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:

A quantitative and multiplexed approach to uncover the fitness landscape of tumor suppression in vivo

Nature Methods volume 14, pages 737–742 (2017)Cite this article

Subjects

Abstract

Cancer growth is a multistage, stochastic evolutionary process. While cancer genome sequencing has been instrumental in identifying the genomic alterations that occur in human tumors, the consequences of these alterations on tumor growth remain largely unexplored. Conventional genetically engineered mouse models enable the study of tumor growth in vivo, but they are neither readily scalable nor sufficiently quantitative to unravel the magnitude and mode of action of many tumor-suppressor genes. Here, we present a method that integrates tumor barcoding with ultradeep barcode sequencing (Tuba-seq) to interrogate tumor-suppressor function in mouse models of human cancer. Tuba-seq uncovers genotype-dependent distributions of tumor sizes. By combining Tuba-seq with multiplexed CRISPR–Cas9-mediated genome editing, we quantified the effects of 11 tumor-suppressor pathways that are frequently altered in human lung adenocarcinoma. Tuba-seq enables the broad quantification of the function of tumor-suppressor genes with unprecedented resolution, parallelization, and precision.

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

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

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

Figure 1: Tuba-seq combines tumor barcoding with high-throughput sequencing to allow parallel quantification of tumor sizes.
Figure 2: Tuba-seq precisely and reproducibly quantifies tumor sizes.
Figure 3: Massively parallel quantification of tumor sizes enables probability distribution fitting across multiple genotypes.
Figure 4: Rapid quantification of tumor-suppressor phenotypes using Tuba-seq and multiplexed CRISPR–Cas9-mediated gene inactivation.
Figure 5: Tuba-seq uncovers known and novel tumor suppressors with unprecedented resolution.

Similar content being viewed by others

Accession codes

Primary accessions

Gene Expression Omnibus

References

  1. Lawrence, M.S. et al. Discovery and saturation analysis of cancer genes across 21 tumour types. Nature 505, 495–501 (2014).

    Article  CAS  Google Scholar 

  2. Kandoth, C. et al. Mutational landscape and significance across 12 major cancer types. Nature 502, 333–339 (2013).

    Article  CAS  Google Scholar 

  3. Hahn, W.C. & Weinberg, R.A. Modelling the molecular circuitry of cancer. Nat. Rev. Cancer 2, 331–341 (2002).

    Article  CAS  Google Scholar 

  4. Lawrence, M.S. et al. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature 499, 214–218 (2013).

    Article  CAS  Google Scholar 

  5. Chin, L., Hahn, W.C., Getz, G. & Meyerson, M. Making sense of cancer genomic data. Genes Dev. 25, 534–555 (2011).

    Article  CAS  Google Scholar 

  6. Van Dyke, T. & Jacks, T. Cancer modeling in the modern era: progress and challenges. Cell 108, 135–144 (2002).

    CAS  Google Scholar 

  7. Sánchez-Rivera, F.J. et al. Rapid modelling of cooperating genetic events in cancer through somatic genome editing. Nature 516, 428–431 (2014).

    Article  Google Scholar 

  8. Chiou, S.H. et al. Pancreatic cancer modeling using retrograde viral vector delivery and in vivo CRISPR/Cas9-mediated somatic genome editing. Genes Dev. 29, 1576–1585 (2015).

    Article  CAS  Google Scholar 

  9. Xue, W. et al. CRISPR-mediated direct mutation of cancer genes in the mouse liver. Nature 514, 380–384 (2014).

    Article  CAS  Google Scholar 

  10. Annunziato, S. et al. Modeling invasive lobular breast carcinoma by CRISPR/Cas9-mediated somatic genome editing of the mammary gland. Genes Dev. 30, 1470–1480 (2016).

    Article  CAS  Google Scholar 

  11. Bhang, H.E. et al. Studying clonal dynamics in response to cancer therapy using high-complexity barcoding. Nat. Med. 21, 440–448 (2015).

    Article  CAS  Google Scholar 

  12. Levy, S.F. et al. Quantitative evolutionary dynamics using high-resolution lineage tracking. Nature 519, 181–186 (2015).

    Article  CAS  Google Scholar 

  13. Naik, S.H. et al. Diverse and heritable lineage imprinting of early haematopoietic progenitors. Nature 496, 229–232 (2013).

    Article  CAS  Google Scholar 

  14. Nguyen, L.V. et al. Barcoding reveals complex clonal dynamics of de novo transformed human mammary cells. Nature 528, 267–271 (2015).

    Article  CAS  Google Scholar 

  15. Sun, J. et al. Clonal dynamics of native haematopoiesis. Nature 514, 322–327 (2014).

    Article  CAS  Google Scholar 

  16. Venkataram, S. et al. Development of a comprehensive genotype-to-fitness map of adaptation-driving mutations in yeast. Cell 166, 1585–1596.e22 (2016).

