Like most large pharmaceutical companies, Merck invested heavily in gene expression profiling before paying much notice to proteomics. In fact, Merck’s proteomics group may have an even bigger chip on its shoulder: With the company’s acquisition of Rosetta Inpharmatics in May 2001, “there’s definitely a bias toward gene expression,” said Patrick Griffin, senior director of proteomics for Merck in Rahway, NJ.
But that doesn’t mean Griffin’s protein analysis group hasn’t had the resources to build a comprehensive suite of proteomics technologies. “Merck has a pretty good appreciation of when to include proteomics as a key driver, [and] clearly we have programs where proteomic profiling makes the most sense based on the target or what’s actually trying to be accomplished in the experiment,” he said. To that end, Griffin has assembled a fleet of ion trap, MALDI-TOF, Q-TOF, and FT-ICR mass spectrometers, and has recently begun trying out a few tricks to assist in the study and identification of proteins.
Griffin’s newest toy is the beta version of the NanoMate nanoelectrospray nozzle from Advion BioSciences, a 100-nozzle array that he uses to individually inject samples from a 96-well plate directly into Thermo Finnigan LCQ ion trap mass spectrometers. In addition, the proteomics group at Merck has outfitted its LCQ spectrometers with the AP/MALDI source sold by MassTech, allowing the group to transfer MALDI targets directly to the LCQ for MS/MS analysis of peptides that could not be easily identified via MALDI-TOF mass spectrometry. Griffin has also taken the plunge into using the ICAT reagents, because the technology is the only commercially available method for quantifying protein expression using LC/MS, he said.
In the future, Griffin said he’s interested in integrating lab-on-a-chip technology into Merck’s proteomics platform, as well as protein chip technologies as they reach maturity. “It doesn’t necessarily have to be mass spectrometry-based,” he said. “We’re looking at everybody right now, [but] there’s interest in the rolling circle technology [from Molecular Staging] for sensitivity reasons, although it’s too early to say for sure.”
Building the Proteomics Platform
More generally, Griffin’s proteomics group at Merck represents just one element of the company’s “molecular profiling” program, an effort to combine various gene and protein analysis experiments with information from databases, sample preparation data, and bioinformatics to answer specific biological questions. Within that larger thrust, Griffin is responsible for directing the company’s attempts at quantifying protein expression analysis, identifying surrogate markers in tissue and fluids, studying protein-protein interactions, and quantifying pharmacologically active peptide hormones.
Developing a platform to do this didn’t happen overnight. When Griffin joined Merck in 1991, after a postdoc in Lee Hood’s lab at CalTech, where he worked with John Yates, an old colleague from his graduate years in Don Hunt’s lab at the University of Virginia, he initially began working in the department of immunology and inflammation, where he applied his protein mass spectrometry skills to studying the mechanisms of action of anti-inflammatory drug leads.
But Griffin felt protein mass spectrometry could put its stamp on a much wider variety of drug discovery programs, including small molecule development. When combinatorial chemistry became all the rage two years later, his bosses agreed to move him into the department of basic chemistry, where his group resides now.
By 1995, Griffin’s group had grown from three to 20 people, and he had taken on new responsibilities supporting mass spectrometry technology within the department of basic chemistry.
Over the next few years his responsibilities continued to increase, as the potential for applying proteomics to drug discovery began to rise along with the anticipation that the human genome sequence would soon be complete. In 1999, he began working with Bill Hanlon, a colleague from the department of immunology and inflammation, to build Merck’s proteomics platform.
Together they developed a platform that utilizes both 2D gels and multidimensional liquid chromatography coupled with mass spectrometry. The 2D-DIGE technology, available exclusively from Amersham Biosciences, is used to compare protein expression levels among up to three samples, and a MDLC platform, which relies on standard reverse phase and ion exchange methods, to separate proteins from sera and other biological fluids .
A Plug-on Approach
Griffin has taken a modular, or “plug-on,” approach to building the group’s mass spectrometry capabilities, and many of his instruments are compatible with more than one type of ionization source. In addition to his MALDI-compatible Thermo Finnigan ion traps, his lab contains two Applied Biosystems Voyager MALDI-TOF spectrometers, as well as a Micromass Q-TOF I for performing de novo peptide sequencing, and a Bruker Daltonics FT-ICR spectrometer compatible with both MALDI and electrospray sources for looking at differentially labeled intact proteins.
“If we know a little bit more about the compound and we want complete coverage of the peptide in MS/MS, the Q-TOF gives us the advantage of seeing the ammonium ions,” he said, adding that the MALDI-TOFs are most effective when used as a “quick and dirty sample characterization to see just how good the digest looks.”
But Griffin said he has no immediate interest in purchasing a MALDI TOF-TOF from either ABI or Bruker because his current set-up allows his group to operate at relatively high throughput on high-end instruments.
“We think by using MALDI with the AnchorChip technology [on the Bruker FT-ICR] we can be fairly automated and get MS/MS data in a pretty efficient manner,” he said.
“The combination of the Voyager and the AP/MALDI [source on the ion traps] gives us a cheap version of the TOF/TOF with the manual step of moving the target. But that’s the only manual step.”
Drawing on the bioinformatics prowess of the Rosetta and Merck team, Griffin has taken pains to build a robust computation platform to handle the resulting mass spectral data. The group has software for comparing Mascot data with SEQUEST data, and the required database correlations are performed on a large cluster of CPUs [Griffin declined to specify the number] managed by LSF — Load Share Facility — a load-sharing system that distributes jobs across all available nodes on a defined system.
“It''s dynamic so as soon as a new multiprocessor server is added to the system we have access to those CPUs,” Griffin said. Griffin''s group stores its data on a customized Oracle database interface to SEQUEST and Mascot and the protein databases are updated nightly, he said.