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.

  • In the Laboratory
  • Published:

Accurate determination of relative messenger RNA levels by RT-PCR

Nature Biotechnology volume 17, pages 720–722 (1999)Cite this article

Abstract

The increasing focus on functional genomics spurs new techniques for accurately quantitating differences in mRNA levels.

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

Access options

Buy this article

  • Purchase on SpringerLink
  • Instant access to the full article PDF.

USD 39.95

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

Figure 1: The measuring range of the assay when coamplifying 0, 103, 104, and 105 copies of an internal detection limit (IDL) standard with 102–109 copies of SCCA1 and SCCA2 cDNA in a 1:1 ratio.
Figure 2: The interassay accuracy displayed as 99% confidence intervals of the means of two test samples included in 10 individual sample series and analyzed in separate experiments as single samples.

References

  1. Wang, A.M., Doyle, M.V. & Mark, D.F. Proc. Natl. Acad. Sci. 86, 9717 –9721 (1989).

    Article  CAS  Google Scholar 

  2. Becker-André, M. & Hahlbrock, K. Nucl. Acids Res. 17, 9437–9446 ( 1989).

    Article  Google Scholar 

  3. Gilliland, G., Perrin, S., Blanchard, K. & Bunn, H.F. Proc. Natl. Acad. Sci. 87, 2725–2729 (1990).

    Article  CAS  Google Scholar 

  4. Ikonen, E., Manninen, T., Peltonen, L. & Syvänen, A.C. PCR Methods Appl. 1, 234–240 (1992).

    Article  CAS  Google Scholar 

  5. Piatak, M., Jr., et al. Science 259, 1749– 1754 (1993).

    Article  CAS  Google Scholar 

  6. Pannetier, C., Delassus, S., Darche, S., Saucier, C. & Kourilsky, P. Nucl. Acids Res. 21, 577– 583 (1993).

    Article  CAS  Google Scholar 

  7. Alard, P. et al. BioTechniques 15, 730– 737 (1993).

    CAS  Google Scholar 

  8. Holland, P.M., Abramson, R.D., Watson, R., & Gelfand, D.H. Proc. Natl. Acad. Sci. 88, 7276–7280 (1991).

    Article  CAS  Google Scholar 

  9. Gibson, U.E., Heid, C.A. & Williams, P.M. Genome Res. 6, 995– 1001 (1996).

    Article  CAS  Google Scholar 

  10. Livak, K.J., Flood, S.J., Marmaro, J., Giusti, W. & Deetz, K. PCR Methods Appl. 4, 357– 362 (1995).

    Article  CAS  Google Scholar 

  11. Wittwer, C.T., Herrmann, M.G., Moss, A.A. & Rasmussen, R.P. BioTechniques 22, 130–131, 134–138 (1997).

    Article  Google Scholar 

  12. Karttunen, L., Lönnqvist, L., Godfrey, M., Peltonen, L. & Syvänen, A.C. Genome Res. 6 , 392–403 (1996).

    Article  CAS  Google Scholar 

  13. Kubo, Y., Klimek, F., Kikuchi, Y., Bannasch, P. & Hino, O. Cancer Res. 55, 989– 990 (1995).

    CAS  Google Scholar 

  14. Schütze, K. & Lahr, G. Nature Biotechnol. 16, 737–742 ( 1998).

    Article  Google Scholar 

  15. Ihalainen, J., Siitari, H., Laine, S., Syvänen, A.C. & Palotie, A. BioTechniques 16, 938– 943 (1994).

    CAS  Google Scholar 

  16. Chen, X., Zehnbauer, B., Gnirke, A. & Kwok, P.Y. Proc. Natl. Acad. Sci. 94, 10756–10761 (1997).

    Article  CAS  Google Scholar 

  17. Tyagi, S. & Kramer, F.R. Nature Biotechnol. 14, 303–308 (1996).

    Article  CAS  Google Scholar 

  18. Kato, H., Nagaya, T. & Torigoe, T. GANN 75, 433– 435 (1984).

    CAS  Google Scholar 

  19. Syvänen, A.C., Aalto-Setälä, K., Harju, L., Kontula, K. & Söderlund, H. Genomics 8, 684–692 (1990).

    Article  Google Scholar 

  20. Suomalainen, A. & Syvänen, A.C. Methods Mol. Biol. 86, 121–131 ( 1998).

    CAS  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. department of clinical chemistry University of Helsinki and laboratory department of the Helsinki University, Central Hospital, Helsinki, Finland

