Nanoscale imaging of clinical specimens using pathology-optimized expansion microscopy

Abstract

Expansion microscopy (ExM), a method for improving the resolution of light microscopy by physically expanding a specimen, has not been applied to clinical tissue samples. Here we report a clinically optimized form of ExM that supports nanoscale imaging of human tissue specimens that have been fixed with formalin, embedded in paraffin, stained with hematoxylin and eosin, and/or fresh frozen. The method, which we call expansion pathology (ExPath), converts clinical samples into an ExM-compatible state, then applies an ExM protocol with protein anchoring and mechanical homogenization steps optimized for clinical samples. ExPath enables 70-nm-resolution imaging of diverse biomolecules in intact tissues using conventional diffraction-limited microscopes and standard antibody and fluorescent DNA in situ hybridization reagents. We use ExPath for optical diagnosis of kidney minimal-change disease, a process that previously required electron microscopy, and we demonstrate high-fidelity computational discrimination between early breast neoplastic lesions for which pathologists often disagree in classification. ExPath may enable the routine use of nanoscale imaging in pathology and clinical research.

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: Design and validation of expansion pathology (ExPath) chemical processing.
Figure 2: ExPath reduction of tissue autofluorescence.
Figure 3: ExPath imaging of a wide range of human tissue types.
Figure 4: ExPath analysis of kidney podocyte foot process effacement.
Figure 5: ExPath improvement of computational diagnosis of early breast lesions.

References

  1. 1

    Huang, B., Bates, M. & Zhuang, X. Super-resolution fluorescence microscopy. Annu. Rev. Biochem. 78, 993–1016 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  2. 2

    Huang, B., Babcock, H. & Zhuang, X. Breaking the diffraction barrier: super-resolution imaging of cells. Cell 143, 1047–1058 (2010).

    CAS  Article  Google Scholar 

  3. 3

    Hell, S.W. Far-field optical nanoscopy. In 2010 23rd Annual Meeting of the IEEE Photonics Society (ed. Novak, D.) 3–4 (IEEE, 2010).

  4. 4

    Hell, S.W. Toward fluorescence nanoscopy. Nat. Biotechnol. 21, 1347–1355 (2003).

    Article  CAS  Google Scholar 

  5. 5

    Phelps, P.E., Houser, C.R. & Vaughn, J.E. Immunocytochemical localization of choline acetyltransferase within the rat neostriatum: a correlated light and electron microscopic study of cholinergic neurons and synapses. J. Comp. Neurol. 238, 286–307 (1985).

    Article  CAS  Google Scholar 

  6. 6

    Nixon, R.A. et al. Extensive involvement of autophagy in Alzheimer disease: an immuno-electron microscopy study. J. Neuropathol. Exp. Neurol. 64, 113–122 (2005).

    Article  Google Scholar 

  7. 7

    Gaietta, G. et al. Multicolor and electron microscopic imaging of connexin trafficking. Science 296, 503–507 (2002).

    Article  CAS  Google Scholar 

  8. 8

    Chen, F., Tillberg, P.W. & Boyden, E.S. Expansion microscopy. Science 347, 543–548 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. 9

    Tillberg, P.W. et al. Protein-retention expansion microscopy of cells and tissues labeled using standard fluorescent proteins and antibodies. Nat. Biotechnol. 34, 987–992 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. 10

    Waldman, M. et al. Adult minimal-change disease: clinical characteristics, treatment, and outcomes. Clin. J. Am. Soc. Nephrol. 2, 445–453 (2007).

    Article  CAS  Google Scholar 

  11. 11

    Elmore, J.G. et al. Diagnostic concordance among pathologists interpreting breast biopsy specimens. J. Am. Med. Assoc. 313, 1122–1132 (2015).

    Article  CAS  Google Scholar 

  12. 12

    Chen, F. et al. Nanoscale imaging of RNA with expansion microscopy. Nat. Methods 13, 679–684 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. 13

    Leproust, E.M., Chen, S. & Ruvolo, M. Synthesis of long fish probes. US patent 20140256575 A1 (2014).

