High-throughput proteomics takes on new meaning as liquid chromatography and mass spectrometry join hands, but with such a powerful coupling comes limitations.

Once upon a time, there was liquid chromatography (LC) and there was mass spectrometry (MS) and people used both, separately, to fractionate and then identify, respectively, large numbers of proteins from complex biological samples. And there were those that were happy enough with these techniques to let them be. However, some people had gripes with the way sample loss during the chromatography step was decreasing the sensitivity of detection by downstream MS applications. The result: Multidimensional Protein Identification Technology, or MudPIT.

The MudPIT technique involves digesting the protein sample into its constituent peptides, which are then separated with two liquid column chromatography steps: The first being a strong cationic exchange, and the second being reversed-phase high performance liquid chromatography (HPLC). As the peptides elute from the second column, they are sprayed into a linear ion trap mass spectrometer. The first MS scan assigns each peptide a mass/charge ratio. The most intense peptide signals are then fragmented in a second MS/MS scan, which assigns each peptide a unique “fingerprint.” The fingerprints are then fed into bioinformatics databases that reveal the protein’s identity.

But the development of MudPIT was met with obstacles. For example, with ion exchange, there is a need to use salts like sodium chloride and potassium chloride in wash and elution buffers, but these cannot be introduced into the mass spectrometer, says John Yates, PhD, professor of cell biology at the Scripps Research Institute, La Jolla, Calif. So the Yates lab had to find a way around this problem, eventually discovering that ammonium chloride is a more MS-compatible salt. Although many of the kinks had been worked out before MudPIT became commercially available, some obstacles remain, which can hinder the optimal use of this important proteomics technology.

A Useful Tool
People make the comparison of MudPIT to 2D gel electrophoresis, but is it a fair comparison? “It is like comparing apples to oranges,” says Yates. He cites two major problems with 2D gels. First, 2D gels don’t have a great dynamic range. So the abundance of proteins that can be seen from a complex mixture is not as good as people had previously thought. The other problem is that the number of proteins identified by a 2D gel is much lower than the number identified by MudPIT. Yates does point out, however, that with MudPIT, the researcher is not looking at intact proteins, making it especially difficult to identify isoforms. According to Yates, another strategy that may be employed involves running intact proteins on an SDS gel, cutting out each section of the gel, and then running one-dimensional chromatography on it. However, this is not the best approach, either.

click to enlarge 
The workflow for MudPIT is shown. Proteins are prepared, digested into constituent peptides, which are then separated by 2D chromatography and analyzed via tandem mass spectrometry. SCX=strong cationic exchange; RP=reversed-phase. (Source: Emily Chen, PhD)
The problem is that you’re biased with a gel, says Ahna Skop, PhD, assistant professor of genetics and medical genetics at the University of Wisconsin, Madison. “When I was a postdoc, we did an experiment in which we cut out an area of the gel that was not stained with dye but still contained two or three proteins in it.” So the low-abundance proteins were not detected because the dye staining method was not sensitive enough. But MudPIT technology is sensitive enough to detect low-abundance proteins, and the results are not biased.

“The advantage of using MudPIT is that we don’t need to use a gel-based approach,” says Emily Chen, PhD, postdoctoral fellow in the Yates lab at Scripps. “There are a lot of disadvantages in the gel-based approach including solubilization and just getting the protein into the gel.” She goes on to say that amyl urea or urea could be used to make the protein more soluble, but if the protein does not get into the gel, it cannot be detected. And these solubilization problems are not an issue with MudPIT because denaturation is not an issue. As such, peptides stay in solution.

Chen is applying MudPIT to problems in cancer biology, specifically looking at protein regulation; that is, how long proteins remain in the cell once they are expressed. Of course, this is important for many cellular processes and is thought to be one of the mechanisms in the development of tumors and cancer. “We found that we can use MudPIT to study the proteasome system, which is regulated by ubiquitin. We know that in our system we are dealing with ubiquitination and degradation, so we do time courses,” says Chen. With MudPIT, she can not only identify ubiquitin-modified proteins, but also determine how that modification occurs over time.

The mass spectrometry component allows MudPIT to be quantitative, a significant attribute over 2D gel electrophoresis. “These [mass spectrometry] instruments are developed with such precision and care that they sample information in a very quantitative fashion,” says Michael Washburn, PhD, director of proteomics at the Stowers Institute for Medical Research, Kansas City, Mo. However, he explains that it is because MudPIT produces a lot of information that has to be dealt with in a computational manner that the bioinformatics side of MudPIT is its most challenging aspect.

With MudPIT, researchers can view 2,000 to 3,000 proteins at the same time. That’s heavy-duty bioinformatic data! Luckily, the computer program lists all of the proteins and then compares each fraction, determining which proteins are unique to that fraction, and which appear in other fractions as well. These proteins can then be organized into functional groups based on gene ontology or by using other databases such as BioCarta, which organizes proteins according to the biological pathway(s) in which they participate.

One of the weaknesses of MudPIT is that bioinformatics has not caught up with the technology, says Chen. “How do you go from say 1,000 proteins to organizing the data and being able to present a solution or approach a biological question?” Chen asks. She goes on to say that in order to make sense of all of the data, it takes months to do a literature search to match what’s out there and to look through all of the databases because the search process itself is not automated. The reason for this slow process is that there are a lot of databases to accomplish this bioinformatics feat but that they are not integrated.