    Jakob Stenman, Patrik Finne, Ulf-Håkan Stenman & Arto Orpana

  2. The Haartman Institute, University of Helsinki, Helsinki, Finland

    Anders Ståhls

  3. department of otolaryngology, Turku University Central Hospital, Turku, Finland

    Reidar Grénman

  4. department of pathology and laboratory medicine, UCLA, Los Angeles, CA

    Aarno Palotie

Authors
  1. Jakob Stenman
    View author publications

    Search author on:PubMed Google Scholar

  2. Patrik Finne
    View author publications

    Search author on:PubMed Google Scholar

  3. Anders Ståhls
    View author publications

    Search author on:PubMed Google Scholar

  4. Reidar Grénman
    View author publications

    Search author on:PubMed Google Scholar

  5. Ulf-Håkan Stenman
    View author publications

    Search author on:PubMed Google Scholar

  6. Aarno Palotie
    View author publications

    Search author on:PubMed Google Scholar

  7. Arto Orpana
    View author publications

    Search author on:PubMed Google Scholar

Corresponding author

Correspondence to Jakob Stenman.

Supplementary information

Abstract

The increasing focus on functional genomics spurs new techniques for accurately quantitating differences in mRNA levels.

Quantitative RT-PCR has evolved into a widely used tool for sensitive detection and quantitation of low-abundance RNA species. As the focus of genomic research is shifting from the location of genes towards functional genomics there is a growing demand for techniques capable of accurately quantitating differences in mRNA levels in different settings.

The first truly quantitative RT-PCR techniques used a synthetic internal standard consisting of in vitro generated RNA1, 2 or DNA3. In these competitive methods the wild-type cDNA is coamplified with the internal standard and the initial mRNA copy number is estimated from the ratio of the endpoint wild-type and internal standard PCR products. Based on this technique several methods for quantitation of the end point PCR products have been described4-7.

More recently, real time PCR techniques have been developed in which the accumulating PCR products are fluorometrically monitored during amplification. Quantitation can be done competitively against an internal standard by measuring the relative decrease in fluorescein quenching by rhodamine after exonuclease cleavage of dual-labeled probes (TaqMan, Perkin-Elmer)8, 9 or by resonance energy transfer of fluorescein to Cy5 between adjacent probes (FRET principle)10. As these techniques are dependent on specific hybridization of the fluorescent probes, there will be inherent inaccuracies in the quantitative detection of templates differing at single nucleotide positions because of misannealing of the hybridization probes.

Alternatively, quantitation can be done against a constitutively expressed house-keeping gene by monitoring the fluorescence of a double strand-specific dye at separate product-specific melting temperatures during cycling (LightCycler, Roche) 11. Real time PCR offers fast analysis of samples as quantitation is done during amplification and the risk of post amplification cross contamination of samples is abolished. However, current techniques typically allow accurate quantitation to a factor of two. This does not yet offer an increase in accuracy compared to well-optimized competitive techniques using endpoint PCR product detection.

Another approach for quantitation of mRNA levels by RT-PCR is to use two closely related mRNAs occurring in the same cells as internal standards for each other12. We describe here an RT-PCR-based technique for accurate determination of the relative levels of two closely related mRNAs. The technique was developed in order to study the differential expression of the squamous cell carcinoma antigen (SCCA) genes 1 and 2 in microscopic frozen sections of normal and malignant epithelium. SCCA13 is widely used as a serum marker in squamous cell carcinomas of the uterine cervix, and the head and neck region. SCCA is encoded by two separate serpin genes SCCA1 and SCCA214. SCCA1, showing cross class inhibition of cathepsins15 is the dominant form expressed in normal epithelium whereas SCCA2 displaying more traditional serpin properties16, is mainly responsible for the increased serum SCCA levels seen in malignant conditions17.

