For the mass spec buyer the selection can be mind-boggling. Here’s some help.
By Aaron J. Sender
When GeneProt furnished its Geneva proteomics facility last April, it installed 45 ion-trap and six MALDI-TOF mass spectrometers, because neither could analyze all proteins well. But the company stuck with a single supplier — Bruker Daltonics. “It helps a bit to have both mass specs from the same manufacturer,” explains CSO Keith Rose. In fact, Bruker returned the favor by placing an engineer on-site to service its loyal customer.
Yet less than a year later, when making purchases for its North Brunswick, NJ, site, set to open later this year, GeneProt turned to Waters subsidiary Micromass for the bulk of its instruments — 45 QTOFs — along with six Bruker TOF-TOFs.
Why the change of heart? “They were bought at different times,” says Rose. And within that short period of time the technology available for mass spectral analysis had advanced enough for GeneProt to opt for a different configuration of instruments.
With vendors continually leaping over each other, mass spec technologies for proteomic analyses are advancing faster than you can say “Fourier transform ion cyclotron resonance.”
And unlike the relatively simple task of outfitting a genome sequencing center, building a proteomics shop can be a complicated, confusing choice: because of the complexity of proteins there is no one machine that can do it all. One instrument might measure the mass of peptides more accurately, but will not be as good as others in detecting peptides of low abundance. Another may have good sensitivity, but slow throughput, and so on.
“There are never any easy answers,” says Tom Neubert who heads the New York University core mass-spec lab. “There are so many different machines out there and so many advantages and disadvantages to each.” And to get the best of each instrument required for a complete proteomics facility usually means assembling mass specs from several vendors. “Each company comes up with improvements every single year. So it’s a moving target. There is never anything that just blows everybody away,” says Neubert.
Although mass spectrometrists may relish the opportunity to fiddle around with instruments and various configurations, most biologists would prefer a black box in which to load their sample and get their results. Because no such instrument exists, GT has surveyed some 2,000 users to help you wade through the constantly shifting landscape of mass spectrometry technologies, and to provide some food for thought in the search for the right balance between price, reliability, and performance.
Don’t worry, even seasoned mass spectrometrists don’t have it easy. “Proteomics is a field that’s changing incredibly fast. Mass spectrometry is changing incredibly fast. And the two of them together make for a nightmare for people working in this area,” says head of the state of Michigan proteomics consortium Phil Andrews. “It’s a real challenge to keep up and make good decisions based on the technology.”
A Tale of Two Ionization Methods
Simply put, a mass spectrometer is an expensive, high-tech scale. The instruments, in their various forms, have one thing in common: They each have an ionization source that vaporizes and electrically charges the proteins; and each has at least one mass analyzer that applies an electromagnetic field to the resulting ions and sorts them by mass.
Two competing and overlapping ionization methods dominate protein analysis: Electrospray ionization (ESI) and matrix-assisted laser desorption ionization (MALDI). Each has its advantages. ESI, for example, is more sensitive, while MALDI is faster. But most importantly, “some peptides ionize better in MALDI mode than electrospray and vice versa,” says Rose. And it is often hard to predict which method will work better for a particular peptide. In short, for a complete picture you need both.
What’s more, each ionization source can be mixed and matched to a dizzying array of mass analyzers. Historically researchers used a MALDI coupled with a single time-of-flight analyzer to first try to identify proteins by molecular weight, matching them with ones in the databases. This on average takes care of 60 to 70 percent of the proteins in a sample. Then for those that are still unidentified, you needed to go to ESI: A second machine with tandem mass analyzers smashes peptides to smaller fragments, providing more detailed information or amino acid sequence. Traditionally, an ESI source coupled with a quadrupole or ion trap analyzer performed this task.
“Think about at least two instruments,” advises Catherine Fenselau of the University of Maryland and former president of the American Society of Mass Spectrometry, “one for molecular weights and one for sequences. The molecular weight machine would probably be a MALDI and the other one would be electrospray.”
Absent was the ability to get the richer MS/MS data with the easily automatable MALDI source. But that is no longer the case. The MALDI-TOF/TOFs recently unveiled by both Bruker and ABI are changing that equation. Now tandem MS can be performed at the speed of MALDI. Instead of the 30 minutes per experiment on an ESI-ion trap, a MALDI-TOF/TOF can do it in seconds.
Here’s a quick tour through the major instrument categories:
For the simplest case of protein identification, the MALDI-TOF is generally the instrument of choice. In GT’s recent survey of mass spec users, this category was the overwhelming choice for identifying proteins from 2D gels.