    Article  CAS  Google Scholar 

  17. Grüner, B.M. et al. An in vivo multiplexed small-molecule screening platform. Nat. Methods 13, 883–889 (2016).

    Article  Google Scholar 

  18. DuPage, M., Dooley, A.L. & Jacks, T. Conditional mouse lung cancer models using adenoviral or lentiviral delivery of Cre recombinase. Nat. Protoc. 4, 1064–1072 (2009).

    Article  CAS  Google Scholar 

  19. Jackson, E.L. et al. Analysis of lung tumor initiation and progression using conditional expression of oncogenic K-ras. Genes Dev. 15, 3243–3248 (2001).

    Article  CAS  Google Scholar 

  20. Cancer Genome Atlas Research Network. Comprehensive molecular profiling of lung adenocarcinoma. Nature 511, 543–550 (2014).

  21. Jackson, E.L. et al. The differential effects of mutant p53 alleles on advanced murine lung cancer. Cancer Res. 65, 10280–10288 (2005).

    Article  CAS  Google Scholar 

  22. Ji, H. et al. LKB1 modulates lung cancer differentiation and metastasis. Nature 448, 807–810 (2007).

    Article  CAS  Google Scholar 

  23. Caswell, D.R. et al. Obligate progression precedes lung adenocarcinoma dissemination. Cancer Discov. 4, 781–789 (2014).

    Article  CAS  Google Scholar 

  24. Chuang, C.H. et al. Molecular definition of a metastatic lung cancer state reveals a targetable CD109–Janus kinase–Stat axis. Nat. Med. 23, 291–300 (2017).

    Article  CAS  Google Scholar 

  25. Callahan, B.J. et al. DADA2: high-resolution sample inference from Illumina amplicon data. Nat. Methods 13, 581–583 (2016).

    Article  CAS  Google Scholar 

  26. Norton, L. A Gompertzian model of human breast cancer growth. Cancer Res. 48, 7067–7071 (1988).

    CAS  PubMed  Google Scholar 

  27. Newman, M.E.J. Power laws, Pareto distributions and Zipf's law. Contemp. Phys. 46, 323–351 (2005).

    Article  Google Scholar 

  28. Liu, G. et al. Chromosome stability, in the absence of apoptosis, is critical for suppression of tumorigenesis in Trp53 mutant mice. Nat. Genet. 36, 63–68 (2004).

    Article  CAS  Google Scholar 

  29. Feldser, D.M. et al. Stage-specific sensitivity to p53 restoration during lung cancer progression. Nature 468, 572–575 (2010).

    Article  CAS  Google Scholar 

  30. Dudgeon, C. et al. The evolution of thymic lymphomas in p53 knockout mice. Genes Dev. 28, 2613–2620 (2014).

    Article  Google Scholar 

  31. Imielinski, M. et al. Mapping the hallmarks of lung adenocarcinoma with massively parallel sequencing. Cell 150, 1107–1120 (2012).

    Article  CAS  Google Scholar 

  32. Schuster, K. et al. Nullifying the CDKN2AB locus promotes mutant K-ras lung tumorigenesis. Mol. Cancer Res. 12, 912–923 (2014).

    Article  CAS  Google Scholar 

  33. Ho, V.M., Schaffer, B.E., Karnezis, A.N., Park, K.S. & Sage, J. The retinoblastoma gene Rb and its family member p130 suppress lung adenocarcinoma induced by oncogenic K-Ras. Oncogene 28, 1393–1399 (2009).

    Article  CAS  Google Scholar 

  34. Hernández, J. et al. Tumor suppressor properties of the splicing regulatory factor RBM10. RNA Biol. 13, 466–472 (2016).

    Article  Google Scholar 

  35. Park, I.Y. et al. Dual chromatin and cytoskeletal remodeling by SETD2. Cell 166, 950–962 (2016).

    Article  CAS  Google Scholar 

  36. Yoh, S.M., Lucas, J.S. & Jones, K.A. The Iws1:Spt6:CTD complex controls cotranscriptional mRNA biosynthesis and HYPB/Setd2-mediated histone H3K36 methylation. Genes Dev. 22, 3422–3434 (2008).

    Article  CAS  Google Scholar 

  37. Edmunds, J.W., Mahadevan, L.C. & Clayton, A.L. Dynamic histone H3 methylation during gene induction: HYPB/Setd2 mediates all H3K36 trimethylation. EMBO J. 27, 406–420 (2008).

    Article  CAS  Google Scholar 

  38. Fang, D. et al. The histone H3.3K36M mutation reprograms the epigenome of chondroblastomas. Science 352, 1344–1348 (2016).

