A Swedish-led team has published details on an in situ sequencing method that uses sequencing-by-ligation on short nucleotide fragments to track specific messenger RNAs in cell collections or tissue samples.
As reported in Nature Methods this past weekend, the researchers' method involves using padlock probes to detect mRNA transcripts of interest within the individual cells that make up a tissue sample. After letting the probes hybridize to target sequences, which leads to circularization of the DNA, the probes are amplified by rolling circle amplification and identified by using fluorescently labeled nucleotides to sequence four-base barcoded tags in each probe.
The in situ sequencing approach is compatible with either gap-fill padlock probes or barcode padlock probes, the group demonstrated, and can be used to detect the expression of dozens of different transcripts per cell — all while maintaining information about the location of each cell in the tissue sample.
"It's really massively parallel sequencing," co-corresponding author Mats Nilsson, a researcher with the Science for Life Laboratory based at Stockholm University and Uppsala University, told In Sequence. "It's a next-generation sequencing concept … The only thing is that we also give the spatial information in the sequencing reaction. So we can localize the RNA molecule."
In proof-of-principle experiments included in the new paper, for example, he and his colleagues used the method to find cells carrying mutated forms of the KRAS gene. They also applied the approach for assessing expression of multiple genes across cells in human breast cancer tissue sections, uncovering the sorts of co-expression profiles that may offer prognostic information or treatment clues in some cases.
The quest for methods to profile gene expression and DNA sequences in situ stems from researchers' desire to understand cells in settings that are as close to their natural contexts as possible, rather than looking at homogenized samples or single cells from samples lacking spatial information.
Consequently, several research groups have been trying to come up with approaches to accomplish in situ DNA sequencing and/or expression profiling. As part of the Archon Genomics X Prize competition, for instance, George Church and his colleagues are pursuing in situ sequencing approaches that bring together published DNA cloning and amplification schemes with fluorescent sequencing-by-synthesis (IS 10/9/2012).
Church is also working with a team using Illumina single-cell RNA sequencing to pluck individual cells out of their larger context as part of the ongoing effort to put together a three-dimensional map of gene expression in the human brain (IS 11/27/2012).
The type of fluorescent in situ sequencing schemes being developed for such applications are primarily focused on producing reads on the order of 30 base pairs each, Church told IS in an email message, with the goal of monitoring complete transcriptomes in spatially-defined cells.
In contrast, the method published by Nilsson and colleagues focuses on transcripts targeted by predefined probes, noted Church, who interacted with members of Nilsson's lab as they developed the newly described approach.
For the targeted in situ RNA analysis approach, researchers start by fixing a tissue sample or collection of cells to a microscope slide. They then convert RNA in these cells into complementary DNA so that gene transcripts and/or mutations of interest can be nabbed using one of two padlock probing methods: a gap-fill approach or a barcoded approach.
When using the gap-fill method, the team designs DNA probes that interact with bases flanking the targeted sequences. Then, when hybridization occurs, the target sequence is copied by a DNA polymerase and ligated to form a circular probe.
For the barcoding approach, on the other hand, the entire probe molecule gets produced synthetically. Those linear probes only become circularized when they encounter the target sequence of interest in a given cell, since sequences at each end of the probe interact with the targeted cDNA sequences.
Regardless of whether gap-filled or barcoded probes are used, interactions with the targeted cDNA of interest prompts the probes to circularize, he said. From there, the DNA gets amplified using a rolling circle amplification method before four- or five-base barcodes in the probes are sequenced with the help of fluorescently-labeled oligonucleotides.
"The next step is to perform next-generation sequencing chemistry on these rolling circle products localized in the tissue," Nilsson said. "The next-generation sequencing chemistry we used was sequencing-by-ligation — the same [chemistry] that Complete Genomics uses in their approach."
"We perform the very same kind of sequence experiment that Complete Genomics does," he said, "but we do it in a targeted way and we do it within fixed cells and tissues."