For a MALDI experiment, a robot (or a grad student) cuts out the protein spots from a 2D gel and deposits digested peptides on a plate coated with a UV-reactive substance. A pulse laser shoots the samples and vaporizes and ionizes the proteins. The most common experiment performed on this instrument, with the help of a computer, is a peptide-mass-fingerprint database search. An enzyme slices the protein at particular amino acids, forming a pattern of peptides of varying mass, each measured by the instrument. Software, such as Matrix Science’s Mascot or ProteoMetrics’ ProFound, theoretically digests proteins in a database and tries to find a match.
The MALDI-TOF is in some sense a commodity instrument. Bruker, Applied Biosystems, Micromass, Kratos, and others offer choices here. Most manufacturers also offer an ESI version for those proteins where MALDI doesn’t cut it. Single TOFs range in price from less than 100 grand to about $400,000. So this is an area where customer service, interoperability with other instruments, and cost weigh in heavily.
According to the GT survey, to which 189 mass spec users responded, Applied Biosystems’ MALDI-TOFs are the most commonly used. But the tide is changing, says Bruker VP and general manager John Wronka. “In terms of new sales, we are number one,” he says. Bruker is selling more than 200 MALDI-TOFs a year. Part of the incentive for buying a Bruker is that it offers the option to upgrade to a TOF/TOF. The Bruker TOF lists for $375,000 in the US, and for an extra $200,000, you can add another TOF at some later date.
There is no need to become a mass spec geek to operate MALDI-TOFs, either. “You can be trained on it in half an hour,” says David Arnott, a mass spec scientist in John Stults’ lab at Genentech, who uses an ABI Voyager DE Star MALDI-TOF.
University of Michigan’s Andrews, who has Micromass’ [email protected] R has the same experience. “Any MALDI-TOF, like the [ABI Voyager] DE Pro and the [email protected] R, are simple enough and cheap enough that you could use those instruments as walk-up instruments,” he says. “We have several users that just walk in, sign up for time, and use it.”
And that’s precisely the audience Amersham Biosciences, the newest player in the MALDI-TOF market, is going after. “Occupying the high-end systems with easy questions is not the most economical way,” says Pieter Noordeloos, Amersham’s VP of marketing for proteomics. Instead, Amersham promises a benchtop MALDI-TOF that biologists can use to identify their favorite proteins instead of bothering the experts in the core lab. So far, Amersham has installed 15 to 20 units in a pre-release of its Ettan MALDI-TOF Pro and has “orders in for 10 to 17 units,” says Noordeloos.
But the instrument has been in the works for four years and has already had one failed release. “We have the old version, but they were never able to get it up and running,” says Tracy Andacht, who manages the University of Georgia proteomics resource facility. “You might be able to acquire a few spectra, and after that you were never able to acquire anything more.” The problems began when Amersham engineers arrived to train the lab members in preparation for the facility’s grand opening in January. “I don’t think they actually understood exactly what the issue was,” says Andacht.
“There were only a few [instruments] that actually came over to the United States and there were issues with most of those,” she says. Meanwhile, Andacht is waiting for her replacement, the MALDI-TOF Pro. But “they had problems with the previous version and I’m not confident that all of these issues are going to be resolved,” she says.
Among the more than 143 mass specs that the GT survey respondents own, 11 Amersham instruments were reported installed. None of the users considered the instrument to be reliable.
An Amersham representative explains that Andacht’s machine was a beta version, but buying any newly released instrument is a risk. “We had an early Finnigan ion trap which never worked,” says Hank Fales, a mass spectrometrist who has been with the NIH for 50 years. “That was kind of a disaster.” Today Thermo Finnigan ion traps are among the most robust mass specs around.
Thermo Finnigan invented the ion trap and has clearly captured the market for that instrument, perhaps at the expense of further development in other categories of mass specs. Although not the best choice in terms of performance, ion traps are certainly the best bang for the buck and the workhorse in most proteomics labs. They are quite easily automated, offer high sensitivity, provide MS/MS data and are ranked highest by GT survey respondents for ease of use. The one category where ion traps fall behind other tandem mass specs, such as the QTOF, is in mass accuracy. But for most cases that doesn’t really matter (see sidebar, “Critical Mass?” p. 58).