Materials and methods

Sample preparation. Tumor (n=80) and control (n=37) samples consisted of five 10 µm microscopic frozen sections of squamous cell carcinoma tumor specimens and normal epithelium surgically removed from the head and neck region. The area of the sections varied from approximately 1 to 50 mm2. The frozen sections were added directly to 750 µl of RLT lysis buffer (Qiagen, Hilden, Germany) containing guanidine isothiocyanate and stored at -70ƒ C. The proportion of the different celltypes in the sample was estimated from an adjacent toluidine blue stained section by an experienced pathologist (A.S.). The samples were homogenized by centrifuging the lysate through a QIAshredder spin column (QIAgen). Total RNA was extracted using RNeasy Mini spin columns (QIAgen) and treated with 2 U of RQ1 RNase-free DNase (Promega, Madison, WI) at 37ƒC for 30 min according to the manufacturers' instructions. One fifth of the total RNA was reverse transcribed with Superscript II reverse transcriptase (Life Technologies, Rockville, MD) according to the instructions, using the common outer antisense PCR primer.

Wild-type templates and IDL standard. The SCCA1 and SCCA2 templates differ within the PCR amplicon at positions 206, 211 and 212 of the SCCA2 cDNA sequence (Genebank, accession number: U19557). An internal standard was constructed by generating an additional A to T mismatch at position 206 of the SCCA2 cDNA using modified primers and PCR. The generated sequence was amplified with the outer PCR primers, cloned into the pCR II vector (Invitrogen, Carlsbad, CA) and sequenced from both ends on an ABI 310 genetic analyzer (Perkin-Elmer, Norwalk, CT) using the ABI Prism dRhodamine terminator cycle sequencing kit (Perkin-Elmer). One thousand copies of the nonlinearized vector containing the modified SCCA2 sequence was included in the PCR reactions as an IDL standard.

PCR and minisequencing primers. To optimize stringency during the initial cycles of the outer PCR reaction the primers were designed to have binding energies at the 3' end between -7.5 and -8.0 kcal/mol and a relatively stronger -9.0 to -11.0 kcal/mol binding domain close to the 5' end. The aim was to ensure that elongation takes place only after a major portion of the primer has annealed to the template18. The formation of primer dimers was eliminated by designing all of the PCR primers so that any possible primer-primer duplexes with protruding 5' end enabling formation of amplified dimers possessed considerably weaker binding energies than the 3' ends of the primer to template duplexes. A minimum of three adjacent mismatches in the 3' ends of possible primer-primer duplexes was required. Oligo 5.0 primer analysis software (National Biosciences, Plymouth, MN) was used as an aid in designing and analysing PCR and minisequensing primers. The outer PCR primers were: 5'-TTCTATTCCCCTATCAGCATC-3' (sense), 5'-TTGCAGCTTTTTCTGTGGT-3' (antisense) and the inner were: 5'-biotin-GCATCACATCAGCATTAGG-3' (sense) and 5'-GCTTTTTCTGTGGTGTT- CTC-3' (antisense). The PCR amplicon spans the exon 2-3 junction of the SCCA1 and SCCA2 genes. The minisequencing primers: 5'-GTGAAGAACCTTCTTAAT-3' (SCCA1) and: 5'-GTGAAGAACCTTGCTAAT-3' (SCCA2 and the internal standard) both bind symmetrically on the exon 2-3 junction.

Amplification and detection of PCR products. One µl of the cDNA was transferred to a 40 µl PCR reaction containing 0.25 mM of each dNTP, 20 pmol of the outer sense and antisense primers, 1.6 U of Dynazyme II polymerase (Finnzymes, Espoo, Finland) and 1 x PCR buffer (10 mM Tris-HCl, pH 8.8, 1.5 mM MgCl2, 50 mM KCl, 1 ml/l Triton X-100) supplied with the enzyme. The IDL standard was added to the master mix to give one thousand copies per PCR reaction. Thirty cycles of amplification at 94ƒC for 30 s, 57ƒC decreasing by 0.1ƒC for each cycle for 1 min and 72ƒC for 30 s was performed on a Geneamp 2400 cycler (Perkin-Elmer). Two µl of the PCR product was transferred to a 100 µl nested PCR reaction containing 20 pmol of the 5'-biotinylated inner sense primer and 100 pmol of the inner antisense primer and 4 U of Dynazyme II polymerase. Thirty additional amplification cycles were performed at 94ƒC for 1 min, 55ƒC for 1 min and 72ƒC for 30 s. Following amplification, the biotinylated sense strand of the nested PCR product was captured on a Scintistrip streptavidin coated scintillating microtitration plate (EG&G Wallac, Turku, Finland). The three endpoint PCR products were separately labeled with 3H-labeled nucleotides by solid-phase minisequencing and the incorporated radioactivity was measured in a beta counter and expressed as counts per minute (CPM) 19-21.

