RP Columns in Series?

[juddc]

Good morning,

I'm currently working on developing a method to separate approximately 24 analytes along with assorted isomers and impurities (total ca 35 major peaks). With quite a bit of work, I've managed to separate 21-22 of these from everything else several times with different columns, but thus far I've not gotten all 24 cleanly separated.

Most of the analytes are neutrals and five analytes have poor chromophores, so pH isn't a terribly viable tool and ACN is my preferred organic phase in order to optimize sensitivity, though I've been working with MeOH, trying to keep it's concentration as constant as possible across my gradient (i.e ternary gradient).

I'm working with relatively short columns, typically 100mm with 3uM particles from a trusted supplier, and what I'm wondering is whether any of you have successfully developed any methods by running a pair of different RP columns in series.

As I said, I was able to cleanly separate 21 or so analytes on a C18 and different set of 21 or so on a Phenyl, both from the same manufacturer, in the same line of columns with the same particle size, and I'm thinking of buying 75mm columns of each type and running them in series just to see what I get.

Yes, I know, I can make two methods and get all 24, but I'd rather not. About 16 of them can also be derivatized (pre-column), but I don't want to go there unless I have to, either.

I've checked the literature on this pretty thoroughly and so far I'm closer to it than anything I've seen thus far, but I'm not there yet. It's the dreaded last 10%, and I can see the finish line, but can't get there quite yet.

Is there any reason I shouldn't try this other than you think it won't work? Has anyone done it and made it work? Any pitfalls other than system volume issues? (Remember please that I'd use well made, extremely reproducible columns from a single reputable supplier in a single column line to keep reproducibility issues to a minimum).

Many thanks!

[Mark Tracy]

Yes I have: http://www1.dionex.com/en-us/acclaim_li ... 28434.html shows DNPH derivatives separated on coupled columns and compared to plain C18. I have also done these derivatives with other combinations using 3x75mm columns. When you couple dissimilar columns, you will get fewer plates than the equivalent length of one stationary phase, but if it improves the selectivity enough, it will work.

You might also look at Bischoff's POPLC literature: http://www.poplc.de/

[juddc]

Cool, so I'm not completely out of my mind...yet

If we knew what we were doing, we wouldn't call it R&D...

Thanks!

CJ

[tom jupille]

You could get a good idea whether it will work by using any of the chromatography modeling programs (DryLab, ACD, Chromsword, etc.). I'm most familiar with DryLab, and I know it will work on that; I think the others are similar.

You "lie" to the program by assigning the retention times on one column to "100%A" and the retention times on the second column to "100%B". Fudge the plate numbers as necessary to make the resolutions match up. On the laboratory window, double the column length and set the mobile phase to "50%B". That should give you a pretty good idea of the selectivity for the coupled columns. As Mark suggested, you should probably adjust the plate number down a bit.

[Mark Tracy]

For a simple binary combination of two columns, k'12 = (k'1 * L1 + k'2 * L2) / (L1 + L2). You can plot on one graph a line between k'1 and k'2 for each compound, and estimate the ratio of column lengths that will give the best selectivity.

I see Tom and I were answering the same question at nearly the same time, but Tom got there first.

[Uwe Neue]

A bit of caution: The additivity of retention times works well when you combine columns in isocratic separations. It is more complicated in gradients. If the two columns have very similar retention, then there is a reasonable chance that you get new chromatograms with some form of scrambled selectivity. Also, in your studies switch the sequence of the columns. This can give you different retention patterns also. But don't be surprised if 1 + 1 does not equal 2 in gradients with different columns. Good luck!

[tom jupille]

That's a good point, Uwe. I think Mark and I were assuming isocratic. If nothing else, with a gradient, the selectivity will depend on the order of the columns.

[HW Mueller]

This makes me wonder whether it isn´t more straight forward to do two methods after all, but automatically in two steps (conventionally called two dimensions).

[Alex Buske]

Hi,

For a simple test you could sum up retention times for each analyte.
Its not a big surprise that you failed to get 35 peaks separated on a 10 cm column. I didn't manage to get less tha 20 compounds separated on a 10 cm. As you choose ACN already, try to use longer columns. I know people who couple 25cm 3µm columns to get enough separation.
What would be the the maximum number of peaks separable on a 10cm column?

Alex

[juddc]

Thank you for all of your comments!

Tom - My run is indeed a gradient, so I don't know exactly how I'd set that up in Drylab.

Uwe - I had thought of the possibility of "scrambled selectivity", but thought it worth a try. A pair of columns is cheap compared to the time being poured into this.

Alex - I may go with a longer column, but would prefer to keep things as short as possible for the sake of sensitivity. I also would like to optimize for time. Also, I'm not trying to separate all 35. I just want 24 clean peaks - I don't much care about the other 11 as long as they don't interfere.

This would probably be a fine application for UPLC, but alas I have no such unit.

[Mark Tracy]

While the calculation I mentioned did assume isocratic, the method in my link is a gradient. The elution order is not drastically different than on C18.

[Uwe Neue]

I recently found a rule of thumb that may be good for checking, how far you can get with the setup that you have. The peak capacity that you need to separate a bunch of compounds is the number of peaks to the 1.5th power. With other words, you need a peak capacity around 120 to do what you desire. If the peak capacity of your gradient is much less, your chances are slim.

[tom jupille]

number of peaks to the 1.5th power.

Interesting! I would have thought it would be higher. Can you give us any background on the rationale?

In any case, it depends on the selectivity (which was the original topic of the thread). In the ideal circumstance, the peak capacity required equals the number of peaks (yes, I know that's unlikely in real life!).

[Uwe Neue]

It is based on data from Lloyd Snyder. I took his graph and added one of our data points (physiological amino acids). It fitted nicely. Then I just drew a line through the major data points...

[MDB]

coupling columns for optimising selectivity on a isocratic way is indeed already well described (cfr. Sz. Nyiredy and Bischoff POPLC).

The past year, I did some research to find strategies to expand to gradient analysis, but accurate prediction is still a limitation.
I think for so far, there is no valid model known for gradient analysis that accurately predicts the retention (and selectivity) on coupled columns. However, I heard from the possibility to optimise with ChromSword, but I can't hardly find some information on the internet (on another topic on this forum I've read that somebody does it that way), and I just can't find some separation examples. So, if somebody has experience with Chromsword and optimising selectivity with coupling columns in combination with gradient analysis, feel free to post some example chromatograms. I'm curious.

Cheers, Maarten