Single-Cell Epigenomics In Cancer Research

A single tumor mass is comprised of many different subpopulations of cells. During evolutionary trajectory of a tumor, cells with different molecular features are likely to evolve and interactions between these heterogenous cell subpopulations in tumor are highly dynamic [1]. This intratumoral heterogeneity (ITH) is regulated in multiple levels and each cell subpopulation in a particular tumor can have distinct genomic, epigenomic, transcriptomic and spatial characteristics, effecting their metastatic potential and chemoresistance [2]. How the complex contributions of each subgroup in a complete population of cells in tumor affects the clinical progression of the disease is not completely understood. This is mostly due to the fact that cancer research previously has been limited to the analysis of bulk cells without dissecting heterogenous cell populations into subgroups with similar molecular features, therefore these studies have only reflected average profile of complex subgroups of cells, called subclones. Tumors are characterized by dysregulated epigenomes. Epigenetic modifications contribute to tumor cell heterogeneity resulting in, for example, differential responses to targeted therapies [3].


Introduction
A single tumor mass is comprised of many different subpopulations of cells. During evolutionary trajectory of a tumor, cells with different molecular features are likely to evolve and interactions between these heterogenous cell subpopulations in tumor are highly dynamic [1]. This intratumoral heterogeneity (ITH) is regulated in multiple levels and each cell subpopulation in a particular tumor can have distinct genomic, epigenomic, transcriptomic and spatial characteristics, effecting their metastatic potential and chemoresistance [2]. How the complex contributions of each subgroup in a complete population of cells in tumor affects the clinical progression of the disease is not completely understood. This is mostly due to the fact that cancer research previously has been limited to the analysis of bulk cells without dissecting heterogenous cell populations into subgroups with similar molecular features, therefore these studies have only reflected average profile of complex subgroups of cells, called subclones.
Tumors are characterized by dysregulated epigenomes. Epigenetic modifications contribute to tumor cell heterogeneity resulting in, for example, differential responses to targeted therapies [3].
Epigenetic regulation encompasses many biological layers analyse the single-cell data, which are still rapidly emerging in the field. In addition to methods with which single-cell genome and single-cell transcriptome sequencing can be performed, methods to profile epigenome of a single-cell have also been developed and continue to be improved. With these single cell epigenomics tools, it is now possible to study a single cancer cell in terms of all forms of epigenetic regulation (for example, DNA methylation, chromatin accessibility, histone modifications and 3D chromatin topology). In this article, we will review recent single cell epigenomics methods with a particular focus on their applications in cancer research.

DNA Methylome Analysis at Single-Cell Resolution
In cancer, DNA methylation aberrancies is highly evident and it is considered as a hallmark of human cancers [4]. These epigenomic alterations, for instance, can result in silencing of tumor suppressors due promoter hypermethylation or in upregulation of oncogenes due to gene-body hypermethylation. The regulation of DNA methylome is highly complex in cancer and this topic was reviewed elsewhere [4][5][6]. Here, we touch upon some single-cell methods developed in recent years, which enable us to study methylomes of single cancer cells.

scRRBS (Single-Cell Reduced Representation Bisulfite
Sequencing) The first single-cell methylome analysis was reported in 2013 [7]. In this study, authors modified the original bulk RRBS method [8] and entire protocol was performed in a single tube prior to bisulfite conversion, minimizing DNA losses arising from multiple purification steps. In bisulfite conversion, unmethylated cytosines are converted to uracils due to deamination events, whereas methylated cytosines remain unconverted, thus allowing to study DNA methylation at single-base resolution after generating sequencing libraries of bisulfite-converted DNA template. In this method, unmethylated lambda DNA was used as spike-in to control for the false-positives which are due to non-conversion of unmethylated cytosines. It should be noted that this method does not discriminate between 5-methylcytosine (5-mC) and less and multiplexed scRRBS results [9]. They demonstrated that the correlation between DNA methylation and transcription is lower in CLL cells when compared to the same correlation for normal B cells in bulk analyses, however they found a higher correlation between DNA methylation and transcription in CLL cells when analyses were performed in single cells, reflecting epigenetic diversity in CLL cells.
They concluded that this epigenetic heterogeneity in leukemic cells may facilitate the emergence of novel cell states with a possibly higher fitness potential, resulting in, for example, higher resistance to therapy.

