This story originally ran on Oct. 31 and has been updated to include comments from an industry executive and outside researcher.
A team of scientists led by Northwestern University researcher Neil Kelleher has developed a workflow for high-throughput top-down proteomics, using it to identify more than 3,000 protein species in an analysis of HeLa S3 cells.
According to Kelleher, this number of IDs represents a 20-fold increase over any previous top-down experiments in mammalian cells and is one of the first demonstrations of the feasibility of large-scale, discovery-style intact protein work.
In the study, which was detailed in a paper published this week in Nature, the researchers used a four-dimensional separation platform attached to either a Thermo Scientific 12 T LTQ FT Ultra or Thermo Scientific Orbitrap Elite mass spectrometer to identify a total of 1,043 proteins and more than 3,000 protein species.
Construction of the separations platform was the key to the effort, Kelleher told ProteoMonitor, noting that previous methods had suffered from "a lack of separation and of a lack of recovery."
"The mass spec hardware was probably third" in terms of importance, "and the informatics was probably secondary," he said. "But the primary thing was putting together this high-yielding, integrated, high-efficiency separations system that is compatible with LC-MS."
The separations workflow comprised three steps: an initial stage of solution isoelectric focusing, which was performed on a custom instrument built by Northwestern researcher John Tran – an author on the paper; gel-eluted liquid fraction entrapment electrophoresis on a Gelfree fractionation system from Protein Discovery; and nanocapillary liquid chromatography.
Use of Gelfree fractionation was "a critical step," Kelleher said. The technique, in contrast to conventional 2D gels, allowed the researchers to transfer proteins from the initial separation steps to the nano-LC system with "95 percent recovery or more," he noted.
"If you're just handling intact proteins and you want to separate them, a 2D gel is a venerable way to do that," Kelleher said. "But you can't get the proteins out of the gel unless you digest them, so you just can't interface it with top-down mass spectrometry."
The system also allowed the researchers to eliminate from the samples detergents like SDS, which can interfere with mass spec analyses. In an April interview with ProteoMonitor discussing top-down proteomics, Andreas Huhmer, proteomics marketing director at Thermo Fisher Scientific, noted that removal of these detergents had proven a bottleneck in high-throughput intact protein research (PM 4/8/2011).
Bottom-up, peptide-based proteomics has dominated the field for the last decade, but interest in top-down methods has grown as improvements in instrumentation have made analysis of intact proteins easier and researchers have become increasingly aware of the importance of protein isoforms and post-translational modifications.
Speaking to ProteoMonitor this week, Huhmer called the Nature study a "breakthrough" in top-down work, comparing it to early bottom-up proteomics studies done a decade ago.
"I would make the analogy to when [Scripps researcher] John Yates came out with his paper in 2001 showing that you could detect 1,500 proteins in a matter of 24 hours using a MudPIT [separations system]," he said. "I see it in the same league as a breakthrough."
Princeton researcher Ben Garcia, whose lab also focuses on top-down work, seconded Huhmer's remarks, calling the paper a "tremendous advancement."
"Previously, on average, large-scale analyses using top down [mass spec] produced proteome coverage of a few hundred protein isoforms, with most of these proteins being very abundant ones like histones," he told ProteoMonitor. "These results easily represent a 20 to 30-fold increase over previous top-down MS results."
The study opens up some "very exciting scientific angles," Huhmer said. "You now actually have the ability to observe the biology of proteins and how they are processed. So you can really confirm hotspots of phosphorylation, not just infer them [from bottom-up data]. You can see the regulation of individual splice variants. So you create a new frame of reference for the complexity of the proteome."
"From a scientific perspective this is really significant," he said. "And I think that as a vendor we look at it [as a sign] that there are going to be a lot of opportunities [in top-down proteomics] in the future because people will have to understand these networks and how these protein isoforms are processed … [and with top-down] you can directly observe and measure them."
In the study, Kelleher's team demonstrated the potential of top-down methods for investigating such phenomena, identifying a number of previously undetected isoforms of endogenous human proteins and monitoring over several days the modification states of proteins linked to stress-induced senescence in H1299 lung cancer cells and B16F10 melanoma cells after the induction of DNA damage.
In the study, Kelleher's team demonstrated the potential of top-down methods for investigating such phenomena, identifying a number of previously undetected isoforms of endogenous human proteins and monitoring over several days the modification states of proteins linked to stress-induced senescence in H1299 lung cancer cells and B16F10 melanoma cells after the induction of DNA damage.
Kelleher said he hopes the Nature paper will further accelerate progress in the field by showing vendors and researchers that proteome-scale intact protein work is more practical than generally thought.
"I think the major limitation has been one of perception and investment," he said. "People didn't think it was possible.Whether it’s a single [principal investigator] not wanting to get into top down and invest their resources or vendors not building instruments that can handle [top-down studies], this has been the major issue, so I'm hopeful that this can get us to the point where people invest more and the field can move further faster."
Currently, all of the workflow components except the solution isoelectric focusing instrument are commercially available, Kelleher said. That wasn't the case when the researchers started work on the system several years ago, he noted.
Improvements in mass spec instrumentation have also made the system easier to implement, Kelleher said. The FT-ICR LTQ FT Ultra machine the researchers used for a portion of the work is a custom instrument, but the Orbitrap Elite, which Thermo Fisher released at this year's American Society for Mass Spectrometry meeting in June, has proven "even better" for his lab's top-down work, he said.
In a July interview with ProteoMonitor he noted that, while Orbitrap machines historically have been relatively weak performers for intact protein work, the Elite's high-resolution and speed make it a good fit for such research (PM 7/1/2011).
Obtaining an MS-MS spectrum for one protein on the FT-ICR machine typically takes his lab around seven to eight seconds, Kelleher said. By comparison, working with the Elite his team found they could get MS-MS spectra for three different proteins in 1.5 seconds.
"That [jump in speed] is a big deal when you're talking about intensities and building up signals for top-down [proteomics] on a chromatographic timescale," he said.
In the study, the researchers identified proteins as large as 105 kDa, with the system functioning optimally for proteins in the 50 kDa and below range, Kelleher said, noting that platform's "main bias" is size.
"Between zero and 50 [kDa], that's our wheelhouse," he said. "The 50 to 150 [kDa] range is where you start to see a gap in terms of what we detected and what is predicted for the human proteome."
This bias toward smaller proteins likely skewed the number of protein isoforms detected by the platform, Kelleher said, noting that the researchers detected on average three isoforms per protein – fewer than he had expected.
"I think that as we get higher in mass, we'll actually get more [isoforms] on average," he said, adding that another estimate in the paper suggests the average may be closer to four or five isoforms per protein.
"There's clearly work left to do," he said. "But I hope the effect [of the study] will be symbolic – like, 'Hey, this is possible.'"
Have topics you'd like to see covered in ProteoMonitor? Contact the editor at abonislawski [at] genomeweb [.] com.