Single-cell biology: resolving biological complexity, one cell at a time

Unravelling the function of the many cells that make up complex organisms has been one of the main challenges of the last century. Research tools used during the past few decades have allowed the discovery and characterisation of many cell types that are fundamental for health and disease, but it is now clear that they are unsuited to fully capture the heterogeneous nature of cell states. In order to untangle this heterogeneity, the interest of the scientific community has shifted from the characterisation of cell populations to the study of single cells. This was made possible due to rapid technological advances that include single-cell transcriptomics and genomics, imaging and computational tools to process the large amounts of generated data. In March 2018, researchers from all over the world met at the Wellcome Genome Campus, home of the Sanger Institute and the European Bioinformatics Institute, to discuss the latest developments in the study of single cells. The Single Cell Biology conference brought together more than 250 delegates and gave an excellent snapshot of the exciting research involving single-cell approaches, as well as identifying key challenges and opportunities for the future. The inclusive scientific programme was put together by local organisers Sarah Teichmann (Wellcome Sanger Institute, UK) and Berthold Göttgens (University of Cambridge, UK), together with Ellen Rothenberg (Caltech, Pasadena, CA, USA) and Timm Schroeder (ETH Zurich, Switzerland). Among the many topics covered, three main themes emerged that are of interest to the developmental biology community and will be the focus of the present report. The first theme was focused on the technological advances in single-cell biology, with a particular emphasis on imaging and tracing of single cells. The second theme was focused on the opportunity that single-cell technologies offer to generate developmental atlases of entire organisms. The final theme focused on the efforts to resolve the heterogeneity of cell states, with particular attention to the haematopoietic and immune system as well as cancer.

The past decade has seen an extraordinary advance in technologies that allow the interrogation of the molecular content of individual cells. Since the first report of single-cell RNA-seq data in 2009 (Tang et al., 2009), many platforms have been developed. At the same time, the fall in the cost of sequencing means that the broader research community can apply single-cell methods to answer their research questions in diverse scientific fields. However, many challenges still persist, one of which is the complete loss of spatial information in the single cells analysed. To highlight the importance of this challenge, the meeting opened with an exciting keynote presentation by Xiaowei Zhuang (Howard Hughes Medical Institute and Harvard University, Boston, MA, USA), who notably developed the MERFISH (multiplexed error-robust fluorescent in situ hybridisation) method. MERFISH uses combinatorial labelling and sequential imaging, coupled with an error-robust encoding scheme to multiplex single-molecule FISH measurements, allowing the detection of the expression of thousands of genes at the same time, both in cells in culture and within complex tissues (Chen et al., 2015; Moffitt et al., 2016). Her lab also took a similar approach in developing imaging platforms to map the spatial organisation of genomic domains, such as topologically associating domains (TADs), within individual chromosomes. From the imaging of several tens of TADs in single chromosomes, her work showed that active chromatin and inactive chromatin tend to be organised into spatially segregated compartments in single autosomes (Wang et al., 2016). The work also showed that the active and inactive X chromosomes greatly differ in the organisation of chromatin arrangements (Wang et al., 2016). These results suggest an additional level of gene regulation mediated by the spatial localisation of chromatin domains.

Scott Fraser (University of Southern California, Los Angeles, USA) presented an account of the progress made in multimodal and multiplex molecular imaging of single cells. His talk highlighted the need to strike a balance between high-throughput analysis, multiplexing and spatial analysis, and focused on the most-recent developments of qHCR (quantitative hybridisation chain reaction). qHCR uses DNA probes carrying initiators that interact with fluorophore-labelled hairpins that assemble into fluorescent polymers. When the DNA probe hybridises to its mRNA target, the fluorescent signal scales in a linear manner, ensuring accurate quantitative analysis of mRNA expression (Trivedi et al., 2018). In multiplex experiments, one of the main hurdles in using multiple labels is the overlap between different spectra. As an improvement to the conventional Spectral Phasors (Fereidouni et al., 2012), which apply the Fourier transform to represent the spectrum of a pixel as a single dot on the phasor plane, Francesco Cutrale in the Fraser lab recently introduced HySP (Hyper-Spectral Phasor). HySP enables the unmixing of dozens of spectrally overlapping signals, with sufficient speed to be used for time-lapse imaging of zebrafish embryos (Cutrale et al., 2017).

The spatial contextualisation of single-cell data was also the focus of an interesting talk from Antti Lignell (University of Helsinki, Finland/Caltech, Pasadena, USA), who presented his data on the characterisation of the chick neural crest stem cell niche using Spatial Genomics Analysis, a newly developed image-based toolkit. By combining multiplex single-molecule fluorescence in situ hybridisation and cell segmentation based on a machine learning algorithm, he examined the simultaneous expression of 35 genes in vivo at single-cell resolution. This analysis allowed the identification of five different cell populations in the chick dorsal tube and the identification of a neural crest stem cell niche with a specific pluripotency signature (Lignell et al., 2017). Guo-Cheng Yuan (Harvard University, Boston, MA, USA) presented exciting and yet unpublished data from his lab focused on decomposing the cellular heterogeneity within the mouse visual cortex region. Using a new approach, which integrates sequencing and image-based single-cell transcriptome analyses, Yuan's lab identified previously unknown spatially associated cell populations within the astrocyte and glutamatergic cell compartments, suggesting unappreciated interactions between these cell types and their environment.

Lineage tracing represents an invaluable tool in the field of developmental biology, as it allows researchers to follow the fate of individual cells over time. At present, most lineage-tracing methods are limited to early developmental stages and small subsets of cells due to technical limitations. Jan Philipp Junker (Max Delbrück Centre for Molecular Medicine, Germany) presented LINNAEUS, a strategy for massively parallel lineage tracing and cell type identification at single-cell resolution (Spanjaard et al., 2018). The method relies on the CRISPR/Cas9 system to introduce double-strand breaks that, when repaired, result in unique genetic scars that vary in length and position, and act as heritable barcodes that are useful for lineage tracing. LINNAEUS was applied to reconstruct complex lineage trees by analysing gradual accumulation of genetic scars in the zebrafish embryo as well as in dissected organs in the adult. Simultaneous measurement of single-cell transcriptomes allowed cell type identification in addition to lineage tracing. In the embryo, genetic scars were introduced in the first 8 h of development, which enabled the detection of key developmental events such as gastrulation. Applying this method to haematopoiesis allowed the reconstruction of the entire haematopoietic lineage tree, including the detection of a separate branch composed of erythrocytes only, which possibly represents the primitive wave of haematopoiesis in zebrafish.

The convergence of recent advances in single-cell technologies and data analysis have prompted ambitious efforts to create reference maps of every cell within an entire organism over a defined developmental period. These so called ‘developmental atlases’ provide snapshots of the diversification of cell types at different stages of development and can be used to assess cellular heterogeneity and lineage relationships, and to deconstruct the pivotal steps in cellular differentiation and development. Four talks described efforts to generate developmental maps in Caenorhabditis elegans, zebrafish, mouse and human. Itai Yanai (New York University School of Medicine, USA), in his keynote lecture, presented the efforts of his lab in generating a transcriptomic map of C. elegans embryogenesis at the single-cell level. His data revealed the expression of all genes across over 100 types of cell during embryogenesis, and the analysis suggests that the C. elegans embryo is patterned by a juxtaposition of distinct lineage-specific gene regulatory programs, each with a unique encoding of cell location and fate. Daniel Wagner (Harvard Medical School, Boston, MA, USA) presented the assembly of a single-cell state landscape that delineates the first 24 h of zebrafish development; this covers key events, including axis patterning, germ layer formation and somitogenesis. His work includes RNA-seq data of more than 90,000 individual cells and captures the rapid increase of cell types in early vertebrate development (Wagner et al., 2018). This unbiased map of early embryonic development reveals a tree-like shape comprising three main branches representing the neural, epidermal and mesendodermal cell states. Berthold Göttgens (University of Cambridge, UK), reported an initial analysis of more than 90,000 single-cell transcriptomes from mouse embryos, generated as part of a UK-based consortium effort funded by the Wellcome Trust. Covering a window of development from E6.5 and E8.5, temporal analysis of the RNA-seq data from 350 embryos shows that from just a few cell types at the first sampling, 25 main cell clusters could be identified at the E8.5 stage. Efforts like these constitute an invaluable tool for the developmental biology community to use as references to develop new hypotheses and to better understand key events in early development. In 2016, the Human Cell Atlas consortium was established with the ambitious aim of mapping all the cells of the human body, both in the adult and during development (Regev et al., 2017). Sarah Teichmann (Wellcome Sanger Institute, Hinxton, UK), co-chair of the Human Cell Atlas organising committee, presented her exciting, and as yet unpublished, work carried out as part of the Developmental Human Cell Atlas initiative. Her talk focused on the characterisation of the maternal-foetal interface during early pregnancy. Single-cell RNA-seq of ∼50,000 cells from samples of foetal placenta, maternal decidua and matched maternal peripheral blood revealed three novel subsets of decidual natural killer (dNK) cells with different regulatory, inhibitory and cytotoxic capacity. Moreover, her lab developed a new repository of ligands and receptors, named ‘cellphoneDB’, and combined this with the single-cell transcriptomics data to predict the partners of the novel dNK subsets.