    Article  CAS  Google Scholar 

  39. Efeyan, A. et al. Limited role of murine ATM in oncogene-induced senescence and p53-dependent tumor suppression. PLoS One 4, e5475 (2009).

    Article  Google Scholar 

  40. Gan, R.Y. & Li, H.B. Recent progress on liver kinase B1 (LKB1): expression, regulation, downstream signaling and cancer suppressive function. Int. J. Mol. Sci. 15, 16698–16718 (2014).

    Article  Google Scholar 

  41. Wang, T., Wei, J.J., Sabatini, D.M. & Lander, E.S. Genetic screens in human cells using the CRISPR-Cas9 system. Science 343, 80–84 (2014).

    Article  CAS  Google Scholar 

  42. Welford, S.M. & Giaccia, A.J. Hypoxia and senescence: the impact of oxygenation on tumor suppression. Mol. Cancer Res. 9, 538–544 (2011).

    Article  CAS  Google Scholar 

  43. Jones, R.G. & Thompson, C.B. Tumor suppressors and cell metabolism: a recipe for cancer growth. Genes Dev. 23, 537–548 (2009).

    Article  CAS  Google Scholar 

  44. Pickup, M.W., Mouw, J.K. & Weaver, V.M. The extracellular matrix modulates the hallmarks of cancer. EMBO Rep. 15, 1243–1253 (2014).

    Article  CAS  Google Scholar 

  45. Kirsch, D.G. et al. A spatially and temporally restricted mouse model of soft tissue sarcoma. Nat. Med. 13, 992–997 (2007).

    Article  CAS  Google Scholar 

  46. Meuwissen, R. et al. Induction of small cell lung cancer by somatic inactivation of both Trp53 and Rb1 in a conditional mouse model. Cancer Cell 4, 181–189 (2003).

    Article  CAS  Google Scholar 

  47. Rogers, Z. Barcoding lentiviral Cre vectors for use in experiments involving downstream Tuba-seq analysis. Protocol Exchangehttp://dx.doi.org/10.1038/protex.2017.090 (2017).

  48. Rogers, Z. Genomic DNA isolation from tissue samples and library prep for Tuba-seq barcode analysis. Protocol Exchangehttp://dx.doi.org/10.1038/protex.2017.091 (2017).

  49. Madisen, L. et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 13, 133–140 (2010).

    Article  CAS  Google Scholar 

  50. Jonkers, J. et al. Synergistic tumor suppressor activity of BRCA2 and p53 in a conditional mouse model for breast cancer. Nat. Genet. 29, 418–425 (2001).

    Article  CAS  Google Scholar 

  51. Nakada, D., Saunders, T.L. & Morrison, S.J. Lkb1 regulates cell cycle and energy metabolism in haematopoietic stem cells. Nature 468, 653–658 (2010).

    Article  CAS  Google Scholar 

  52. Cerami, E. et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2, 401–404 (2012).

    Article  Google Scholar 

  53. Gao, J. et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci. Signal. 6, pl1 (2013).

    Article  Google Scholar 

  54. Rizvi, N.A. et al. Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 348, 124–128 (2015).

    Article  CAS  Google Scholar 

  55. Ding, L. et al. Somatic mutations affect key pathways in lung adenocarcinoma. Nature 455, 1069–1075 (2008).

    Article  CAS  Google Scholar 

  56. Brinkman, E.K., Chen, T., Amendola, M. & van Steensel, B. Easy quantitative assessment of genome editing by sequence trace decomposition. Nucleic Acids Res. 42, e168 (2014).

    Article  Google Scholar 

  57. Edgar, R.C. & Flyvbjerg, H. Error filtering, pair assembly and error correction for next-generation sequencing reads. Bioinformatics 31, 3476–3482 (2015).

    Article  CAS  Google Scholar 

Download references

Acknowledgements

We thank P. Chu and R. Ma for technical support; A. Orantes for administrative support; C. Murray, C. Kim-Kiselak, J. Lipsick, B. Callahan, J. Sage, and members of the Petrov and Winslow laboratories for helpful comments; J. Xuhuai and the Stanford Functional Genomics Facility (S10OD018220) for advice and technical assistance; and S. Chan for sequencing expertise. I.P.W. and Z.N.R. were supported by the National Science Foundation Graduate Research Fellowship Program (GRFP). Z.N.R. was also supported by a Stanford Graduate Fellowship. C.D.M. was supported by NIH grant no. E25CA180993. D.P. is the Michelle and Kevin Douglas Professor of Biology. This work was supported by NIH grant nos. R01CA175336 and R21CA194910 (to M.M.W.), R01CA207133 (to D.P. and M.M.W.), and in part by the Stanford Cancer Institute support grant (NIH grant no. P30CA124435).