  14. 14

    Schnell, S.A., Staines, W.A. & Wessendorf, M.W. Reduction of lipofuscin-like autofluorescence in fluorescently labeled tissue. J. Histochem. Cytochem. 47, 719–730 (1999).

    Article  CAS  Google Scholar 

  15. 15

    Viegas, M.S., Martins, T.C., Seco, F. & do Carmo, A. An improved and cost-effective methodology for the reduction of autofluorescence in direct immunofluorescence studies on formalin-fixed paraffin-embedded tissues. Eur. J. Histochem. 51, 59–66 (2007).

    CAS  PubMed  Google Scholar 

  16. 16

    Neumann, M. & Gabel, D. Simple method for reduction of autofluorescence in fluorescence microscopy. J. Histochem. Cytochem. 50, 437–439 (2002).

    Article  CAS  Google Scholar 

  17. 17

    Mendez, M.G., Kojima, S. & Goldman, R.D. Vimentin induces changes in cell shape, motility, and adhesion during the epithelial to mesenchymal transition. FASEB J. 24, 1838–1851 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. 18

    Maier, J., Traenkle, B. & Rothbauer, U. Real-time analysis of epithelial–mesenchymal transition using fluorescent single-domain antibodies. Sci. Rep. 5, 13402 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. 19

    Pease, D.C. Fine structures of the kidney seen by electron microscopy. J. Histochem.Cytochem. 3, 295–308 (1955).

    Article  CAS  Google Scholar 

  20. 20

    Ranganathan, S. Pathology of podocytopathies causing nephrotic syndrome in children. Front Pediatr. 4, 32 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

  21. 21

    Hirose, T. et al. An essential role of the universal polarity protein, aPKClambda, on the maintenance of podocyte slit diaphragms. PLoS One 4, e4194 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. 22

    Dandapani, S.V. et al. Alpha-actinin-4 is required for normal podocyte adhesion. J. Biol. Chem. 282, 467–477 (2007).

    Article  CAS  Google Scholar 

  23. 23

    Mundel, P. et al. Synaptopodin: an actin-associated protein in telencephalic dendrites and renal podocytes. J. Cell Biol. 139, 193–204 (1997).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. 24

    Sim, J. & Wright, C.C. The kappa statistic in reliability studies: use, interpretation, and sample size requirements. Phys. Ther. 85, 257–268 (2005).

    PubMed  Google Scholar 

  25. 25

    Degnim, A.C. et al. Gene signature model for breast cancer risk prediction for women with sclerosing adenosis. Breast Cancer Res. Treat. 152, 687–694 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. 26

    Allison, K.H. et al. Understanding diagnostic variability in breast pathology: lessons learned from an expert consensus review panel. Histopathology 65, 240–251 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

  27. 27

    Dong, F. et al. Computational pathology to discriminate benign from malignant intraductal proliferations of the breast. PLoS One 9, e114885 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. 28

    Chozinski, T.J. et al. Expansion microscopy with conventional antibodies and fluorescent proteins. Nat. Methods 13, 485–488 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. 29

    Ku, T. et al. Multiplexed and scalable super-resolution imaging of three-dimensional protein localization in size-adjustable tissues. Nat. Biotechnol. 34, 973–981 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. 30

    Lubeck, E. & Cai, L. Single-cell systems biology by super-resolution imaging and combinatorial labeling. Nat. Methods 9, 743–748 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. 31

    Lubeck, E., Coskun, A.F., Zhiyentayev, T., Ahmad, M. & Cai, L. Single-cell in situ RNA profiling by sequential hybridization. Nat. Methods 11, 360–361 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. 32

    Chen, K.H., Boettiger, A.N., Moffitt, J.R., Wang, S. & Zhuang, X. RNA imaging. Spatially resolved, highly multiplexed RNA profiling in single cells. Science 348, aaa6090 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. 33

    Jungmann, R. et al. Multiplexed 3D cellular super-resolution imaging with DNA-PAINT and Exchange-PAINT. Nat. Methods 11, 313–318 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. 34

    Germroth, P.G., Gourdie, R.G. & Thompson, R.P. Confocal microscopy of thick sections from acrylamide gel embedded embryos. Microsc. Res. Tech. 30, 513–520 (1995).