Seemingly homogeneous cell populations, which are conventionally identified by sets of surface markers, can often hide exceptional heterogeneity. This was nicely exemplified in a talk from Connie Eaves (BC Cancer Agency, Vancouver, Canada) in which she discussed using single-cell technologies to interrogate the molecular heterogeneity of human cord blood stem and progenitor phenotypes that display different cell outputs in vitro and in xenotransplanted immunodeficient mice in vivo. Focusing on the CD49f⁺ cell subset, which is enriched in long-term repopulating HSCs (Notta et al., 2011), her results revealed high molecular heterogeneity even within this cell compartment based on an unbiased mass cytometric quantification of 40 proteins in thousands of cells and RNA-seq and methylome analyses performed in hundreds of single cells. Importantly, this was confirmed by in vitro and in vivo analysis of different lineage outputs. Overall, the results showed that the conventional phenotypes used to subdivide the CD34⁺ (progenitor) population are both heterogeneous and overlapping in their molecular make-up and lineage output abilities. At the end of her talk, Eaves introduced a new complex topology of the human haematopoietic compartment that visualised the different trajectories that cells might follow in this system. The heterogeneity of the CD49f⁺ compartment was also the focus of a short talk from Serena Belluschi from Elisa Laurenti's lab (University of Cambridge, UK). Her research identified a molecular polarisation within the continuous transcriptional profile of human CD49f⁺ cord blood cells which correlated with functional outputs. She identified two cell subsets at opposite extremes, one multipotent and the other lymphoid committed and lacking erythroid differentiation capacity, thus showing that lineage restriction is already in place in the CD49f⁺ HSC compartment. In another short talk, Marijn van Loenhout (University of British Columbia, Vancouver, Canada) presented his research aimed at overcoming the challenges of expanding human HSCs in vitro. Focusing again on the CD49f⁺ compartment of cord blood HSCs, his research combined continuous live-cell microscopy and single-cell micromanipulation to obtain clonal lineages of thousands of cells for up to six generations in vitro. For each clone, data on the time of cell division, transcriptome and long-term culture up to 8 weeks were recorded and overall revealed similar patterns for sister cells and growing divergence with more distant kinship. Ana Cvejic (University of Cambridge, UK) discussed her research focusing on the understanding of haematopoietic lineage progression in zebrafish at the single-cell level. Ordering individual haematopoietic cells along their differentiation trajectory according to their transcriptional changes (without the use of cell-surface markers) revealed a gradual transition from multipotent to lineage-restricted cells. The computationally generated lineage tree highlighted a continuous differentiation path of haematopoietic cells. Interestingly, the population of HSCs and progenitor cells within similar transcriptional profiles showed considerable variation in the probability of differentiating into distinct cell types (Athanasiadis et al., 2017).

The nervous system is one of the most complex organs, comprising an astonishing number of highly specialised cells. The advent of single-cell technologies represents a new tool in the characterisation of this highly heterogeneous landscape. Sten Linnarsson (Karolinska Institute, Solna, Sweden) described the efforts of his lab in the identification and characterisation of the different cell types in distinct areas of the brain. Gene expression signatures were used to build a map of brain cell types, revealing a hierarchical taxonomy based on cell class, developmental origin and type of neurotransmitter. With this approach, hundreds of neuronal cell types were identified, including several previously unknown types of astrocytes. In addition, Linnarsson's lab have developed a method to quantify RNA velocity, which is a derivative of RNA abundance, an indicator of the state of an individual cell. RNA velocity can be measured by dividing spliced and unspliced transcripts, and enables one to extrapolate the future state of an individual cell in a timescale of hours. Aparna Bhaduri (University of California San Francisco, USA) gave an exciting talk focused on the development of the human cortex. Gene expression analysis of the developing cortex reveals a variety of spatiotemporal trajectories involved in neuronal differentiation (Nowakowski et al., 2017). In particular, across cortical areas, small transcriptional differences result in clearly distinct mature neurons forming the prefrontal cortex and visual cortex.

