Search

The field of cancer genomics has been empowered by increasingly sophisticated inference tools to distinguish driver mutations from the vastly greater number of passenger mutations. Epigenetic alterations such as promoter DNA hypermethylation have been shown to drive cancer through inactivation of tumor suppressor genes (TSGs), but growing malignant populations also accrue pervasive stochastic epigenetic changes in DNA methylation (DNAme), most of which likely carry little functional impact. Unlike with somatic mutations, we have limited ability to robustly differentiate driver DNAme changes (DNAme drivers) from stochastic, passenger DNAme changes. To address this challenge, we developed MethSig, a statistical inference framework that accounts for the varying stochastic hypermethylation rates across the genome and between samples. MethSig estimates expected background DNAme changes, thereby allowing the identification of epigenetically disrupted loci, where observed hypermethylation significantly exceeds expectation, potentially reflecting positive selection (Fig. 1a). We applied MethSig to reduced representation bisulfite sequencing (RRBS) data of chronic lymphocytic leukemia (CLL) cohorts, which include 304 CLLs collected in a prospective clinical trial (CLL8) and 103 CLLs in a previously published study (CLL-DFCI, Landau et al., 2014), as well as other malignancies where RRBS data is available, including ductal carcinoma in situ (Abba et al., 2015) and multiple myeloma. Area under the receiver operating characteristic curve (AUROC) was used to evaluate sensitivity and specificity of methods in the inference of likely DNAme drivers. We identified two key features that are likely to be strongly associated with true candidate DNAme drivers: gene silencing in relation to promoter hypermethylation and association with clinical outcome. MethSig qualitatively improved ROC across those clinical and biological read outs (0.955 of MethSig, 95% confidence interval [CI] 0.945 - 0.965, versus 0.703 of benchmarked methods, 95% CI 0.669 - 0.737, Fig. 1b used CLL8 as an example). We identified 189 candidate DNAme drivers in CLL, which include known TSGs, and are enriched in genes hypermethylated or inactivated across cancer types. To further validate MethSig's inferences, selected CLL candidate DNAme drivers (DUSP22, RPRM) underwent CRISPR/Cas9 knockout (KO) in CLL cells and stable KO clones were generated through single-cell cloning to eliminate genetic heterogeneity effect. The RPRM and DUSP22 KO clones showed faster growth without treatment (Fig. 1c) and superior fitness in ibrutinib/fludarabine treatment compared with controls (Fig. 1d). Notably, we observed a gene dose effect in the RPRM KO clones (Fig. 1c-d, greater growth of the bi-allelic compared to mono-allelic KO). Elastic net regression with a Cox proportional hazards model was used to evaluate DNAme drivers' contribution to the prediction of failure-free survival after treatment (FFS; failure defined as retreatment or death) and a rigorous training (CLL8) and validation (an independent cohort CLL-DFCI) cohort study design was implemented to safeguard from overfitting and poor generalizability. DNAme drivers were found to be associated with shorter FFS in independent CLL cohorts (Fig. 1e-f). A regression model including established CLL risk indicators (IGHV unmutated status, del[17p] or TP53 mutation) showed an adjusted hazard ratio of 2.3 (95% CI 1.6 - 3.3, P = 2 × 10-6) in CLL8 cohort and 3.2 (95% CI 1.2 - 8.8, P = 0.02) in CLL-DFCI cohort for patients with high risk. Application of MethSig to CLL relapsed after chemoimmunotherapy further identified relapse-specific DNAme drivers, enriched in TP53 targets as well as DNA damage pathway, which indicates that CLL relapse after chemotherapy may follow an alternative path compared to CLL progression in the absence of therapy, offering novel insights for therapeutic strategies to address drug-resistant or relapsed cancer. Collectively, our data support a novel framework for the analysis of DNAme changes in cancer to specifically identify DNAme drivers of disease progression and relapse, empowering the discovery of epigenetic mechanisms that enhance cancer cell fitness. This work addressed a central gap between cancer epigenetics and cancer genetics, where such tools have had a transformative impact in precision oncology and cancer gene discovery. Disclosures Tausch: Janssen-Cilag: Consultancy, Honoraria, Research Funding; Roche: Consultancy, Honoraria, Research Funding; AbbVie: Consultancy, Honoraria, Research Funding. Fink:AbbVie: Other: travel grants; Janssen: Honoraria; Celgene: Research Funding. Fischer:AbbVie: Honoraria; F. Hoffmann-La Roche: Honoraria, Other: travel grants. Gnirke:FL67 Inc.: Consultancy. Moreaux:Diag2Tec: Consultancy. Hallek:Celgene: Consultancy, Research Funding; Pharmacyclics: Consultancy, Research Funding; AbbVie: Consultancy, Research Funding; Gilead: Consultancy, Research Funding; Mundipharma: Consultancy, Research Funding; Janssen: Consultancy, Research Funding; Roche: Consultancy, Research Funding. Stilgenbauer:Janssen-Cilag: Consultancy, Honoraria, Other: travel support, Research Funding; Genentech: Consultancy, Honoraria, Other: travel support, Research Funding; Gilead: Consultancy, Honoraria, Other: travel support, Research Funding; Genzyme: Consultancy, Honoraria, Other: travel support, Research Funding; Pharmacyclics: Consultancy, Honoraria, Other, Research Funding; F. Hoffmann-LaRoche: Consultancy, Honoraria, Other: travel support, Research Funding; Celgene: Consultancy, Honoraria, Other: travel support, Research Funding; Boehringer-Ingelheim: Consultancy, Honoraria, Other: travel support, Research