Periodic Reporting for period 1 - SCISSORS (Single cell profiling of breast tumors and the residing immune cells leveraged by integration with multi-dimensional molecular data from thousands of tumors)
Reporting period: 2021-01-01 to 2022-12-31
Breast cancer is the most common type of cancer in Europe, responsible for the highest women cancer mortalities each year. Characterisation of breast tumours teased them apart to distinct subtypes and facilitated targeted treatments that improved survival rates significantly, yet some aggressive subtypes remain difficult to treat. New therapies that harness the patient own immune system to fight the tumour show remarkable success in several cancers but limited one in breast cancer. Single cell genomics methods that profile the tumour and its surrounding cells hold the potential to better understand the dynamics of tumour development and its resistance to therapy and improve current cancer treatment.
The SCISSORS project aimed to profile normal and cancerous breast tissue samples on the single cell level to elucidate the interplay between genomic clones, their functional profile, and the response of the stromal environment as they all evolve. The objectives were 1) analyse a comprehensive single cell expression atlas of healthy human breast tissue, 2) develop a computational framework to infer somatic clones from single cell RNA-seq data, 3) profile the complete environment of normal and malignant breast tissues at single cell and clonal levels, and 4) extend results by integration with existing data on thousands of multi-dimensional molecularly profiled tumour samples.
The technical developments made in the project facilitated the analysis of a large cohort of breast tumours from various types. Using a normal breast cell atlas as reference enabled the classification of single cells to phenotypes. We observed single cells mixing several normal epithelial breast phenotypes, a phenomenon confirmed to be unique to tumours. However, a minority of the tumours evaded this phenotypical plasticity and remained extremely homogenous. We were able to infer the genomic clones by combining several published and in-house tools. Integration of these genomic clones with the phenotypic classifications shed light on the dynamics of clonal evolution and the corresponding transcriptional changes.
The SCISSORS project aimed to profile normal and cancerous breast tissue samples on the single cell level to elucidate the interplay between genomic clones, their functional profile, and the response of the stromal environment as they all evolve. The objectives were 1) analyse a comprehensive single cell expression atlas of healthy human breast tissue, 2) develop a computational framework to infer somatic clones from single cell RNA-seq data, 3) profile the complete environment of normal and malignant breast tissues at single cell and clonal levels, and 4) extend results by integration with existing data on thousands of multi-dimensional molecularly profiled tumour samples.
The technical developments made in the project facilitated the analysis of a large cohort of breast tumours from various types. Using a normal breast cell atlas as reference enabled the classification of single cells to phenotypes. We observed single cells mixing several normal epithelial breast phenotypes, a phenomenon confirmed to be unique to tumours. However, a minority of the tumours evaded this phenotypical plasticity and remained extremely homogenous. We were able to infer the genomic clones by combining several published and in-house tools. Integration of these genomic clones with the phenotypic classifications shed light on the dynamics of clonal evolution and the corresponding transcriptional changes.
Following analysis of an in-house pilot dataset of normal breast tissue single-cell RNA-seq (scRNA-seq), several similar yet larger scale datasets were published. I analysed them and developed a method to use them for classification of breast tumour cells. In addition, I applied my cell atlas experience to a new collaboration aiming to generate a single cell atlas of the healthy fallopian tube.
I used several breast patient derived tumour xenografts (PDTXs) datasets generated in our lab to develop and test a computational framework to infer somatic clones from scRNA-seq data. A genomic clone is defined predominantly by the set of copy number alterations and single nucleotide variants its cells harbour. I spent considerable effort in adjusting existing tools that infer both and devised a novel method to integrate both clonal outputs and improve the accuracy of the clonal inference.
The developments described above enabled me to analyse a large cohort of diverse breast PDTXs. My aim was to classify each cell by three different classifications: genomic clone, cell identity, and transcriptional cell state. Integration of these classifications allowed the assessment of intra-tumour transcriptional plasticity and the transcriptional plasticity within a genomic clone. There are three types of epithelial cells in the healthy breast – basal, luminal progenitor and mature luminal. Surprisingly, we found that tumour cells assume intermediate phenotypes, mixing several normal cell types in the same cell. The observed phenotypes corresponded to the originating tumour types but were not fully determined by them. Although transcriptional plasticity was pervasive, we detected a minority of tumours that maintained extremely homogenous transcriptional profile in all their cells. We are currently focusing on this intriguing group of tumours to understand the mechanism that maintains tumour homogeneity. Integration of the inferred genomic clones with the phenotypic classifications elucidates the dynamics of clonal evolution and transcriptional changes and revealed examples where newly emerged clones manage to span the full transcriptional repertoire of the parent clone, and where new clones introduce new phenotypes to the tumour. A comprehensive literature scan yielded many breast tumour datasets that I analyse aiming to see if and how the above results constitute in larger cohorts of human breast tumours scRNA-seq datasets.