    Article  CAS  Google Scholar 

  35. 35

    Irshad, H., Veillard, A., Roux, L. & Racoceanu, D. Methods for nuclei detection, segmentation, and classification in digital histopathology: a review-current status and future potential. IEEE Rev. Biomed. Eng. 7, 97–114 (2014).

    Article  Google Scholar 

  36. 36

    Lowe, D.G. Distinctive image features from scale-invariant keypoints. Int. J. Comput. Vis. 60, 91–110 (2004).

    Article  Google Scholar 

  37. 37

    Vedaldi, A. & Fulkerson, B. Vlfeat - an open and portable library of computer vision algorithms. In Proc. 18th ACM International Conference on Multimedia (eds. Bimbo, A.D. et al.) 1469–1472 (ACM Press, 2010).

  38. 38

    Sternberg, S.R. Biomedical image processing. Computer 16, 22–34 (1983).

    Article  Google Scholar 

  39. 39

    Zuiderveld, K. Contrast limited adaptive histogram equalization. In Graphics gems IV (Ed. Heckbert, P.S.), 474–485 (AP Professional, 1994).

  40. 40

    Fan, J. Notes on Poisson distribution-based minimum error thresholding. Pattern Recognit. Lett. 19, 425–431 (1998).

    Article  CAS  Google Scholar 

  41. 41

    Al-Kofahi, Y., Lassoued, W., Lee, W. & Roysam, B. Improved automatic detection and segmentation of cell nuclei in histopathology images. IEEE Trans. Biomed. Eng. 57, 841–852 (2010).

    Article  Google Scholar 

  42. 42

    Haralick, R.M., Shanmugam, K. & Dinstein, I. Textural features for image classification. IEEE Trans. Syst. Man Cybern. 3, 610–621 (1973).

    Article  Google Scholar 

  43. 43

    Galloway, M.M. Texture analysis using gray level run lengths. Comput. Graph. Image Process. 4, 172–179 (1975).

    Article  Google Scholar 

  44. 44

    Friedman, J., Hastie, T. & Tibshirani, R. Regularization paths for generalized linear models via coordinate descent. J. Stat. Softw. 33, 1–22 (2010).

    Article  PubMed  PubMed Central  Google Scholar 

  45. 45

    Tibshirani, R. Regression shrinkage and selection via the lasso. J. R. Stat. Soc. B 58, 267–288 (1996).

    Google Scholar 

  46. 46

    Dormann, C.F. et al. Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography (Cop.) 36, 27–46 (2013).

    Article  Google Scholar 

  47. 47

    Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).

    Article  Google Scholar 

  48. 48

    John, G.H. & Langley, P. Estimating continuous distributions in Bayesian classifiers. In Proc. of the 11th Conference on Uncertainty in Artificial Intelligence (eds. Besnard, P. & Hanks, S.) 338–345 (Morgan Kaufmann Publishers, 1995).

Download references

Acknowledgements

For funding, E.S.B. acknowledges the MIT Media Lab; the Open Philanthropy project; the HHMI-Simons Faculty Scholars Program; the US Army Research Laboratory and the US Army Research Office under contract/grant number W911NF1510548; the MIT Brain and Cognitive Sciences Department; the New York Stem Cell Foundation-Robertson Investigator Award; NIH Transformative Award 1R01GM104948; NIH Director's Pioneer Award 1DP1NS087724; NIH 1R01EY023173; NIH 1U01MH106011; NIH 1R01MH110932; and the MIT McGovern Institute MINT program. A.H.B. acknowledges support from Harvard Catalyst, the Harvard Clinical and Translational Science Center (National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health Award UL1 TR001102), and financial contributions from Harvard University and its affiliated academic healthcare centers. A.H.B. and O.B. also acknowledge support from the Ludwig Center at Harvard and the Ludwig's members J. Brugge and J. Staunton. O.B. acknowledges the Lady Tata Memorial Trust, London. F.C. acknowledges the NSF Fellowship and Poitras Fellowship.