There are 2,000 Thermo Finnigan ion traps installed worldwide, according to Bill Hancock, the company’s VP of proteomics. “Last year was the best year ever in Thermo Finnigan’s history and the growing proteomics market had a fair bit to do with it,” he says. For years, the only ionization option for the ion trap has been ESI. But just a few months ago, Finnigan released a version with a MALDI source.
Another trap maker worth a mention is Bruker, which is generally respected in the community. It has only a small share of the overall market, but has made some bulk sales to highly visible proteomics facilities such as GeneProt.
Yet for the complete picture, ion traps are not enough. “There are going to be instances where the QTOF would be able to solve problems that the ion trap wouldn’t,” says Ron Orlando, a mass spectrometrist at the University of Georgia’s Center for Complex Carbohydrate Research. For example, an ion trap won’t nail down post-translational modifications or de novo sequencing, where accurate mass measurements are pivotal. But for the run-of-the-mill experiments, a QTOF doesn’t match the throughput of traps.
The QTOF is the instrument you want when the accuracy of the trap just doesn’t cut it. “If you’re in the game, a QTOF is something you want to have,” says Northeastern University mass spectrometrist Barry Karger.
When Micromass first introduced the concept at the 1996 ASMS meeting in Portland, Ore., “it was fairly low key. We simply had a presentation on one poster, no oral presentation,” QTOF co-inventor Bob Bateman recalls. “But it was difficult to get to see the poster. It attracted an awful lot of interest.”
Bateman got the idea for the instrument when Michael Guilhaus of the University of New South Wales arrived at Micromass’ Manchester, UK, HQ in 1992 to show off his new TOF analyzer. “It’s different to what was then normal time-of-flight in that it was arranged orthogonally,” says Bateman. Until that time, ions flew from the source in a straight line to the mass analyzer. In the orthogonal arrangement, the source is off to the side. “If you have a source that produces a continuous stream of ions, such as the electrospray, then this is a much more efficient way of analyzing the ions,” says Bateman. “A much higher percentage of the ions are analyzed.”
After combining the TOF with a quadrupole analyzer, it became clear to Bateman and Micromass technical manager John Hoyes that “this is an improvement over anything that existed before,” says Bateman.
One day in 1995, “I went off to the coffee machine and by the time I got back, there was this beautiful spectrum that just built up over a period of 20 minutes,” says Hoyes. “It was at this point we realized what wonderful signal-to-noise you got from this instrument.”
Over the next few years the QTOF became one of mass spec’s biggest success stories and gained a reputation for being able to identify a protein when all other instruments failed. “It was a great leap forward. It was one of these quantum leaps,” says GeneProt’s Rose. “It was like going from floppy disks to having a hard disk on your PC.”
By the end of 1996, Micromass had shipped only two or three instruments. In 1997 it shipped nearly 30, and the year after it approached 100 sales. In 2001 the company sold more than 200 instruments and sales continue to rise.
Micromass dominates the $150 million QTOF market with an 80 percent share, according to a 2001 research report by Strategic Directions, but not just because it was the first to manufacture a QTOF. The ABI/MDS Sciex QSTAR is a distant second with 15 percent.
“What bugged me was that the QSTAR had just come out, they were both about the same price,” — around $500,000 — “yet the QSTAR was clearly deficient as far as the software,” explains University of Georgia’s Orlando. Many customers who went with the Micromass instrument concede that perhaps the QSTAR performs a bit better. Still, “we’ve made decisions based solely on software, when we could have gotten a slightly better machine that had really awful software,” says Andrews, whose Micromass instrument was delivered in December.
ABI concedes that software hasn’t been its strongpoint. “We recognized a number of years ago that this was something people were saying about our platform,” says ABI marketing manager Dave Hicks. “They love the hardware, but that isn’t sufficient.” ABI has since poured a substantial amount of resources into improving its software. Some insiders say it is catching up.
Bristol-Myers Squibb’s director of proteomics Stanley Hefta says that as the amount of data increases, the quality of the software becomes more important. “It used to be that the hardware was the big seller for the mass spec. Now software is probably more important,” he says. “Software that gives you a capacity to automate things, to refine your database searches, and deal with integration is going to be very important.” In other words, machine and code can no longer be viewed as separate entities, but as two parts of a protein analysis system.
Yet for some, software, while important, is not the deciding factor. “The software on the Micromass was more mature, but I decided it was easier to improve software than it was to improve analyzer physics,” says Fenselau, who bought a QSTAR three years ago.
