Investigating Breast Cancer Evolution with Single Cell Genomics

Transcription

1 Investigating Breast Cancer Evolution with Single Cell Genomics AACR Drug Sensitivity & Resistance June 21 st, 2014 Nicholas E. Navin MD Anderson Cancer Center Innovation in Breast Cancer Madrid, Spain

2 Breast Cancer Subtypes Breast cancers can be classified into 6 major subtypes by receptor staining (ER, PR and Her2) Subtypes correlate with gene expression signatures (luminal A, luminal B, normal-like, basal-like, her2, claudin-low) Sorlie et al. 2001, PNAS Sorlie et al. 2003, PNAS Luminal A (ER+/PR+/Her2-) Patients have the highest 5-year survival rates and harbor the lowest number of somatic mutations Triple-Negative (ER-/PR-/Her2-) Patients have the lowest 5-year survival rates and show the highest number of somatic mutations and genomic heterogeneity Maxmen, Nature 2012

3 Inter-Patient Heterogeneity in Breast Cancer Extensive genetic heterogeneity exists between breast cancer patients Breast cancer patients typically only share 1-2 mutations and harbor a large number (N=20-70) of different nonsynonymous mutations Molecular Portraits of Breast Cancer. TCGA, Nature 2012

4 INTER- and INTRA- Tumor Heterogeneity in Cancer Patients Inter-patient heterogeneity TCGA ICGC Intra-tumor Heterogeneity

5 Selective Pressures immune system hypoxia Intratumor heterogeneity chemotherapy

6 Intratumor heterogeneity can be used to reconstruct genetic lineages by assuming that mutational complexity increases with time

7 General Models of Tumor Evolution Navin & Hicks (2010) Mol Onc; Russnes et al. (2011) JCI

8 Cytological Methods for Resolving Intratumor Heterogeneity Karyotyping Trent et al, 1985 Mitelman et al Tsarouha et al Teixeira et al Teixeira et al Teixeira et al Immunohistochemistry Allred et al. 2008, Clin Canc Res. Nassar et al. 2010, App Immunohist. Potts et al. 2012, Lab Inv. Jakobsen et al. 2013, EJC FISH Trinh et al. 2014, Genome Biology Navin et al. 2010, Genome Research Almendro et al. 2014, Cell Press Park et al. 2010, JCI Shipitsin et al. 2007, Cancer Cell Farabegoli et al. 2001, Cytometry Roka et al. 1998, Breast Cancer Res. HER2 IHC, Flagship Biosciences Inc. Farabegoli et al. 2001

9 Sequencing Methods for Resolving Intratumor Heterogeneity Spatial Exome Sequencing Gerlinger et al. (2012) NEJM Gerlinger et al. (2014) Nat. Gen. Zhang et al. (2014) Science De Bruin et al. (2014) Science Kumar et al. (2014) Genome Bio. Deep-Sequencing and Clustering Mutation Frequencies Single Cell Sequencing Leung et al. (2015) Genome Bio. Wang et al. (2014) Nature Hou et al. (2012) Cell Zong et al. (2012) Science Baslan et al. (2012) Nature Prot. Navin et al. (2011) Nature Nik-Zainal et al. (2012) Cell Shah et al. (2012) Nature Carter et al. (2012) Nat. Biotech Campbell et al. (2008) PNAS Miller et al. (2014) PLOS Comp Bio Ha et al. (2014) Genome Res Roth et al. (2014) Nature Methods Eirew et al. (2014) Nature

10 Advantages & Limitations of Single Cell Sequencing Advantages: 1. Complex admixtures (intratumor heterogeneity) can be fully resolved with single cell data 2. The combination of mutations in each cell can be determined 3. Phylogenetic lineages can be reconstructed by comparing multiple cells 4. Rare tumor cells (CTCs,DTCs,CSCs) can be sequenced and studied Limitations: 1. Limited number of cells (N) can be sequenced at a reasonable cost 2. Technical errors are more extensive and thus orthogonal validation is required

