Technology has expanded the scope of sequencing to encompass entire genomes or transcriptomes, and pushed sequencing resolution down to the level of single cells. What has been missing is spatial context—the ability to 'read' sequence directly from intact tissue. In situ sequencing could provide a deeper understanding of the relationship between a cell's genotype or gene expression program and its morphology and local environment. Although it is too early to predict the ultimate form and potential of such a technology, some initial steps towards in situ sequencing have already been taken.

In situ sequencing maps sequence data directly onto morphology. Credit: Katie Vicari

Methods such as in situ hybridization have long been used to pinpoint sequences in intact cells and tissues, but they are limited by the need to know the target sequence. On the sequencing side, many technologies use a light-based readout, suggesting that they could be compatible with imaging in intact tissues. But amplification and sequencing reactions require a special substrate or need to be separated from each other in an emulsion.

An alternative strategy is rolling circle amplification, which has been used to boost signal for in situ genotyping in cells. The method relies on a 'padlock' probe that hybridizes on either side of a target sequence to form a circular template that can be copied repeatedly as a long string. Because the product is tethered to the template, it provides reliable localization and is amenable to in situ sequencing by successive rounds of ligation-based oligonucleotide probe incorporation (Nat. Methods 10, 857–860, 2013).

This form of in situ sequencing was used to interrogate four-base sequence variation in transcripts directly in tissue sections; however, it is a targeted approach that still requires knowledge of flanking sequence. It will be important to increase the number of transcripts and length of sequence that can be interrogated in a single cell, as well as balance image resolution with the ability to image tissue-wide. Image registration methods are a key component of this scaling up, and tools to quantify associations with morphology will be needed.

The ultimate goal is to sequence many loci—and even a significant fraction of the transcriptome or genome—but this will require a breakthrough in solving the signal-density problem: cells contain too much information to be imaged in such a small space. We look forward to further steps that join sequencing with biological context.