First, we set up the method to recover both mRNA and histone mark position in single cells at high sensitivity. After establishing protein-A Micrococoal nuclease-based approach to study the distribution of histone modifications in single cells before the start of the fellowship that we termed sortChIC (due to its FACS compatibility), we first optimised conditions for efficient recovery of the cells RNA. Classical approaches using PFA proofed to be problematic for Chromatin accessibility in repressive regions, while the nuclear preparation used by most alternative single cell chromatin methods lead to a nearly complete loss of cytoplasmic RNA. We set up an alcohol-based fixation approach, that had neglectable effect on accessibility while keeping most of the mRNA. To further increase the transcriptome complexity, we used an RNA sequencing approach that recovers full length and also non-polyadenylated transcripts. Specificity as well as sensitivity of the approach were further proven in cell line experiments and a computational approach was established to remove potential mis-annotations of reads from transcription to chromatin.
Adaptation to different antibodies was performed by selection and titration of ChIP validated antibodies.
Aiming to identify suitable enrichment markers for the enrichment of bone marrow differentiation intermediates, we looked at previously generated high resolution sortChIC data for H3K4me1, H3K4me3. Despite the enrichment of immature cells in these experiments, cell type transitions were found to be rather discrete, not allowing the observation of differentiation intermediates. We therefore switched the biological model to one with an even higher fraction of differentiating cells. Specifically, we looked at in vitro gastrulation in a system called Gastroloids, that starts with the aggregation of around 200 ES cells and undergoes over a timeframe of 7 days Apical/basal axis formation, gastrulation, tissue formation and somatogenesis. In this fast-developing system, we looked at the distribution of H3K4me3 and H3K27me3 in around 2000 cells from each day and histone mark from day3 to day7 (FigA). Analysing this complex dataset, we identified 4 major differentiation trajectories with many cell type intermediates. Analysing the changes of relative changes in chromatin and transcription identified numerous genes where the switch from H3K27me3 silent to actively transcribed state could be followed, including intermediate stages that already missed H3K27me, yet did not start expressing the gene yet (example in FigB). With this we generated the first high resolution Chromatin and transcript method in single cells that enables the recovery of full-length cytoplasmic RNA and generated a detailed dataset of the early steps of in vitro gastrulation.
Following steps aim to finalise the analysis of transient chromatin states during gene activation and silencing. As the gastroloid protocol is in vitro, we aim to use the well-established inhibitor for the H3K27 methyl transferase Ezh2 GSK126 to proof the consequence of transient removal of this pathway during gene silencing and activation.
