By Tony Fong
This story originally ran on June 24.
Name : Iuliana Lazar
Position: Assistant professor, department of biological sciences/Virginia Bioinformatics Institute, Virginia Polytechnic Institute and State University, 2003 to current
Background: Postdoc, Oak Ridge National Laboratory, 1998 to 2000; postdoc, Sensar-Larson-Davis, 1997 to 1998.
As a strategy for screening panels of protein biomarkers, researchers at the Virginia Bioinformatics Institute have developed a method that allows for the detailed profiling of complex cellular extracts of proteins.
Described in an article published in the online version of the Journal of the American Society for Mass Spectrometry, the method is based on iTRAQ technology combined with pulsed Q dissociation, or PQD, a novel approach for precursor ion activation/dissociation in ion trap mass spectrometers that enables the trapping of low m/z ions.
The one-step approach was developed for differential expression profiling of complex proteomic cellular extracts, the authors said in the study, and was evaluated on the MCF-7 breast cancer cell line cultured in 17-beta estradiol and tamoxifen. Using their method, the VBI team detected 530 proteins and quantified 230 of them. Eventually, they selected 16 differentially expressed proteins, "demonstrating the potential of iTRAQ/PQD-MS for biomarker discovery applications," according to the researchers.
ProteoMonitor spoke this week with Iuliana Lazar, the corresponding author for the study, on the development of the method. Below is an edited transcript of the conversation.
Briefly describe your work for me.
Well, the main purpose was to develop and evaluate a methodology for protein differential expression analysis in cancerous cells.
We know that as a result of a disease, cancer cells could express different proteins than normal cells, or could overexpress them … and we should have a way of evaluating that.
Now, our detection tool is a mass spectrometer. This is a very sensitive detector. It enables us to detect hundreds and thousands of proteins or their corresponding peptides, and basically this is good enough for a preliminary, qualitative profile.
Then, once that is accomplished, we need to do differential analysis, and that was the purpose of this particular paper.
There are a variety of methods for looking at differential expression analysis — they can be non-labeled methods where we quantify peptides, or peak areas, or spectral counts, or peak intensities in two different samples.
Or [we can use] labeled methods where the samples, the cellular extracts, are labeled with isotopic reagents, and during analysis, the relative ratios of these isotopically labeled peptides are monitored. Based on these ratios, we can estimate whether proteins go up in abundance or drop in abundance, for instance, in cancerous cells versus normal cells.
Now, this is theoretical. Practically, there is a large variability in these measurements, so most of the work that we performed was focused on evaluating the methods.
What we did in this work … is culture breast cancer cells in the presence of estradiol and in the presence of tamoxifen. … After culturing, we processed the cellular extracts, [and] we labeled them with isotopic labels. In this case, we used the iTRAQ technology … and we took two replicates of each cell state, or two replicates of cancer cells cultured in the presence of estradiol, and two replicates of cell cultured in the presence of tamoxifen.
Each replicate was labeled separately and the four samples were mixed, and we ran them through liquid chromatography and mass spectrometry.
Overall, we were able to identify about 530 proteins, so we are confident about the identification of these proteins. But then, to increase our confidence in quantitation, we qualified only about half, only about 255 proteins for looking at the protein-expression levels.
In the first place, we looked for proteins that had at least two peptide matches. This dropped [the number of protein identifications] from 530 to about 300 proteins, and then we manually eliminated some other proteins, and we were left with a final list of 255 proteins.
[ pagebreak ]
For these proteins, we generated four ratios — the two estradiol cultures were used as controls, and the tamoxifen-treated cells were used for the treatment [samples], and we were able to generate four ratios for each of the peptides, [and] respectively, for the proteins.
And ultimately after statistical treatment of data, it turned out that we felt confident in changes that were at least two-fold in protein-expression level in order to select these proteins that were up- or down-regulated.
How did you choose this two-fold change as the threshold?
The way [we did it] is we spiked the samples with standards, with a set of nine bovine proteins, and based on the variability of these proteins, it worked out that we need at least a two-fold change in order to claim that the protein, indeed, changed expression levels.
In addition … we empirically decided that at least three of the four ratios should also show a two-fold change in expression level.
And based roughly on the criteria, about 16 proteins qualified as up- or down-regulated.
If you decided to have all four ratios meet this two-fold change as a criteria, what would that have meant to your results?
That could have been done also, and this is something that we still need to look at. This was our first testing of our methodology. We decided for three mainly by looking at the raw data that is not included in the paper.
But we are actually looking to evaluate this method in more detail and possibly make it even more restrictive, and in that case, we could look at [all] four ratios.
From the top of my head I do not remember [the specific details] but I think if pretty much three of the ratios worked out, it's almost the same as with four.