11 Single Cell Sequencing Publications are Growing Rapidly development Publications on single cell sequencing have grown rapidly over the last 5 years, with most studies focusing on cancer research and developmental biology Navin & Wang (2015) Mol. Cell, in press

12 SINGLE 2013 CELL SEQUENCING Method of Method the of Year the Year (2013) Single Cell Sequencing

13 Mutational Evolution and Clonal Diversity in Breast Cancer Patients

14 Questions 1. What is the range and extent of clonal diversity in human breast cancers? 2. Which mutations occur early & late in breast cancer progression? 3. Do mutations evolve gradually over time, or in punctuated bursts? 4. Do breast tumor cells have increased mutation rates?

15 Single-Nucleus-Sequencing (SNS) SNS was the first single cell genome-wide DNA sequencing method developed SNS combines flowsorting, DOP-PCR and next-generation sequencing to generate copy number profiles from individual tumor nuclei SNS data revealed punctuated copy number evolution during tumor progression in breast cancer Navin et al. 2011, Nature Baslan et al. 2012, Nat Prot. Navin and Hicks, Gen Med.

16 Improving Genome Coverage Breadth of Single Cell Data Coverage Breadth ** SNS used DOP-PCR which generated sufficient coverage (~10%) for detecting large-scale copy number aberrations (54kb) However DOP-PCR provided insufficient coverage for detecting mutations at base-pair resolution SNS MALBAC NUC-SEQ G1/0 To address this problem NUC-SEQ was developed to achieve high coverage breadth (90%) data from a G2/Msingle mammalian cell (Ni et al. 2013) (Wang et al. 2014)

17 Exploiting the Cell Cycle for Single-Cell Sequencing NUC-SEQ Single cells duplicate their genomes during S-phase with low error rates (1e-10) Single cells are isolated by flow-sorting with 4 copies of their genome doubling their DNA content (12 picograms) Using 4 copies of the genome as input material mitigates technical errors that ocurr during single-cell sequencing 2 N 2N 4 N 4N G2/M single cells with 4N DNA content (12 picograms)

18 NUC-SEQ Time-limited MDA Single-Cell Genome/Exome Sequencing 1. Isolate G2/M single cells by flow-sorting DAPI stained nuclei 2. WGA by time-limited multipledisplacement-amplification (MDA) 3. Quality control using chromosome-specific qpcr panels 4. Library construction using TN5 tagmentation MDA TN5 5. Exome capture 6. Next-generation sequencing Wang et al. 2014, Nature Leung et al. 2015, Genome Bio.

19 Clonal Diversity and Mutational Evolution in an ER+ Breast Tumor

20 Single-Cell Sequencing of an ER+ Breast Tumor FACS Invasive ductal carcinoma from a 53-year old patient Receptor Status: ER+ / PR+ / Her2- Deep-sequencing of normal and aneuploid (3N) tumor cell populations Histopathology: Grade (III) Invasive Ductal Carcinoma Whole-genome sequencing of 4 single tumor cells (6N) at high coverage depth

21 Population Genome Sequencing of the Aneuploid Breast Tumor Cells Population sequencing (50X) detected only 12 nonsynonymous point mutations and 3 frameshift indels Several mutations occurred in known cancer genes: PIK3CA CASP3 MAP3K1 FBN2 PPP2R5E Wang et al. 2014, Nature

22 Single Cell Copy Number Profiling Heatmap Hierarchical clustering of single cell copy number profiles shows that tumor cells have highly similar copy number rearrangements (R 2 = 0.89) The majority of amplifications and deletions are present in every tumor cell

23 Whole-Genome Single-Cell Sequencing Identifies Subclonal Mutations Mutations detected in only 1 cell are excluded Wang et al. 2014, Nature

24 Single-Cell Exome Sequencing of 47 Tumor and 12 Normal Cells Mutations in only 1 cell are excluded

25 Validation by Single-Molecule Ultra-Deep Targeted Sequencing Duplex libraries reduces the sequencing error rate by adding two sets of random 12bp tags to the ends of libraries to label individual DNA molecules with unique identifiers The unique identifiers are then expanded during PCR After sequencing, reads with similar identifiers are grouped to calculate consensus sequences Error rate of Illumina Sequencing = 1e-2 Error rate of Duplex Sequencing = 1e-9 Schmitt et al. 2012, PNAS