Estimation of relative mRNA levels and sample exclusion criteria. A set of six calibrators consisting of three mixtures of the cloned outer PCR amplicons of SCCA1 and SCCA2 cDNA (at 0.33, 1.00 and 3.00 ratios) at two dilutions (104 and 108 copies), and two negative controls containing only IDL standard template, were included in each experiment. The relative SCCA2/SCCA1 mRNA level in the samples was estimated from a standard curve calculated by linear regression of the log of the CPM ratios and the log of the SCCA2/SCCA1 cDNA ratios in the calibrators. The amount of wild-type cDNA template in a sample was considered sufficient for estimation of the SCCA2/SCCA1 mRNA ratio when the amplification (absolute CPM signal) of the one thousand copy IDL standard was competitively suppressed below half of that observed in the negative control samples. This indicates that the wild-type cDNA template copy number exceeds that of IDL standard. Samples were excluded from the study if the CPM signal obtained from the IDL standard exceeded half of that obtained from the negative controls.

Results

Measuring range and effect of the internal standard. The measuring range of the assay was determined by amplifying dilutions of SCCA1 and SCCA2 cDNA in a 1:1 ratio over the concentration range 102 to 109 copies together with 0, 103, 104 and 105 copies of the IDL standard (see Fig. 1). The ratio of the SCCA1 and SCCA2 PCR products remained constant over the whole range when the copy number of the wild-type cDNA templates exceeded that of the IDL standard. In the final assay 1000 copies of the IDL standard was used, limiting the measuring range to 103-109 copies of input wild-type cDNA template. In our material of 117 samples from microscopic frozen sections of normal epithelium and tumor specimen 16 samples were excluded, as the observed SCCA2/SCCA1 ratio was derived from less than 1000 of cDNA template. All observed CPM signals for SCCA1 and SCCA2 in the accepted samples exceeded three times the baseline determined from their respective negative controls. No signal was obtained from genomic DNA template.

Figure 1

(GIF 26.6 KB)

Four figures showing the measuring range of the assay when coamplifying 0, 103, 104 and 105 copies of an internal detection limit (IDL) standard with 102 to 109 copies of SCCA1 and SCCA2 cDNA in a 1:1 ratio. The ratio of the SCCA1 and SCCA2 PCR products remained constant over the whole range when the wild-type cDNA template copy number exceeded that of the IDL standard. In the final assay one thousand copies of the IDL standard was used limiting the measuring range to 103-109 copies of input cDNA wild-type template.

Accuracy of the assay. The accuracy of the assay was determined by preparing two test samples by pooling cDNA from six normal epithelium samples and six squamous cell carcinoma tumor samples from the head and neck region. The test samples were included in ten individual sample series and analysed in separate experiments as single samples. The 99% confidence intervals of the means were 0.23 to 0.27 (mean=0.25) and 0.37 to 0.41 (mean 0.39) for the normal epithelium and squamous cell carcinoma tumor test samples, respectively (see Fig. 2). The inter-assay coefficient of variation (CV) was 9% for the normal, and 5% for the tumor test samples.

Figure 2

(GIF 16.3 KB)

The interassay accuracy displayed as 99% confidence intervals of the means of two test samples included in ten individual sample series and analysed in separate experiments as single samples. The two test samples were prepared by pooling cDNA from six squamous cell carcinoma tumor samples and six normal epithelium samples from the head and neck region.