Author information

Author notes

  1. Zoë N Rogers, Christopher D McFarland and Ian P Winters: These authors contributed equally to this work.

Authors and Affiliations

  1. Department of Genetics, Stanford University School of Medicine, Stanford, California, USA

    Zoë N Rogers, Ian P Winters, Santiago Naranjo, Chen-Hua Chuang & Monte M Winslow

  2. Department of Biology, Stanford University, Stanford, California, USA

    Christopher D McFarland & Dmitri Petrov

  3. Department of Pathology, Stanford University School of Medicine, Stanford, California, USA

    Monte M Winslow

  4. Cancer Biology Program, Stanford University School of Medicine, Stanford, California, USA

    Monte M Winslow

  5. Stanford Cancer Institute, Stanford University School of Medicine, Stanford, California, USA

    Monte M Winslow

Authors
  1. Zoë N Rogers
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Christopher D McFarland
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Ian P Winters
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Santiago Naranjo
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Chen-Hua Chuang
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Dmitri Petrov
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Monte M Winslow
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

Z.N.R. tested sgRNA cutting efficiency; generated barcoded vectors; produced lentivirus; and performed mouse analysis, indel analysis, and analysis of single sgRNA tumor sizes. C.D.M. performed data analysis, including processing sequencing data, designing the tumor-calling procedure, and carrying out all statistical analyses. I.P.W. selected tumor suppressors to investigate, designed sgRNAs, generated Lenti-sgRNA/Cre vectors, tested sgRNA cutting efficiency, produced lentivirus, and performed indel analysis. C.-H.C. performed experiments to assess the function of Smad4. D.P. and M.M.W. oversaw the project. C.D.M., Z.N.R., I.P.W., D.P., and M.M.W. wrote the manuscript with comments from all authors.

Corresponding author

Correspondence to Monte M Winslow.

Ethics declarations

Competing interests

Stanford University has filed a patent (U.S. Provisional Application 62/481,067) based on this work in which Z.N.R., I.P.W., C.M., D.P., and M.M.W. are coinventors.

Integrated supplementary information

Supplementary Figure 1 Frequency of genomic alterations in human lung adenocarcinoma and description of tumor initiation and barcoding

a, The percent of tumors with potentially inactivating alterations (frameshift or non-synonymous mutations, or genomic loss) in each tumor suppressor gene is shown for all tumors (All) as well as in tumors with oncogenic KRAS mutations (KRASmut). The number and percent of tumors with oncogenic mutations in KRAS in each dataset is indicated.

b, Inhalation of barcoded lentiviral-Cre vectors initiate lung tumors in genetically engineered mouse models. Importantly, the lentiviral vectors stably integrate into the genomes of the transduced cells. The relative expansion of each uniquely barcoded cell can be determined by high-throughput sequencing-based methods.

c, Hemotoxilin and Eosin (H&E) staining of lung tissue sections from KrasLSL-G12D/+;R26LSL-Tomato (KT) mice transduced with Lenti-Cre virus. These mice develop small expansions of neoplastic cells as well as larger adenomas. Scale bars = 50 μm.

Supplementary Figure 2 Tuba-Seq pipeline to quantify tumor sizes in vivo.

a, Illumina® sequencing of the DNA barcode region of the integrated lentiviral vectors enables precise measurement of lesion sizes (see Methods). First, reads with poor Phred quality scores or unexpected sequences were discarded. Next, reads were piled-up into groups with unique barcodes. Recurrent Illumina® sequencing errors were delineated from small lesions using DADA2, a model of Illumina® sequencing errors initially designed to identify full read-length deep-sequencing amplicons. Small barcode pileups deemed to be recurrent sequencing errors from the amplified barcode region of large tumors were combined with these larger pileups by this clustering algorithm. Read pileups were translated into absolute cell number using the benchmark controls. Lastly, a minimum cutoff to call lesions was established using both the sequencing information and absolute cell number to maximize reproducibility of the pipeline. b,c, A unique read pileup may not correspond to a unique lesion but rather arise from recurrent sequencing errors of the barcode from a very large tumor. DADA2 was used to merge small read pileups with larger lesions of sufficient size and sequence similarity. The algorithm calculates the sequencing error rates from the non-degenerate regions of our deep sequenced region (i.e. the region of the lentiviral vectors that flank the barcode) (b). The likelihood of every transition and transversion (A to C shown) was calculated for every Illumina® Phred score to generate an error model specific for each run (c). The advertised Phred error rates (red) are generally lower than observed (black; LOESS regression used for regularization). These error models (trained to each Illumina® machine) were then used to determine if smaller read pileups should be bundled into larger pileups with strong sequence similarity (suggesting that the smaller pileup is a recurrent read error) or left as a separate lesion. d-f, We sequenced our first experimental samples (KT, KLT, and KPT from Figure 1) on three different Illumina® machines to vet and parameterize DADA2. A sound lesion calling protocol was expected to show (d) strong similarity in the number of called lesions, (e) good correlation between lesion sizes, and (f) similar mean sizes of each pool across the 3 runs. The three runs naturally varied in sequencing depth (40.1 x 106, 22.2 x 106, and 34.9 x 106 reads after pre-processing) and naturally varied in their expected error rate per base (0.85%, 0.95%, and 0.25%)—offering useful technical perturbations to vet concordance of the method. We found that truncating lesion sizes at 500 cells and truncating the DADA2 clustering probability (omega) at 10-10 (red square) offered a profile of

