Looking for ways to clean up your metabolomic and other omic data? Check out these tips.
No way. You’re back?! You’re obviously either really dedicated or exceptionally bored. The 42 Factors Series may seem daunting, but I’m glad you’re here…we have a lot more to discuss.
The previous article was the first in a series dedicated to increasing awareness of factors that can affect metabolomic (and other omic) data quality. I’m frequently asked whether it’s ok to analyze samples that have been biobanked or stored in a laboratory freezer for various periods of time, so today I’d like to continue the 42 Factors Series with a discussion of sample storage time.
Factor – Sample Storage Time
DNA is generally stable. RNA and proteins are less stable. Metabolites (and lipids) can be highly unstable. Given the complexity of biological systems, that makes sense—life as we know it wouldn’t exist without a subset of biochemical reactions that occurs very rapidly. But frequently in metabolomics, lipidomics, environmental science, and food chemistry, we’d like to measure those unstable small molecules. Is it possible using current techniques?
It is indeed possible to measure reactive lipids (see this example) and other reactive metabolites using derivatization techniques that trap a reactive molecule in a stable state and facilitate detection. But those techniques currently are not omic in scope; they focus on measuring a small number of analytes in many samples rather than measuring a large number of analytes in each sample. We’re still many years away from comprehensive profiling of the reactive oxygen species (ROS)-ome, for example.
A thorough review of the molecules that can be measured in biological samples using current metabolomic techniques is beyond the scope of this article and will require discussion of multiple factors, so today we’ll focus on sample storage time.
We designed a simple non-targeted metabolomics experiment aimed at estimating the effect of storage time on metabolites extracted from cancer cells. We performed a typical cell extraction (using 0.1% formic acid in methanol:water = 1:1) followed by immediate non-targeted profiling by mass spectrometry. Additional aliquots were stored frozen at -80°C and thawed after 1 and 24 hours to assess metabolite degradation during short-term storage. We were shocked to find that over 30% of the ~5,000 metabolite features (peaks detected by liquid chromatography-mass spectrometry (LC-MS)) in the initial sample were absent after 1 hour of storage at -80°C. After 24 h, more than 50% of the features had decreased more than 2-fold. Wow.
There are a number of limitations in that simple experiment, and skeptics will be quick to point out that the precision of non-targeted metabolomics can be very poor; that is, the absence of metabolites in the 1- and 24-hour samples could have been attributable to technical error. But our non-targeted platforms have been remarkably reproducible (e.g., technical replicates yield less than 10% variation in metabolome coverage), as I’ll detail in a future article. For now, it seems reasonable to conclude that something other than technical error caused the observed loss of metabolites during short-term storage.
A few candidate sub-factors that could have contributed to metabolite instability in our experiment included: 1) unquenched enzymes; 2) release of reactive metabolites during sample preparation; and/or 3) atmospheric oxygen.
In this instance, unquenched enzymes were not likely to be an issue, because we added liquid nitrogen to the cells to rapidly quench cellular metabolism prior to extracting metabolites. If you’re not currently using liquid nitrogen to quench, we strongly recommend it.
The second and third sub-factors, by contrast, were likely significant contributors to metabolite instability in our experiment. For example, ROS accumulate in multiple subcellular organelles including mitochondria, lysosomes, endoplasmic reticulum, and peroxisomes. The metabolite extraction process eliminates barriers between those compartments, exposing the entire metabolome to ROS and other reactive metabolites, resulting in numerous artifacts (e.g., see this example of peroxide-mediated conversion of α-ketoglutarate to succinate). Similarly, atmospheric oxygen is ubiquitously present through all stages of sample preparation and sample storage, unless steps are taken to eliminate it, and it can cause oxidation of a large number of metabolites.
We’ve adopted three best practices to address the problem of metabolite instability. First, at The University of Texas MD Anderson Metabolomics Core Facility we invite investigators to conduct their experiment(s) onsite with coaching provided by our staff. That provides the dual benefit of coaching researchers through their experiments and minimizing sample storage time (sometimes loading the samples into the mass spectrometer the same day they were collected).
Second, we displace oxygen by sparging wash buffers, extraction solvents, mobile phases, and the headspace of sample vials with nitrogen gas (or argon gas).
Third, we add the food preservative butylated hydroxytoluene (BHT) during sample collection and preparation.
The three tips noted above can be tedious, so our laboratory doesn’t always perform them, particularly when we’re reasonably confident that the key analytes are sufficiently stable, as is the case for most amino acids.
To summarize, the metabolome consists of a stable subset and an unstable subset. What is the half-life of the stable metabolome stable? 1 year? 10 years? 100 years? We don’t yet know the degradation time course for all metabolites under all sample preparation and storage conditions, so it’s difficult to say. When investigators have 20-year-old samples in the freezer that they’d like to analyze, my instinct is to go for it and see what we can find. Oftentimes, interesting results (e.g., differences between two treatment groups) can still be obtained. Negative results in those cases don’t mean that there weren’t interesting results at the point of initial sample collection, so the conclusions should be framed appropriately.
Metabolite stability is a grand challenge in the metabolomics field; in future articles, we’ll continue to discuss metabolite stability as it relates to other factors.
Take-home messages: 1) use liquid nitrogen during sample preparation to rapidly quench metabolic enzymes; 2) minimize sample preparation and storage time; 3) sparge liquids and the headspace of sample vials with nitrogen or argon gas to displace oxygen; and 4) add BHT during sample collection and preparation to minimize oxidation artifacts.
Next factor—adherent cell detachment. Stay tuned.
Big thanks to Rob Vreeken’s laboratory for pointing out BHT to us, and to present and former members of my team, especially Lin Tan, Leona Martin, and Leslie Silva, who investigated this factor.
Do you have any experience with artifacts introduced during sample storage? Any pressing metabolomics questions? Leave me a comment.
0 comments