Author information

Affiliations

Authors

Contributions

Y.Z., O.B., A.H.B. and E.S.B. all contributed key ideas, designed experiments and analyzed data. F.C. and Y.Z. performed SR-SIM experiments on tissues. Y.Z. and O.B. designed and acquired ExPath data for all tissues. H.I., O.B. and Y.Z. analyzed ExPath data for breast benign neoplasia experiments. H.I. developed the computational image analysis framework for the breast benign neoplasia analysis. A.L.S. performed the expert annotation (ground truth) for the image analysis framework. E.-Y.O., S.J.S. and B.G. performed the selection and annotation of the breast lesions. A.W., M.D., V.T. and I.E.S. participated in the single blinded test for the ExPath kidney experiment. All authors contributed to the writing of the manuscript. A.H.B. and E.S.B. supervised the project.

Corresponding authors

Correspondence to Andrew H Beck or Edward S Boyden.

Ethics declarations

Competing interests

Y.Z., O.B., H.I., A.H.B. and E.S.B. have filed for patent protection on a subset of the technologies here described (US provisional application no. 62/299,754). E.S.B. helped cofound a company to help disseminate expansion microscopy to the community. A.H.B. helped cofound a company to develop artificial intelligence technology for pathology.

Integrated supplementary information

Supplementary Figure 1 The full workflow of Expansion Pathology.

Supplementary Figure 2 Conditions that affect the successful expansion of human tissues.

(A) Images of human skin samples, both taken by a conventional camera (i, iv) and via fluorescence microscopy (ii-iii, v-vi) when stained with DAPI (gray) and antibodies against vimentin (green) and smooth muscle actin (gene name, ACTA2) (magenta). The samples were digested with 8 units/mL proteinase K solution containing 25 mM Tris (pH 8), 1 mM EDTA, 0.25% Triton X-100, and 0.4 M NaCl at 60°C for 0.5 hour (i-iii) or 2 hours (iv-vi). (i and iv) Conventional camera (i.e., non-fluorescent) photographs of human skin-hydrogel hybrid samples in PBS (~2x expansion), after digestion for 0.5 hour (i) or 2 hours (iv). (ii) Widefield fluorescent image from the sample of (i) pre-digestion. (iii) Post-expansion wide-field fluorescent image of the sample of (ii), with a dashed orange box highlighting regions with autofluorescence in the DAPI channel and distorted vimentin networks post-expansion. (v) As in (ii), for the sample in (iv) pre-digestion. (vi) As in (iii) for the sample of (v) with a dashed orange box highlighting regions with autofluorescence in the DAPI channel. (B) As in (A), except that the samples were digested with 8 units/mL proteinase K solution containing 25 mM Tris (pH 8), 25 mM EDTA, 0.25% Triton X-100, and 0.4 M NaCl, at 60°C for 0.5 hour (i-iii) or 2 hours (iv-vi). Note that in Bi and Biv the samples look quite transparent, due to the heavy digestion. (C) Photographs of human liver samples digested with a 1 mM EDTA-based solution – specifically, 8 units/mL proteinase K solution containing 25 mM Tris (pH 8), 1 mM EDTA, 0.25% Triton X-100, and 0.4 M NaCl at 60°C for 0.5 hour (left) or 2 hours (right). (D) Photographs of human liver samples digested with a 25 mM EDTA-based solution – specifically, 8 units/mL proteinase K solution containing 25 mM Tris (pH 8), 25 mM EDTA, 0.25% Triton X-100, and 0.4 M NaCl at 60°C for 0.5 hours (left) or 2 hours (right). Note that the samples of D are significantly more transparent than those of C. (E) Widefield fluorescent image of a human liver sample stained with DAPI (gray) and an antibody against smooth muscle actin (gene name, ACTA2) (magenta) prior to the ExPath process. (F) Post-expansion wide-field fluorescent image of the sample of E, digested with a 1 mM EDTA-based solution for 1 hour. White dashed line outlines an out-of-focus region caused by distortion. (G) Wide-field fluorescent image of a human liver sample stained with DAPI (gray) and an antibody against smooth muscle actin (gene name, ACTA2) (magenta) prior to the ExPath process. (H) Post-expansion wide-field fluorescent image of the same sample as in G. The sample had been digested with a 25 mM EDTA-based solution for 0.5 hour. (I) Post-expansion confocal image of human lymph node tissue with invaded breast cancer stained with DAPI (blue) and an antibody against vimentin (green), and treated with a 1 mM EDTA-based solution for 3 hours. (J) Post-expansion confocal image of a different sample from the same tissue used in I, treated with a 25 mM EDTA-based solution. (K) Post-expansion confocal image of normal human kidney tissue fixed with acetone, stained with an antibody against collagen IV (magenta), and treated with 0.1 mg/ml Acryloyl-X prior to in situ polymerization. Cracks are indicated by white arrows. (L) Post-expansion confocal image of a different sample from the same tissue used in K, treated with 0.03 mg/ml Acryloyl-X prior to in situ polymerization. Scale bars (yellow scale bars indicate post-expansion images: Aii and iii, 9.2 μm (physical size in iii: 40 μm, expansion factor: 4.33). Av and vi, 9.4 μm (physical size in vi: 40 μm, expansion factor: 4.28). Bii and iii, 9 μm (physical size in iii: 40 μm, expansion factor: 4.41). Bv and vi, 8.9 μm (physical size in vi: 40 μm, expansion factor: 4.51). E and F, 119 μm (physical size in F: 500 μm, expansion factor: 4.22); G and H, 109 μm (physical size in H: 500 μm, expansion factor: 4.58). (I-L) 40 μm, physical size.

