (2×, bottom). Images of single-molecule benchmarks were taken under the same acquisition condition of snapshot observation. Background was subtracted from the raw data. f, Representative observed spots in continuous monitoring without deconvolution (raw data). The distribution of fluorescence intensity was fitted to Gaussian function (bottom). The diameter of the spots is comparable to the diameter of the Airy disk (533 nm), indicating that the spots are diffraction limited. Scale bar, 500 nm. g, The signal intensity distributions of observed spots and single-molecule benchmarks for Alexa647 (red, left) and Alexa555 (green, right) in continuous monitoring. Observed, fluorescent spots observed in continuous monitoring (Fig. 3d; WT, +D/R, +Cpn, yellow). Benchmark, the catalytic mutant Ago2 programmed with siRNA duplex, of which the guide and passenger strands were labelled with Alexa647 and 555, respectively (1×, middle) or with the same dye (2×, bottom). Images of single-molecule benchmarks were taken under the same acquisition condition of continuous monitoring. Background was subtracted from the raw data.

Supplementary information

Supplementary Information

This file contains a Supplementary Discussion and additional references. (PDF 137 kb)

PowerPoint slides

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Iwasaki, S., Sasaki, H., Sakaguchi, Y. et al. Defining fundamental steps in the assembly of the Drosophila RNAi enzyme complex. Nature 521, 533–536 (2015). https://doi.org/10.1038/nature14254

Download citation

Further reading

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

Sign up for the Nature Briefing newsletter for a daily update on COVID-19 science.
Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing