Peak eluting before the void volume - Clueless

[mactavish606]

Good morning, and let me start by saying that I am the newest of the new and have only been doing this HPLC thing for 3 weeks now. It has been a baptism-by-fire for me; I have been tasked with the duties of HPLC work for the foreseeable future because the man who has done this small companies HPLC work for the last God-knows-how-many-years was "let go" by the higher ups and courtesy of this lovely economy, the company lacks the funds to hire someone who knows what they are doing. So in the mean time I, a microbiologist with no real chemistry background other than that applicable to microbiology, have been given HPLC duties. Lucky me!

The problem I am experiencing at the moment is that my peak is eluting BEFORE the void volume. I am following the method prescribed in a publication by Agilent Technologies for the analysis of paclitaxel (http://www.chem.agilent.com/Library/sol ... 682635.pdf on page 31). In case the link doesn't work, I am using a 4 x 125mm Hypersil ODS column (5 micron), my mobile phase consists of water (A) and acetonitrile (b) following at 1 ml/minute. The gradient starts at 50:50, proceeds to 10:90 over 10 minutes. Detection is by UV absorbance at 204nm and the injection volume is 5 uL. These are the exact same conditions prescribed by the Agilent document with one exception - I am using a 10 mm C18 guard column and the publication makes no mention of guard columns. In the Agilent document a chromatogram is given showing the paclitaxel peak at 4 minutes, but for some reason my peaks are coming out before the void.

The sample matrix is as simple as it gets - paclitaxel solubilized in a 50:50 v/v of water - acetonitrile. Since these are standard solutions, no sample prep steps are involved.

To check that the column isn't just totally destroyed, I ran a few samples containing docetaxel. Docetaxel and paclitaxel are 99% identical in structure:
Paclitaxel - http://en.wikipedia.org/wiki/Paclitaxel
Docetaxel - http://en.wikipedia.org/wiki/Docetaxel

My docetaxel peak eluted where I expected it to elute - just shy of 5 minutes.

I have tried using different bottles of paclitaxel as well as docetaxel to see if perhaps one was degraded to a point that varied the retention time, but this made no difference. New, old, different suppliers, no change what so ever.

So I pose two question - 3 - What in the world is causing this? 2 - how in the world do I fix this? and 3 - would it be a terrible breach of acceptable scientific HPLC protocol to use this pre-void volume peak for data generation that will eventually be used in a publication? This peak does give extreme consistent and linear results. Thanks so much for your time.

[tom jupille]

So I pose two question - 3 - What in the world is causing this? 2 - how in the world do I fix this? and 3 - would it be a terrible breach of acceptable scientific HPLC protocol to use this pre-void volume peak for data generation that will eventually be used in a publication? This peak does give extreme consistent and linear results. Thanks so much for your time.

I'll take those in reverse order.

It wouldn't be a breach of protocol to quantitate on a peak eluting at t0, but you would never get that method past an auditor or reviewer. The FDA in their "Reviewer Guidance" (http://tinyurl.com/46cr94a) suggest that k' values should be greater than 2. That's not directly applicable to gradients, but the equivalent (that means peaks should elute no sooner than 3 times t0) makes a lot of sense.

The use of a guard cartridge should not be an issue. Assuming it's the same diameter as your analytical column and is packed with the same material, it amounts to lengthening your column by about 8%, so if anything, retention times should be a bit longer.

So what's going on? Several possibilties:

1. You're not seeing the taxol at all, and what you're looking at is, in fact, t0 noise. If you did your calibration by varying the injection volume, it's not unheard of for that baseline upset to be proportional to volume injected (ask me how I learned that! ). You can check by injecting varying volumes of a diluent blank.

2. Your column is less retentive than the one used in the application note. Columns *do* age, and retention often decreases when that happens. The workaround in that case is to start the gradient with a weaker solvent (the app note starts at 50% acetonitrile). It would be a good idea to try at least once a "full range" gradient from 5 to 100% in 19.5 minutes. The taxol peak should elute *somewhere* in that range; once you know where, you can adjust your method accordingly.

[southp]

Has your system backpressure stabilized before you inject? Backpressure tracks with the relative % of organic in your flowing mobile phase because of viscosity changes. The Agilent method specifies a Post Time of 5 minutes, which I think (I'm not a Chemstation user) allows time for the column to re-equilibrate back to initial conditions before starting again. If you inject well before the column is re-equilibrated (say at the 12 minute Stop Time), you may see early eluting peaks because there's too much strong solvent (ACN) in the column yet.

You mentioned you ran several samples of docetaxel and got retention. Were those injections back-to-back, or timed exactly like the paclitaxel runs, or was there some extra time in between docetaxel injections while mobile phase flowed at 50:50?