Discussion

This RT-PCR technique measures the relative levels of two closely related mRNAs occurring in the same cells. Despite variation in RNA degradation during sample handling and storage, the relative proportion of these templates can be expected to remain fairly constant. Two highly homologous mRNA templates are also likely to be subjected to similar variations in efficacy of the RNA extraction and reverse transcription reaction. Thus, the ratio of the cDNA templates inserted in the PCR reaction reflects that of the mRNAs in the sample irrespectively of factors affecting recovery. This ratio can be accurately estimated by extrapolating the PCR product ratio against a standard curve of PCR calibrators, consisting of known ratios of the templates under study. Using this technique, the relative mRNA levels of genes showing cell type specific expression can be accurately determined even in tissue samples containing a heterogeneous mixture of cells of various origin and without prior knowledge of the exact amount of analyzed cells.

In the amplification step the ratio of nearly identical PCR products corresponds linearly to the ratio of the input cDNA template whenever sufficient amounts of the templates are present (see Fig. 1). It is, however, essential to control that the observed PCR product ratio is derived from a sufficient amount of cDNA template molecules. Even single or a few molecules of the cDNA templates will result in detectable PCR products and subsequently give rise to a wildly inaccurate PCR product ratio. To control that the wild-type cDNA templates have been present at representative amounts in the PCR reaction, 1000 copies of an IDL standard was included as a third template in the amplification step. Thus, the nearly identical templates of SCCA1, SCCA2, and the IDL standard were coamplified competitively in the PCR reactions. Samples were excluded from the study if the amount of wild-type cDNA template in a sample was insufficient to competitively suppress the amplification (absolute CPM signal) of the IDL standard to half of that observed from the IDL standard in the negative control samples containing no wild-type template. This ensures that the SCCA2/SCCA1 ratios observed in the samples are derived from a total number of cDNA template copies exceeding that of the IDL standard. Thus, the detection limit of the assay was set to 1000 copies of the wild-type cDNA template. We found that this detection limit resulted in high accuracy and reproducibility. At this sensitivity level the no preamplification cross contamination was detected. The sensitivity was sufficient for determining the relative SCCA2/SCCA1 mRNA levels in 86% of our 117 samples of microscopic frozen sections and no effort was made to improve the detection limit below this level. By lowering the copy number of the IDL standard the detection limit of the assay can be brought down to 10-100 copies of input cDNA template (results not shown), making it potentially useful in combination with laser-mediated microdissection techniques22, 23. The use of an internal standard in this manner also provides a tool to indirectly control the specificity of the PCR reactions. A decrease in amplification of the internal standard not accompanied by corresponding increase in amplification of the wild-type templates indicates inhibition of amplification by PCR inhibitors or nonspecific amplification. This feature is increasingly important as real time and end point detection methods are moving away from techniques based on visual characterization of the PCR products.

A prerequisite for accuracy is that all possible sources of nonspecific amplification in the PCR reactions are carefully excluded. The PCR primers were designed to eliminate the formation of amplified primer dimers (see materials and methods) and the competitive effect imposed by coamplification of genomic DNA was minimized by DNase digestion of the total RNA extracted from the samples. The wide measuring range of this technique (6 orders of magnitude) was achieved by using a nested PCR approach and amplifying the PCR reactions to the plateau. Thus, the PCR products were brought into the measuring range of the endpoint detection method irrespectively of the amount of input cDNA template. Quantitation of the endpoint PCR products by solid phase minisequencing enables highly accurate separation of the PCR products on the basis of a single nucleotide mismatch in the PCR amplicons. An advantage of the 3H label is a significantly higher signal-to-noise ratio compared to fluorescent labels. This technique, requiring post amplification pipetting of samples, is quite laborious compared to real-time quantitation methods. It is, however, very robust and adaption to high-throughput applications by automatic pipetting is possible21. Although not tested in this study, it is reasonable to assume that detection could have been done by any quantitation technique, endpoint or real time, capable of accurately differentiating between the three PCR products24, 25.

In conclusion, this RT-PCR technique enables accurate quantitative measurement of the relative mRNA levels of two highly homologous genes or splicing variants in biological samples consisting of minute amounts of target cells. By using an internal DNA standard, it is possible to control that a sufficient amount of wild-type cDNA template has been present in the PCR reaction. Thus, the relative levels of the two mRNAs can be reliably determined even in samples containing unknown amounts of target RNA. The earlier observation that the relative proportion of SCCA2 to SCCA1 protein is elevated in malignant conditions17 was reflected as approximately 50% higher relative SCCA2/SCCA1 mRNA levels in malignant as compared to normal squamous epithelium. The accuracy of the assay was sufficient for detecting these subtle but consistent differences, which are well beyond the resolution of other currently available RT-PCR techniques.