Online, Offline
However, generating that list of proteins takes time and a lot of chromatography. MudPIT is basically a two-dimensional chromatography technique in which both dimensions can be done online. In other words, the fractions collected are sprayed directly into the mass spectrometer. According to Yates, this online feature prevents sample loss that occurs when peptides in the sample come into contact with the surfaces of the chromatography apparatus. A consequence of this feature is that MudPIT is more sensitive and has greater dynamic range than some of the other methods, such as an offline ion-exchange approach.

“Before we started using MudPIT, we typically did one dimension offline and the second dimension online,” says Richard Somiari, PhD, president and chief scientific
click to enlarge 
The theory of MudPIT separation is shown. A complex peptide mixture is loaded onto the strong cation exchange column (SCX) where charged peptides are separated; uncharged peptides then enter the reversed-phase column (RP) where they are separated. Eluted peptides are then sprayed directly into the tandem mass spectrometer (MS). (Source: Emily Chen, PhD)
officer, ITSI-Biosciences, Johnstown, Pa. So, having both dimensions online can reduce the loss of material that commonly occurs in chromatographic processes. “The challenge is choosing a method that allows you to achieve sufficient resolution in the first dimension,” Somiari says. He adds that even if this is achieved, the rate-limiting step in the whole MudPIT process will still be running all of the chromatographic fractions through the mass spectrometer.

Yates points out that the 24- to 36-hour time demand of MudPIT is one of its weaknesses. And most of this time is taken up by the multiple chromatography steps, with up to 24 hours just taken up for a 12-step gradient. “But this is comparable to offline chromatography systems where people collect and exchange fractions offline and then run them through the autosampler,” says Yates. The bottom line is MudPIT should be faster regardless of its dependence on the notoriously slow ion-exchange chromatography technology.

“Listing the weaknesses of MudPIT is like saying what’s wrong with ion-exchange chromatography,” says Somiari. However, he says that it is faster to use MudPIT than to use two different chromatography systems that are not linked. And since the whole mantra of proteomics is speed and throughput, he goes on to explain that, if an excessively long column or an excessively slow flow rate is employed to produce better resolution, there will be a concomitant decrease in both speed and throughput.

But speed and throughput mean nothing without high resolution, and resolving peptides that are too similar in molecular weight could cause a potential problem in MudPIT. However, Chen explains that although two proteins may be similar in molecular weight or isoelectric point, once they are digested, the peptide properties might not be the same. But some peptides might be too similar in mass to resolve easily using MudPIT, especially if they are derived from isoforms in which there are regions of amino acid homology. And the only solution here might be to modify the chromatographic process to achieve higher resolution.

Improvement Needed
Another problem with MudPIT is that “one size does not fit all,” so all samples have to be looked at on an individual basis, says Chen. One example she points to is the serum sample. “There are some basic procedures you do to resolve complexes. You have cell lysate that is pretty robust. We know how to digest. We know how to resolve. But with serum, for example, you have to optimize this procedure.”

Data validation is one way to optimize a technique. MudPIT results must be validated in some way to make sure that the lists of proteins generated by this technique are really expressed in the biological system under study. But there are different ways to validate these results. Skop purifies the mammalian protein complex of interest, identifies the proteins using MudPIT, looks for homologues in the roundworm Caenorhabditis elegans, and then uses RNAi to knock down their expression. Specifically, Skop works on cytokinesis, studying how membrane cytoskeletal proteins interact during the late stages of cell division. She is continuing to study the proteins identified in her initial MudPIT screens, indicating that validation is a very important part of using MudPIT.

Quantitative data produced by MudPIT should also be validated. By using the spectral counting approach, Washburn is able to normalize MudPIT data. “That gives us a really good way to do quantitative comparisons in a label-free manner,” says Washburn. So-called “labeled” proteomics involve things like labeling the peptides with radioisotopes (13N or 15N) or other non-radioactive isotopes (iCAT or iTRAQ). In contrast, for the “label-free” proteomics approach, the sample is analyzed without any isotopic treatment. “So by using the spectral counting approach and focusing on replicates—both biological and technical replicates—we can ensure that we can do very good label-free proteomics.”

There is another weakness of MudPIT, referred to as the abundance problem. “You are not really able to quantify proteins all that well with MudPIT; it’s just not that sensitive,” says Skop. However, she says, this problem can be remedied by labeling the peptides using the SILAC method before attempting to quantify them. There are many who would like to see an improvement in the sensitivity of MudPIT, so that more low-abundance proteins can be detected. Skop says that the improvements she would like to see with MudPIT in the future depend on the biological questions her lab may study in the future. She thinks that MudPIT is a fantastic technology in which to investigate biological questions, but has its limitations as all proteomics techniques do.

But some have more specific improvements in mind. “For now, I think MudPIT still represents a good proteomics platform that increases resolution compared to an independent LC/MS. But I think it still needs improvement in resolution and speed,” says Somiari.

According to Washburn, there are significant advances being made in HPLC, such as better high pressure systems used for sample delivery. There are also advances in chromatography, in general, where people are making smaller reversed phase particles and smaller strong cationic exchange particles so that they can take advantage of the newly-developed high pressure systems. There are also compelling advances in mass spectrometry as well, he says. “When you add it all together, the biggest need for MudPIT technology is really at the back end, where the field needs some very good comprehensive bioinformatics solutions for assembling all of the data sets and doing proper statistical analysis of data sets.”

In summary, the strengths and weaknesses of MudPIT are clear to see. And there surely needs to be improvement in the separation and detection components of MudPIT technology. Undoubtedly, advances in LC and MS technologies will simply improve what is already a very strong tool.

This article was published in G & P magazine: Vol. 7, No. 1, January, 2007, pp. G4-G8.