lesion sizes at very small scales, while still minimizing variability in our test metrics.

Supplementary Figure 3 Benchmark controls allow calculation of the number of neoplastic cells in each tumor within each lung sample.

a, Schematic of the protocol using three benchmark control cell lines with known barcodes. 5x105 cells of each cell line were added to each lung sample. DNA was then extracted from the lung plus all three benchmark controls, and the barcodes were PCR amplified and deep sequenced. We then calculated the number of neoplastic cells in each tumor within that lung sample by dividing the average % reads associated with the benchmarks by the % reads observed from each tumor (unique barcode) and multiplying by 5x105 to obtain neoplastic cell number.

b, Example of two lungs with very different tumor burdens. These benchmark cell lines can be used to determine the number of neoplastic cells within individual tumors regardless of overall tumor burden. It should also be noted, that the surrounding “normal lung” tissue has no impact on this calculation as this tissue has no lentiviral integration and thus will contribute no reads.

Supplementary Figure 4 The DADA2-based tumor calling pipeline is robust and reproducible.

a, Calculated tumor sizes exhibited a subtle GC-bias. Barcodes with intermediate GC-content appear to be PCR-amplified most efficiently. A 4th-order polynomial fit to the residual bias corrected lesion sizes most effectively (Methods). This correction was applied to all subsequent analyses, which adjusted each lesion size by an average of 5%, and reduced the standard deviation of lesions sizes of each sgID in each mouse by only 2.9% relative to the mean—suggesting that, while measurable, variability introduced by GC-bias was minimal.

b, The random barcode exhibited a high-degree of randomness across the intended nucleotides.

c, Number of lesions called per mouse using Tuba-seq. Numbers of tumors above two different cell number cutoffs (1000 and 500) are shown as the average number of tumors per mouse ± the standard deviation. KT mice transduced with a high titer (6.8x105) (used in the main text) and a lower titer (1.7x105; KTlow) of Lenti-mBC/Cre virus. There was no statistically significant difference in the number of tumors observed per capsid at either cell cutoff suggesting that barcode diversity is still not limited above half a million tumors and that small tumors are not caused by tumor crowding.

d, Unsupervised hierarchical clustering of the KT, KTlow, KPT, and KLT mice based on the total least-squares distance between tumor sizes at defined percentiles (linkage determined by Ward’s Incremental algorithm.) Mice of the same genotype, but different viral titers, cluster together, suggesting that size profile differences are determined primarily by tumor genotype, not viral titer.

e, f, Lesion sizes are not dramatically affected by differences in read depth. The barcode region from the tumor-bearing lungs of an individual mouse was sequenced at very high depth and then randomly down-sampled to typical read depth. (e) The tumor size distributions of the full (x-axis) and downsampled (y-axis) data sets were very similar, indicating that our analysis parameters are unbiased by, and fairly robust to, read depth. (f) The percentile calculations are also reproducible upon downsampling.

g, KT, KLT, and KPT mice with Lenti-mBC/Cre initiated tumors (from Figure 1) have tumors initiated with six unique Lenti-sgID-BC/Cre viruses (each harboring a unique sgID and naturally varying barcode diversity). This allowed us to quantify the variation in DADA2-called tumor sizes with six replicates within each mouse. Tumor size distributions are reproducibly called when using all tumors from each mouse and when using each subset of tumors with a given sgID. The size of the tumors at the indicated percentiles are plotted for KT (left), KLT (middle), and KPT (right) mice. Each dot represents the value of a percentile calculated using tumors within a single sgID. Percentiles are represented in grey-scale. The six replicate percentile values of tumor size with differing sgIDs are difficult to distinguish since their strong correlation means that markers for each sgID are highly overlapping.