26 Validation using Targeted Ultra-Deep Duplex Sequencing A custom capture platform was designed to validate a subset of the mutations detected by single cell sequencing Targeted deep-sequencing of the bulk tumor tissue was performed using duplex libraries at 116,952X coverage depth (15,431X molecular depth) Validation Rate = 94.44% Validation Rate = 90.47% Validation Rate = 19.40% Wang et al. 2014, Nature

27 Investigating Mutational Evolution in a Triple- Negative Breast Tumor

28 Triple-Negative Breast Tumor Triple-Negative (ER-/PR-/Her2-) Ductal Carcinoma (Grade III) was resected by lumpectomy from 66- year old woman at the MD Anderson Cancer Center imaging size: 2.4cm x 2.3cm x 1.8cm frozen tumor sample size: 2.2cm x 0.8cm x 0.9cm No chemotherapy or hormonal therapy before the sample was collected Grade: III Invasive Ductal Carcinoma No metastatic lesions detected Wang et al. 2014, Nature

29 Isolation of Single Tumor Nuclei by Flow-Sorting by Ploidy Nuclear suspensions were prepared from the frozen tissue and stained with DAPI Single nuclei were flowsorted for amplification and exome-sequencing Experiments: Bulk exome sequencing of the tumor cell population (63X) Bulk exome sequencing of the normal tissue (43X) Single-Cell Exome Sequencing: 8 cells from matched normal tissue 8 cells from the D population 8 cells from the H population 8 cells from the A population

30 Bulk Sequencing of the Triple-Negative Breast Tumor Exome sequencing was performed on normal tissue (43X) and tumor tissue (63X) 360 nonsynonymous mutations and 50 indels were identified 15 significant mutations were identified in cancer genes, including: PTEN JAK1 NOTCH3 ARAF MAP3K4 Many genome-wide deletions of whole chromosomes were detected Wang et al. 2014, Nature

31 Single-Cell Copy Number Profiling Reveals Stable Subpopulations Single cell copy number profiles are highly stable within each subpopulation: H (mean R 2 =0.87) and A (mean R 2 = 0.91)

32 Single-Cell Exome Sequencing of a TNBC Patient

33 Multi-dimensional Scaling of Single-Cell Exome Mutations Multi-dimensional scaling identified 3 major tumor cell subpopulations in addition to the normal cells The H Subpopulation shows the highest variance from the mean cluster distance

34 Single-Molecule Deep Sequencing of Bulk Tumor Tissue A custom capture platform was designed to validate the mutations detected by single cell exome sequencing DUPLEX Sequencing was performed at 118,743X coverage depth (6,634X single molecule depth) mean freq = Mean Freq = 0.05 Mean Freq = Validation Rate = 99.73% Validation Rate = 64.83% Validation Rate = 26.99%

35 Many Subclonal Mutations are Predicted to Damage Protein Function SIFT and POLYPHEN were used to predict the damaging impact of rare mutations identified by single-cell sequencing on protein function A large number of rare point mutations (37/118) were predicted by both methods to have a damaging effect on protein function, possibly affecting the phenotype

36 Gradual vs. Punctuated Genome Evolution Point Mutations Copy Number Aberrations Sorted mutation frequencies of single cells show that point mutations evolve gradually over extended periods of time In contrast copy number evolution occured in punctuated bursts, followed by stable clonal expansions Wang et al. 2014, Nature

