NEW YORK (GenomeWeb) – Trypsin has long been the go-to protease for proteomics experiments, but recent work by researchers at the University of Texas Southwestern Medical Center suggests that the use of alternative enzymes could offer advantages including improved sequence coverage, lower detection limits, and enhanced isoform differentiation.
Alternative proteases could prove particularly useful in the case of targeted quantitation experiments where, UT Southwestern researcher Hamid Mirzaei told ProteoMonitor, for many proteins conventional trypsin digestion fails to generate peptides ideally suited to analysis by methods like selected-reaction monitoring mass spec.
SRM mass spec is a highly specific, highly sensitive method of protein quantitation. However, peptides are not all equally suited to such analysis, and so, when constructing SRM assays, a significant amount of work goes into selecting the best peptides for quantitation of a target protein.
Selection is based on a variety of criteria, including how well a given peptide generates signal in the mass spectrometer and whether it elutes with other peptides that could complicate or suppress its signal. Also a consideration for quantitative experiments is the expense and feasibility of creating the required synthetic peptide standards.
While trypsin is overwhelmingly used for digestion in SRM experiments, the enzyme does not always generate suitable peptides, Mirzaei said, noting that, in his experience, out of a list of 10 SRM candidate peptides, typically "you will end up with one that you can really use."
Using proteases besides trypsin, he said, "gives you a better chance of finding an ideal peptide." Indeed, in a study published this month in Molecular & Cellular Proteomics, Mirzaei and his colleagues found that for 25 percent of the 6,056 proteins they identified in HeLa cell digests, the peptides best suited to SRM analysis were generated by a protease other than trypsin. Looking at a smaller subset of eight proteins, they found that for five of the eight, peptides generated by the protease Asp-N produced higher mass spec signal than tryptic peptides.
In the paper, the researchers also presented a web tool named Confetti that aids SRM assay design by reporting the most commonly observed peptides per proteins for various protease digests and allowing researchers to filter for attributes like protein-sequence exclusivity.
Formerly a post-doc in the lab of Swiss Federal Institute of Technology researcher Ruedi Aebersold, Mirzaei was involved in some of the early explorations of SRM techniques. It was not until he joined UT Southwestern, however, that he began looking into the potential of alternative proteases to improve targeted proteomics experiments.
His interest in this area stemmed largely from his work as director of the center's proteomics core where, he said, many of his customers are "biochemists and biologists who mainly study just one protein or a family of proteins, and they want to study those particular proteins to death."
This sort of exhaustive analysis presents problems for trypsin-based SRM, Mirzaei said, due to the fact that not all portions of a target protein will be cut by trypsin into portions suitable for mass spec detection.
"A lot of these peptides, for instance, are never seen either because they are too large or too small – the lysines and arginines are too close together," he said. "Or because it carries multiple PTMs, so it doesn't fly well."
"So, the idea was that if we use alternative enzymes to digest these proteins, we would have a lot more flexibility to cut the protein exactly as we wanted," he said.
A main concern regarding use of alternative proteases has been reproducibility, Mirzaei said, noting that this was a concern of his, as well.
"One of the advantages of running a core facility, though, is that you see a lot of samples," he said. "We do this over and over again for people who bring their proteins to the core and say 'I want to use this enzyme or that enzyme.'"
"We see them bringing this sample over and over again because they are studying this specific protein," he said. "And when we apply that [enzyme], we always get the same results."
Mirzaei and his colleagues set out to show this reproducibility in a limited way in their MCP study, demonstrating similar coefficients of variance for trypsin and Asp-N across replicate digests of the eight proteins they quantified using their two enzymes.
One complicating factor with alternative enzymes like Asp-N is that, unlike trypsin, it is difficult to predict in silico where they will cut a given protein. "They can be promiscuous and will sometimes cleave a site that is not predicted," Mirzaei said. However, he noted, they are relatively reproducible, in that once you have experimentally determined the cleavage sites for a given protein, they will cut at those sites each time.
Expanding the repertoire of potential peptide targets holds a variety of benefits for SRM assay development, Mirzaei said. It could allow, for instance, better analysis of PTMs that cannot be suitably isolated using trypsin. Or it could offer improved ionization efficiency by generating peptides that elute in less crowded portions of an LC gradient.
It could also, he noted, help researchers choose peptides less prone to degradation during the digestion process – a potentially complicating factor in SRM-based quantitation. Additionally, an expanded range of potential peptide targets could prove useful for immunoenrichment-based workflows like SISCAPA, where the ability to generate antibodies against peptide targets is key.
Mirzaei said he anticipates that researchers using SRM to look at large protein panels – 50 or 100 analytes, for instance – will by and large continue to rely primarily on trypsin given the enzyme's robustness and broad suitability to mass spec analysis.
However, he said he believed that researchers focused intensely on a small number of proteins, including clinical biomarker researchers, could benefit from using a range of proteases in their SRM assays.
"When it comes to people, especially in the field of biomarkers, who want to measure a particular protein in a very sensitive, very reproducible, very quantitative manner, for those kinds of applications, this could be very valuable," he said.
"One of the biggest challenges to SRM is that finding a good target is hard," Mirzaei added. "But if there are enough options and enough peptides for people to look at [as] targets for assay development, SRM can be really powerful. You need two really good, robust peptides to do good quantification. And those peptides may not come from trypsin."
If you were to choose a single enzyme, trypsin provides the highest number of identifiable sequences, said University of Victoria researcher Christoph Borchers, who was not involved in the study. But, he said, as described in the MCP paper, alternative enzymes could prove useful for cutting PTMs or stretches of sequence without tryptic cleavage sites.
He said that cost would likely prove a consideration for researchers, however, noting that while proteomics grade trypsin costs around $2 per µg, Asp-N costs around $92 per µg. For experiments looking at a small number of samples or a small number of proteins, this might not be an issue, he said, but "substituting a much more expensive enzyme for a relatively inexpensive one would be a significant factor in any large-scale SRM project."
Perhaps most significant, Borchers said, was the HeLa proteome digestion information contained in the researchers' Confetti database. Providing the peptides generated by 48 single, double, and triple enzyme combinations, the database "will tell people which enzymes to be used to generate a specific target sequence," he said, adding that this "will be particularly useful for SRM studies targeting a specific peptide," for instance a particular PTM.
Also of note, he said, is the database's data on the sensitivity resulting from the different enzyme combinations used.
"It's good to have this relative sensitivity information, because sometimes it might be preferable to choose a combination that is less sensitive by a factor of two but has a cost lower by a factor of 1,000," he said.