Among the many challenges facing mass spec-based proteomics as it tries to move out of the laboratory and into the clinic, the need for higher-throughput assays dominated discussion this week at the Association for Mass Spectrometry's third annual Applications to the Clinical Lab meeting in San Diego.
In particular, attendees highlighted the need for improved automation of potentially clinically useful mass spec-based approaches like SISCAPA, with several presenting automated sample prep and LC-MS processes that could significantly streamline such workflows.
Leigh Anderson, CEO of the Plasma Proteome Institute and one of the inventors of the SISCAPA method – which combines antibody-based peptide enrichment with MRM mass spec – presented work his lab has done devising an automated SISCAPA workflow using an Agilent Bravo liquid handling system for sample prep attached to an Agilent 1200 series LC system and an Agilent 6490 triple quadrupole mass spec. Protocol development for the system "is well underway, and in six to nine months we should have all of the protocols for all of the stages of sample preparation for analysis on a system of this type," he said.
Matthew Rosenow, a researcher at the Translational Genomics Research Institute, introduced an automated system for SISCAPA sample prep capable of processing roughly 1,000 samples every 24 hours. The system, which the TGen scientists began testing at the beginning of the year, uses a Tecan robotic workstation and a Thermo Scientific KingFisher Flex liquid handling instrument to take plasma samples from cryovials and run them through the SISCAPA prep process, leaving them in a 96-well format ready for LC-MS analysis.
The hope, Rosenow noted, is that such automated workflows can help close the gap between the demands of discovery-based proteomics, which typically looks at a large number of analytes in a relatively small set of samples, and the verification and clinical validation worlds, which focus on smaller sets of analytes, but in much larger sample sets.
"In the discovery phase we're trying to find everything that we can find. The number of analytes is usually in the thousands," he said. "As we move to the verification and validation phases where we've already defined candidate biomarkers, our number of analytes goes down. But your number of samples increases. With the verification and validation phases we're talking about hundreds and thousands of samples. So it's obvious that a need exists to have a high-throughput quantitative method."
The fact that protein biomarker research typically confines itself to relatively small sample sets is one of the reasons so few candidate markers have thus far translated to the clinic, Anderson suggested.
"I'm a little bit embarrassed to say that in the 5,000 or 10,000 papers on biomarker proteomics, I don't know of a single one in which anybody has actually run 1,000 samples," he said. "And it's well known in the diagnostics community that if you can't run a few thousand samples, you can't know if a biomarker is clinically relevant. So this is a huge limitation that we need to overcome. Robustness and automation are becoming major barriers to applying this kind of [proteomic] methodology to solving our problems."
As Andy Hoofnagle, director of clinical mass spectrometry at the University of Washington's Department of Laboratory Medicine, told ProteoMonitor several months ago, automation is also needed to bring the price of mass spec assays down to where they can compete with a typical immunoassay (PM 10/22/2010). Currently, a typical immunoassay from Hoofnagle's lab costs between $10 and $15, with less frequently performed immunoassays running in the $40 to $50 range. A SISCAPA assay, on the other hand, runs in the $70 to $90 range.
SISCAPA assays aren't "going to cost less than $70 unless we have an automated platform doing everything for us," he said. "Until we have true hands-free automation, it's going to be pretty expensive in terms of labor."
In addition to automated sample prep platforms like the one developed at TGen and the one under development by Anderson's group, speedier LC run-times are also needed to improve the throughput of mass spec-based proteomics workflows like SISCAPA. The TGen system is able to produce 10 LC-MS-ready 96-well plates of sample every 24 hours. Anderson's platform, he estimated, will take about one hour to process a 96-well plate. But with nano-LC runs – frequently used for SISCAPA work – taking 35 minutes to 40 minutes per sample, the assay's throughput remains limited.
To get around this problem, Anderson has investigated using standard HPLC, which offers roughly a thousand-fold increase in flow rate compared to nano-LC. In SISCAPA work done with an Agilent 1200 series LC and an Agilent 6490 triple quadrupole mass spectrometer, his team was able to achieve sensitivities equivalent to previous assays they had done using nano-LC attached to an AB Sciex 4000 Q TRAP.
That, Anderson said, suggests that "although we may be able to push [SISCAPA] further going forward with the nanoflow, we're at least at the point where we can do the assays that we believe are important with this technology at high flow rates."
"That gives us the opportunity then to increase the cycle time," he added. "In this case the [nano-LC] cycle times which have historically taken 35 to 40 minutes per sample have been decreased, so far, to about 10 minutes per sample with a single channel LC system," meaning that an automated system with one LC channel could run "on the order of 150 samples a day, and two LC systems could do about 300 samples a day."
Such a processing rate, Anderson said, "begins to get to a regime in which it's extremely important to automate the generation of the samples, and to automate it in a way that you get a continuous flow of sample through the system."
According to Gustavo Salem, vice president and general manager of Agilent's Biological Systems division, the company established its relationship with Anderson several years ago through work it was doing on applying its chip LC-MS systems to the protein quantitation portion of the SISCAPA method. More recently, as the two parties began considering the method's commercial viability, the notion of automating it arose, Salem told ProteoMonitor.
The system as currently configured is built around Agilent's Bravo workstation, which automates the plate- and liquid-handling prep steps required prior to introduction of the sample into the LC-MS system. Anderson and Agilent are also evaluating the use of the company's new AssayMap liquid dispensing head technology, Salem noted, which, he said, could significantly reduce the amount of time required for the trypsin digestion portion of the process, potentially bringing it from hours down to minutes.
Automated workstations are a product area that "Agilent has invested a lot of time and energy into," Salem said, noting that rather than other mass spec vendors, its competition in the automation space comes from companies like Tecan and PerkinElmer
"None of these guys are in the mass spec game as it were. So it puts us in a nice position to be able to provide a complete workflow solution where we provide the automation, we provide the LC, we provide the MS," he said. "We could provide a customer complete access to the workflow, with one vendor to go to to make sure it all works. So you don't get the automation guy saying, 'No, it's the LC, no it's the mass spec.'"
The automation space isn't entirely empty of other mass spec vendors, however. TGen's SISCAPA sample prep system employs a Thermo Scientific KingFisher Flex machine for liquid handling, and Danaher's recently announced purchase of Beckman Coulter will put that company's Biomech line of automation tools under the same corporate umbrella as AB Sciex.
In addition to improving throughput and bringing down costs, automation should also reduce variability – another issue that has hindered mass spec-based proteomics' move to the clinic.
"Using [SISCAPA] there are multiple steps of liquid handling. There are timed steps and those steps – those incubation and digestion times – need to be consistent," Rosenow said. "Manual processing is much likelier to be error prone."
"Every experiment I've ever seen done where you've got to compare somebody pipetting samples versus an instrument doing it, the instrument is always more precise," Salem said. "And certainly in this market, where you're talking about a clinical setting, you want that reproducibility to be as high as possible because these are presumed future diagnostic tests. So automation is really key."
Have topics you'd like to see covered in ProteoMonitor? Contact the editor at abonislawski [at] genomeweb [.] com.