NEW YORK (GenomeWeb) – Researchers from the BC Cancer Agency, the University of British Columbia, and elsewhere have used whole-genome sequencing and targeted single-cell sequencing to understand pretreatment disease progression in high-grade serous ovarian cancer (HGSOC) cases.
In a study published in Nature Genetics this month, they described their efforts to assess cell migration patterns and clone diversity in samples from seven HCSOC patients. As part of the project, they also released software to analyze single-cell sequencing data, called Single Cell Genotyper , which they described in Nature Methods this month.
For their study, the researchers sampled tissues collected from different sites during debulking surgeries. Specifically,they analyzed 68 samples from seven HGSOC patients — five to 13 samples per patient. This included samples from the ovary, omentum, fallopian tube, peritoneal sites, and other metastatic sites. They then used whole-genome and single-cell sequencing to identify clones present at each site and quantify their abundance. Examining evolutionary features such as mutation loss, they evaluated the relationship between clones in the primary ovary site and those at distant anatomic sites and reconstructed the migration pathway of the cells. They also assessed the clonal mixture present at each of the sampled sites.
High-grade serous ovarian cancers are one of the most lethal and difficult-to-treat subtypes of the ovarian cancer family, with poor outcomes for most patients. The standard treatment for the disease is primary site debulking surgery followed by treatment with platinum-based chemotherapy. Patients could also receive neo-adjuvant therapy followed by surgery, or they might be treated with PARP inhibitors like AstraZeneca's Lynparza (olaparib). Still, some 80 percent of women diagnosed with the disease relapse after an initial response to treatment and many succumb to resistant iterations of the disease.
Part of the difficulty with treating HGSOC is that the disease is characterized by widespread metastases to different organs in the peritoneal cavity. Even after debulking, there could still be clones deposited in different sites in the body that these surgeries miss, Sohrab Shah, a senior scientist at the BC Cancer Agency and associate professor at the University of British Columbia, and a senior author of the study, explained. Furthermore, mutations in genes like TP53 cause genomic instability in most HGSOC cases, which is accompanied by disruptions in homologous recombination DNA repair pathways, according to the paper. These changes support the growth of diverse clones from a single source that spread to other organs in the abdominal cavity, the researchers wrote.
For the study, Shah's team wanted to evaluate clonal diversity at the time of diagnosis, as well as assess how these clones settle and thrive in different regions of the body, he told GenomeWeb. They also wanted to study the spread patterns of clones and try to reconstruct those migration pathways. "Is it a unidirectional seeding from one site to another, or is there a lot of intermixing?" he said. With HGSOC, "you have a cavity-type space where there is lots of potential for intermixing of clones, so you might expect that there would be lots of populations co-existing in multiple different [sites], but that wasn't what we observed at all."
What they saw, according to the Nature Genetics paper, was that the clones mostly spread in a unidirectional pattern away from a single site — typically in the ovary — that contained diverse clones. When they sampled metastases, they also found that these sites contained clones from a single phylogenetic clade. This suggests that there may be microenvironments within the peritoneal cavity that select for specific clones, Shah said. They are now planning follow-up studies to try to figure out why some sites can support multiple clones, while others are monoclonal.
Other follow up studies will try to assess mechanisms of drug resistance in the clones. One of the HGSOC patients enrolled in the Nature Genetics study relapsed about 13 months after treatment and testing showed metastasized cells in her brain. She then relapsed a second time after 19 months due to metastases in her pelvis. What that suggests is that there is a minor population within the initial multi-clonal site that is resistant to therapy, Shah said. If researchers could identify those clones at the time of diagnosis, they might make different therapy decisions. For example, patients might get more aggressive, multi-drug treatment from the start to reduce their risk of relapse.
Shah's team plans to create migration maps for cells from sampled sites in relapse cases and compare them to migration maps created for cells drawn from the primary tumor site to try to understand what properties help this subpopulation resist treatment. He and his collaborators are currently recruiting and consenting a new set of patients to participate in these studies, he said. It is not clear how many patients they will enroll but the goal is to evaluate far more individuals than for the first study, he said. The researchers will also grow the resistant cells in model systems.
As part of the Nature Genetics study, the researchers also characterized the genotypes of each of the clones at the sampled sites. They used single-cell sequencing to identify mutational signatures that uniquely mark cells as belonging to a particular clone. However, inferring these signatures from single-cell sequencing data using current computational tools is challenging, according to Shah, because current methodologies generate very noisy datasets — due to missing data and allelic dropout, among other factors — that do not always faithfully represent the genotypes of the individual cells. "We set out to try to really understand the technical properties of those data and see if we could develop methods that are more robust to determine these genotypes than standard methods," Shah said.
As a result, they developed the Single Cell Genotyper (SCG), a software tool that uses a probabilistic model and algorithm to denoise input data, cluster cells with shared genotypes into groups, and infer a genotype for each group.
Full technical details are provided in the paper, but the basic idea behind the software is that "one needs to borrow strength across multiple measurements in order to have much more certainty about the genotype that one is inferring," Shah explained. "So we borrow strength across cells and then try to assign each cell to a particular genotype." Compared to two existing methods — hierarchical clustering and the BitPhylogeny model — SCG showed "increased accuracy in clustering cells and inferring genotypes" when applied to both real and synthetic datasets.
The main input for SCG is data from targeted multiplex single-cell sequencing but Shah's group is also working on separate software for profiling and analyzing single-cell whole-genome sequencing data that they will release in the future, he said.