What I want to talk to you about today is the work that we have--that we do using the MRM analysis, which is used for the quantation of proteins in this case

And this is the work that I've been involved with at the proteomics facility of the University of Victoria that is lead by Christophe Borchers [sp]

Most of my talk will focus on the development of an MRM multiplexed assay for the quantitation of 67 candidate cardiovascular disease biomarkers. And if you would like to have the details of this, they have been just published last week in this issue of Proteomics, which has focused on targeted proteomics

At the Evac GMBC [sp] Proteomics Center, we have focused on these multiplex MRM assays for the precise quantitation of targeted proteins, of specific proteins that we are interested in

And specifically, we focus on the absolute quantitation or the determination of the concentration of these proteins

And this is thanks to the stabilized up-labeled [sp] peptide standards that we use in all of our assays

At the center, we're expanding these--we are synthesizing more of these peptides for specific proteins. We have accumulated a large base of these, and we can classify these into disease-specific peptide panels

In the past year and a half, we've switched from nanoflow [sp] to standard flow UHPLC applicate [sp] methodology. And this has definitely increased our robustness of the analysis and has increased throughput as well as the sensitivity, as I will show you of the analysis

Part of this has been thanks to the Agilent 6490 Triple Quadrupole, which we found has really good sensitivity. And I will show you how we've applied these technologies to this CVD plasma protein assay as well as I will show you some work that was done by Andrew Percy [sp] in our group comparing standard flow and nanoflow methods

And our final goal right now, our approaches right now are to expand this--these plasma protein MRM assays to the top 500 proteins in plasma

And this is not only to be able to use these assays for validation, which is usually what has been happening at the facility where groups have come to us and they've initially done some global type of analyses studies, where they've found some proteins that go up and down. They would like to validate these

So, this work usually leads from studies such as iTrack into MRM analyses

But, by expanding these assays to such large amount such as these 500 proteins, we would also like to apply these two biomarker discovery type of applications and to--in the future to be able to translate these into clinical type of research

Of course, we are strong believers in the use of these peptide standards, not only for the absolute quantitation of proteins, but the use of these standards also greatly increases specificity. This is shown in this example where these are MRM chromatograms

And although they're very specific, very often, they can have multiple peaks. And without the presence of a standard, you would not know which of these is your specific target of interest

These are--although we define our data in the end as absolute quantitation, having this standard in here gives us--allows us to have a relative measurement, which is the area of the natural target over the area of the peptide standard

And this type of normalization also increases the reproducibility, which is the second advantage of using standard peptides

Here's an example that was published, was done by--in our group by Kozik [sp]. This is analysis of 45 targets and 12 replicate analyses

And in blue here is shown the variation of just the natural targets if you just measured the absolute signal intensity of the natural analyte

But, if you normalized this data using standard peptides, the coefficient of variation drops below 20 percent. And this can be further reduced by balancing the peptide standards to their natural levels

A third advantage of using cis peptides is to--is for the detection of interference. And this can be done by comparing the relative response of the cis peptides to the natural analytes across multiple transitions for that specific peptide

And any types of signal addition or depression would be revealed as a change in this relative response

So, we have been expanding these MRM assays at the facility. We've--we're able to target over 700--over 300 proteins or so. And these peptide panels can be grouped into specific disease panels

But, I will focus on one of them, which is the MRM CVD plasma protein assay, which is composed of 67 candidate bio--CVD biomarkers

This has initially been taken from the list generated by--under Lee Anderson [sp] in 2005, where he grouped 177 proteins as possible cardiovascular and stroke disease biomarkers

And he represented these by multiple peptides from one to six peptides per protein for a total of 135 peptide targets in this single assay

We synthesized these stabilized up-label peptides at the Protein Center. We purified these. And we determined their absolute concentration using capillaries and electrophoresis as well as amino acid analyses

And this allows us to know how much we add to our sample and determine the absolute quantitation of our protein based on that

And the protein--the peptides are, of course, used as the surrogate indicators of the protein

We then optimized the signal that we obtained in our MRM analysis. And we do this by injecting--by doing direct infusion to mass spectrometer using these pure synthesized cis peptides

We determined what the optimized precursor ion charge data is. We also empirically determined which fragment ions are the ones that give us the highest signal. And we empirically optimized the collision energy that will give the highest signal for these specific fragment ions