But any comparison is just a temporary snapshot of rapidly evolving choices. ABI has since improved its software and Micromass has improved its analyzer. “The manufacturers are always beavering away, making the next one even better,” says Rose. Yet waiting for the market to stabilize before making a purchase may be a bit like holding off on a PC purchase. “Mass specs are getting a bit like that. Every 18 months or so there is a pretty significant improvement,” says Rose.
One thing to consider if you’re thinking about purchasing a Micromass QTOF in the US: Earlier this year, a federal court ruled that Micromass infringed a patent owned jointly by Applied Biosystems and MDS Sciex that covers the way samples are introduced into some of the QTOF instruments. Micromass stopped shipping the infringing instruments and has begun shipping an alternative product without the offending technology. “If you look at the sensitivity, resolution, and so on, it’s almost indistinguishable,” says Micromass marketing manager Mark McDowell of the reengineered instrument. It’s too early to tell if customers will agree.
The MALDI-TOF/TOF has been the subject of much hype over the last couple of years. But like any other revolutionary mass spec that has hit the market, it won’t solve every problem. “We do not subscribe to the idea that one system can replace every other system for the entire proteomic application space,” says ABI’s MALDI marketing manager Paul Danis.
Although the MALDI option is now available on the QTOF too, a MALDI-TOF/TOF offers features such as increased throughput and higher energy fragmentation. But it is too early to tell whether these are true advantages.
“One of the issues for the TOF/TOF is, how do you feed that beast?” says Andrews, whose lab has the ABI TOF/TOF, whose trade name is the 4700 Protein Analyzer. Most labs do not have the capability to generate enough samples to take advantage of a potential throughput of 10,000 to 20,000 samples a day. The question is whether extreme high throughput is worth the extra cash. “I can argue both ways,” says Andrews. “I’ve advised people that rather than waiting a year and trying to raise more money to buy the TOF/TOF they may be able to get by with one of the QTOFs.”
On the other hand, “I’m a firm believer that even a small facility, if they thought about it carefully, could find a way to use that excess capability in very effective ways,” he says.
Jan van Oostrum, head of proteome sciences at Novartis, just put in an order for an ABI TOF/TOF because he was impressed with its throughput. “The throughput in the MS mode alone is a 20-fold increase over a MALDI-TOF. In MS/MS mode it is 50-fold over the QSTAR approach,” he says. Van Oostrum also likes that the TOF/TOF is like two instruments in one: “The MALDI-TOF/TOF allows MS and MS/MS on the same instrument.”
Whether high-energy collisions in the second TOF analyzer will really help proteomics researchers is a matter of debate within the community. The idea is that breaking a peptide into smaller fragments can generate more details and coverage of the peptide. “Is that an advantage?” says Fenselau. “Well, a lot of us don’t think so. It makes the spectra really hard to interpret.”
Andrews argues that high-energy fragmentation may be a big time saver. “It may ultimately change our strategy,” he says. Because the TOF/TOF can create small fragments, Andrews’ lab is experimenting with skipping the time-consuming step of enzymatically digesting the proteins before loading them into the instrument.
One difference between the two TOF/TOFs on the market is about $200,000. The Bruker UltraFlex lists for about $575,000 while the ABI 4700 Protein Analyzer goes for about $750,000. Another is the speed of their lasers. ABI’s 200 hertz laser accounts for its instrument’s high throughput. But Bruker’s John Wronka says his company’s instrument is faster even with a 20 hertz laser. “It’s not just the laser rep rate, it’s how fast you can get an MS/MS spectrum,” he says. “We can get an MS/MS spectrum with 10 to 100 laser shots at 20 hertz. Whereas the other instrument on the market, while the laser runs at 200 hertz, you need 1,000 or more shots.”
The two instruments also differ in their strategies for generating the peptide fragments necessary for MS/MS spectra. The first TOF of both instruments selects “precursor” peptides to shuttle to the second TOF where they collide with an inert gas and split into smaller peptide fragments — a process called collision-induced decomposition, or CID. The laser, however, gives some peptides enough energy to fall apart on their own before entering the second chamber. This is called laser-induced decomposition, or LID. The ABI instrument excludes LID fragments, and only uses CIDs in its analysis. The Bruker instrument, on the other hand, takes both into account. The jury is still out on which method gives better data.