Supplementary Figure 5 Efficient genome editing in lung tumors initiated with Lentiviral- sgRNA/Cre vectors in mice with an H11LSL-Cas9 allele.

a, Schematic of the experiment to test somatic genome editing in the lung cancer model using a Lenti-sgTomato/Cre (Lenti-sgTom/Cre) viral vector and the H11LSL-Cas9 allele. All mice were homozygous for the R26LSL-Tomato allele to determine the frequency of homozygous inactivation.

b, Fluorescence dissecting scope images of a lung lobe from a KPT;Cas9 mouse with Lenti-sgTom/Cre-initiated tumors. Tomato-negative tumors are outlined with dashed lines. Top scale bars = 5 mm; bottom scale bars = 1 mm.

c, Immunohistochemistry for Tomato protein uncovered Tomato-positive (Pos), Tomato-mixed (Mixed), and Tomato-negative (Neg) tumors. Tumors are outlined with dashed lines. Scale bars = 200 μm.

d, Quantification of Tomato expression in four KPT;Cas9 mice with Lenti-sgTom/Cre-initiated tumors indicates that approximately half of the tumors have CRISPR/Cas9-mediated homozygous inactivation of the targeted gene in at least a fraction of the cancer cells. Percent of Tomato positive, mixed, and negative tumors is shown with the number of tumors in each group indicated in brackets.

e, Schematic of the experiment to test somatic genome editing in the lung using Lenti-sgLkb1/Cre virus and the H11LSL-Cas9 allele.

f, Fluorescence dissecting scope images of lung lobes of KT and KT;Cas9 mice with Lenti-sgLkb1/Cre initiated tumors show increased tumor burden in the KT;Cas9 mouse. Lung lobes are outlined with white dashed lines. Scale bars = 2 mm.

g, Tumor burden, represented by lung weight, is increased in KT;Cas9 mice with Lenti-sgLkb1/Cre-initiated tumors relative to KT mice, consistent with successful deletion of the tumor suppressor Lkb1. Normal lung weight is indicated by the dotted red line. * p-value < 0.02. Each dot represents a mouse and the bar is the mean.

h, Western blot showing that Lenti-sgLkb1/Cre initiated tumors in KT;Cas9 mice express Cas9 and lack Lkb1 protein. Hsp90 shows loading.

Supplementary Figure 6 In vitro sgRNA cutting efficiency.

a, Schematic of the experiment to assess the in vitro cutting efficiency of each sgRNA by transducing Cas9-expressing cells with lentiviral vectors carrying each individual sgRNA. We tested three individual sgRNAs for each targeted loci and we report the cutting efficiency of the best sgRNA.

b, Cutting efficiency of the best sgRNA for each targeted tumor suppressor. Cutting efficiency was assessed by Sanger sequencing and TIDE analysis software (Brinkman et al., Nucl. Acids Res., 2014).

c, Schematic of the experiment to assess the in vitro cutting efficiency of each sgRNA by transducing Cas9 cells with Lenti-sgTS-Pool/Cre. Cells were harvested 48 hours after transduction, genomic DNA was extracted, the 14 targeted regions were PCR amplified, and the products were Illumina sequenced. By calculating the % of indels at each region, and normalizing to both the representation in the pool and Setd2 indel %, a relative cutting efficiency was determined for each sgRNA within the pool.

d, Relative cutting efficiency of each sgRNA including the inert Neo-targeting controls.

Supplementary Figure 7 Selection and characterization of sgRNAs targeting eleven known and candidate tumor suppressor genes.

a, sgRNAs were selected based on their location within each gene, their proximity to splice acceptor/splice donor (SA/SD) regions, whether they were upstream of (or within) annotated functional domains, whether they were upstream of (or adjacent to) documented human mutations, as well as their predicted ontarget cutting efficiency score (the maximum score is 1.0; higher score = greater activity) and off target cutting score (the maximum score is 100.0; higher score = greater specificity) (Doench et al., Nature Biotechnology, 2014; Hsu et al., 2013).

b, Summary of data from published studies in which these tumor suppressor genes were inactivated in the context of KrasG12D-driven lung cancer models.

c, Each vector has a unique sgID and was diversified with random barcodes. The sgID for each of the vectors and the estimated number of barcodes associated with each sgRNA is indicated.

d, Schematic of the experiment to assess the initial representation of each sgRNA within Lenti-sgTS-Pool/Cre.

e, The percent of each sgRNA within Lenti-sgTS-Pool/Cre, as determined by sequencing of samples from three replicate transductions. Mean +/- SD is shown. The percent of each vector in the pool deviated only slightly from the expected representation of each vector (red dashed line).