We then choose the top three MRM transitions, which are free of interference per peptide to form the assay. And then we balance these cis peptides in the mixture. And they are balanced through the natural analyte levels of our target

We also use liquid handling robotics, where we found this clearly increases the throughput of sample preparation. It increases reproducibility

And everything is--the whole process of plasma sample denaturation reduction, alkylation, tripsin [sp] digestion, the addition of the cis peptides posed digestion as well as the solid phase extraction is done automatically

In our multiple play, we can also generate--we also often generate a standard curve, which is used for the determination of the absolute concentration of the protein targets

So, then we use the Agilent 1290 UHPLC and the 6490 Triple Quadrupole to analyze the samples. We found that the UHPLC definitely adds a lot of robustness to the whole analyses

In the top graph, you have the absolute signal stability. This is the absolute signal of the cis peptides over 85 different analyses for the 135 peptides. And this is without normalizing the signal. This is without normalization

And you can see that there's basically no fluctuation in the signal like you would see in the nanoflow type of analyses. Everything's basically below 20 percent CV

The retention time stability is extremely high on the UHPLC. Over the 85 sample analyses, there's basically virtually no shift in the retention time. Everything is below half a percent CV

This type of retention time stability really allows us to decrease--really allows us to increase the throughput because all of our MRM transitions are scheduled

Basically, we have small retention time windows where we have--we try to fit as many transitions. By having stable transition, by having stable retention times, we're able to maximize the amount of transitions and therefore the number of targets that we can see within a given time

This is an--you can see here that, within a 30-minute HPLC method, all of these 135 peptide targets are within the 25-minute time

You can also see here the great dynamic range. The standard peptides are shown in red. The analytes are shown in blue

The analytical reproducibility of the assay is also really high. We can see that, by doing multiple injections of the same samples, the analytical reproducibility basically for all the 135 peptides is below 20 percent CV. And for 112 of these 135 peptides, it's actually below 10 percent CV

And we've also observed very high, very large linear dynamic range. And this is probably due to the new technology of the dual ion funnel

We did standard curves where we had the standard curve composed of nine samples. Each of these samples had the same amount of matrix. So, each of these had 10 micrograms in the column of a digested plasma sample

And we spiked different amounts of the cis peptides spanning at 500,000-fold range. And we can see that the--we observed the linearity up to five orders of magnitude for 12 of these targets. Most of these you can see in this graph are 10 to the four, 10 to the three and 10 to the four, which is quite high

These graphs also show the lowest level of quantitation was determined from these graphs. And that is--this is--these are just four examples out of these 67 proteins, spanning the entire range, all the way from albumin to low-level targets, such as engal [sp] DLOQs

We defined our LOQ as based on the reproducibility of the measurement below 20 percent CV and with an accuracy of 20 percent of what is expected. And that is shown by the black arrows. The natural--the observed natural concentrations are shown with the red arrow

We also tested the intra- and the interday assay reproducibility of the assay. And we see that, if we analyze 85 separate sample digests within a single day, 118 of the 135 peptides fall below 20 percent CV. And if we do this on three separate days, it is also quite similar

This graph shows the observed protein concentration in blue for the 67 protein targets. And in green, we've just compared these to the published literature values, the normal plasma concentrations in published literature

But, because these can vary quite a bit, anywhere from 10-fold to 400-fold for proteins, such as apolypo [sp] protein A, these are just an indication

And you can see that over 50 percent of them were under--within these--within the twofold of the expected literature value

But, what I wanted to draw your attention to is the LOQ of these 67 protein assays. And these are shown in the brown squares

And we've observed that, for 81 of the 135 peptides, the LOQ on column was actually in the atomal [sp] range or under one femtomole

And for--if we translate this to protein concentrations, we can observe the LOQ for 40 percent of these 67 protein targets is actually between two and 100--between the two to 100 nanogram per plasma--per milliliter of plasma

And this could've been even better, but the standard curves for such proteins, such as serum albumin, were centered on the analyte's concentration. So, the lowest level was not tested. So, this LOQ number could be even higher

But, what this leads us to is that it is possible to--using this type of a setup where we've actually--this is nonfractionated plasma that has not been depleted that we can possibly quantitate down to around 10 nanograms per mL for certain proteins

