SRM calibration concerns

[lmh]

I'd really appreciate some second opinions:

I'm setting up an LC-MS method for "B", which is an aromatic thing approx 200Da with heterocyclic rings, no bonds of purely single-bond character, and no obvious leaving groups. My instrument is an ion trap. Not surprisingly, "B" doesn't fragment at reasonable energies, and is detected dreadfully at very high energies.

Question 1: is it OK for me to quantify a "fragement" ion that is actually the precursor? I've never done this before, but because the molecule is so tough, doing SRM of M -> M is actually adding selectivity and removing noise. I have about 6 peaks in full-scan MS2 of the precursor ion in real samples, but if I select the parent mass and use an energy at which everything else fragments, I have only one peak remaining, with good S/N.

Question 2 is about internal standard calibration. My client has added an internal standard "A" which is closely related to salicylate, and which I think was originally put in the method because the method started out as an assay for salicylate. I recover about 60% of A (external standard calibration) from all his samples. He spiked two of his samples with B, of which I recover about 50% (by external standard calibration). My client naturally assumed that he could scale up my reported amounts of B by 100/60, and get quite a happy result.

The problem is that when I do internal standard calibration "properly", with a set of standard containing increasing B and constant A, plotting cali-curve of area ratio versus amount of B, the answer for recovery of spiked "B" is still only 50%. The reason is that in the standards, when I add a lot of B, the signal for A starts to fall. It can't be cosuppression because the two chemicals do not coelute. I suspect it must be something chemical going on. On the other hand, the spiked samples show the same signal for A as all the others, rather than showing an extra-low signal for A. Therefore the effect is either not additive, or the real samples contain something that stops B from removing A. (Or something very strange happened and the two samples that were spiked just happened to have a much better recovery of A before they were spiked... highly unlikely!). At the level my client chose to spike, the signal of A has already fallen to about 60%, so the calibration curve assigns the 60% signal of "A" to being caused entirely by the presence of "B", and concludes there was no loss during extraction.

My feeling is that I can't trust internal standard calibration using "A" on these samples for measuring "B". I don't really believe that there is no loss during extraction, or no loss of efficiency of measuring A in real samples (based on what I see in all the un-spiked samples). Since the calibration curve for B tries to behave otherwise, I don't really believe it's working. The bottom line is that I'm concerned B isn't sufficiently chemically similar to A.

I'd appreciate any thoughts on (1) what might be going on, chemically/physically, and (2) am I right to abandon this internal standard? What would you do, faced with this sort of situation? I'm a bit puzzled.

Thanks for reading this far!

[Don_Hilton]

Can you try to improve recovery of analyte and internal standard? Small changes in matrix may make the ratio recovered vary slightly - or more than slightly - which will affect the accuracy of your results.

The best internal standard is one that is chemically similar to the compound of interest. Variations in matrix will affect both more similarly - or so the theory goes. What you have sounds risky to me.

On the mass spec technique - I'm not an MS/MS guy, but what you have sounds intriguing. And if it will validate, why not? Nothing like an after column cleanup!

[zokitano]

Hello Imh,

Regarding the internal calibaration issue, i would also suggest to work on your extraction procedure to increase the recovery and/or to find close-structurally similar compound as IS for your analyte (as suggested by Don) and if you're able to purchase a isotope-labeled analogue of your analyte it would be the best approach.

Regarding the SRM question, i can say people are using same product ion (after no fragmentation in the collision cell) as precursor ion. In one article recently published in Journal of Chromatography A:

Journal of Chromatography A
Volume 1217, Issue 3, 15 January 2010, Pages 312-328
Title: Metabolic profiling of intracellular metabolites in fermentation broths from β-lactam antibiotics production by liquid chromatography–tandem mass spectrometry methods
Authors: Beatrix Preinerstorfera, Simone Schiesela, Michael Lämmerhofer and Wolfgang Lindner

in this article authors are using same ions as product and precursor, naming the SRM mode in this case as "pseudo MRM" or "psuedo SRM", because they are detecting at the end the same precursor ion after passing through the collision cell (no fragmentation). I think that they were using this "pseudo MRM" approach because the product ions of some of their analytes were so small in m/z values, close to the low cut-off of the filtering quadrupole. If they used the small m/z product ions, they could not have obtained so sensitive detection of the low MW analytes (such as oxalic acid, if i remember good).
I think this issue is very close to your question.

Best regards

[lmh]

Thanks enormously both replies! I'm relieved that using a molecular ion as a fragment is OK, and very grateful for the reference (sorry, I should have done more literature searching before asking, but wasn't quite sure how to search for that one...). The sample prep is out of my control, unfortunately, but I will definitely impress on my client that it would be a good idea to see if we can improve recovery. I will also check recovery internally to my lab by spiking some samples on arrival, just to make sure there's no losses in anything I'm doing. I'll recommend we look for an internal standard that shows strong similarity to the analyte. As usual, cash is going to be a problem for the labelled version; I don't think the chemical is commercially available even in unlabelled form.

[Alp]

I was unable to get a mental picture of exactly what you have done to try to discover the source of your problem. I am not clear on if you have tried the following...

I have found it useful to extract blanks, with no internal standard and no analyte(s) and then reconstitute or fortify the blank sample with concentrations of the internal standard and/or the analytes to theoretical 100% recovery level.

Run these samples with some normally extracted samples that contain only internal standard or only analyte and samples that contain both. Also inject pure solutions at theoretical 100% recovery concentration. If you have ionization suppression or matrix effect from the matrix of the biological samples it will be evident when compared to the pure solution samples.
Running only one (internal standard or analyte) in extracted samples will also give you an idea if the really is some sort of chemical reaction going on. You may even want to try running some samples prepared in water (if possible) through the entire extraction process to see if there are differences.

You might also wish to look into whether your compounds "stick" to any plastic or glass containers used during sample prep. and make sure you have benchtop stability and autosampler stability for your extractions.

What happens if you dilute a solution with high conc of B with the usual amount of A? Say 1:1 dilution and 1:10 dilution. Does the ratio of A to B change?

Good luck

Alp

[lmh]

Thanks for the suggestions. All I've got so far is natural material containing little or no natural "B", spiked at the begining of extraction (by my client) with a very large overdose of B, and the typical internal standard spike of A. At my end, I only have a variety of standard curves made up in the absence of any natural material whatsoever. Yes, it would be interesting to try blank samples extracted without A or B and add them before analysis; it's something I can do without expecting my client to put much extra work in, and it will differentiate between problems during extraction and problems caused by chemistry in the actual analysis. I'll also do the benchtop and autosampler stability (that's simple enough; I can just leave some standards hanging around for a bit and re-analyse them).
I'm also trying to find a more appropriate internal standard.

[Alp]

I think you should keep in mind about diluting your samples (say 1 to 10) and injecting again to check if the A/B ratio changes. If it does you have an ionization problem (You are using some sort of "ionspray"?) It's a very quick and easy test.

Alp

[lmh]

I'm having problems believing it's an ionisation problem as the two peaks elute at very different times in the chromatogram, so "A" and "B" are never in the mass spec at the same time. I'm using electrospray, so yes, it would worry me otherwise.
I'll try the dilution anyway, next time I run it.