Developing and democratizing tools for spatially resolved profiling of genome and transcriptome

Abstract

The common thread of the work presented in this thesis is the development and democratization of tools that facilitate the spatial profiling of genome and transcriptome.

Paper I describes iFISH, a resource that facilitates the large-scale design and costeffective, flexible production of probes for DNA FISH. More specifically, we created a database of oligos targeting the human genome and a probe design tool which we made publicly available through a user-friendly web interface, which we have named iFISH4U. We used iFISH4U to design hundreds of small probes (<10kb span), targeting equally spaced landmark loci, evenly distributed along all chromosomes. We validated a large fraction of those probes individually and then combined them to create “spotting probes” in order to profile chromosomal territories (CTs) in an unbiased way in 3D fixed cells. We were able to visualize up to 6 different chromosomes in thousands of terminally differentiated fibroblasts and embryonic stem cells and to compare their CTs. In ESCs we observed less distinct CTs, and, in a subset of cells, a remarkable lack of chromosome territoriality. We also assessed whether the degree of chromosomal intermingling differs between pluripotent and differentiated cells. All the tools and probes used in paper I were made available for the community, making iFISH a resource that can greatly facilitate genome architecture studies.

In paper II, we developed genomic loci positioning by sequencing (GPSeq), a next generation sequencing based method for profiling of the radial arrangement of chromatin in fixed cells. In GPSeq, we control the incubation time of a restriction enzyme in order to digest chromatin in a gradual manner, starting from the periphery and moving towards the center. By ligating adapters and sequencing the DNA fragments around the cut sites, we can generate high-resolution, genome-wide radiality maps upon computing of a radiality score. In order to define the best radiality score we compared the different estimates with 3D DNA FISH measurements taking advantage of the tools we developed in iFISH. We used GPSeq maps to profile the radiality of various levels of chromatin organization such as chromosomes, A and B compartments and their subcompartments. Moreover, we integrated those maps with different annotation tracks in order to study the radial localization of features and marks of active and inactive chromatin. We also used GPSeq to assess the radial arrangement of SNVs, SNPs and DNA breaks. Finally, we developed a 3D genome reconstruction algorithm that combines GPSeq and Hi-C data, to predict the chromatin structure in single cells. Altogether, GPSeq is a versatile assay that can be used for radial profiling of the 3D genome.

Paper III introduces Deconwolf (DW), a publicly available and easy-to-use software for image deconvolution. Even though DW can be used for different type of images, we tested its performance on data generated by imaging-based spatially resolved omics, technologies that (depending on the method) suffer from 2 different limitations, the requirement for imaging with high magnification objectives and the optical crowding problem, or in other words the inability to resolve densely packed targets. As a proof of concept, we applied DW to images from smFISH experiments and we were able to resolve crowded signal as well as to quantify smFISH signals in tissue sections imaged with 20x magnification. We then applied DW to In Situ Sequencing (ISS) and OligoFISSEQ images and we found that DW greatly increases the sensitivity of transcript calling in hybISS and the barcode detection efficiency in OligoFISSEQ. Our results highlight the potential of DW to improve image deconvolution and consequently the amount of information yielded upon imaging-based spatially-resolved molecular profiling