REFERENCES

  1. 1

    Wang, A.M., Doyle, M.V. & Mark, D.F. Proc. Natl. Acad. Sci. 86, 9717-9721 (1989).

  2. 2

    Becker-AndrÈ, M. & Hahlbrock, K. Nucl. Acids Res. 17, 9437-9446 (1989).

  3. 3

    Gilliland, G., Perrin, S., Blanchard, K. & Bunn, H.F. Proc. Natl. Acad. Sci. 87, 2725-2729 (1990).

  4. 4

    Ikonen, E., Manninen, T., Peltonen, L. & Syv”nen, A.C. PCR Methods Appl. 1, 234-240 (1992).

  5. 5

    Piatak, M., Jr. et al. Science 259, 1749-1754 (1993).

  6. 6

    Pannetier, C., Delassus, S., Darche, S., Saucier, C. & Kourilsky, P. Nucl. Acids Res. 21, 577-583 (1993).

  7. 7

    Alard, P. et al. BioTechniques 15, 730-737 (1993).

  8. 8

    Holland, P.M., Abramson, R.D., Watson, R., & Gelfand, D.H. Proc. Natl. Acad. Sci. 88, 7276-7280 (1991).

  9. 9

    Gibson, U.E., Heid, C.A. & Williams, P.M. Genome Res. 6, 995-1001 (1996).

  10. 10

    Livak, K.J., Flood, S.J., Marmaro, J., Giusti, W. & Deetz, K. PCR Methods Appl. 4, 357-362 (1995).

  11. 11

    Wittwer, C.T., Herrmann, M.G., Moss, A.A. & Rasmussen, R.P. BioTechniques 22, 130-131, 134-138 (1997).

  12. 12

    Karttunen, L., L–nnqvist, L., Godfrey, M., Peltonen, L. & Syv”nen, A.C. Genome Res. 6, 392-403 (1996).

  13. 13

    Kato, H. & Torigoe, T. Cancer 40, 1621-1628 (1977).

  14. 14

    Schneider, S.S. et al. Proc Natl Acad Sci. 92, 3147-3151(1995).

  15. 15

    Schick, C. et al. Biochemistry. 37, 5258-5266 (1998).

  16. 16

    Schick, C. et al. J. Biol. Chem. 272, 1849-1855 (1997).

  17. 17

    Kato, H., Nagaya, T. & Torigoe, T. GANN 75, 433-435 (1984).

  18. 18

    Rychlik, W. BioTechniques 18, 84-86, 88-90 (1995).

  19. 19

    Syv”nen, A.C., Aalto-Set”l”, K., Harju, L., Kontula, K. & S–derlund, H. Genomics 8, 684-692 (1990).

  20. 20

    Suomalainen, A. & Syv”nen, A.C. Methods Mol. Biol. 86, 121-131 (1998).

  21. 21

    Ihalainen, J., Siitari, H., Laine, S., Syv”nen, A.C. & Palotie, A. Biotechniques 16, 938-943 (1994).

  22. 22

    Kubo, Y., Klimek, F., Kikuchi, Y., Bannasch, P. & Hino, O. Cancer Res. 55, 989-990 (1995).

  23. 23

    Schütze, K. & Lahr, G. Nature Biotechnol. 16, 737-742 (1998).

  24. 24

    Chen, X., Zehnbauer, B., Gnirke, A. & Kwok, P.Y. Proc. Natl. Acad. Sci. 94, 10756-10761 (1997).

  25. 25

    Tyagi, S. & Kramer, F.R. Nature Biotechnol. 14, 303-308 (1996).

Rights and permissions

Reprints and permissions

About this article

Cite this article

Stenman, J., Finne, P., Ståhls, A. et al. Accurate determination of relative messenger RNA levels by RT-PCR. Nat Biotechnol 17, 720–722 (1999). https://doi.org/10.1038/10942

Download citation

  • Issue date:

  • DOI: https://doi.org/10.1038/10942

This article is cited by

Search

Advanced 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