Advances in single-cell genomics are pushing forward our understanding of gene expression and regulation. New developments include progress in single-cell RNA-seq for profiling RNA and ATAC-seq for profiling DNA. In addition, new computational approaches are assisting the analysis of single-cell nucleic acids, as well as facilitating predictions of intercellular interactions. Here is a look at some of the developments in single-cell genomics that were discussed at the Single Cell Genomics Day Workshop at the NYU Center for Genomics and Systems Biology earlier this year.
Modalities informing each other
Emerging single-cell multiomics are facilitating both a higher throughput and a richer degree of information by pairing RNA and DNA data. One technique, known as SNARE-Seq, profiles DNA and RNA from the same nucleus by performing snRNA and ATAC-seq simultaneously in high-throughput. This is accomplished by using a “splint oligo” that captures both the DNA and RNA on a bead, which is labeled by a molecular barcode linking the two modalities. A similar technique, known as Paired-Seq, uses a different method to analyze RNA and DNA from single cells in high-throughput. Instead of droplet microfluidics, Paired-Seq performs ATAC-seq and RNA-seq in multiwell plates using primers with the same barcodes for each cell. After analysis, the RNA and DNA modalities can be linked by their barcodes.
An advantage of multiomics is that the information learned about one modality can help you to improve your analysis of the other. Indeed, the ability to see whether changes in chromatin accessibility and in transcription are happening together will help researchers to understand gene expression going forward. Triple-omics profiling—analyzing RNA expression, chromatin accessibility, and DNA methylation—has also emerged, but is not yet as high-throughput as SNARE-Seq or Paired-Seq.
Saturation loading for drop-seq
Methods to increase the number of sequenced cells are evolving from drop-seq, a technique that isolates single cells using microfluidics for highly parallel analysis of numerous individual tiny droplets; droplets are diluted such that each contains either zero or one barcoded cell, and rarely two cells. Typically, the RNA from each cell is analyzed using barcoded primers. Drop-seq efficiently isolates single cells, but the genomic analysis can incur a cost per droplet regardless of cell number, which can add up quickly.
New methods have adapted the drop-seq technique to result in no empty droplets. The dsciATAC-seq method measures chromatin accessibility without empty droplets, using drop-seq combined with combinatorial barcoding. A similar method for RNA profiling, a technique known as scifi-RNA-seq, uses multiple rounds of cell barcoding and a creative addition of a scATAC-seq kit to generate scRNA-seq profiles with no empty droplets. By adding combinatorial indexing to droplet microfluidics, these new methods can increase throughput of future gene expression studies.
Genotype-free multiplexing
These methods are good news for researchers studying human genetic variation. Pooling barcoded cells, then “demultiplexing” them with computational methods, can help to support larger scale experiments. Sequencing the DNA from hundreds or thousands of people is especially valuable for studying rare genetic variations. Usually these studies require subjects to have been previously genotyped, which can become expensive, but new methods are overcoming the need for prior genotyping.
Vireo is a computational method that uses single-cell RNA-seq data to differentiate cells from about eight different people. This allows researchers to pool tissue samples from multiple people without prior genotyping. The soupercell and scSplit methods both use statistical modeling and demultiplexing to detect cells of different genotypes. Further progress in these areas will facilitate greater multiplexing of human genetic studies.
Prokaryote and microbial sequencing
Single-cell sequencing of bacteria and prokaryotes has revealed different levels of activation of various metabolic pathways. However, these cell types have characteristics that make profiling nucleic acids difficult, such as a low mRNA content, lack of poly-adenylated tails on mRNA (which prevents the use of convenient tools), and thick prokaryotic cell walls that are difficult to lyse.
Two new methods aim to overcome these kinds of hurdles. PETRI-Seq uses split pool combinatorial barcoding (an adaptation of the SPLiT-Seq method) and gains access to cell interiors using optimized fixation and permeabilization. Instead of relying on poly(A) primers, the method uses random hexamers for reverse transcription. Another new method, microSPLIT, first treats the mRNA with poly(A) polymerase (essentially adding poly(A) tails), then performs reverse transcription using random hexamers and poly(A) primers. Along with methods such as these, better sequencing tools may help researchers to understand the sources of observed heterogeneity in prokaryotic gene expression.
Increased sensitivity
New methods are making single-cell genomics more sensitive, allowing researchers to detect and sequence more genes. The SMART-Seq3 method can profile more than 100,000 molecules from individual cells, increasing sensitivity by optimizing reverse transcription and adding a 5’ UMI. The method can sequence almost 4,000 genes from individual human T cells derived from blood samples, although it is currently a lower throughput protocol. Another method, Seq-Well S^3, increases sensitivity by capturing more of the molecules that are normally lost during second strand synthesis. The S^3 method is high-throughput, using barcoded beads on the Seq-Well platform.
Cellular interactions
Computational advances in single-cell genomics are also shedding light on cell-cell interactions. The CellPhoneDB method aims to predict which cells were communicating with each other, based on RNA-seq data from a mixture of cells. The method uses a repository of cellular interactions consisting of known ligand-receptor pairs, and secreted and cell surface molecules. CellPhoneDB identifies patterns of expression of both ligand and receptor in specific cell types, which suggests an interaction between those cells. It uses a permutation-based statistical framework to predict cell-cell interactions and communications.
Another method, NicheNet, also uses a repository of cellular interactions, including associations of ligand-receptor pairs with their downstream signal transduction and gene regulatory networks. NicheNet predicts cell-cell interactions, as well as influences from the environment, such as effects of ligands, identities of signaling molecules, and downstream targets.
By linking information about DNA, RNA, and sometimes other information such as DNA methylation, all from individual cells, single-cell multiomics can help us understand which genes are regulated by particular sequences, in response to which environmental factors, and possibly other interacting cells. As single-cell genomics continues to advance, we are likely to see the possibilities expand, the pace quicken, and our scope of understanding broaden.