Background

Understanding the molecular mechanisms that drive cancer initiation, progression, and therapeutic resistance is crucial for devising effective therapies for cancer treatment. In addition to the genetic drivers of cancer, such as somatic mutations, copy number variations, and genomic instability, new research has shown epigenetic mechanisms as potential drivers of cancer progression and therapeutic resistance [1–3]. These epigenetic features modulate transcriptional programs and cellular functions through a variety of mechanisms without modifying the underlying DNA sequence. Underlying these epigenetic mechanisms is ultimately the binding of transcription factors (TF) and chromatin remodelers at specific non-coding regions of the DNA known as regulatory elements. These elements, such as promoters, enhancers, or silencers, regulate gene expression when engaged by specific TFs [4]. Using chromatin accessibility [5] to infer TF binding at these regulatory regions allows for a more comprehensive understanding of expression regulation.

The single-nucleus assay for transposase-accessible chromatin by sequencing (snATAC-seq) [5,6] performs high-throughput profiling of chromatin accessibility at single-cell resolution, allowing us to identify cell-specific epigenetic patterns. By combining this information with publicly available ChIP-seq and CUT&RUN data for currently available DNA-binding factors (DBFs) (TFs and chromatin regulators) [7], we can predict the enrichment of potential binding events of each factor in any given cell using in silico methods such as ChromVAR [8] and ReMapEnrich. This approach allows us to characterize unique epigenetic signatures that link chromatin accessibility to DBFs, enabling the discovery of key transcription factors or chromatin regulators as targetable epigenetic biomarkers [9,10].

In our study aimed at characterizing the chromatin landscape and transcriptional features of drug-tolerant persister cells in high-grade serous ovarian cancer (HGSOC), we performed single-nucleus multiomic profiling (snRNA+snATAC seq) of treatment-naïve and neoadjuvant chemotherapy (NACT)-treated tissues from patients with HGSOC [11]. By comparing the open-chromatin signatures between the treatment-naïve patients who later developed resistance to adjuvant chemotherapy and the NACT-treated patients, we identified a distinct epigenetic signature of persister cells that distinguished the chemotherapy response in treatment-naïve tumors.

In this paper, we describe the computational protocol that was developed to identify this persister cell signature using chromatin accessibility data from snATAC-seq. The protocol starts with snATAC-seq data and a database of DNA-binding information (ChIP-seq/CUT&RUN data) formatted as a four-column BED file. We used data downloaded from ReMap 2022 for this purpose. We assume that the snATAC-seq count matrices from all the samples in the cohort are already merged/pooled and that the cell types are previously identified and included as a metadata table. In the current example of HGSOC, we have three specific groupings of samples. First, the samples are grouped as naïve and NACT. The naïve group is further subdivided into two groups: sensitive and resistant. The snATAC-seq data is initially processed using Signac [12–14] with TF-IDF normalization and singular value decomposition for dimensionality reduction. We then incorporate two different methods to combine DNA-binding information with snATAC-seq data. The first method uses ChromVAR to calculate the enrichment score (bias-corrected deviation score) for each DBF and for each cell. These scores can be used to calculate differential enrichment between any two groups of cells. In the second method, we use ReMapEnrich to calculate the ranked enrichment of DBFs in a single set of open-chromatin regions. This method is specifically used on the open-chromatin regions that were simultaneously up-regulated in the naïve-resistant cohort and in the NACT cohort (both comparisons using the naïve-sensitive cohort as control), to define the persister cell signature. We believe that the protocol will serve as guidance for researchers interested in multimodal analysis of snATAC-seq data in combination with a database of DNA-binding data, such as ChIP-seq or CUT&RUN. In addition, we also describe using gene expression data from snRNA-seq to filter DBFs that are lowly expressed in the cell population of interest.