scBS-seq / scPBAT (Single-Cell Bisulfite Sequencing / Single-Cell Post-Bisulfite Adaptor Tagging)
This protocol uses a modification of post-bisulfite adaptor tagging (PBAT) used in bulk samples and it minimizes DNA loss from single cells [10,11]. In this method, sequencing adaptors are Similar to scBS-seq, this technique also takes advantage of postbisulfite adaptor ligation protocol with which relatively low amount of DNA is lost [12]. This method does not focus on CpG islands as scRBBS method does; but rather covers CpGs in a cumulative manner. In contrast to scBS-seq method, library complexity in scWGBS is relatively low, because this method does not require any pre-amplification step. Gkountela et al. applied this method to single tumor cells that are dissociated from circulating tumor cell (CTC) clusters [13]. They show that CTC cluster dissociation into single constituent cells leads to DNA methylation remodelling at some critical loci (binding sites for OCT4, SOX2, NANOG, and SIN3A) in these single circulating tumor cells. This study highlights the importance of single cell methylome research in providing essential information on epigenetic dynamics of single circulating tumor cells which pose an increased risk for metastasis when they form multicellular clusters. There are multiple other methods developed in recent years which allow studying DNA methylation profiles in single cells [14][15][16].
In one study, Li et al. used single ovarian cancer cells isolated from formalin-fixed and paraffin-embedded (FFPE) human ovarian cancer tissue using laser capture microdissection to explore DNA methylome heterogeneity between these single cells from the same subpopulation of tumor tissue [15]. We anticipate that many other single-cell DNA methylation analysis methods will be developed, and existing tools will be improved in the near future, reducing the time and cost associated with these techniques.
Also, it is very likely that number of cancer studies reporting the use of these protocols will exponentially increase, providing novel insights in the understanding of methylome regulation in cancer, which were previously missed due to bulk analyses. It is also worth noting that DNA modifications other than 5-cytosine methylation (such as 5-formylcytosine, 5-hydroxymethylcytosine and 5-carboxylcytosine) can be measured in a single-cell level using recently developed techniques [17][18][19]. Thus, these relatively low abundance DNA modifications which form multiple epigenetic layers of molecular connectivity between genome and its functional output can be studied in single cancer cells, providing a great multitude of new data.

scCHIP-seq (Single-Cell Chromatin Immunoprecipitation followed by Sequencing)
Bulk CHIP-seq is used in cancer research to map histone modifications and protein-DNA interaction (such as TF binding sites) genome-wide. However, the limitation of this method is that Targets and Tagmentation) [22]. It also maps transcription factor (TF) binding and accessible DNA in parallel, similar to multi-omics approaches that we will mention below. In this study, they reported that chromatin profiling is sufficient to discriminate single cell types.

scATAC-seq (Single-Cell Assay for Transposase-Accessible Chromatin using Sequencing)
This method is used to identify active regulatory regions of genome characterized by lower density of nucleosomes (therefore, accessible sites of genome) in single cells [23,24]. It uses prokaryotic Tn5 transposase to insert sequencing adaptors to nucleosome-free elements of the genome where presence of nucleosomes does not block its activity. scATAC-seq allows mapping the accessible genome of individual cells with the use of microfluidic devices, enabling the study of epigenetic regulatory variation between these cells. In one study, the authors first created a mixed cell population using two oesophageal adenocarcinoma cell lines and one non-cancer cell line in equal proportions [25]. Using scATAC-seq data and an algorithm which they developed (named Scasat, single-cell ATACseq analysis tool), they were able to cluster these heterogenous cell

scDNase-seq (Single-Cell DNase Sequencing)
Similar to scATAC-seq, scDNase-seq can be used to map active regulatory regions in the genome with low nucleosome presence (therefore, sensitive to DNase-I cleavage due to reduced protection from nucleosomes) in single cells [29]. These DNase-I hypersensitive sites (DHS) reflect chromatin accessibility patterns and activity of regulatory genomic elements, thus providing important data on the epigenome of a single-cell. In the same study where they reported the development of scDNase-seq, the authors applied this technique to follicular thyroid carcinoma (FTC) cells dissected from formalin- One can imagine the potential application of this method in characterizing heterogeneity in chromosomal conformation within cancer cells which are seemingly identical, but phenotypically diverse. Analyses at this epigenetic layer will provide insights into functional consequences of topologically associating domains (TADs) in single cancer cells.
scHi-C will also be valuable in studying the spatial organization of chromosomes in rare cell types which cannot be obtained in

Single-Cell Multi-Omics Approaches
Analysis of only one molecular feature (such as only methylome) from a single-cell provides limited and incomplete information, because a cellular state is determined by the highly complex interplay of multiple molecular layers in a cell. Therefore, in addition to single-cell omics approaches mentioned above, several single-cell multi-omics techniques with great applicability potential in cancer research have been developed in recent years.
Here, we review single-cell multi-omics approaches which include single-cell epigenomics analyses, and exclude those which provide data only on genomic, transcriptomic and/or proteomic levels.

