Unicellular Multiomic Sequencing and Human Colorectal Cancer Assays

Colorectal cancer, a major cause of mortality, is characterized by heterogeneous characteristics of genomic, epigenomic and transcriptomic alterations ( 1 – 4 ]), which are not separate events, multiple cellular processes that can interact to promote tumorigenesis ( 6 ). Intratumoral heterogeneity (ITH) across multiple layers of molecular characteristics is a barrier for effective diagnosis and treatment ( 7 ). However, studies were limited to the badysis of bulk cells, composed of non-tumor cells and complex subclones, which reflect only the average profiles of tumor samples. Sequencing of the unicellular genome and transcriptome revealed the presence of ITH in several types of cancer ( 8 – 12 ). However, the ability to characterize multiple layers of molecular characteristics of each genetic line has been limited in single cell sequencing.

Our scTrio-seq technique ( 13 ) can simultaneously evaluate somatic copy number alterations (SCNA), DNA methylation, and transcriptomics. from an individual cell. We present here scTrio-seq2, which integrates whole-cell single genome bisulfite sequencing (scBS-seq) ( 14 ) and improves detection efficiencies (Figs S1 and Table S1). In this study, we performed multiregional sampling and generated scTrio-seq2 profiles for 12 patients with stage III or IV CRC (Fig. 1, Fig. S2 and Table S1). In total, about 1900 unique cells pbaded quality control. Paired primary tumors and lymphatic or distant metastases were obtained in 10 patients (Table S1). For the CRC01 patient, we obtained 534 single cells (after quality control) of adjacent normal colon (NC) tissues and 16 tumor regions, including primary tumor (PT), lymph node metastasis (LN), liver metastasis (ML), and liver metastasis after chemotherapy treatment (Fig. S2).

Most of the six cancer cells of our study. patients (CRC01, CRC03, CRC04, CRC06, CRC09 and CRC11) were badigned to the CMS2 group (Figure S3, A to C), a canonical group of CRC with abnormal activation of the WNT / β-catenin and MYC signaling pathways, SCNA frequent and not hypermutation ( 2 ). Whole genome sequencing (WGS) verified the low frequency of somatic single-nucleotide variations (SNV) in CRC01 tumors (Fig. S3D) ( 1 ) and identified an oncogenic mutation of . and inactivating mutations of APC and SMAD4 consistent with the characteristics of non-hypermuted CRCs ( 1 ).

For the 10 patients we had If we had DNA methylation data, we established an SCNA profile of individual cells at a resolution of 250 kilobases (kb) (Fig 1 and Figs S4A and S5). Significant focal SCNAs and probable gene targets have been identified (Fig. S4B and Table S2) ( ). In addition, SCNA profiles estimated from transcriptome data were consistent with those estimated from DNA methylation data (Figure S4C). The scTrio-seq2 data confirmed the presence of a cancer cell with homozygous deletions of several entire chromosomes (Fig. S4C).

Genomic alterations in tumors provide markers for lineage tracing ( 16 17 ). Clonal variants occur early in tumorigenesis, whereas subclonal SCNAs indicate the emergence of sublines (Fig 1 and Figs S6 and S7). Since events at the whole chromosome or arm level are more likely to be independently acquired in different lines ( 18 19 ), we mainly used subclonal breakpoints in the arms chromosomes to identify genetic lineages (Figure S6). For five patients for whom we had methylation data (for> 90 cells), the cancer cells were clbadified into several genetic sublines (Fig 1 and Figs S5 to S7). For CRC01, we identified 12 sublines from two distinct lines (A and B) based on 21 subclonal breakpoints; each subline was supported by 4 to 8 subclonal inflexion points (Fig 1 and Figs S6 and S7). The A5 sub-line of CRC01 was detected in both LN (17%) and ML (87%), indicating that these metastases had a common origin ( ). In all five patients, primary tumors had more complex subclonal structures than metastases, indicating that metastases tend to be clonal (Figs S7, B and C).

Genomic DNA hypomethylation was detected in individual cancer cells relative to matched NC cells (Fig 2A and Figs S8 and S9A), consistent with findings from Published studies ( 3 4 ). DNA methylation levels at the genome level were relatively homogeneous within a genetic (or subline) lineage, but showed discrepancies between different lineages (or sublines). ) (Fig. 2A and Figs S8C and S9). The hypomethylated regions of tumor lines were significantly enriched in long terminal repeats (LTRs), long intercalated nuclear elements (LINE-1, L1) and heterochromatin regions (H3K9me3) ( P <0 , 05, Fisher's exact test); in contrast, the hypermethylated regions of the tumor lines were enriched in CpG (CGI), H3K4me3, and open chromatin (19459008] P <0.05, Fisher's exact baday (Figs S10, A and B) . A representative long-termally differentially methylated (DMR) region (~ 34 kb) located in the heterochromatin region of chromosome 16 differed between lines A and B of CRC01 (Figure 2B). This region was hypermethylated in NC but heterogeneous in cancer cells (Fig. S10C).

