Cell culture profiling involves analysis of the growth medium for markers of cell health, productivity, viability, expansion, or other characteristics of interest. Profiling begins before cells and media meet: for media, during quality control to quantify baseline concentrations of media components and nutrients; and on cells through short tandem repeat analysis for cell line authentication. Media component concentrations serve as a basis for quantifying culture- or process-related depletion of media or feed ingredients, which is where the lion’s share of cell culture profiling occurs. Theoretically, this information has the potential to guide process and culture optimization.
Among the ‘omics methods for quantifying molecules of significance in living systems, metabolomics is of particular interest for monitoring the health and productivity of cell cultures. Metabolomics has been touted as a snapshot or picture into the instantaneous state of organisms, a reflection of all extant biochemical activity, including gene expression, protein and enzyme activity, and regulation/dysregulation. In some ways metabolome is synonymous with phenotype.
The NMR approach
Chenomx, which specializes in NMR software for the life sciences, is among a handful of companies using NMR spectroscopy for culture profiling. Their basic method involves measuring metabolites within fresh and depleted media (hence the importance of establishing a medium’s initial ingredient concentrations). “This approach uncovers which components are consumed, and a representative sample of compounds that have exited cells,” says Neil Taylor, president. “Some of these molecules are ‘anti-nutrients’ that restrict cellular growth and proliferation.”
The most common example of metabolite control over a process is inhibition of fermentation by alcohol during wine or beer production.
In one study, a Chenomx customer identified nine intermediates, or byproducts of amino acid metabolism, with anti-nutrient status. When investigators re-engineered their cells not to express these molecules, culture times increased from 9 days to 14 days. “Since the additional fermentation time occurred during the culture’s most active period, productivity doubled.
With NMR, sensitivity and peak resolution depend on magnetic field strength. Chenomx, which does not promote or support a specific NMR instrument platform, recommends using fairly high-field NMR systems, in the 600 MHz to 700 MHz range. While culture medium components are present at concentrations high enough for quantitation using conventional instrumentation, metabolites often exist at extremely low levels requiring high-field instruments. Chenomx methods use single-dimension proton NMR and techniques based on the nuclear Overhauser effect.
“We measure small molecule metabolites, and we’re not too particular about what cells we support,” Taylor adds. CHO, E. coli, yeast, and human cells are all suitable for our method.”
Sample preparation is simple as well, involving centrifugation and application of 3 kDa cutoff filtration to remove larger molecules, such as lipids and proteins, which may obscure the small molecule signals. Note that both proteomic and lipidomic methods are available for cell culture profiling as well—the subject, perhaps, of another article. At this stage a standard solution is added to aid in calibrating the locations of spectral peaks of interest, and to measure the concentration of the metabolites when they are discovered in the mixture.
No metabolomics happens, at least not at high throughput, without the assistance of a spectral library for rapid identification of relevant chemical markers.
“We have three hundred-plus metabolites in our spectral library; a finite number—around fifty or sixty—show up in sufficient concentration to be measured in a typical sample,” Taylor explains. “And the spectral libraries are fairly consistent across cell types. Users will occasionally discover a new metabolite or something not previously noticed in their media. These are quite easy to add to the library.
Chenomx software makes no assumptions regarding which metabolites it will detect, so the product operates firmly in discovery or untargeted mode. One could argue that, by virtue of its lack of sensitivity compared, say, with mass spectrometry, NMR is inherently sensitivity-limited.
“That’s probably the most serious limitation of NMR,” Taylor explains. “Some undetectable metabolites are probably more significant than those that are easily picked up. But we pick up some additional relative sensitivity by excluding large biomolecules. Also, while profiling a particular metabolite in a sample we compare our detailed signature to the experimental spectrum and, where it matches, we subtract out those signals.”
Improvements in LC/MS
All ‘omics methods must deal with the colossal concentration dynamic range of target molecules, as high as 1014. Proteomics, lipidomics, and genomics focus on one molecule type, which somewhat simplifies the separation of analytes from everything else. Metabolomics faces the same concentration conundrum, plus a molecular diversity challenge. These are exacerbated by broad, unpredictable interactions among these molecules, sample matrix effects, liquid chromatography retention times for anionic metabolites, and high-salt samples.
For these samples Agilent Technologies has developed an LC/MS method, based on InfinityLab’s Poroshell HILIC-Z polyethylether ketone (PEEK)-lined columns, to achieve reliable sensitivity and retention time reproducibility for high-salt samples. The column allows one-pass separation of anionic and hydrophobic metabolites, and, according to the company, “could be used to monitor biomanufacturing processes.”
Agilent’s method tracks both metabolite generation and the consumption of media components, both of which are critical to cell culture health.
“High salt is one of the main hurdles to overcome as salty samples cause retention time shifts, and can alter signal intensity in hydrophilic interaction chromatography (HILIC),” explains Jordy Hsiao, R&D scientist at Agilent. Common culture media contain approximately 100 mM of sodium chloride plus other salts. “Another hurdle that needs to be addressed is the ability to obtain optimal chromatography for the secreted metabolites, specifically the chelating organic acids. These metabolites are sensitive to metal interaction, which leads to poor peak shapes and inconsistent signal intensity.” Agilent’s method uses medronic acid as an additive to overcome these effects.
De-salting the sample could achieve these goals, but as Hsiao notes “any additional sample-preparation steps, e.g., desalting, could introduce operational error, sample loss, and additional labor and time. The simplicity of this approach is that minimal sample prep is needed and cell culture media content can be quickly analyzed using the method described.”
The future
It’s obvious that the future of cell culture profiling lies in metabolomics. The past decade has seen an explosion in new hardware, software, and methods for unraveling the metabolomes of numerous organisms and cells. The Human Metabolome Database, for example, catalogs more than 100,000 metabolites, with detailed MS and NMR spectra for almost 20,000 of them. “Major developments in analytical methodologies now allow assignment and even quantification of several hundred compounds,” says Miroslava Cuperlovic-Culf, senior research officer at the National Research Council Canada. “NMR now allows analysis of smaller sample sizes as well as newer methods for spectral alignment and quantification. Finally, there is the increasing presence of hyphenated technologies, combining MS and NMR for larger metabolomics coverage.”
While these methods provide better metabolite coverage than ever, metabolomics is still years from being a truly high-throughput science.
That should change, says Cuperlovic-Culf, as more informatics tools are adapted to metabolomics. “Semi-automated assignment and quantification, plus developments in genome-scale metabolism modeling are becoming available. Further, machine-learning methods now enable improved selection of features and data modeling.”