The above work is shaping into two papers I now wrap up and plan to submit in the upcoming months. I also presented this work in the annual conference of the European Association of Cancer Resarch (EACR 2022) in a selected talk, and in the Computational Cancer Genomics (CCG2022) conference in a poster.
I used several breast patient derived tumour xenografts (PDTXs) datasets generated in our lab to develop and test a computational framework to infer somatic clones from scRNA-seq data. A genomic clone is defined predominantly by the set of copy number alterations and single nucleotide variants its cells harbour. I spent considerable effort in adjusting existing tools that infer both and devised a novel method to integrate both clonal outputs and improve the accuracy of the clonal inference.
The developments described above enabled me to analyse a large cohort of diverse breast PDTXs. My aim was to classify each cell by three different classifications: genomic clone, cell identity, and transcriptional cell state. Integration of these classifications allowed the assessment of intra-tumour transcriptional plasticity and the transcriptional plasticity within a genomic clone. There are three types of epithelial cells in the healthy breast – basal, luminal progenitor and mature luminal. Surprisingly, we found that tumour cells assume intermediate phenotypes, mixing several normal cell types in the same cell. The observed phenotypes corresponded to the originating tumour types but were not fully determined by them. Although transcriptional plasticity was pervasive, we detected a minority of tumours that maintained extremely homogenous transcriptional profile in all their cells. We are currently focusing on this intriguing group of tumours to understand the mechanism that maintains tumour homogeneity. Integration of the inferred genomic clones with the phenotypic classifications elucidates the dynamics of clonal evolution and transcriptional changes and revealed examples where newly emerged clones manage to span the full transcriptional repertoire of the parent clone, and where new clones introduce new phenotypes to the tumour. A comprehensive literature scan yielded many breast tumour datasets that I analyse aiming to see if and how the above results constitute in larger cohorts of human breast tumours scRNA-seq datasets.
The above work is shaping into two papers I now wrap up and plan to submit in the upcoming months. I also presented this work in the annual conference of the European Association of Cancer Resarch (EACR 2022) in a selected talk, and in the Computational Cancer Genomics (CCG2022) conference in a poster.
I plan to continue the work laid out in the SCISSORS project in the same lab, where I am now employed. The major next steps will be to analyse the large number of human breast tumour datasets with similar methodology I applied on our in-house dataset. These datasets will add another axis to the exploration of the tumour – the response of the tumour microenvironment as the tumour evolves. Analysing both normal, malignant, and normal tissue adjacent to tumours will hopefully shed light on the dynamics of early tumour development. These analyses of published datasets will help us plan the profiling of our in-house breast tumours paired with their adjacent normal tissues.
The last aim is to analyse large cohorts of breast tumour profiled in bulk in light of the project findings. This will bring the project results one step closer to clinical relevance, as bulk methods are considerably cheaper than single cell genomic methods and are therefore more routinely performed during patient diagnosis.
The detection of highly homogenous tumours was a striking result. Tumour plasticity is a hallmark of cancer and is a major driver of therapy resistance. Therefore, elucidating a mechanism that restrain the transcriptional plasticity of a tumour will be of major clinical importance. We are in the early stages of exploring this phenomenon and still treat it with caution, but the potential impact pushes us to focus on it with a set of follow-up and validation experiments.
The last aim is to analyse large cohorts of breast tumour profiled in bulk in light of the project findings. This will bring the project results one step closer to clinical relevance, as bulk methods are considerably cheaper than single cell genomic methods and are therefore more routinely performed during patient diagnosis.
The detection of highly homogenous tumours was a striking result. Tumour plasticity is a hallmark of cancer and is a major driver of therapy resistance. Therefore, elucidating a mechanism that restrain the transcriptional plasticity of a tumour will be of major clinical importance. We are in the early stages of exploring this phenomenon and still treat it with caution, but the potential impact pushes us to focus on it with a set of follow-up and validation experiments.