Supplementary Figure 8 Identification and validation of tumor suppressors at multiple time points using Tuba-seq.

a, Histology confirms that KT mice have hyperplasias and small tumors, while KT;Cas9 mice have much larger tumors. Viral titer is indicated. Top scale bars = 3 mm. Bottom scales bars = 500 μm.

b, Analysis of the relative tumor sizes in KT mice (which lack Cas9) 12 weeks after tumor initiation with Lenti-sgTS-Pool/Cre identified essentially uniform tumor size distributions. Relative tumor size at the indicated percentiles represents merged data from 10 mice, normalized to the average of sgInert tumors. 95% confidence intervals are shown. Percentiles that are significantly different from sgInert are in color.

c, Estimates of mean tumor size, assuming a lognormal tumor size distribution, showed expected minor variability in KT mice. Bonferroni-corrected, bootstrapped p-values are shown. p-values < 0.05 and their corresponding means are bold.

d, Tumor sizes at the indicated percentiles for each sgRNA relative to the average of sgInert-containing tumors at the same percentiles. Merged data from 3 KT;Cas9 mice 15 weeks after tumor initiation with Lenti-sgTS-Pool/Cre is shown. Dotted line represents no change from Inert. Relative tumor size at the indicated percentiles represents merged data from 10 mice, normalized to the average of sgInert tumors. 95% confidence intervals are shown. Percentiles that are significantly different from sgInert are in color.

e, Estimates of mean tumor size, assuming lognormality, identified sgRNAs with significant growth advantage in KT;Cas9 mice. Bonferroni-corrected, bootstrapped p-values are shown. p-values < 0.05 and their corresponding mean estimates are in bold.

f, Percent representation of each Lenti-sgRNA/Cre vector in KT mice 12 weeks after tumor initiation (calculated as 100 times the number of reads with each sgID/all sgID reads). As there is no Cas9-mediated gene inactivation in KT mice, the percent of each sgID in these mice represents the percent of viral vectors with each sgRNA in the Lenti-sgTS-Pool/Cre pool. Mean +/- standard deviation of sgID representation is shown.

g, Percent representation of each Lenti-sgRNA/Cre vector in KT;Cas9 mice 12 weeks after tumor initiation (calculated as 100 times the number of reads with each sgID/all sgID reads). Mean +/- standard deviation of sgID representation is shown.

h, Fold change in overall sgID representation in KT;Cas9 mice relative to KT mice (ΔsgID Representation) identified several sgRNAs that increase in representation, consistent with increased growth of tumors with inactivation of the targeted tumor suppressor genes. ΔsgID Representation is the fold change in percent of reads with each sgID in KT;Cas9 mice versus KT mice, normalized such that ΔsgID Representation for sgInert = 1. Means and 95% confidence intervals are shown.

Supplementary Figure 9 Analysis of tumor size distributions demonstrates that Lkb1- and Setd2- deficiencies are lognormally distributed.

a,b, Size of tumors at the indicated percentile (%ile) with sgLkb1 (a) or sgSetd2 (b) versus sgInert tumor size at the same percentile. Each percentile was calculated using all tumors with each sgRNA from all KT;Cas9 mice with Lenti-sgTS-Pool/Cre initiated tumors analyzed 12 weeks after tumor initiation (N=8 mice). The size relative to sgInert-initiated tumors is indicated with dashed lines.

c, Probability density plot for tumors initiated with Lenti-sgSetd2/Cre in KT;Cas9 mice with Lenti-sgTS-Pool/Cre initiated tumors shows lognormally distributed tumor sizes very similar to those seen in KLT mice (Figure 3d).

Supplementary Figure 10 Independent methods validate Setd2 as a potent suppressor of lung tumor growth.

a, Fluorescence dissecting scope images and H&E of lung lobes from KT;Cas9 mice infected with Lenti-sgSetd2#1/Cre, Lenti-sgSetd2#2/Cre, or Lenti-sgNeo2/Cre analyzed 9 weeks after tumor initiation. Lung lobes are outlined with white dashed lines in the fluorescence dissecting scope images. Upper scale bars = 5 mm. Lower scale bars = 2 mm.

b, Quantification of percent tumor area by histology shows a significant increase in tumor burden in KT;Cas9 mice infected with Lenti-sgSetd2#1/Cre or Lenti-sgSetd2#2/Cre compared to KT mice infected with the same virus. Each dot represents a mouse and the bars are the mean. * p-value < 0.05. NS = not significant.

c, Representative fluorescence dissecting scope images of lung lobes from KT;Cas9 mice with tumors initiated with Lenti-sgNeo2/Cre (left), Lenti-sgSetd2#1/Cre (middle), or Lenti-sgSetd2#2/Cre (right) analyzed 9 weeks after tumor initiation. Lung lobes are outlined with white dashed lines. Scale bars = 5 mm.