Using just one probe will provide information on the expression of a single transcript in cells from a tissue sample, while adding cocktails of probe molecules allows a look at the expression of several transcripts in parallel.
In the current study, for instance, the team started by showing that it was possible to use gap-targeted sequencing of a nucleotide quartet to correctly identify transcripts encoded by the human beta-actin gene ACTB.
From there, the researchers turned their attention to mixed cell line samples, using the in situ sequencing method to differentiate between cells with or without mutations in codon 12 of the KRAS gene.
With breast cancer tissue section samples, meanwhile, they highlighted the feasibility of finding cells that expressed either HER2 and/or ACTB, the actin control. Cells expressing the former gene typically turned up in cancerous tissue, while the latter gene cropped up in tumor cells and in cells from the neighboring stroma.
With sequencing-by-ligation chemistry, Nilsson noted that the in situ sequencing method is limited to five bases or less. Even so, he said, that combination of bases is enough to generate a wide range of barcodes.
Based on the combinations possible with four bases, for example, it should be possible to distinguish between up to 256 different barcodes by sequencing four bases. The number of barcodes jumps to more than 1,000 when five bases per probe are sequenced, Nilsson noted, and increases further still if researchers opt to do sequencing-by-ligation in both directions rather than one.
"From the decoding or encoding perspective this is unlimited," he said. "You can basically do this to as many transcripts as you wish."
In experiments included in the current study, the group came up with probes targeting dozens of different transcripts in breast cancer tissue, showing that it was possible to gauge expression of 39 genes in the same sample.
That set included transcripts for some of the same genes that are included on Genomic Health's OncoType DX diagnostic panel for breast cancer, Nilsson noted, pointing to possible clinical applications of the in situ RNA sequencing approach.
Because it allows individual cells in a tissue to be considered separately in a well-maintained spatial setting, such in situ sequencing approaches should make it possible to learn more about the biology of tumor samples and other tissue types.
In the case of gene expression-based diagnostic tests such as OncoType DX, for example, in situ sequencing may be used to reveal exactly which cells are expressing prognostically informative genes — be it in cancer cells, cells from the stroma, or both. And that, in turn, may eventually help to further refine the information that can be gleaned from the expression data, Nilsson predicted.
For in situ RNA sequencing experiments in their own lab, which are currently focused on breast, lung, colon, and prostate cancer tissue samples, Nilsson and his colleagues have upgraded their microscope's camera so that it scans more quickly during sequencing.
There are currently limitations to the number of transcripts that can be accurately detected in each cell, on the other hand — around 90 transcripts per cell, in an average sample. That's because "we cannot squeeze more sequencing substrates into a cell than this," Nilsson said.
At the moment, researchers are also relying on a fairly simple fluorescence microscope system that lacks automation. "We use a fluorescence microscope as the scanner," Nilsson said, "but we don't have a liquid handling system for executing the wet part of the sequencing reaction."
"It's not that important, since you only read four or five bases, so we only have to do the sequencing chemistry four or five times," he added.
Nevertheless, Nilsson said he is keen to see a commercial version of the in situ RNA sequencing system at some point. It remains to be seen whether his group will try to develop such an instrument or whether they may try to work with an existing sequencing vendor to produce a platform that can do the analysis routinely.
"I really think this is a useful approach, but we haven't really started looking into how to commercialize this," Nilsson said.
Compared with existing sequencing instruments, he predicted that a commercial platform capable of performing in situ RNA sequencing would likely need slightly modified flow cells as well as optical systems that can accommodate the sometimes uneven terrain of tissue sections.
"You have some topology on tissue sections," Nilsson noted. "They are not entirely flat, unlike the flow cells of next-gen sequencing instruments."
"With tissue you would have to work a little bit on [optical] focusing," he added, "but this is not really rocket science, I think."
More generally, Nilsson said that it would theoretically be possible to adapt the approach so that it uses sequencing-by-synthesis chemistry rather than sequencing-by-ligation.
"That really should work, but we don't have it in hand," he said. "What we do have is the sequencing-by-ligation and that works very robustly."