Cellular differentiation is the process by which a clonal cell population organizes into a collection of distinct yet interacting states (commonly referred to as “cell types”) and is a canonical feature of higher organisms. While we accept that cells containing the same genome can occupy different states in complex “multicellular organisms” such as humans, it is important to remember that differentiated cell types have also been observed in clonal bacterial populations using the original “single-cell” tool – the microscope. In fact, some of the founders of microbiology,  including Robert Koch and Ferdinand Cohn, discovered morphologically distinct cell types, such as spores, in the late 1800s. Similarly, intricate multicellular structures were observed in social bacteria like myxobacteria by Roland Thaxter and others. In recent years, studies of bacterial cell-cell variability have relied heavily on fluorescent reporters and molecular markers to differentiate between isogenic bacterial cells that occupy different cell states. Using these methods, researchers have made great strides into discovering that clonal bacterial populations can contain cells that display distinct physiological traits involving metabolic processes, structural arrangements, antibiotic resistance, and many other biological processes.

While the use of reporters has enabled fundamental measurements defining cell type features (“markers”) in both prokaryotic and eukaryotic organisms, one area in which bacterial research has been lagging behind eukaryotic studies in the past few years is obtaining high-dimensional transcriptional data from individual cells. In mammalian systems, high throughput single-cell RNA sequencing (scRNA-seq) has been used to identify new cell types in a robust and unbiassed way, catalogue cell-atlases of whole organs or even entire organisms, and identify the developmental trajectories that cells undergo during differentiation and in diseases like cancer. A similar approach for bacteria, where investigators can obtain transcriptome-wide information of individual cells across thousands of cells within a population could be transformative for microbiology. While working on our own approach to solve this problem, several other groups developed techniques to address this missing piece in the toolbox of microbiological research. The group of Jörg Vogel (Helmholtz Institute) sorted individual bacteria into 96 well plates to obtain high quality transcriptional data from several hundred cells. Shortly after them, the groups of Georg Seelig (University of Washington) and Saeed Tavazoie (Columbia University) used a multiplexed approach that relies on combinatorial barcoding to measure single-cell transcriptomes across thousands of cells.

While both of these approaches offer welcomed advances to the field of bacterial scRNA-seq, we reasoned that we can design a microfluidic method to measure bacterial transcriptomes in a completely different way – circumventing several of the technical challenges posed by working with bacteria and providing an easy to use assay that achieves high per-cell transcriptional coverage. To do so, our group chose to rely on large pools of DNA probes designed to cover the complete transcriptome of bacteria of interest. This approach, analogous in some ways to microarrays, allows designing multiple probes for a given gene, which increases the chance that lowly expressed transcripts will be detected. In addition, since probes are not designed against rRNA targets, we avoid sequencing uninformative RNA molecules, and dramatically reduce the cost of the experiments.

Our method, called ProBac-seq (probe-based bacterial single-cell RNA sequencing) works as follows: bacterial cells are first fixed to retain their RNA. This is to avoid RNA degradation through turnover which happens at a significantly higher rate in bacterial cells compared with eukaryotic cells. After fixation, the cell walls of the bacteria are permeabilized to allow probes to penetrate the cells. Next, probes are added and allowed to hybridize with their target mRNA. Unbound probes are washed off in a series of washes. Notably, all of these initial steps are applied to the bacterial population in bulk (i.e. one tube). Finally, cells containing their payload of probe-bound transcripts are microfluidically encapsulated into individual droplets using a commercial system commonly used for mammalian single cell encapsulation (the 10X Genomics Chromium Controller). In these microfluidic droplets, our probes bound to their target transcripts are themselves captured and tagged with cell-identifying-barcodes. The resulting libraries of cell-barcoded probes are sequenced to obtain the entire repertoire of probes from each single cell.

Using this approach, we were able to capture highly resolved full transcriptomes of cells from well-known cellular states in bacteria, such as sporulation, and also identify new cell states for biological processes that were not known to be conducted in specialized subpopulations of cells – such as a cell state for arginine synthesis. To demonstrate one potential use of this new technology, we tracked toxin production in the enteric pathogen Clostridium perfringens and showed that a subset of cells within a culture of C. perfringens differentially upregulates toxin expression. Importantly, having a transcriptome-wide view of this subpopulation, as well as the other subpopulations that had substantially lower expression of the toxin, allowed us to predict that a metabolite (acetate) can be used to favor the cell states that produced low levels of the toxin. Indeed, when we added acetate to the culture, we observed that overall toxin levels are reduced, and this reduction owes to both a reduction in the number of cells in the toxin-producing state, and an overall decrease in transcription of the toxin-encoding gene within these cells. The ability of our method to identify a cellular state of interest and predict environmental perturbations that can alter cellular behavior is an exciting application that, we believe, will enable discoveries in other infectious diseases.  

While our probe-based approach to prokaryotic single-cell RNA sequencing requires an initial investment in probe design and amplification, probes can now be ordered for a relatively inexpensive price from DNA synthesis vendors such as TWIST Biosciences, IDT, and Genscript. Once ordered, probes can be amplified cheaply and indefinitely to run additional experiments. Using this approach, once a probe set exists and an in-situ hybridization protocol is optimized, experiments for a given bacterial species can be run quickly and easily. Because probes increase the chances of capturing transcripts with low expression, we hope that this approach will be useful for understanding heterogeneity in natural environments where cells are generally less active, and thus contain fewer transcripts. Altogether, our approach to prokaryotic single-cell RNA sequencing offers a high throughput, high resolution, and low cost method for comprehensive studies of transcriptional heterogeneity in bacteria.