Supplementary Figure 3 Distortion evaluation of post-expansion HeLa cells processed by 25 mM EDTA-assisted proteinase K digestion.

(A) Super-resolution structured illumination microscopy (SR-SIM) image of a HeLa cell stained with DAPI and an antibody against α-tubulin. Blue, DAPI; green, α-tubulin. (B) Post-expansion image of the same sample acquired with a spinning disk confocal microscope. (C and D) Root mean square (RMS) length measurement error as a function of measurement length for ExPath vs SIM images (blue solid line, mean of DAPI channel; green solid line, mean of α-tubulin channel; shaded area, standard deviation; n = 3 samples from separate cultures). Scale bars: (A) 4 μm, (B) 4 μm (for yellow scale bar, physical size post-expansion, 19.5 μm; expansion factor: 4.89).

Supplementary Figure 4 Fluorescence of unstained tissue vs. H&E stained tissue.

(A) Brightfield image of unstained formalin-preserved human lymph node tissue with invaded breast cancer. (B-E) Widefield fluorescent images of the same sample as in A in four fluorescent channels (listed below). (F) Bright field image of tissue from the same block as in A, stained with H&E. (G-J) Widefield fluorescent images of the sample of F in four fluorescent channels. Filters for each fluorescent channel (Semrock, Rochester, NY): B and G: excitation 628 ± 20 nm, emission 692 ± 20 nm; C and H: excitation 586 ± 10 nm, emission 628 ± 19 nm; D and I: excitation 482 ± 9 nm, emission 520 ± 14 nm; E and J: excitation 387 ± 6 nm, emission 409 nm and longer. Images from a given fluorescent channel are adjusted to have the same contrast. (K) Brightfield and widefield fluorescent images of tissue from the same block as in A, stained with H&E, after each step of ExPath processing. Images from a given fluorescent channel are adjusted to have the same contrast. Samples may be especially bright after the heat treatment and gelation steps due to the buffer compositions present. Scale bars: A-J, 500 μm; K, 40 μm (for yellow scalebars, post-expansion, physical size 184 μm; expansion factor: 4.58).

