NEW YORK – Researchers from the Broad Institute, Massachusetts Institute of Technology, and Harvard University have improved an in situ RNA sequencing method to provide higher resolution and enable both targeted and untargeted studies.
Dubbed expansion sequencing (ExSeq), the technique combines fluorescent in situ sequencing (FISSEQ), developed in George Church's lab at the Wyss Institute, and expansion microscopy, a technique from Ed Boyden's MIT lab that swells samples in order to separate molecules for easier detection while preserving the spatial relationships between them.
"When you separate these molecules in the expanding sample, and move them away from each other, that gives you more room to actually perform the chemical reactions of in situ sequencing," Adam Marblestone, a neuroscientist now at the Federation of American Scientists who is a senior author of the paper, said in a statement. In addition, higher resolution allows spatial mapping at subcellular scales.
The researchers described ExSeq in a paper published Thursday in Science, providing proof-of-concept studies and comparisons with RNA-seq and FISSEQ. The paper contained data from mouse brain, roundworm, fruit fly, and HeLa cells. Using an untargeted version of the technique, the researchers said they could sequence between 20 and 50 percent of the genes expressed in a sample.
Many of the study's authors are also inventors on three US patents covering expansion microscopy and ExSeq. Boyden is cofounder of and consultant to Cambridge, Massachusetts-based Expansion Technologies, a spinout aiming to commercialize expansion microscopy and develop tools for researchers. "We see expansion sequencing being part of that," said Jamie Zhan, the firm's head of strategy and business. The firm has licensed a family of patents from MIT related to expansion microscopy, including the three covering ExSeq.
As of press time, the study authors had not provided comment on their work.
ExSeq starts with FISSEQ, a method published in 2014 that conducts the RNA-seq chemistry within a tissue. Sequencing is often done by ligation but can be performed with other chemistries. FISSEQ is the basis for the technology underlying ReadCoor, a Church lab spinout that was acquired by 10x Genomics last year for $350 million in cash and stock. FISSEQ for genomic DNA has also been paired with Illumina short-read sequencing in a recent study of chromosome organization in situ.
ExSeq builds on FISSEQ by adding Boyden's method for expanding tissue samples to better image them. Embedding water-absorbent polymers into a tissue sample can expand the volume one hundred times or more while preserving the overall organization of cells and molecules. Samples are fixed and RNA molecules are chemically anchored to the gel, which is then expanded.
One challenge the researchers faced was that the expansion chemistry resulted in charged carboxylic acid groups throughout the gel, which suppressed FISSEQ's enzymatic reactions. They overcame this by re-embedding samples in uncharged gels and then chemically treating them to reach a neutral environment.
ExSeq comes in two flavors, untargeted and targeted. For the untargeted version, transcripts are amplified and converted to cDNA that form nanoballs containing many copies of an RNA sequence.
The authors noted that in the past, in situ sequencing methods had been limited to reads of about 30 bp, which are difficult to align to a genome. To address this, the team processed the sample to capture amplicons and resequenced them outside the sample with Illumina technology. "The random nature of untargeted sequencing results in the creation of distinct molecular identifiers from the in situ sequenced region of the amplified cDNA," the authors explained. The different read types were matched, "augmenting the effective in situ read length," the authors said, however, they did not provide a measurement for how long those reads were.
In addition to longer effective read lengths and higher resolution, ExSeq has more advantages. "In FISSEQ, highly abundant genes were underrepresented — for example, genes involved in translation and splicing," the authors wrote. "By contrast, we did not observe this detection bias with ExSeq."
The method also compared favorably with standard bulk RNA-seq approaches. "The expression levels of well-annotated genes (genes from the RefSeq database) using RNA-seq and ExSeq were highly correlated," the authors wrote. And while most of the RNAs were ribosomal RNAs, ExSeq produced a slightly higher percentage of coding RNAs: 4 to 9 percent versus approximately 2 percent with RNA-seq.
The targeted flavor of ExSeq uses oligonucleotide padlock probes bearing barcodes to hybridize to transcripts. Following circularization and rolling circle amplification, the barcodes are sequenced in situ. "The inefficient reverse transcription step required by untargeted in situ sequencing is circumvented by the binding and ligation of padlock probes on each targeted transcript using PBCV-1 DNA ligase (also known as SplintR ligase)," the authors wrote. "This enzyme can ligate DNA on an RNA template ~100× faster than T4 DNA ligase."
In HeLa cells, targeted ExSeq had detected approximately 62 percent of mRNAs relative to expansion microscopy fluorescence in situ hybridization, which has a detection efficiency of approximately 70 percent in tissue, the authors said.
In one proof-of-concept study for targeted ExSeq, the researchers designed a 42-gene panel marking excitatory and inhibitory neuron types in a mouse model. The spatial distribution of reads "recapitulated spatial distributions in the Allen [mouse brain] in situ hybridization atlas," the authors said, and "transcripts known to express in the same cell type appeared in similar positions."
They were also able to show "nanoscale RNA compartmentalization" in mouse hippocampal neurons. “We know that the location of RNA in these small regions is important for learning and memory, but until now, we didn't have any way to measure these locations because they are very small, on the order of nanometers," co-first author Shahar Alon said in a statement. The team found mRNA containing introns, normally edited out of mRNA, in dendrites, the branchlike extensions of neurons. And they discovered mRNA molecules encoding transcription factors in the dendrites, which may help with novel forms of dendrite-to-nucleus communication.
"These are just examples of things that we never would have gone looking for intentionally, but now that we can sequence RNA exactly where it is in the neuron, we're able to explore a lot more biology," co-first author Daniel Goodwin said in a statement.
In addition to neuroscience, the method could be applied in immuno-oncology. The researchers looked at breast cancer metastases, revealing that expression in B cells varied depending on their location within a tumor.
"The tumor microenvironment has been studied in many different contexts for a long time, but it's been difficult to study it with any depth," co-first author Anubhav Sinha, a doctoral student in Boyden's lab, said in a statement. "A cancer biologist can give you a list of 20 or 30 marker genes that will identify most of the cell types in the tissue. Here, since we interrogated 297 different RNA transcripts in the sample, we can ask and answer more detailed questions about gene expression."
Expansion Technologies remains an "early-stage company," Zhan said, and is still working on R&D, including optimizing the technology on its own and improving imaging data analysis. The team is "small," but she declined to disclose the company's headcount as well as how much fundraising it has completed. Zhan said the company's focus is on evaluating "high-value applications of the expansion microscopy family of technologies." Once those applications are identified, the firm will "optimize it for scale and unleash it for a broader audience," she said. "We're always looking for partners, in different forms," she noted.
ExSeq could be useful in any field where different cell types make up complex tissues, including developmental biology, immunology, and aging, the authors said, as well as any applications of indexed RNA barcodes that are not conducive to a FISH approach.
They noted they plan to develop the method to also detect proteins — potentially with tagged antibodies — as well as nucleic acids, including endogenous DNA.