37 Tumor Evolution at Single-Cell Genomic Resolution Wang et al. 2014, Nature

38 Estimating Mutation Rates from Single-Cell Mutation Frequencies Calculating Mutation Rates from Single Cell Mutation Frequencies Binary Branching Tree Process Simulations For a series of mutation rates (M R ) branching tree simulations are performed, by incorporating cell death rates, cell birth rates, tumor cell doubling time and the total number of tumor cells. 10,000 cells were tracked over many generations, and cells were randomly sampled to generate mutation frequency distributions. Empirical Parameters: ER+ Tumor Total Number of Tumor Cells: 12,451,945 (flow-sorted cells) Cell Birth Rate: (24.2% Ki-67 index) Cell Death Rate: (1.1% Caspase-3 index) Tumor cell doubling time: 168 days TNBC Total Number of Tumor Cells: 17,719,218 (flow-sorted cells) Cell Birth Rate: (43.5% Ki-67 index) Cell Death Rate: (10.4% Caspase-3 index) Tumor cell doubling time: 168 days Franziska Michor Rui Zhao

39 Mathematical Modeling of Mutation Rates ER+ TNBC ER+ M R = 0.6 M R = 0.9 M R = 8 Mathematical modeling of single cell branching using empirical parameters was performed at different mutation rates The resulting distributions were compared to the empirical single-cell mutation frequency distributions and smallest square difference was calculated Data suggest that the ER+ tumor has similar mutation rates to normal cells (1e-10), while the TNBC has a 13.3X increase in the mutation rate

40 Summary We developed a whole-genome and exome single cell sequencing method called NUC-SEQ that uses G2/M cells to achieve high coverage data with low technical error rates In both breast tumors, single-cell sequencing identified hundreds of rare and de novo mutations that were not detected by sequencing the bulk tumor en masse Targeted deep-sequencing using molecular barcodes showed that many subclonal mutations occur at low frequencies ( < 1% ) in the tumor mass Copy number rearrangements occurred early, in punctuated bursts of evolution, followed by stable clonal expansions, while point mutations evolved gradually generating extensive clonal diversity The TNBC displayed an increased (13X) mutation rate, while the ER+ tumor had an error rate similar to normal cells

41 Implications for Therapy Copy number amplifications and deletions are good targets for therapy since they are stable throughout the tumor mass and occur in many tumor cells Many point mutations are restricted to minor subpopulations and targeting these mutations may not eliminate all tumor cells A large number of mutations that confer resistance to therapy are likely to be pre-existing in the tumor mass at low frequencies (rather than being spontaneously induced in response to the therapeutic agent) Measuring a genomic diversity index is likely to have prognostic value in predicting which breast cancer patients will respond poorly to therapy and which patients will have poor survival

42 Translational Applications of SCS in the Clinic 1. Non-invasive monitoring - sequencing circulating tumor cells in the blood or bodily fluids to monitor response to therapy 2. Prognostic Diversity Index increased diversity of tumor cell populations is expected to predict poor response to therapy and poor overall survival 3. Early detection - of tumor cells in clinical samples (blood, sputum, urine) 4. Scarce clinical samples genomic analysis of limited clinical samples Navin & Hicks (2011) Genome Medicine Early detection non-invasive monitoring diversity as a prognostic

43 SCS Applications in Cancer Research Single cell DNA and RNA sequencing can delineate complex biological processes in cancer, including transformation, invasion, clonal evolution, metastasis and chemoresistance SCS can provide genomic data on rare tumor cell populations, such as circulating tumor cells (CTCs), disseminated tumor cells (DTCs) and cancer stem cells (CSCs), to understand their role in cancer progression Navin (2014) Genome Biology

44 Acknowledgements Navin Laboratory Yong Wang Emi Sei Marco Leung Anna Unruh Xiuqing Shi Annalyssa Long Charissa Kim Naveen Ramesh Ruli Gao Pei-Ching Tsu Alex Davis Jill Waters Clinical Collaborators Funda Meric-Bernstam Amy Zhang Mary Edgerton Statistical Collaborators Ken Chen, Paul Scheet Franziska Michor (Dana-Farber) Rui Zhao (Dana-Farber) MDA Sequencing Core Erika Thompson, Khadan Kahnov Louis Ramagli, Hongli Tan Funding Agencies Nadia s Gift Foundation Texas STARS Center for Genetics & Genomics