Supplementary Figure 5 Heat treatment significantly improves immunostaining of vimentin and actinin-4 on human kidney samples.

Left panel, widefield fluorescent images of acetone fixed human kidney sections with and without heat treatment in citrate buffer. Right panel, zooms into regions corresponding to the regions indicated by the dashed yellow rectangles in the left panels. Scale bars: 1 mm (left panel), 200 μm (right panel). Abbreviations: ACTN4, actinin-4; ColIV, collagen IV.

Supplementary Figure 6 Anti-actinin-4 specifically stains tertiary podocyte foot processes.

(A-D) Post-expansion widefield images of a human acetone fixed kidney section stained with DAPI and antibodies. Blue, vimentin; green, actinin-4; red, synaptopodin; gray, DAPI. (E) Merged image of (A-D). (F and G) Magnified regions showing actinin-4 (F) and synaptopodin (G) zoomed into the white dashed squares of B and C. (H) Profiles of fluorescent intensity taken along the white dashed line cuts of F and G. Green, actinin-4; red, synaptopodin. Scale bars: 1 μm (4.5 μm physical size, expansion factor 4.5).

Supplementary Figure 7 Immunostaining images of FFPE kidney samples.

A. Post-expansion widefield image of a FFPE normal human kidney sample treated with a citrate antigen retrieval method (20 mM sodium citrate, pH 8.0), with highlighted region (boxed line) magnified in panel B. C. Post-expansion widefield image of a FFPE normal human kidney sample treated with a Tris-EDTA antigen retrieval method (10 mM Tris base, 1 mM EDTA solution, 0.05% Tween 20, pH 9.0), with highlighted region (boxed line) magnified in panel D. E. Post-expansion confocal image of a FFPE normal human kidney sample treated with a citrate antigen retrieval method, with highlighted region (boxed line) magnified in F. G. Post-expansion confocal image of a FFPE human kidney sample with minimal change disease treated with a citrate antigen retrieval method, with highlighted region (boxed line) magnified in H. All the samples were stained with DAPI (gray) and antibodies against vimentin (blue), actinin-4 (green) and collagen IV (magenta). Scale bars: A, C, E and G, 40 μm (physical size). B, D, F, and H, 8 μm (physical size).

Supplementary Figure 8 Training and and test images for single-blinded nephrotic kidney disease diagnosis.

Frozen kidney sections fixed with acetone were stained with DAPI and antibodies. Blue, vimentin; green, actinin-4; magenta, collagen IV; gray, DAPI. Scale bars are indicated in images. For ExPath images, only physical size is indicated. Expansion factors are listed in Supplementary Table 4.

Supplementary Figure 9 Examples of computational detection and segmentation of the nuclei in pre-expansion H&E images and post-expansion fluorescent images (ExPath).

Computational detection and segmentation of the nuclei is significantly more accurate in expanded samples as compared to pre-expanded samples: examples of normal breast, usual ductal hyperplasia (UDH), atypical ductal hyperplasia (ADH) and ductal carcinoma in situ (DCIS). For the “expert annotation” and “automated segmentation” columns: green filled nuclei are nuclei segmented by the expert or the automated segmentation algorithm, respectively (red circles indicate nucleus outlines, which are not visible in the ExPath row because the resolution is too high and thus the outline is barely visible). In the “automated vs expert” column: green filled nuclei, true positives; red filled nuclei, false negatives; blue filled nuclei, false positives (note that when the automated segmentation yielded larger outlines than the expert, this is expressed as a blue “halo” around the green).

Supplementary information

Supplementary Text and Figures

Supplementary Figures 1–9 and Supplementary Tables 1–15 (PDF 3190 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Zhao, Y., Bucur, O., Irshad, H. et al. Nanoscale imaging of clinical specimens using pathology-optimized expansion microscopy. Nat Biotechnol 35, 757–764 (2017). https://doi.org/10.1038/nbt.3892

Download citation

Further reading

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