The impact of methanol fixation on your scRNA-seq dataset

Benchtop Technical Tips

Blog impact methanol

Methanol fixation is among the most widespread technique to preserve a cell sample that cannot be processed immediately. However, few in-depth studies exist about its molecular impact on genetic material, particularly for single-cell experiments. To fill this gap in knowledge, Prof. Angela Wu and her team at the Division of Life Science, Hong Kong University of Science and Technology, investigated the influence of methanol fixation on the quantity and diversity of RNA transcripts during a standard scRNA-seq experiment.

A scRNA-seq study initiated from a need for robustness

Methanol dehydrates cells, denaturing and precipitating proteins but preserving nucleic acids. A simple rehydration step is needed to recover the sample later for further processing, making methanol fixation a fast and standard technique to store samples for weeks or months before processing.

“I kept seeing single-cell papers trying out different fixation techniques and saying, ‘we use methanol and it worked fine’, or ‘we use something else, and it works better’,” explains Wu. “But nobody really characterized its potential impact.”

“We do a lot of single-cell work in our lab,” Wu continues. “I wanted to be very certain of what I was doing. If we are working with patient samples that need to be transported, we need to know which preservative is best and if we are inducing a systematic bias in our resulting datasets.”

A common concern with the use of preservatives is the cross-linking with RNA molecules, which might alter chemical bonds and cause degradation, with the transcripts breaking down in shorter pieces. Methanol is also concerned as it possesses an active hydroxyl group that might affect the integrity of RNA molecules.

Prof. Wu and her team processed two batches of cells in parallel, from a human colon cancer cell line (HCT-116) and a human liver cancer cell line (HepG2). Both cell lines were selected because of the group’s expertise with them and the existence of reference datasets. For each cell line, a batch of living cells was immediately processed while the other was fixed with methanol and stored for a week before rehydration. In total around 200 cells were used, with 96 sequenced in the end.

Wu clarifies their methodology: “We wanted to see the effect of methanol on the full length of mRNA transcripts, so we used Smart-seq2 as you get the full-length readout with both ends of the transcripts.”

Methanol fixation preserves nucleic acid integrity and does not affect scRNA-seq data interpretation

The outcome revealed great news: there were no obvious shifts between the transcriptomic profiles of the live and fixed cells from the resulting transcript libraries. Even better, after performing standard single-seq data processing methods, Wu and her team figured that methanol did not significantly impact cell type identification or clustering.

They performed Principal Component Analysis and looked in turn at each of the first principal components, representing the largest degrees of variance in the dataset. The two largest variances, linked to the cell type and stage of the cell cycle respectively, were not significantly affected by methanol fixation. After normalizing for cell cycle effects, however, the effect of fixation became noticeable.

Taking each cell line separately, they investigated the subset of genes that were mostly affected by methanol to elucidate the molecular mechanism of its impact.

Length and GC content are primary factors for methanol fixation impact on full-length transcripts

After determining that the concerned genes might share similar structures triggering the impact of methanol rather than it being the result of biological activity, the team identified two critical parameters: transcript length and GC content.

“We wanted to look at transcript length specifically because of the hydroxyl group of methanol that might cause fragmentation of RNA molecules,” explains Wu. “Also papers reported a size shift in the RNA library after long periods of fixation, meaning there is an impact on transcript length.”

Wondering about other parameters impacting your scRNA-seq data? Have a look at our article listing common technical artefacts in sample preparation

Analyzing the GC content of RNA transcripts was the following logical step, as higher GC content usually implies higher stability of a transcript.

The outcome indicates that long, GC-rich transcripts are the most affected by methanol fixation. The hypothesis suggests that those long mRNA transcripts might form super-structures that could be degraded by the hydroxyl group from methanol molecules, and that subsequent post-rehydration processing such as reverse transcription and amplification might be impeded compared to the content of live cells.

Considering the effect of methanol is greater on longer transcripts, Prof. Wu and her team presumed the effect would be different with other scRNA-seq methodologies based on 3’ sequencing rather than full-length. Simulating data from a Drop-seq experiment with the same cell lines, the distinction between fixed and live cells was much less clear.

The study concludes that “for users of scRNA-seq who wish to investigate differences in splice isoforms, live samples will be more reliable since the full-length information of transcriptome is better preserved; methanol fixation will result in skewed abundance readouts from those transcripts with high GC and long length.”

Methanol fixation is not introducing significant bias, but other options should also be considered for scRNA-seq

Methanol fixation seems to be confirmed as an effective fixation method with little impact on the genetic material of a cell, particularly for 3’ – based single-cell studies, but it might be different for non-mammalian cells.

“We started doing some marine research as well, working on cyanobacteria, and we wanted to test different fixation conditions. And we got quite a surprise,“ tells Wu. “For mammalian cells, methanol seems to work a bit better. If you use formaldehyde for example, it seems to cause a lot more RNA degradation. But in cyanobacteria, it seemed to be the opposite. Methanol was really not good as a preservative, but low levels of formaldehyde turned out to be the best.”

“If you can use fresh samples that are stable, then it is obviously still the best,” suggests Wu. “When fixation is absolutely necessary, I would say that for most purposes, methanol is still a viable option. But low concentrations of formaldehyde are really not bad either, I think it manages to keep the transcriptome stable for a longer period of time than methanol.

Wu recommends doing a “mini-bulk” test prep of the fixation, using just around 100 cells, followed by library construction. A quick check of the fixation effect by running a gel or fragment analyzer trace on the library size will show whether there is a shift in the library size distribution compared to the live sample, which could indicate degradation.

“If you can even bypass fixation altogether, for example if you’re just looking at cell type, just extracting the nucleus and perform single-nucleus RNA-seq might even be a more efficient way.”