Funding; Amgen: Consultancy, Honoraria, Other: travel support, Research Funding; AbbVie: Consultancy, Honoraria, Other: travel support, Research Funding; GlaxoSmithKline: Consultancy, Honoraria, Other: travel support, Research Funding; Novartis: Consultancy, Honoraria, Other, Research Funding; Mundipharma: Consultancy, Honoraria, Other, Research Funding. Wu:BionTech: Current equity holder in publicly-traded company; Pharmacyclics: Research Funding. Elemento:Acuamark: Current equity holder in private company, Membership on an entity's Board of Directors or advisory committees; OneThree Biotech: Current equity holder in private company, Other: Cofounder; Owkin: Current equity holder in private company, Membership on an entity's Board of Directors or advisory committees; Sanofi: Research Funding; Eli Lilly: Research Funding; Volastra Therapeutics: Current equity holder in private company, Other: Cofounder; Freenome: Current equity holder in private company, Membership on an entity's Board of Directors or advisory committees; Genetic Intelligence: Current equity holder in private company, Membership on an entity's Board of Directors or advisory committees; Janssen: Research Funding. Landau:Bristol Myers Squibb: Research Funding; Illumina: Research Funding.

Genetic, epigenetic and transcriptional heterogeneity cooperate to fuel cancer's ability to evolve and adapt to therapy. In chronic lymphocytic leukemia (CLL), we have shown through bulk DNA methylation (DNAme) sequencing that the growing CLL populations diversify through stochastic DNAme changes (epimutations), impacting transcriptional heterogeneity, clonal evolution and clinical outcome. To directly integrate across epigenetic, genetic and transcriptional intra-leukemic cell-to-cell variation, we developed a high-throughput multi-modality platform that jointly interrogates the methylome, transcriptome and genetic driver mutations from the same single cell (Fig. A). We applied it to >2,000 B cells from 6 healthy donors and 12 CLL samples. We found that the common clonal CLL origin resulted in an elevated but uniform epimutation rate (i.e., low cell-to-cell epimutation variability). In contrast, in index sorted B cell subsets, ranging in maturity from naïve B cells (CD27-IgM+IgD+++IgG-) to memory B cells (CD27+IgG+), variable epimutation rates reflect cells with diverse evolutionary ages across the B cell differentiation trajectory (Fig. B). Thus, we posited that epimutation can serve as a "molecular clock", enabling high-resolution lineage reconstruction, applicable directly to patient samples. CLL lineage tree topology revealed earlier branching, and longer branch lengths than normal B cells, consistent with rapid drift after transformation, and a greater proliferative history (Fig. C). In contrast to CLL topologies, non-clonal normal B cell trees provided a smaller increase in clustering accuracy compared with parsimony-based trees (P < 0.001; Fig. D). To validate the inferred tree topology, we leveraged our multi-modal capture of DNAme and genetic drivers in single cells. Indeed, genetic subclones mapped accurately to distinct clades inferred solely based on epimutation information [e.g., a clade composed of SF3B1 mutants and another clade composed of SF3B1 wild-type cells (P = 7 x 10-9; Fig. E, F)]. Using the joint single-cell transcriptional profiling, we found that cells in SF3B1 mutated clade also displayed higher alternative 3' splicing than their wild-type counterparts (P = 0.01526; Fig. G). Cells belonging to SF3B1 mutated clade were marked by expression changes in genes related to DNA damage (e.g., KLF8) and Notch signaling (e.g., DTX4) (P < 0.05; Fig. H, I). The direct linking of single-cell transcriptional data with lineage identity also showed that transcriptional similarity between cells decreases as a function of their lineage distance (P < 0.05; Fig. K). Notably, the molecular clock feature of epimutations enabled precise timing of subclonal divergence event in the CLL's evolutionary history, estimated to have occurred 2180±219 days after the emergence of the parental clone (Fig. J). To examine potential lineage biases during therapy, we performed serial multimodal single-cell profiling of a CLL patient without subclonal genetic drivers, prior to and during ibrutinib-associated lymphocytosis. The lineage trees revealed a distinct clade of cells preferentially expelled from the lymph node, marked by a distinct transcriptional profile, likely representing ibrutinib-sensitive cells. Lastly, we hypothesized that frequent epigenetic modifier mutations seen in hematological malignancies may increase the epimutation rate promoting intra-leukemic cellular diversity. To test this hypothesis, we applied our multi-modality single-cell platform to cells from TET2 knock-out (KO) mouse models, and observed higher epimutation rates, closely associated with higher cell-to-cell transcriptional heterogeneity compared with wild-type cells (P < 0.0001; Fig. L, M). In summary, we revealed that CLL show uniformly elevated epimutation rates, reflecting the common evolutionary age of these cells. By leveraging the heritable information captured through epimutation data, we provided a native lineage tracing applicable directly to patient samples. With this approach, we showed that CLL lineage topology exhibit early branching and long-branch length, consistent with exponential growth. Finally, we demonstrated that multimodal single-cell profiling enables projection of genetic and transcriptional identity onto the lineage tree, providing a direct measurement of the heritability of transcriptional profiles as a function of lineage distance. Figure. Figure. Disclosures Wu: Neon Therapeutics: Equity Ownership.

