DNA microscopy: Optics-free spatio-genetic imaging by a stand-alone chemical reaction

Background
The word “microscope” was coined in 1625, to describe a device with a glass lens made by Galileo. The centuries since have seen considerable improvements, with dramatic increases in the resolution at which cellular structures can be resolved. Nevertheless, the fundamentals are relatively unchanged: these techniques still involve bombarding a sample with particles (photons or electrons) and observing the result of their interactions using a series of lenses. By reconstructing the positions from which photons have emerged, we can understand spatial organisation within the sample.

Can DNA sequencing provide an alternative technology? In recent years DNA has begun to be used for far more than simply understanding the natural genetic code of living systems. DNA barcodes are now used in chemistry laboratories to uniquely tag billions of chemical compounds, and as a deterrant marker for forensics. In biology, such genetic barcodes have been used in high-throughput screens, and even as a ticker tape to record the developmental trajectories of cells.

In this preprint, the authors develop a new method to encode data on the spatial relationship between molecules into sequences of nucleotides, and use this information to do away with the need for an optical microscope, creating images with a DNA sequencing machine alone.

Key findings:

To be able to record spatial information within sequences of DNA the authors employ an innovative approach based on overlap-extension PCR. They first label molecules of interest (in this case RNA transcripts inside fixed cells) with short sequences of DNA. These DNA labels have a degenerate section which serves as a unique identifier for each labelled molecule in downstream analysis.

The labelled DNA is then amplified, producing many copies of these molecular identifiers in the local area of the original label. But these amplified molecules are not static – they drift around by diffusion, and it is this process that allows spatial information to be recorded. As the molecules drift, they encounter amplicons from other nearby labelled molecules, and complementary 5′ sequences on their primers cause the amplification of a concatenated version of the two initial sequences. These concatamer sequences record in a single piece of DNA the molecular identifiers from the two molecules. The researchers create a hydrogel in the sample which slows down the diffusion process so that concatameters will predominantly form between amplicons from nearby molecules. Each concatenated sequence also gets its own unique molecular identifier, created from degenerate sequences in the amplification primers. As molecules drift they create concatamers with molecules further and further from their initial location until the cycles of the PCR reaction are complete, and sequencing libraries are prepared from the concatenated molecules.

These sequencing libraries are now read using an Illumina sequencing machine. Inspecting this data allows a long list to be made of the molecular identifiers of all the molecules that were labelled, and the type of probe from which they originated, and next the number of contamers formed between each pair of molecules can be counted. In practise the number of such potential pairs is so vast that the researchers adopt sophisticated approaches to streamline their analyses, but their data essentially takes the form of a vast matrix recording which molecules are close to each other, rather like a table of driving times between cities.

 

 

From such a table, one could try to reconstruct a map of a country, and similarly the researchers analyse their data to reconstruct a 3D image of the positions of molecules in the cells they are studying.  In this experiment they looked at a co-culture of cells expressing RFP and GFP, and by visualising the position of RFP transcripts in red and GFP transcripts in green they form an image reminiscent of microscopy, but created entirely from DNA data. The mutually exclusive positions of GFP and RFP seen across cells demonstrates the success of their technique, and shows that this first demonstration has resolution sufficient to clearly resolve cells. They also directly compare their data to a light microscopy image and see similar cellular arrangements.

 

Figure 5E from the preprint, showing a DNA-microscopy image with GFP transcripts labelled in green and RFP transcripts labelled in red.

 

What I like about this preprint
This work opens up a range of opportunities for investigating spatial properties of biological systems in new ways. Light microscopy is fundamentally limited in how many channels can be imaged at the same time. Although multiple fluorophores can be used simultaneously, spectral overlap limits experimenters to using a handful in each experiment. By contrast, a 10 base pair sequence of DNA could potentially encode a million “channels”, in an approach like this. The experimenters demonstrate this scalability by resolving 20 different transcripts in parallel.

The preprint also does a good job of acknowledging current limitations of this technology. Whereas optical microscopy finds it easiest to resolve sources of material which are sparse and well-separated, DNA microscopy can actually struggle to resolve empty space, since this can only be inferred by reference to labelled molecules. The authors also highlight that the technique need not be limited to RNA transcripts. Almost anything can be tagged with a piece of DNA, whether using DNA-tagged antibodies or aptamers.

 

Future directions
Many of the images obtained in this work could have been acquired with a light microscope using fluorescent in-situ hybridisation, likely with higher resolution. The preprint presents strong proof of a concept and it will be exciting to see what refinements can be made to the technology, and what new biological questions can be asked using these approaches to look at a wider range of probe molecules.

Interestingly, in the same week that this work was posted on bioRxiv, two similar preprints were also posted. “From space to sequence and back again” uses ligation rather than overlap-extension and recovers patterns from 2D surfaces, and  “A Computational Framework for DNA Sequencing-Based Microscopy” presents a computational method, also described in a 2D setting, which uses a similar chemistry as the amplification reaction in Illumina sequencing. With a wide range of exciting work being carried out, this nascent field looks like one to watch.

 

Questions for the authors

1. In traditional light microscopy there is a fundamental limit to resolution associated with the wavelength of light used. Do you have a sense of what would drive the analogous limits for DNA microscopy, and what their order of magnitude might be?
2. In what areas of biology do you think DNA microscopy will first have tangible practical applications?

 

Posted on: 28th November 2018

doi: https://doi.org/10.1242/prelights.5903

Read preprint (4 votes)