Drift in LC-MS despite isotope internal standard

[leka]

Hi,
Have anyone noticed a drift in an LC-MS-MS method despite the use of an isotope as internal standard (IS)? I run a method with several analytes - all with a corresponding isotope as internal standard. When I recently tried to run the method on another (equivalent) LC-MS-MS system, I noticed a drift of two of the analytes. The concentration increased in both cases, but in the first case it was due to decreasing IS area (the analyte area stayed the same) and in the other case it was due to increasing analyte area (the IS area stayed the same). Can anyone explain this? Several of the other analytes are OK. The only (obvious) thing that unite these two analytes is that they both are the first channels in their respective MRM-function (I use two MRM-functions). Is ion suppression or ion enhancement possible, despite a co-eluting internal standard? The ion source was cleaned the same day - may a unoptimal vacuum give rise to these problems?

[sassman]

I have also seen this in analysis of hormones by LC/MS. Even though you are using an isotopic internal standard, the precursor ion (and possibly product ion) mass of the internal standard is not the same as the mass of your analyte. Therefore, it is possible for some noise or co-eluting compound to interfere with the internal standard mass and not with the analyte mass or vice versa. That's my best guess, anyway.

[Kostas Petritis]

What Sassman says is possible but you can probably figure that out by doing the analysis without the internal standard but still following the transition in LC-MS-MS.

Another possibility can be cross-talk. Can you indicate the parent and daughter masses that you are following and their respective order?

[sassman]

Kostas, can you explain how cross-talk would cause drift over the course of many runs. From my understanding of cross talk, I would think that any analyte-analyte interference should be the same for each run.

[Kostas Petritis]

Sassman,
It is my understanding from the leka's description that he speaks about quantitation/area counts "drifts" and not m/z drifts.

Assuming this is the case, here is how cross-talk could affect quantitation: If the back to back transition of three ions are let's say 200->150 and 250->150 and 350->260.

In a lot of cases the third quadrupole won’t be able to purge all the 150 ions before the next transition…

This leads to the phenomena called cross-talk as now the 250 to 150 transition will be affected by the 200 to 150 (especially if the intensity of the latter is much higher of the former). This can lead to similar problems described by Leca...

PS: I had a more detailed explanation on cases where the third quadrupole won't be able to purge all the ions but for some reason I was getting the "Forbidden to post bug" -it took me 20 min to rewrite this in a form the forum will accept-

[sassman]

That is also how I understand cross-talk, but I still don't see how it can lead to drift across a batch of samples. When I say drift, I am talking about the response of an analyte decreasing from 100 to 40 over the course of a 100 sample batch. I would expect cross-talk to be affected by the relative concentrations of the two analytes that are involved in the phenomenon. This would typically be more random; although, if his samples were arranged in order of decreasing concentration we could see a similar effect. Am I still missing the point?

[leka]

Thank you both for your answers! My apologies for not replying earlier.

The analysis I´m experience drift in from time to time is analysis of acylcarnitines. I´m monitoring 30 different transitions divided into two MRM functions. 22 of these are analytes with corresponding isotope. All of them have the same product ion (m/z 85) and the precursor ion range between 162 and 428. The isotopes are either 3 or 9 Da higher.

The situation I described in october concerned carnitine (162-->85, isotope 171-->85) and octanoylcarnitine (288-->85, isotope 219-->85). Carnitine (C0) area was fairly stable while its isotope decreased. Octanoylcarnitine area increased while its isotope was stable. To make things more complicated I repeated the test on the same instrument 11 days later. This time the area of C0-isotope increased (C0 still fairly stable) and the area of octanoylcarnitine was stable while its isotope decreased.

A few days ago I experienced a similar problem on the isntrument we always use for this analysis. This time it was acetylcarnitine (C2). It was not a drift as much as a "sudden shift". The area of both analyte and isotope increased, but the area of the isotope increased more, leading to false negtive concentration.

[Kostas Petritis]

Actually is nice that you re-posted as I missed Sassman post. The fact that you have homologues and you are monitoring the same Q3 transition makes the situation very interesting and you might be experiencing several effects. I do not know the fragmentation of the higher homologues of carnitine but if they can fragment down to carnitine you might experience in-source fragmentation of higher homologues down to carnitine and then subsequently give the 85 transition (in order for this to happen your analytes needs to be co-eluted). Depending on the mass and exact atoms that you have isotopes the same phenomena may or may not affect your carnitine isotope. If you include C13 isotopic distributions things can be further complicated...

I would suggest that you inject your higher homolgoues standards and see what is happening for carnitine transitions (do not add carnitine for this test). If you experience effects like these, the areas can go up or down depending on the concentration of other analytes in your samples...

It would also be interesting if you can provide the exact transitions and their order of your analytes as well as their retention time. It would be nice if you can post these in the forum. If this is not possible, send me an e-mail so I can have a look. It will also be helpful if you indicate the full fragmentation obtained by all carnitines that you analyze.

If you can figure everything yourself, please re-post so that we know what the problem was...

[leka]

Thank you for your reply!

I have daughter/product scans of my "major" acylcarnitines and carnitine. All creates a fragment with m/z 59 (which corresponds to the amine group). All (except carnitine) creates a framgment with (ca) m/z 144 (which could be the carnitine "skeleton": acylcarnitine without the acylgoup AND the oxygen in the ester bond). C8 creates in addition m/z 127 and C3 and C6 creates m/z 99. All these fragments are only about 5 % of m/z 85.