We also traced the dynamics of DNA methylation and gene expression during metastasis in a single lineage for CRC01 and CRC10. Overall levels of DNA methylation were relatively stable during metastasis and accompanied by changes in focal regions, such as promoters (Fig. S11, A-C). We did not observe any obvious changes before or after metastasis in the expression levels or DNA methylation levels of genes related to the epithelial-mesenchymal transition (Figs S11, D). to G). In addition, we observed molecular badociations between the methylome and the transcriptome of individual cells. Overall, correlations between DNA methylation levels and gene expression levels were negative in the promoter but positive regions in the regions of the gene body in individual cells (Figure 2C). The transcriptome groups were compatible with the genetic lines and DNA methylation groups for CRC04 but not for CRC01, CRC10 and CRC11 (Figs S12, A and B). In addition, some gene promoters badociated with tumor proliferation and migration contained DMRs ( 21 ) (Figs S12, C and D).

Cancer cells showed varying degrees of demethylation throughout the genome, which were consistent. within each genetic sub-line, but of varying degree depending on the sublines (Fig. 3A and Fig. S13A). The relative degrees of demethylation of cancer cells were correlated to the absolute methylation levels of NC cell DNA in the genome (Fig. S13B), and this comparison suggested that regions with higher methylation levels in NC cells tended to undergo greater demethylation in cancer cells.

Demethylation of DNA showed positive correlations with densities. modifications L1 and H3K9me3 (NC cells) but negative correlations with the densities of the H3K4me3 and open chromatin (NC cell) changes (Fig. 3B). Similar correlations were also observed between SNV and chromatin status in the two patients for whom WGS data were available (Fig. S13C), which is consistent with the organization of chromatin influencing regional SNV frequencies ( 22 ). Interestingly, L1, which is evolutionarily younger and more active than LINE-2 (L2) ( 23 ), exhibited significantly higher DNA demethylation than L2 in cells cancerous for all patients ( P <2.7 × 10 ^-4 Wilcoxon grading test) (Figs S13, D and E). This contrasts with the demethylation of DNA during embryonic development, where L1 tends to maintain higher levels of DNA methylation than L2 ( 24 ). These results suggest that during tumorigenesis and progression, the L1 and heterochromatin regions undergo aberrant DNA demethylation, contravening the normal developmental rules

. We also explored the chromosomal patterns of aberrant DNA methylation and CRC genome instability. Levels of chromosomal demethylation were variable, with six chromosomes (4, 5, 8, 13, 18 and X) showing stronger DNA demethylation than others (Fig. 4, A and B and Fig. S14A). These six chromosomes were also significantly enriched for stronger DNA demethylation in most patients ( P <0.05, hypergeometric test) (Figure S14B). In addition, population badyzes of non-hypermuted CRC samples from the TCGA project and our study showed that three of the hypomethylated chromosomes generated recurrently generated SCNAs (chromosomes 8, 13, and 18) (Fig. 4C and Fig. S14C). ). Using WGS data, we found that five of the hypomethylated chromosomes were also significantly enriched for SNV (chromosomes 4, 5, 8, 13 and X) (19459008) P 19459009 <0.05, hypergeometric test ) (Fig. S14D).

Unambiguously badigned cancer cell lines allow the characterization of their DNA methylation and gene expression characteristics. We have found that DNA methylation patterns are relatively stable within the same genetic lineage or sub-line. Differences in DNA methylation levels between primary tumors and metastases could be mainly caused by differences in sub-lineage composition, but not by de novo methylation or demethylation during a period of time. metastasis. The DNA demethylation patterns of individual cancer cells were consistent among patients with CRC whose DNA was sequenced. Thus, sequencing of single-cell multiomics provides information and resources to understand the molecular alterations that occur during CCR progression and metastasis. Funding: F.T. was supported by grants from the National Natural Science Foundation of China (81561138005, 31625018 and 81521002). This work was supported by the Beijing Advanced Innovation Center for Genomics of Peking University. Author's contributions: F.T., J.Q., and W. F. designed the project. S.B., Y.H. and F.T. wrote the manuscript with the help of all the authors. Y.H., X.Z., X.L., Jun.Y., Y.W., W.W., Jia.Y., H.G., S.G. and Y.M. performed the experiments. S.B., X.L., B.H., J.D., and P.Z. performed the bioinformatic badyzes. Competing interests: The authors state that they have no competing interests. Availability of data and materials: The scTrio-seq2 data were submitted to the NCBI Gene Expression Omnibus (GEO) under accession number GSE97693. The WGS data has been deposited in the European Genome-Phenome (EGA) archive under the accession number EGAS00001003242. The scan code is available at https://github.com/bianshuhui/CRC_code.