Our understanding of complex developmental processes has been advanced mainly at the cell population level. Now, single-cell technologies allow the detection of molecular changes in individual cells, bringing increased resolution to our understanding of development. A number of exciting talks described how single-cell technologies can be applied to answer developmental questions in different systems. Prisca Liberali (Friedrich Miescher Institute, Basel, Switzerland) gave an interesting talk on the application of single-cell technologies to understand how individual cells self-organise and break symmetry during the development of mouse intestinal organoids, despite the fact that they are all exposed to the same environment. A combination of single-cell imaging using light sheet microscopy and gene expression analysis was used to dissect the molecular mechanisms of intestinal organoid formation and to show that organoids develop by following a regeneration program. Furthermore, gene expression profiling during organoid development suggested that variability in expression of Yap1, which is a part of the Hippo signalling pathway, is required to induce symmetry breaking and to allow development of full organoids. Moving to a different tissue, Fiona Watt (King's College London, UK) presented the latest research from her lab on the characterisation of cellular state transitions in the mouse epidermis. The study focused on understanding the effect of indirect exposure of epidermal cells to Wnt/β-catenin signalling. Single-cell RNA-seq was performed in keratinocytes that were co-cultured in vitro with and without beta-catenin activated cells. The data revealed that a shift towards a proliferative stem cell-like phenotype originated from Wnt activation of neighbouring cells. In addition, Smad4 and Bcl3 were identified as the main transcription factors responsible, in a contact-dependent fashion, for cell fate transition, providing additional insights into how Wnt signalling coordinates self-renewal and regeneration in the mammalian epidermis (Ghahramani et al., 2018). Also using the epidermis as a model, Klaas Mulder (Radboud University Nijmegen, The Netherlands) developed RAID (RNA and immuno-detection by sequencing), a novel and unpublished method that relies on antibody-RNA conjugates to allow simultaneous detection of intracellular proteins and mRNA in a single cell. The Mulder lab generated a panel of 70 conjugates covering the major developmental signalling pathways, which revealed previously unknown signalling states involved in human epidermal stem cell differentiation. Dominic Grün (Max-Planck Institute of Immunobiology and Epigenetics, Freiburg, Germany) presented exciting data on the characterisation of γδ T cells, a rare population of T cells involved in both adaptive and innate immune responses. Using a combination of flow cytometry strategies to enrich for this rare population, and single-cell RNA-seq of thousands of murine thymocytes, his research revealed previously unknown cellular subtypes and identified key regulators of the differentiation of γδ T cells. Finally, Kirsten McEwen (Imperial College, London, UK), presented her research focused on network reconstruction from single-cell transcriptome data. Applying information theoretical approaches, which measure statistical relationships between pairs of genes, to a dataset of mouse pluripotent stem cells, this method confirmed known interactions and allowed the identification of novel possible interactions between genes involved in pluripotency networks.

A number of insightful talks focused on the close relationship between cancer and evolution, and how looking at these events at a single-cell resolution can improve our understanding of cancer biology. In his inspiring keynote lecture, Itai Yanai (New York University School of Medicine, USA) described recent unpublished research on the evolutionary aspect of cancer using zebrafish as a model system to follow tumour progression. Single-cell transcriptome and principal component analyses (PCA) on regularly sampled melanoma biopsies revealed a transcriptional program suggestive of four co-existing main states, or tumour archetypes, each characterised by the expression of specific genes and functions. Over time, the clones seemed to diverge towards these specialised states, suggesting that the cell population is adaptive towards the tumour archetypes. Alba Rodriguez-Meira, from Adam Mead's lab (University of Oxford, UK), introduced TARGET-seq, a yet unpublished method that enables the simultaneous detection of targeted mutations at the genomic DNA and cDNA level, coupled with whole-transcriptome analysis of the same single cell. TARGET-seq has a broad range of applications in cancer research, because it allows the distinction of normal and clonal cancer cells at the single-cell level. Philip Jones (Wellcome Sanger Institute, Hinxton, UK) presented his research on the study of the mutational landscape of normal epithelium as a tool to understand early cancer transformation. Ultradeep sequencing of 74 cancer genes in biopsies of normal sun-exposed skin revealed that the skin is a patchwork of pre-cancerous mutations. Interestingly, many cancer genes, including those that are drivers of carcinomas, were found to be under positive selective pressure even in normal skin (Martincorena et al., 2015). Itay Tirosh (Weizmann Institute of Science, Rehovot, Israel) shared his research on deciphering cellular heterogeneity in different types of primary and metastatic tumours using RNA-seq. His analysis revealed that, while non-cancer cells cluster together evenly across different patients, cancerous cells are much more diverse, not only between different patients but also within the same tumour (Puram et al., 2017). Finally, Dana Pe'er (Sloan Kettering Institute, New York, USA) described her latest and yet unpublished research, performed jointly with Ashley Laughney and Joan Massagué, on latent metastases of lung adenocarcinoma. Sequencing the transcriptome of around 50,000 individual cells from primary and metastatic adenocarcinomas allowed the team to build a landscape of the cell types of human lung adenocarcinoma. Interestingly, her research showed that most cells found in the tumour were of immune origin and cancer cells represented only a minority. In addition, data revealed that the progression of metastasis resembles proximal-distal lung development and mechanisms of regeneration.

Single-cell technologies have introduced a new level of understanding in biology and allowed researchers to investigate biological questions with new, unprecedented resolution. The Single Cell Biology conference was a success and gave a great example of the breadth of high quality research in different fields that is currently benefitting from progress in single-cell technologies. At the same time, interrogation of individual cells is revealing a previously unappreciated level of complexity in diverse biological systems. It is therefore an exciting time to work in single-cell biology and great advances, both in technological development and in biological understanding, are expected for the next meetings in the field.

This conference is funded and organised by the Wellcome Genome Campus, Advanced Courses and Scientific Conferences programme.