These are some of the--these are some data from the lowest targets for 91 nanograms per mL and 10 nanograms per mL for insulin-like growth factor binding protein. This is definitely the limit

But, maybe between 10 to 100 nanograms is what you could quantitate accurately using just the single analyses of nonfractionated, nondepleted plasma

Therefore, if you--this is a graph showing the 177 CVD and stroke candidate biomarkers as defined by Lee Anderson in that 2005 paper

These range from picograms per mL for cytokines all the way to serum albumin. And you can see that, as shown in red, these are the targets that we've detected. We can quantitate seven out of the 11 orders of magnitude. And the LOQ's are even one LOQ--one magnitude lower

We also believe that we should be using multiple peptides per protein. For 29 of these 67 CVD markers, we had multiple--we have two or more peptides

The data is shown here. The arrow bars show the standard deviation

Some of these protein--some of these peptides agreed quite well, and some did not

Of course, these proteins, once they're digested, the peptides are at least at different efficiencies. Some are close to 100 percent. And some are really not. And this can be observed only when multiple peptides are compared per protein

Here's an example where not all peptides agreed for alpha 1 antitripsin. One of these peptides actually differed by almost a 10-fold difference if we determine the absolute concentration

However, this is not--this is still useful data, and this can be shown in this analysis of 90 CVD patient samples. And you can see that, despite the absolute concentrations not agreeing between the peptides, the relative response or the relative data is actually still quite valid, as is shown by this linear correlation between the analysis of these different samples

So, using multiple peptides, we could look at a list of the peptides from that protein. We could discard some that are not efficiently digested

Another possible effect could be that there is something happening for this peptide, such as a PTM that is unknown that is decreasing its amount in this MRM analyses

Another advantage of using multiple peptides per protein is that it could reveal potential sample interferences or problems in sample digestion, as is shown by this red arrow, where among these 90 samples, we could see that one sample was off based on multiple peptide analyses

But, without using this standard flow, this higher flow rate, this would come at a cost of sensitivity. And Andrew Percy in our group did this work, which has been published just a couple weeks ago in Analytical Biochemistry, where he basically compared the standard flow analysis, as has been done --as I showed previously, to nanoflow LC

And he used the same detector, the same Agilent 6490. So, for the nanoflow LC, he used an HPLC trip cube [sp]. And he developed 30-minute methods for each, obviously at different flow rates

And what he observed--he then optimized both platforms. He optimized the liquid chromatography steps. He optimized the MRM transitions for each platform. And he optimized the maximum loading capacity. And this is what is shown here as the determination of the absolute loading capacity

He determined that the optimal loading capacity for standard flow rate would be 10 micrograms, as was used for the previous experiment. And one microgram was the optimal loading amount for the nanoflow

Using these optimized platform settings, he then showed that, actually, the linear dynamic range for the standard flow was higher than the nanoflow

The retention time stability, as I showed before was not a surprise at all. And the CVs were much, much lower. This was especially evident in--where different samples were being analyzed at--in different concentrations

But, what's surprising, was good to see is that the LOQ was actually better on the standard flow method than the nanoflow method

So, if we load 10 times more, which is what we did, on the UHPLC method, if you defined the LOQ on a per volume of plasma basis, so if you define it on how many nanograms of your target can you see in a milliliter of plasma, the UHPLC, the standard flow method actually came ahead

So, with these ideas, the next steps of the Evac Proteomics Center are to expand these assays. We know that around 2,000 proteins can be identified in human plasma using MS. And over 500 of these are above the 10 nanogram per mL range

So, the next steps at the center are to target the top 500 proteins with four peptides per protein. And possibly in the future by using specific smaller protein panels and also with the addition of some fractionation, we could target below the 10 nanogram per mL group

I also wanted to invite you to a talk that's being given by Andrew Chambers [sp] from our group. He will be talking about how to use these methods of MRM analysis to drive blood spots [sp]. So, I invite you to go and see that

I want to thank our funding agencies, Genome Canada, Genome BC, WD, our collaboration with proof, and of course, Agilent, who has provided the--who's generated some great equipment, and they've been very helpful in helping us with some software problems--not problems, but demands that we've had from them. So, thank you very much