Supplementary Figure 11 Identification of p53-mediated tumor suppression in KT;Cas9 mice with Lenti-sgTS-Pool/Cre initiated tumors at two independent time points.

a,b, Analysis of the relative tumor sizes in KT;Cas9 mice 12 weeks (a) and 15 weeks (b) after tumor initiation with Lenti-sgTS-Pool/Cre identify p53 as a tumor suppressor using power-law statistics at both time points. Relative tumor size at the indicated percentiles is merged data from 8 and 3 mice, respectively, normalized to the average of sgInert tumors. 95% confidence intervals are shown. Percentiles that are significantly larger than sgInert are in color. Power-law p-values are indicated. Note that in this experimental setting only the very largest sgp53 initiated tumors are greater in size than the sgInert tumors. This is likely partially explained by the relatively poor cutting efficiency of sgp53 (Supplementary Fig. 7d).

c-f, Percent of each size indel at the p53 locus (from ten nucleotide deletions (-10) to three nucleotide insertions (+3)) were calculated by dividing the number of reads with indels of a given size by the total number of reads with indels. In-frame indels are shown in grey. We assessed the spectrum of indels at the p53 locus generated in vitro, in a Cas9 expressing cell line transduced with Lenti-sgTS-Pool/Cre. (c) There is no preference for out of frame mutations in cells targeted in vitro. We then analyzed three individual KT;Cas9 mice with Lenti-sgTS-Pool/Cre initiated tumors after 15 weeks of disease progression (d-f). In these lung samples, there were fewer in-frame indels (-9, -6, -3 and +3) consistent with selection for out-of-frame loss-of-function alterations in tumors that expand. Collectively, these data are consistent with the tumor suppressive function of p53 and our data from KPT mice.

Supplementary Figure 12 Confirmation of on-target sgRNA effects.

a, The percent of reads containing indels at the targeted locus was normalized to the average percent of reads containing indels in 3 independent Neomycin loci. This value is plotted versus the size of the 95th percentile tumor for each sgRNA for three individual mice. We demonstrate a high frequency of indels in Setd2, Lkb1, and Rb1 consistent with selection for on-target sgRNA cutting. Each dot represents an sgRNA from a single mouse. sgNeo dots are in black and all other dots are colored according to Figure 4b.

b,c, Percent of each indel (from ten nucleotide deletions (-10) to four nucleotide insertions (+4)) were calculated by dividing the number of reads with indels of a given size by the total number of reads with indels within each top tumor suppression gene. (b) Average percentage and standard deviation of three KT;Cas9 mice with Lenti-sgTS-Pool/Cre-initiated tumors are shown for Setd2, Lkb1, Rb1, and the average of the three targeted sites in Neo (Neo1-3). Inframe mutations are shown in grey. Average and standard deviations for Neo1-3 was calculated by averaging all three mice and all three Neo target sites as a single group. In general, there were fewer inframe indels (-9, -6, -3 and +3) in Setd2, Lkb1, and Rb1 than in Neo controls consistent with selection for out-of-frame loss-of-function alterations in these genes in tumors that expand. (c) We also assessed the spectrum of indels generated in vitro, in a Cas9-expressing cell line transduced with Lenti-sgTS-Pool/Cre. We detected no preference for inframe mutations in any of these genomic locations, suggesting that the distribution seen in the KT;Cas9 mice is most likely due to advantageous expansion of tumors with out-of-frame indels.

d, Kaplan-Meier survival curve of KT and KT;Cas9 mice with Lenti-sgSmad4/Cre-induced tumors. CRISPR/Cas9-mediated inactivation of Smad4 in the presence of oncogenic KrasG12D does not reduce survival, suggesting limited, if any, increase in tumor growth from Smad4 inactivation.

e, The majority of tumors in Lenti-sgSmad4/Cre transduced KT;Cas9 mice had lost Smad4 protein expression compared to KT mice transduced with the same virus, consistent with indel creation at the Smad4 locus. Scale bars = 50 μm.

f, Several tumors in KT;Cas9 mice with Lenti-sgTS-Pool/Cre-initiated tumors had a distinct papillary histology, uniformly large nuclei, and were Sox9 positive, consistent with the published phenotype of Apc-deficient, Kras-driven lung tumors (Sanchez-Rivera et al., Nature, 2014). Representative Sox9-negative and Sox9-positive tumors are shown. Scale bars = 100 μm (top) and 25 μm (bottom).

Supplementary information

Supplementary Text and Figures

Supplementary Figures 1–12, Supplementary Tables 1–4 and Supplementary Note 1.

Supplementary Protocol

Barcoding lentiviral Cre vectors for use in experiments involving downstream Tuba-seq analysis.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Rogers, Z., McFarland, C., Winters, I. et al. A quantitative and multiplexed approach to uncover the fitness landscape of tumor suppression in vivo. Nat Methods 14, 737–742 (2017). https://doi.org/10.1038/nmeth.4297

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/nmeth.4297

This article is cited by

Search

Advanced search

Quick links

Nature Briefing: Cancer

Sign up for the Nature Briefing: Cancer newsletter — what matters in cancer research, free to your inbox weekly.

Get what matters in cancer research, free to your inbox weekly. Sign up for Nature Briefing: Cancer