Genetic and epigenetic heterogeneity of cancer cells fundamentally shapes cancer progression and relapse. In chronic lymphocytic leukemia (CLL), we previously reported that intra-leukemic epigenetic diversity in DNA methylation (DNAme) follows a stochastic pattern reminiscent of genetic 'trial and error' in cancer evolution. We measured stochastic DNAme from bulk sequencing by observing the proportion of discordantly methylated sequencing reads (PDR), and found that higher PDR was associated with greater cell-to-cell transcriptional heterogeneity and adverse clinical outcome. However, bulk DNAme sequencing does not allow to phase stochastic DNAme across distant genomic loci for a single cell, as it is limited to length of a short sequencing read. Thus, it can only provide the average PDR for a population of cells, rather than the cell-to-cell variation in this important epigenetic feature. Therefore, to define stochastic DNAme changes at the single cell level, we optimized a multiplexed single cell reduced representation bisulfite sequencing (MscRRBS) protocol to allow high-throughput single cell DNAme sequencing. This protocol significantly improves scalability by multiplexing cells with shorter inline-barcodes at an early stage and utilizing SPRI beads purification to eliminate adapter-dimers. We applied MscRRBS to 393 single CD19pos B cells from two healthy volunteers and 111 single cells from a CLL sample. 88% of cells were evaluable with greater than 100,000 covered CpGs (average of 436,230 CpGs per cell). We achieved bisulfite conversion rates of 99.7%+/-0.0001 (mean+/- SD), without a significant reduction in coverage. A downsampling analysis showed that 2.1 million reads per cell provided 85% of CpG coverage with only marginal increase in coverage with further sequencing. Biallelic coverage was observed in 4.6+/-2% of germline SNPs. With MscRRBS, we measure the PDR of each individual cell. As expected from our prior bulk RRBS analysis, we found that the average PDR across cells was higher in CLL compared with B cells from healthy adult volunteers (0.39+/-0.01 vs. 0.26+/-0.08, P The higher uniformity of PDR in CLL may reflect the relationship between epi-mutation rate (measured through PDR) and the evolutionary age of the cells. As additional stochastic DNAme changes are generated with each generation, we hypothesize that PDR estimates the number of generations in a cells' evolutionary history. Thus, CLL cells have uniformly high PDR reflecting a high but uniform number of generations in their history, consistent with a single common cell of origin. In contrast, normal B cells have diverse histories with newly formed naive cells intermixed with long-lived memory cells. Consistent with this hypothesis, healthy donor CD27neg cells showed lower PDR and less PDR cell-to-cell variation compared with CD27pos cells (mean absolute PDR difference 0.008+/-0.01 vs. 0.03+/-0.03, respectively, P As MscRRBS allows complete phasing of DNAme across distant genomic loci, we further calculated the odds ratio of concordance in methylation state between any two neighboring CpG as a function of their genomic distance, compared with any two randomly paired CpGs. Through this procedure we quantified the properties of DNAme concordance decay and uncovered different genomic scales of stochastically disordered methylation. Finally, the MscRRBS also allowed to reconstruct phylo(epi)genetic relationship between the cells, and provided accurate estimate of the rates of stochastic epi-mutation across the genome. Thus, single cell DNAme is a powerful novel tool to define epigenetic diversification and its impact on CLL evolution. Disclosures No relevant conflicts of interest to declare.