Skip to main content
Premium Trial:

Request an Annual Quote

Single-Nucleus Sequencing Holds High Risk, High Reward for Translational Transcriptomics


NEW YORK – In 2017, Luciano Martelotto was trying to solve a riddle. He wanted to sequence RNA in single pancreas cells, but the organ, which produces digestive enzymes, was filled with RNases that could ruin the very molecules he was hoping to study.

Cells, he decided, would not do, but isolated nuclei might. 10x Genomics had a so-called "demonstrated protocol" for isolating peripheral blood mononuclear cell nuclei, so he tried throwing his pancreas samples into it.  

"It was a bust, a disaster," Martelotto said. Most of the samples failed outright, and even in ones that yielded data, there was a lot of background RNA, which he attributed to RNA that had leaked out of the frozen cells and been converted to cDNAs in the 10x droplets. "That was $20,000 in two days and nothing, just nothing," Martelotto said. Moreover, the samples were part of a patient study and irretrievable. "Those things I have nightmares about," he said.

Undaunted, Martelotto came up with his own protocol, affectionately known as the "Frankenstein protocol," as it borrowed bits and pieces from other methods. It was ungainly, "to the point that 10x did not really believe it would work," he said. But it did.

"We went from not getting anything to profiling 100 tumors with 95 percent success," he said.

Now, maturing sample preparation methods are delivering more reliable single-nucleus RNA sequencing (snRNA-seq), letting researchers analyze cells in previously intractable tissues. Single-cell analysis specialists like 10x Genomics have embraced single-nucleus methods; some products, like 10x's ATAC-seq assays, even require nuclei, rather than cells. In addition to PBMCs and embryonic and adult mouse brain, 10x customers have published snRNA-seq protocols and studies for a wide range of sample types, including human brain, kidney, and heart, a company spokesperson said in an email, as well as several tumor types.

And startups such as Dolomite Bio and LevitasBio are keying in on nuclei isolation or cleanup as applications for their hardware.

These developments come as single-cell methods are increasingly applied in translational or clinical research.

"We clearly caveat that if someone can use whole-cell transcriptome sequencing, then they should," Thomas Ayers, senior technical applications specialist at Blacktrace Holdings, Dolomite's parent company, said in an email. "Single nucleus sequencing really becomes applicable when the user cannot for whatever reason isolate their cells without damaging them," such as with fixed or frozen samples, he added.

Even Martelotto, one of the foremost advocates of snRNA-seq in his new position as scientific director of Harvard Medical School's single-cell core, doesn't recommend it for every study. But for tissues that are hard to dissociate, or for samples with an unknown preservation history, "what you do is hit it with lysis, get the nuclei, and away you go," he said.

A nucleus contains only about 1 percent of total RNA inside a cell, but single-nucleus transcriptomics works because that 1 percent is enough to detect the majority of cell populations in a sample. A 2013 study from researchers at the J. Craig Venter Institute and the Salk Institute was the first to hint that snRNA-seq was feasible as an alternative to a whole cell.

Single-cell methods took off in the next few years, but it took until 2017 for snRNA-seq to arrive. In addition to Martelotto's Frankenstein protocol, that year researchers at the Broad Institute from the lab of Aviv Regev, now head of Genentech Research, published DroNc-seq, a modified version of Drop-seq for nuclei.

Single-nucleus methods aren't without drawbacks. Cell types that naturally have low RNA will be less represented. And the inherent riskiness means researchers should do a lot of due diligence on protocols first.

But in some cases, especially cryopreserved or post-mortem tissues where some cells die or burst before others, analyzing nuclei leads to a less biased experiment, offering a more complete picture of the cell population at the time of sample collection.

Even with tissues that can be prepared for single-cell sequencing with enzymes, "everything the cell feels is going to be represented in the transcriptome which is what we're using as cell readout," Martelotto said. "Then our data becomes pestered with a lot of signals that have nothing to do with the biology at the time of harvesting those cells. That's why the less steps you do in the preparation, the better."

Debris remains a problem and Martelotto tries to remove as much as possible. The Frankenstein Protocol uses fluorescence-activated cell sorting (FACS) to pull down nuclei. Magnetically labeled separation also works, as does using a gradient; however, magnetic enrichment is not as precise as with FACS and samples with low cellularity don't lend well to gradients, as too many nuclei are lost.