Despite the number of choices already available, “so far nobody has invented the perfect instrument that’s good for everything,” says Genentech’s Arnott, who chairs the ABRF proteomics research group. So mass spec developers are feverishly at work on new combinations of ionization techniques and analyzers, putting as much as 20 percent of revenue into R&D. ABI/MDS Sciex has just jumped into the trap market with quadrupole/ion trap, or QTrap. Kratos has just created a MALDI-ion trap/TOF. Thermo Finnigan is experimenting with various other TOF combos. Yet if history has anything to say, newfangled mélanges are likely not to produce an all-around winner, just more choices. “Maybe we’re just all boys playing with toys,” says John Cottrell, a former general manager at Thermo, who now runs proteomics informatics company Matrix Sciences.
The instrument generally regarded with the most promise is the Fourier transform ion cyclotron resonance, in which ions trapped in an electromagnetic box continuously circle around in an orbital, resulting in extremely high mass accuracy. “I can’t wait to get my hands on an FTMS,” says Andrews.
“Without doubt in the near future we’ll probably have to think of buying an FT machine,” says Novartis’ van Oostrum. Bruker already sells an FTMS instrument that gets 50 times the mass accuracy of a QTOF and is working with Dick Smith of the Pacific Northwest National Laboratory on a model optimized for proteomics, expected to hit the market within two years. “Because of the mass accuracy and the resolution, you can take a mixture of 100 proteins, digest them all together, and identify all 100,” says Wronka. Thermo Finnigan is also developing an FTMS for proteomics.
But don’t expect it to be the be-all, end-all that quells the massive mass spec confusion. FTMS instruments are resistant to automation and difficult to use. “The world divides into those who believe it’s an industrializable machine and those who believe you sit one in the corner for the occasional really super-duper sample that needs a bit more resolution,” says Rose. “They’re not what I would imagine having 40 of.”
Instead, mass spec users should get used to the idea of a hodgepodge of mass specs for proteomics analysis. “There is no one machine that gives you high mass range, high resolution, high sensitivity, and reasonable robustness. And I don’t think there ever will be one,” says Rose. “The physics won’t let you.”
Here are some specs mass spectrometrists use when comparing the performance of various instruments.
Mass Accuracy: How precisely the instrument can measure a peptide’s mass. Often given in +/- Daltons or parts per million.
Resolution: Often related to mass accuracy. Literally, how well the instrument can distinguish two peptides of similar mass.
Mass Range: The upper and lower limits of mass that the instrument can measure.
Sensitivity: The lowest concentration of protein that the instrument can detect. Usually measured in moles or parts per million.
Throughput: The number of samples that can be analyzed in a given time.
Among the 189 mass spec users who responded to a GT survey, sensitivity was named the most important factor in purchasing an instrument. As one mass spectrometrist put it: “If you can’t see the ions, it doesn’t matter what accuracy you’ve got.” Reliability was second, followed by mass accuracy and resolution. The least important factor was the machine’s laboratory footprint. — AS
One performance characteristic often touted by vendors and users alike is mass accuracy. In fact, it is often the motive for the purchase of a QTOF over an ion trap, which can be a third of the price. But according to Ron Orlando of the University of Georgia’s Complex Carbohydrate Research Center, mass accuracy is often superfluous.
Orlando ran samples extracted from Arabidopsis through a MALDI-TOF and then adjusted the resulting spectra so that they were accurate down to five decimal points. “We gave them the exact mass they were supposed to have,” says Orlando. “No mass spec will ever get this data.”
Next he introduced mass errors by stuffing them through a random number generator, creating numerous sets of data, each with different amounts of errors: one part per million, five ppm, 10 ppm, 50 ppm, 100 ppm, 200 ppm, and so on. He used these data sets to search the databases to see how much accuracy is critical in order to get a confident ID.
“Much more than 50 ppm mass accuracy is not really going to help you a whole lot,” says Orlando. “And actually there was only a marginal difference between 50 and 200 ppm mass accuracy.” There was one exception: database hits did improve when the search was done with ProteoMetrics’s protein identification software.
Orlando is now doing similar experiments with tandem mass spec data, as well. “It doesn’t seem to make a whole lot of difference,” he says.
Although, the work has yet to be peer reviewed — “We’ve been doing this for two years and we’ve yet to publish any of it” — it makes one think twice before shelling the big bucks for better mass accuracy. “This is why ion traps can play the game.”
Except for exceptionally difficult proteins, ion traps, with a mass accuracy in the range of 150 ppm, can do the job just as well as the QTOF, which gets about five ppm. “You need to have some accuracy, but it doesn’t have to be a great degree of accuracy,” says Orlando.