Volume 148, Issue 5, Pages 873-885 (March 2012) Single-Cell Exome Sequencing and Monoclonal Evolution of a JAK2-Negative Myeloproliferative Neoplasm  Yong Hou, Luting Song, Ping Zhu, Bo Zhang, Ye Tao, Xun Xu, Fuqiang Li, Kui Wu, Jie Liang, Di Shao, Hanjie Wu, Xiaofei Ye, Chen Ye, Renhua Wu, Min Jian, Yan Chen, Wei Xie, Ruren Zhang, Lei Chen, Xin Liu, Xiaotian Yao, Hancheng Zheng, Chang Yu, Qibin Li, Zhuolin Gong, Mao Mao, Xu Yang, Lin Yang, Jingxiang Li, Wen Wang, Zuhong Lu, Ning Gu, Goodman Laurie, Lars Bolund, Karsten Kristiansen, Jian Wang, Huanming Yang, Yingrui Li, Xiuqing Zhang, Jun Wang  Cell  Volume 148, Issue 5, Pages 873-885 (March 2012) DOI: 10.1016/j.cell.2012.02.028 Copyright © 2012 Elsevier Inc. Terms and Conditions

Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure 1 Graphic Representation of the Whole Genome of Two Single YH Cells (A) Karyotype of the human reference genome (Hg18). (B) GC content distribution of the reference genome (height of orange rectangles ranges from 0%–70%, bin = 1 Mb). (C) Whole-genome coverage of YH-2 (height of blue rectangles ranges from 0×–40×, bin = 1 Mb). (D) Whole-genome coverage of YH-1 (height of blue rectangles ranges from 0×–40×, bin = 1 Mb). (E) Gene density across the reference genome (Hg18) (gradually changing green represents from 0 to 30 genes per 100 kb). See also Figure S1. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure 2 Characteristics of Artifacts in the Two YH Cells Induced by Single-Cell Sequencing (A) Distribution of ADO and false discovery bases across chromosome 1. (B) Evaluation of ADO on the four different base types. (C) Evaluation of false discovery on different types of base changes. (D) The fraction of genes that failed to amplify in all captured genes. (E) Relative distribution of amplification failure ratio of biological categories of genes. See also Figure S2. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure 3 Schematic of the Somatic Mutation Calling Method in the Single-Cell Population and the Somatic Mutant Allele Frequency Correlation between Single-Cell Data and the Tissue of ET (A) A normal pair of chromosomes is shown with an A/G SNP and two homozygous sites. Following WGA, PCR, and sequencing, a genotype is assembled from the reads. In the vicinity of the SNP, the region surrounding the A allele failed to amplify, and the result is an allele dropout (ADO) identifiable by a homozygous call in a single normal cell and a heterozygous call in the normal bulk tissue. The “A” in red represents a false discovery mutation in the normal single cell. The false discovery ratio can be estimated from these sites in normal single cells as compared to the normal bulk DNA. (B) The same region of the tumor pair of chromosomes is shown with a putative somatic mutation detected in the genome of bulk tumor DNA as well as the genome of single tumor cell. (C) Shown are ten hypothetical single tumor cells with putative somatic mutations. A mutation found in only a single cell is more likely to be a false discovery and is removed from consideration. Assuming a false discovery rate (π) = 10−5, the probability of finding the same mutation in a third or fourth single cell is extremely low and also significantly lower than mutation, as depicted in the graph. (D) Correlation between single-cell sequencing and tissue sequencing. The somatic mutant allele frequency in ET single-cell sequencing is indicated as an unfolded (that is, each genotype is taken as two alleles for a diploid-genome cell) site frequency of mutant alleles. The somatic mutant allele frequency in ET tissue sequencing is indicated as a read frequency of mutant alleles. R2 here refers to the square of the correlation coefficient. See also Figure S3 and Tables S1, S2, and S3. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure 4 Genetic Characteristics of the ET Patient (A) Principle component analysis (PCA) of the mutations in the ET cells (red) and the matched normal tissue (green). (B) Somatic mutant allele frequency spectrum (SMAFS) of ET. The SMAFS of ET was calculated for each mutation site of the ET by taking the oral mucosal epithelium mix as a control. The mutations are indicated as synonymous (green), nonsynonymous (purple), and noncoding (blue) mutations. (C) Monoclonal and polyclonal evolution simulation of ET. The SMAFS of simulated one (orange), two (yellow), three (purple), four (green), and five (blue) clonal evolutions was calculated and compared with the SMAFS of ET (red) in the coding regions. See also Figure S4 and Table S4. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure 5 Key Gene Identification of the ET Patient The driver gene prediction analysis of the 18 ET candidate genes is indicated as Q score. The vertical axis is the Q score, and the circle size (diameter) indicates the cell mutation frequency. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure S1 Single-Cell Isolation and Genome Representation by Whole-Genome Single-Cell Sequencing of Two Cells from Lymphoblastoid Cell Line, Related to Figure 1 (A) Micromanipulation-based isolation of the single cell from the YH lymphoblastoid cell line. Before isolation, the sample should be sufficient-dispersion and cascade-dilution. (B and C) Comparison between electrophoresis results of MDA (B) and DOP PCR (C) on single cells. NC represents the negative control (cell stored buffer); PC represents the amplicon from YH genomic DNA; YH gDNA represents the genomic DNA isolated from millions of cells from YH lymphoblastoid cell line. (D) Genome recovery by SCS. Whole genome and exome (here is the coding sequence obtained from the genome sequencing data, and other panel figures in Figure S1 is the same) recovery between SCS and a million-cell sample sequencing with YH blood as control. (E) Cumulative distribution of sequencing fold coverage of SCS of the coding sequence. We calculated the cumulative distribution of sequencing fold coverage (from 0 to 36 ×) of the coding sequence in YH-1 (in blue) and YH-2 (in red) and pooled YH-1 and YH-2 together to generate the YH-1+YH-2 (in green). The standard Poisson Cumulative Distribution (from 0 to 36; λ = 18) was present as a control (in gray). (F) Distribution of sequence representation as a function of GC content. Note that the horizontal axis of GC content was determined using 1k sliding windows, with the corresponding median of depths as the vertical axis. Tissue sequencing from YH blood was used as a control. (G) Distribution of GC content. The GC content of genome was determined using 1kb sliding windows along whole genome, and the coding region was determined by each exon, box plot of whole genome, coding regions and genome, coding regions of 0 coverage in both single cells were presented according to their GC content and labeled as “Genome,” “Coding region,” “0 Genome” and “0 Coding region” respectively. (H) Amplification bias with repeat. Note that the horizontal axis was repeat content determined by sliding 1k windows, with the corresponding median of depths as vertical axis. Tissue sequencing from YH blood was taken as a control. (I) Amplification bias with chromosome position. Note that the horizontal axis was chromosome position determined from centromere by sliding 1k windows, with the corresponding median of depths as vertical axis. Tissue sequencing sample was taken as a control. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure S2 Evaluation of Artifacts and Amplification Failure Genes in Two YH Cells, Related to Figure 2 (A) False-negative ratio (FNR) per sequencing depth in two YH cells. We calculated the FNR per sequencing depth by taking tissue sequencing as control. (B) False heterozygous allele's rate across X chromosome between single cells and the tissue sequencing data. As the YH was male, any observed heterozygous genotype in X chromosome was false. We calculated the false heterozygous allele's rate with the 1M sliding widows, by looking the mapping reads across X chromosome. (C) The biological processes of 0 coverage (amplification-failure) genes were enriched by an online toolkit Webgestalt (http://bioinfo.vanderbilt.edu/webgestalt/). The left is 0 coverage genes and the right is all captured genes in human genome. The 0 coverage genes were defined as 0 coverage at least in 1 exon. (D) The correlation of the each proportion of biological process category between amplification-failure genes and all captured genes. The correlation coefficient (R2) was 0.9924. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure S3 Clinical Information of Essential Thrombocythemia and Bioinformatics Pipeline of Single ET Cell Sequencing, Related to Figure 3 (A) Histopathology of ET. A Wright's stained bone marrow aspirate smear of the ET patient was showed. (B) Overall bioinformatics pipeline of single ET cell sequencing. This is a whole bioinformatics analysis pipeline designed for our SCS methodology; dotted rectangles represent analyses which were not performed in this ET paper. The prevalence screen was also not presented in the ET manuscript because of lack of large cohort of ET patients. (C) Whole analysis processes of identifying key genes in ET. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions

Figure S4 General Characteristics of Mutation in This ET Patient and Sensitivity of Whole-Exome SCS Compared to Whole-Exome Tissue Sequencing, Related to Figure 4 (A) Mutation spectra of ET by SCS and tissue sequencing. The numbers of each of the six classes of base substitution are shown. (B) Cells need for all SMs captured in ET. The number of cell need for this ET patient was evaluated by SM number of randomly selected cells among all SMs captured in ET. (C) Possibility of loss same sites in 58 qualified ET cells and tissue sequencing. The value per cell number for single cell was calculated by the same sites number in all corresponding cells. (D) Possibility of same sequencing error sites in 58 qualified ET cells and tissue sequencing. (E) Possibility of same allele dropout sites in 58 qualified ET cells. (F) Possibility of amplification and sequencing errors in 58 qualified ET cells. (G) LOH analysis of paired tissue sequencing data. The LOH ratio was calculated by NL/NH, where NL is the number of sites that is homozygous sites in cancer tissue and heterozygous in normal tissue, and NH is the number of sites that are heterozygous in normal tissue. Cell 2012 148, 873-885DOI: (10.1016/j.cell.2012.02.028) Copyright © 2012 Elsevier Inc. Terms and Conditions