FACS works well, but is expensive, and emerging technologies are targeting the snRNA-seq sample prep workflow. LevitasBio, a Stanford spinout that is commercializing an instrument for sample preparation based on the principle of magnetic levitation, recently put together a "crash program" to come up with a way to enrich for nuclei, CEO Martin Pieprzyk said. Getting nuclei is the "number one bottleneck for everyone we speak with," he said.

Their single nucleus assay, which the company plans to launch this quarter, allows the nuclei to drop out of the rest of the debris. The cartridges will cost a little bit more than their other products, Pieprzyk said, but declined to provide the price.

Timothy Semple, a single-cell sequencing expert at Australia's Peter MacCallum Cancer Centre, was able to run LevitasBio's nuclei workflow on a loaner instrument, to compare it to the Frankenstein protocol. "The interesting thing for me was that everything sank — nuclei and debris. Maybe I left the system separating for too long … In the future we'll likely run the separation just long enough to see most of the debris fall to the bottom of the chamber and then collect," he said in an email. "I really like the system, we think it has a ton of potential, we just need to better understand its limitations." He noted that the distributor placed the instrument at another research institute before he was able to run a second attempt.

Dolomite's "sNuc-Seq" protocol is based on DroNc-Seq, but "has undergone plenty of modifications to make the workflow shorter and more ergonomic," Ayers said.

The preparation runs on the firm's Nadia instrument, which is able to produce up to 6,000 sNuc-Seq libraries per sample in a 20-minute run, he said. Cartridges can run up to eight samples in parallel, or up to 48,000 single nuclei per run. "The gene capture in sNuc-Seq is expected to be lower than that in scRNA-Seq at a given read depth," Ayers noted. "In the case of low gene expression, we often advise our customers to use the pressure functions of our machines to make smaller droplets, which increases local mRNA concentration to improve library yield."

Dolomite bundles its sNuc-Seq kits with its scRNA-Seq kits, for one cartridge that can do both. "We've seen a steady increase in sales of these kits since their release in 2018, but it's hard for us to tell how many users are using them specifically for nuclei rather than whole cells since they are multifunctional," Ayers said.

Again, frozen and hard to dissociate tissues  are the main applications of snRNA-seq. Neurobiology is an important application area, since neurons are difficult to isolate as whole cells. Single-nucleus methods could be used instead of pooled methods, such as nuclear-enriched transcript sort sequencing, developed at Rockefeller University.

A benchmarking study from several researchers associated with DroNc-seq, published in Nature Biotechnology in June 2020, noted that skeletal muscle and adipose tissues could also benefit from the method. And in a study being run at the Peter MacCallum Cancer Centre, researchers are using snRNA-seq on rapid autopsy samples from cancer patients, which could be held under a variety of time and temperature conditions that could complicate single-cell analysis.

In his consultations, Martelotto gets a lot of different sample types, from human tissues all over the body to obscure plants and animals. His goal is to prevent other researchers from making the same mistakes he made in his pancreas disaster, blindly going ahead with untested protocols.

"If this is the first time you're running this particular sample, the risk [of failure] is high because you just don't know," he said. He couldn't easily quantify the risk, "but you can mitigate that risk," he said. If researchers don't know anything about their sample, he suggests using snRNA-seq; however, he preaches caution and the use of due diligence.

"My main advice is to always run pilot studies," he said. "Test a couple lysis buffers to make sure they get good suspensions and at least get the nuclei intact." Cleanup remains important, and the method to use depends on how many nuclei will be in the sample, or even how big they are.

Now, every sample gets treated as a clinical sample, even cell lines, Martelotto said. "You can do 100 experiments right, but you mess up one and you remember that forever."

The Scan

NFTs for Genome Sharing

Nature News writes that non-fungible tokens could be a way for people to profit from sharing genomic data.

Wastewater Warning System

Time magazine writes that cities and college campuses are monitoring sewage for SARS-CoV-2, an approach officials hope lasts beyond COVID-19.

Networks to Boost Surveillance

Scientific American writes that new organizations and networks aim to improve the ability of developing countries to conduct SARS-CoV-2 genomic surveillance.

Genome Biology Papers on Gastric Cancer Epimutations, BUTTERFLY, GUNC Tool

In Genome Biology this week: recurrent epigenetic mutations in gastric cancer, correction tool for unique molecular identifier-based assays, and more.