[lmh]

further to southp's comment, if you're using Agilent LC Chemstation with a gradient method, the gradient goes in the table at the bottom of the pump set-up page, but at the top, beside the solvent names things, there is also a percentage box that you'd use in an isocratic method. Between gradient runs, this box takes over. If you forget to set it, you may find the pump pushing a large slug of 100% highly-eluting solvent into the column just before the gradient takes over again, and the next injection happens. This completely ruins retention. Been there, messed up, and took days to realise why...

[mactavish606]

Thanks so much for the quick responses!

1) My injection volume has been constant - 5 uL for every injection I've done. Even with the injection volume the same for all the standard solutions I still get consistent and linear values when I plot the peak area against the known concentration.

2) The software that I'm using is "Chromera." I used the same set of parameters, saved on the computer as a file, for both the paclitaxel and docetaxel that I've run. All I do is load the file containing the computer instructions and click go, so all the gradient steps, flow rates, injection volumes, etc etc etc is the exact same for all injections I've done of both paclitaxel and docetaxel. So if I had an error in my computer instructions it should have affected doceltaxel in the same way it affects paclitaxel, right?

3) The computer instructions include a 10 minute equilibration period prior to each injection, and I double checked the back pressure this morning and it it only changed +/- 10 psi during this equilibration time. Which I assume is an acceptable degree of change since our pump has only a single piston and small mixing cell (or so I'm told, I have no idea what would be considered 'small' or 'large' or even how big the one we have is!)

Thanks again!

[tom jupille]

Okay, the fact that injection volume is constant eliminates the possibility that you were misidentifying the peak.

If this were my problem, the next thing I would do is to run the "full-range" gradient I mentioned (change the initial %B from 50% to 5% and increase the time. Hopefully, you should see a retained taxol peak in there somewhere. If you don't, that suggests that your standards are messed up. If you do, try to figure out what the %B is at the time the taxol peak elutes from the column. Set your initial concentration about 10% lower than that, scale the time so that you get about the same %B/minute gradient rate, and re-run.

If the taxol peak is indeed eluting early, that would suggest that your column needs to be replaced.

[mactavish606]

So here is the update:

I ran the full-range gradient as described, starting at 5% B and going to 90% B over 20 minutes, at which time it held constant at 90% from minute 20 until minute 45..... no peaks to speak of.

I'm starting to entertain the idea that the poor solubility of paclitaxel could be playing a role here. I'm apprehensive about this suggestion for 2 reasons 1) my standard solutions of paclitaxel are at concentrations well within the soluble range of paclitaxel, and 2) docetaxel has a similarly poor solubility and it was retained without a problem. The standard solutions are prepped in water, and when I prepped the solutions in 50:50 water/acetonitrile they did not show any signs of precipitating. Also, I would expect an increase in backpressure if it were precipitating, not a very sharp, well defined peak? Thoughts on this?

[mactavish606]

While I'm asking question I have another that isn't specifically related to paclitaxel, but to integration in general. The software has 3 numeric parameters which are applied to whatever algorithm it uses to define where to start and stop integration. These are the bunching factor, area threshold, and noise threshold. Should these be established and the same values applied to all concentrations in the curve? This is the way I have been doing it so far, but very often this results in poor integration patterns at one end of the concentration spectrum. So if it integrates the lower concentrations well, then it typically will cut off parts of the peaks from the larger concentrations. Or if it integrates the larger concentrations well, it will include large portions of the baseline signal in part of the integration for the lower concentrations.

So should these values be the same for all concentrations, or should I use the values that produce the best integration for all injections of one given concentration?

Thanks again, you have no idea how much help you have provided!

[southp]

Thought I replied yesterday, but it apparently didn't make it to the board.

Seems like it's time to try Tom's suggestion of a new column, even if it's a different brand of ODS at first if that's all you have. Before that, though, I would try the separation without the guard because it's easy. Guards are one of my pet worries when a separation isn't working; I think they often die sooner than expected.

I agree with you that a solubility difference doesn't seem likely given the similarity of the 2 analytes.

As for the integration issue, that's a question that gets discussed a lot; there are probably several threads on this site devoted to it that you could search for. A big consideration is who you have to defend your data to. Government regulators trust computers more than people and like to see the same algorithm applied to all samples in a sequence as much as possible. If no one is ever going to audit your electronic data, then the exact algorithm used to provide valid integrations is of less importance. On a related note, the Oct 2009 issue of LCGC North America (p. 892) has a good article by John Dolan on right/wrong ways to integrate peaks.

[tom jupille]

southp's comment about the guard column is a good one. My suggestion(s) at this point:

1. Double-check your detector wavelength setting (just because I'm paranoid doesn't mean the world *isn't* out to get me!)
2. Run the full-range gradient again without the guard column.
3. Prepare a fresh standard of taxol and run the full-range gradient again
4. Get a new column and run the full-range gradient yet again.