Identities, Characteristics, Environments, Interactions and Integrations

Technology development leverages recent advances in genomics, chemical biology, and nanotechnology to establish cross-disciplinary platforms for in-depth profiling and precise manipulation of cells and their interactions. By applying these platforms, we generate reference datasets (e.g., single cell responses to specific cytokines, pathways influenced by specific cellular interactions) that are essential for properly interpreting the drivers of complex systems-level behaviors.

Identities -- We develop single-cell genomic approaches to comprehensively profile complex biological ensembles. To date, the majority of our work has focused on establishing, validating, and scaling single-cell transcriptomics, often through the development of microdevices to enable genome-wide identification of the cell types/states that comprise functional or dysfunctional biological samples.

Most recently, we have developed Seq-Well, a portable, low-cost platform for high-throughput single-cell RNA-Seq (scRNA-Seq). By providing open access to resources and protocols, we hope to democratize access to cutting-edge approaches in single-cell genomics.

Characteristics -- To complement and inform the analysis of scRNA-Seq datasets, we create methods to simultaneously profile additional cellular characteristics of interest (e.g. genome, epigenome, or proteome), independently, or in combination with, scRNA-Seq. For a given technique or system, we ask what additional information would help us better interpret our scRNA-Seq results and develop methods to collect these data. These novel methods often map ancillary information into a DNA-based readout that can be coanalyzed with cellular mRNA or developing/applying microdevices. Recently, we have developed a method for integrated mRNA and protein detection that leverages proximity extension assays (Genshaft et al. 2016). To extract the information content from these novel datasets more effectively, we also formulate new computational methods and analyses.

Environments -- We explore how the extracellular milieu impacts intracellular decision-making by experimentally controlling the cellular microenvironment or leveraging naturally occurring sources of variation within a tissue. Here, we employ solutions that include controlled culture conditions with cells (Shalek et al., 2014) or organoids, chemical or genetic perturbations (Kumar et al., 2014), and constant microfluidic perfusion. We are also developing in silico approaches that are powered by in-situ cellular tagging techniques. In each instance, we aim to understand the degree to which extracellular environments modulate, and can be used to rationally control, the responses of individual cells or the overall distribution thereof, with an eye toward engineering ensemble responses.

Interactions -- We use microdevices, coupled with functional signal readouts, to create and study defined cell-cell interactions. By explicitly enumerating cell type, number, and additional functional properties (e.g., cytokine secretion), we model ensemble behaviors, looking for synergies and antagonisms­. These genetic signatures, along with those collected via our other platforms, provide a unique and essential reference for deconvolving behaviors in complex ensembles. We are also using genetic tracing strategies to examine differences between interacting and random cell pairs in vivo, and are developing computational methods (Tirosh et al., 2016) to identify putative interactions from scRNA-Seq data.

Integrations -- As the amount of data we have relating to cells, properties, surroundings, and interactions increases exponentially, we are motivated to develop pan-system measurements and analyses to paint comprehensive pictures of immune response in health and disease. Relying on massive transcriptomic datasets generated from complex tissues, like melanoma tumors, inflamed human gut, M. tuberculosis (MTB)-induced granulomas, and healthy or SHIV-infected monkey tissues, we have begun to construct social networks of integrated responses to physiological perturbations. The technologies outlined above uniquely enable us to generate foundational datasets (e.g., transcriptomes from interacting cell pairs) for deconvolving and interpreting the potential drivers of observed ensemble behaviors, as well as for identifying which properties we cannot explain, and thus need to study. To date, our lab has generated over 2 million single-cell transcriptomes across multiple tissues, individuals, and species; we are utilizing this data, paired with metadata and additional characteristics, to look for common cellular network motifs, such as division of labor, quorum sensing, persistence, or bet-hedging.