scMT-seq (Single-Cell Methylome and Transcriptome Sequencing)
This method enables simultaneous profiling of DNA methylome and transcriptome from the same single-cell, providing an opportunity to study the correlation between these two omics data [33]. In scMT-seq, cellular membrane is lysed, however, nuclear membrane is kept intact and isolated by microcapillary picking. mRNA isolated from cytoplasmic component is amplified using a modified version of Smart-seq2 method for transcriptome analysis and nuclear genome is used for methylome profiling in parallel [34,35]. This technique takes advantage of scRRBS method for methylome analysis, whereas a similar method named scM&T uses genome-wide bisulfite sequencing for DNA methylation profiling [36]. Both of these parallel methylome and transcriptome sequencing methods in single cells can help to dissect complex interactions between epigenome and transcriptome, and are powerful tools to study cellular heterogeneity, which is intrinsically strong in cancer, in these different omics layers.
The diversity of drug resistance mechanisms in cancer is not completely known. One potential use of single-cell multi-omics approaches in cancer can be the identification of mechanisms regulating drug insensitivity in both genomic and epigenomic levels. By using methylome and transcriptome data acquired via scMT-seq or scM&T-seq, cancer researchers can compare single chemosensitive and chemoresistance cancer cells of the same type in terms of multiple molecular features and identify (epi) genomic processes driving resistance to therapy. These studies will be invaluable in personalized medicine approaches for cancer treatment. By increasing the resolution and size of data, they will have the potential to provide more effective treatment strategies in cancer. This study also shows the capability of multi-omics single-cell approaches to decipher inter-omics regulation at a single-cell level (for example, the identification of mutual relationships between epigenome and transcriptome).

scTrio-seq (Single-Cell Triple Omics Sequencing)
A similar approach called sc-GEM (single-cell analysis of Genotype, Expression and Methylation) which identifies methylation patterns at specific loci (not globally as in scTrio-seq) was used to cluster single cells from primary lung adenocarcinomas in different patients and single cells from one non-tumor lung tissue, by using DNA methylation state of a panel of genes known to be aberrantly methylated in this cancer type [38]. They were able to group cells

scNMT-seq (Single-Cell Nucleosome, Methylation and Transcription Sequencing)
scNMT-seq combines RNA-seq with chromatin accessibility and methylome profiling performed in the same cell [44]. It is built on scNOMe-seq method with the addition of transcriptome analysis.
As in scNOMe-seq, it allows parallel profiling of two epigenomic molecular layers (chromatin accessibility and DNA methylation) at a single-cell, making it possible to study these dependent features simultaneously. Combined transcriptome analysis adds another omics layer and enables assessing dependencies and associations between epigenome and transcriptome, which is critical for a complete understanding of regulatory mechanisms in cancer. In addition, it will be quite important to see if epigenetic heterogeneity within single cancer cells is coupled or uncoupled between these different omics layers. We previously reported that epigenetic silencing of RGS10 and RGS2 genes by HDAC1 (histone deacetylase 1) and DNMT1 (DNA methyltransferase 1) contributes to cisplatin chemoresistance in ovarian cancer [45,46]. The regulation of this and other genes known to have essential roles in resistance to therapy can be studied genome-wide using scNMT-seq at a single-cell resolution, incorporating data obtained in multiple epigenomic regulatory layers. Another possible use of this multiomics approach can be in identifying cancer cell subpopulations of which epigenetic silencing or activation of genes regulating chemoresistance (such as RGS10) is more profound relative to other cancer cell subpopulations in the same microenvironment for which epigenetic control does not contribute much to drug resistance.

sci-CAR
This method allows joint profiling of gene expression and chromatin accessibility in high-throughput since it uses split pool barcoding for thousands of single cells [47]. In this study, they applied sci-CAR to cells from a human lung carcinoma cell line collected at different time points after dexamethasone treatment.
They were able to cluster untreated and treated cells based on transcriptome or chromatin accessibility data using unsupervised clustering or t-SNE visualization, showing the potential of this technique in scalable profiling of single-cell molecular phenotypes.

Conclusion
Most research on single-cell omics and multi-omics techniques have been consisted of proof-of-concept studies, and publications reporting the application of these methods in various aspects of cancer research are just beginning to emerge. With rapid advances in the field and reducing costs to perform such studies, increasing number of cancer researchers will employ these techniques in their field of study. Another dimension to include in singlecell epigenomics studies is time. Considering the fact that many epigenetic features have different stabilities in time, studying these omics data in temporal dimension will offer interesting insights in single-cell cancer research, such as the change of epigenetic marks over the course of cancer evolution at a single cell level.
Identification of some epigenomic marks which are highly dynamic and short-lived will increase the time resolution in cancer studies and help to offer therapies which are timely administered and more personalized.
Though excluded from this study, single-cell proteomics approaches also provide important findings on tumor individuality and characteristics of the other cells present in tumor microenvironment including immune cells [48]. This type of large-scale studies may be highly useful in precision medicine approaches which target tumor and/or immune cells in the same microenvironment. Another possible outcome of these studies may be more accurate cancer patient classification in the clinic. With the development of more advanced computational algorithms which can process these multidimensional and big data obtained in multiomics approaches, the intricate relationships between all these regulatory layers can be explored. This offers unique perfectives to understand the biology of cancer in more depth (Figure 1).

Figure 1:
Single cell epigenomics methods and the number of omics layers which can be studied using these technologies.