Arthur Moseley, head of GlaxoSmithKline’s Proteomic Technologies group in Research Triangle Park, NC, began his presentation at the recent HUPO conference in Versailles by pointing out why we need proteomics. He showed a photo of a caterpillar and a butterfly: both share the same genome, but different proteomes, and researchers can only get to biological function by studying both.
Moseley went on to provide a rare glimpse of GlaxoSmithKline’s proteomics work within its dedicated Genomic and Proteomic Sciences Division. The division, set up to harness the new technologies for drug development, consists of seven platform technology departments, including Proteomic Technologies, Pathway Discovery, Quantitative Expression, and Expression Proteomics. These groups, though physically located in the US and the UK, work closely together to save costs and minimize duplicate efforts. For example, data from yeast two-hybrid experiments from one group can complement protein interaction data from mass spectrometry, created in the Proteomic Technologies group, in order to build protein interaction networks.
To study protein-protein interactions by mass spectrometry, Moseley’s group pulls down protein complexes using either an exogenous or an endogenous bait protein. To avoid isolating “promiscuous proteins,” the researchers make use of a double tag for two-step affinity chromatography. They then separate the proteins by one-dimensional gel electrophoresis, digest them, and identify the peptides by LC-MS.
Expression proteomics is another area where GSK is active, employing both 2D gel electrophoresis and LC-MS/MS. High sample throughput, Moseley pointed out, is only one way to obtain the maximum amount of information. What is equally important though, is proteome coverage, or how many proteins — as well as their modifications — can be identified from a given sample.
In the debate on the utility of 2D gels, Moseley comes down on the “pro” side, commenting that they are still “the most powerful separation technique” for proteins. Also, posttranslational modifications are easier to detect on a 2D gel than by non-gel based approaches, he said. Contrary to many researchers’ belief, he said, 2D gels are not low-throughput, “often providing higher throughput than non-gel based proteomics,” because they can be easily run in parallel. Also, he said, the method is highly reproducible, allowing for detailed image analysis and accurate quantification of the spots. As evidence, Moseley showed a section from 60 different 2D gels that looked almost identical. Automated spot-cutting further helps to increase the throughput of 2D gels, he said. However, the method reaches its well-known limits when it comes to proteins of extreme isoelectric point or molecular weight, or proteins that are very hydrophobic, he added.
Moseley’s Proteomic Technologies group uses both MALDI-MS and LC-MS/MS for expression profiling and has recently started coupling liquid chromatography with both MALDI-MS/MS and ESI-MS/MS. He has access to more than ten mass spectrometers, ten of which have MS/MS capabilities, as well as more than ten nanoscale capillary liquid chromatography instruments. Furthermore, he has a robotic system for spot cutting and in-gel protein digestion, and a customized LIMS system. His group uses Mascot software to analyze the terabytes of raw data it currently produces in a year. While MALDI-MS is adequate for simple mixtures and samples from organisms whose genomes have been sequenced, Moseley said, MS/MS is necessary to analyze samples from unsequenced organisms and to get information about posttranslational modifications.
Spot Analysis Bottlenecks
As an example, Moseley showed an analysis of changes in protein expression in response to a ligand binding to a receptor. 2D gel analysis identified 117 spots that differed between the two samples. Of these, he has so far analyzed 70 by LC-MS/MS, and identified 97 gene products. However, the LC-analysis “can be the bottleneck,” Moseley said: the analysis of each spot takes about one hour. Recently, he has managed to connect two liquid chromatography instruments to one Q-TOF mass spectrometer, thereby doubling the throughput to two spots per hour, and hopes to get up to four spots per hour soon. This is offset, however, by a decrease in sequence coverage. For those who do not think this is a great achievement, Moseley offered some perspective: “Doubling the throughput on a $500,000 instrument is significant,” he said. Another approach Moseley’s group takes to proteome analysis, which he said is complementary to gel-based methods, is “shotgun proteomics,” in which a complex mixture of proteins is digested and analyzed by LC/LC-MS/MS. This approach is especially useful for “extreme” proteins that do not resolve well on a 2D gel, he said. His group integrates ESI-MS/MS and MALDI-MS/MS, splitting the sample after the second chromatography step such that 20 percent is analyzed by ESI-MS/MS and 80 percent by MALDI-MS/MS. This increases the protein coverage: In a study of bovine mitochondrial ribosomes, 32 of 51 ribosomal proteins were identified by this approach. Eight of those were uniquely identified by the electrospray instrument, and 11 by the MALDI mass spectrometer.
Another project further demonstrated the power of this approach: From a nuclear extract of a human cell line, Moseley’s group identified almost 1,900 unique proteins — this required about 120 hours of mass spectrometry time.
What can still be improved, Moseley said, is the way data is collected and interpreted. Rather than acquiring all the data first, and interpreting it later, researchers in the future might use “intelligent” acquisition methods, where some data is analyzed first, and the method is adapted accordingly.