Why did you choose to address your work to the iTRAQ method when there are so many other methods of quantitation?
We are looking at other methods as well. For example, if you do label-free methods, you have the sample and you just process it.
I can process, for example, this year one sample, next year another sample, and five years later another sample, and then compare them based on certain criteria. That's pretty much a semi-quantitative method. It allows me to do some comparisons, but it's not so reliable.
Since we are not labeling any of the peptides, we are, overall, able to detect a much larger number of proteins. However, the confidence is much smaller. And we use that method for a rough estimation of protein abundances.
This was our philosophy: Let us do something semi-quantitative, then to get more confidence, run a different method such as iTRAQ.
The advantage of [the labeling method] is that … you get more reliable results. On the other hand, due to the way the technology works, we can identify a lower number of proteins, so we are losing somewhat.
In addition, to do the labeling techniques, we need to process the samples together. So every time that I would like to run a comparison, I should always relabel and run together all samples that I want to compare.
[That's a disadvantage] but it's compensated by [the], let's say, reliability.
There is a third method, a targeted analysis method that we are also looking at, where instead of looking for everything that would change in a sample, you are looking only for specific proteins or peptides. And then you can use label methods. It's just that it has the advantage of knowing exactly what you are looking for. You can get much more reliable quantitation.
[ pagebreak ]
But you would run the risk of missing something because this is a biased approach?
Well, it would have a different purpose. For instance, the iTRAQ method is more appropriate for discovery. So … when I'm just interested in finding out what proteins change expression levels [and] I don't know what would change, what I want then is to be able to detect as many proteins as possible and be able to quantify as many proteins as possible.
Let's assume these methods give you an idea about the proteins that would change expression levels in certain cancers. Well, next maybe I would like to see whether those proteins change expression levels in other cancers or in other treatments.
In that case, to speed up the process, instead of looking for everything, I would have a targeted approach and then I would be looking only for those proteins.
In your experiment, did you use a specific plex rate of the iTRAQ?
We used a four-plex. At the time we ran the experiment, only the four-plex was available.
Would your results be different if you ran the eight-plex iTRAQ?
I don't think they would. Every time when you're trying out something new, things happen [but] it should work in the same way.
Is your approach more applicable for an electrospray ion source mass spec than a MALDI, or equally applicable for the MALDI?
I think it would be applicable to a MALDI platform, as well, especially for the iTRAQ. Actually, the original papers with the iTRAQ were performed with MALDI-TOF/TOF instruments, so I guess if anyone wanted to use this approach, they could use it.
Would there need to be significant change to the workflow?
I don't think so. Now, the data sometimes depend on how the manufacturers build their software and … in what form they allow you to extract the data. For instance, the software generates the ratios. If you want to extract raw ion intensities, then probably one needs to build or develop a short Perl script to extract those ion intensities from the raw files.
It can be done but it's not straightforward. Now, I assume that … with the MALDI-TOF/TOF, the software that is supported by the manufacturer will [also] allow you to do certain things. … If they allow you to access the raw data, then probably you need to develop your own software.
Did you do an analysis on these 16 proteins that you found to be differentially expressed? Can you share any data or observations? Were any of these proteins novel?
No, they are not novel. … The proteins that were identified and quantified were processed [using] the Gene Ontology website and were assigned to categories of proteins that are relevant to processes related to mRNA or dRNA processing, to transcription, translation, some to signaling, metabolic processes, apoptosis, and so on.
One thing needs to be taken into account — we are interested in the effect of tamoxifen on cells. Once you add such a drug to a cell, several things can happen, and the change in expression level doesn’t happen necessarily in response to the treatment.
And we identified, actually, four different processes that can result in changes in protein expression levels. One is a direct effect of the [tamoxifen] effect, and that's the intended effect.
Another process is the result of cell response to stress. The other would be the result of cell accumulation in the G1 phase of the cell cycle as it has been shown that the way [tamoxifen] works is that it will … induce apoptosis.
Or it could be an artifact and wouldn't be a change in expression levels but would be a change in post-translational modifications. But since we didn't look for post-translational modifications, it would appear as a change in expression levels.
Now, one of the most interesting proteins that we found and that shows up-regulation was BAG-3 in the [tamoxifen-treated] cell, and it has been shown recently that BAG-3 has anti-apoptotic activity, and it's assumed that BAG-3 could sustain cell survival in response to stress.
But this was observed in other studies and in other treatments of other cancer cells.
It's believed that BAG-3 could be impaired in response to treatment with drugs and eventually could represent a novel therapeutic target.
It was interesting to find this protein was up-regulated in our cells as well, and this is something we would like to further study and perform some biological replicates and see whether we can confirm that this particular protein is indeed up-regulated.