Carnitine also creates m/z 103 (carnitine without the amine group) and this fragment is actually little more intense than m/z 85.

My samples are precipitated serum, which means that they all contain carnitine. This makes it hard to perform your suggested test.

All isotopes (except the ones for C6, C10 and C12) have three or nine hydrogens in the aminegroup replaced with deuterium. The exceptions have instead three hydrogens in the acylgroup substituted.

Neither C0 nor C8 co-elute with any of the monitored compounds (except for the isotopes of course). This doesn't mean that they don't co-elute with any other compound... Since it's serum, other, not monitored, acylcarnitines excist. But the concentrations shouldn't be very high, I think.

I'll mail you the transitions, Kostas.

The order of the transitions are always the same and the two test I've described (the ones with increasing and decreasing C0 and C8 ) contained the same type of sample (i.e. quality controls and calibrators in serum) and the same run order.

[Kostas Petritis]

Hi leka,

From the transition, their order and the retention time I was able to eliminate cross-talk (among compounds you want to quantify) as a possibility. Most of your compounds are separated and those that are not there are several transitions in between them. Furthermore you use a quite high interscan delay (i.e. 100 ms) which should be enough to eliminate most cross-talk even if you had co-eluting compounds.

Another possibility is that you get interferences from other compounds in your matrix. I do not know what the dimensions of your column is but you have a lot of compounds eluting between 1.86 and 4 min which should be pretty close to void volume.

Here is a little experiment that can reveal potential interferences with matrix compounds. Inject your sample and instead of doing MRM transitions, do a precursor ion scan experiment by scanning in the Q1 from 162 to 429 (lower and highest mass you analyze) and then you keep your Q3 at 85. If your procursor ion scan TIC will be much more intense than your MRM transitions TIC this might imply matrix intereferences (look around C0 and C8 retention times where you experience these interferences). If you think that you will have higher carnitine homologues you can scan even higher than 429 (do it in another experiment so that you keep your scann range as small as possible (i.e. do not scan from 162-2000). If really have matrix intereferences, you could eventually find out the intereference mass transitions by doing subsequent experiments where you break up the 162 to 429 zone to different MRM transitions where instead of 1 Da in Q1 you allow 5 or 6 Da per parent mass (i.e. 165 with +/- 3 Dalton will cover 162-168 mass range then you do 171 with +/- 3 Da etc) while you keep Q3 always at 85.

The mystery here of course is why sometimes you have the problem and why sometimes you do not have the problem and in your e-mail you said that the older the columns you get decreasing retention times. If these are the cases where you start to experience problems then it might be that interferences start co-eluting with your compounds during that shift. In any case, there should be a logical explanation...

[leka]

Thank you for your answer, Kostas!
I'll try to get time to do your suggested experiment in the near future.

Just to clearify things:

- It's correct that my interscan delay is 100 ms, but my interchannel delay is "only" 50 ms. But I think this also is enough to eliminate cross-talk.

- I don't think the problem is connected to the age of the column. The column used in the experiments I initially described were OK on our usual instrument. But maybe it's a mixed problem...

- It's correct that the first eluting peak (carnitine) is close to the void. I did an ion suppression test during the validation, which showed that there is an ion suppression region just before carnitine elutes. I'm aware that this region might change with different matrix, but also with the age of the column. This could explain problems with carnitine, but not with octanoylcarnitine. But, as said, I'll do your proposed experiments to see if something else coelutes.

[leka]

The problems with the method have increased or become more strange over the last weeks. We began having quantification problems of carnitine (162-->85, isotope 171-->85) on our "ordinary instrument" and noticed, among other things, that the calibration curve slope has increased over the past months. The quality controls became too high. After some time the calibration slope of octanoylcarnitine (288>85, the first channel in the second function) started to increase as well (!). We also started to see drift of octanoylcarnitine on our "ordinary instrument".

To make a long story short: We've inserted two "dummy" channels in the two MRM-functions and it seems as there is a big difference between the first and the second channel: the area of the analyte in the first channel is 2-4 times higher than the area of the analyte in the second channel. And the difference between the two channels are just 0.01 Da (e.g. we monitor 162.05>84.89 in the first channel and 162.05>84.90 in the second), i.e. the first channel is really just "repeated".

These "dummy" channels make it possible for us to use the second channels for quantification, i.e. the first channels give (most of the time) false results.

We tried to prolong the interscan delay from 100 ms to 200 ms, but the phenomenon persisted.

We have carried out test on another analysis running on the instrument. We "repeated" transistion 301>72 and transition 315>86 in two different experiments: no difference in area between the fist and second channel.

I don't know if the drift problem is correlated with the "channel problem" and I haven't yet had the possibility to test the channels on the instrument there we first noticed the drift.

Does anyone have any experience in this?

Cleaning the source didn't help, but maybe a more thorough cleaning might help? Or could it be a software problem?

[leka]

Hi,
The problem is solved since some time. I forgot to post it here though...I just posted it in the new thread I started (Difference between first and second channel in MRM-function).

We had a thorough cleaning of the mass spectrometry done. This included the collision cell, hexapole and source but also removal of visible marks (ion burns) at the first quadrupole.