Method transfer - lipophilic, apolar analyte; RP18; MeOH ...

[benrock]

Greetings,

I am new to LC, mostly with a theoretical, abstract understanding by reading books, attending seminars and conferences, and asking questions over the last year. You could say that I am living proof that a little knowledge is a dangerous thing... I write to ask for help with transferring a method for quantitative analysis (and ideally also preparing for validation of the method by someone else).

The analyte is a lipophilic, apolar, long-carbon chain polyketide (acetogenin), with a formula of C37 H66 O7. The molecule has a terminal lactone, various hydroxyl groups, and a long carbon chain. It is found in extracts of certain tropical plants. In its purest form, it is a white wax with a melting point below 30 C. An illustration and reference data of the molecule can be found on:

http://www.lipidmaps.org/data/get_lm_li ... PK03000012

Many methods for this analysis have already been published in the scientific journals over the last twenty years. Various methods use an RP18 column and a methanol:water 90:10 solvent. I am trying to make sense of the various methods, plan to repeat the separation, and prepare for a streamlined validation. I understand that validation can be a long, time-consuming process, but this puzzle has more to do with the method transfer than with the validation. I read an article that compares columns (Mac-Mod), and they categorize the following columns as very hydrophobic and apolar:

Intertsil ODS-3
Kromasil C18
Develosil ODS-MG

I would like to develop (redevelop) a method that can be readily validated and also possibly used by a less skilled analyst (like me).

The matrix will contain about 10-15 structurally similar molecules, but we intend to use supercritical fluid extraction (SFE) to extract the oil from the plants, so the matrix will not contain very different molecules (regarding lipophilicity and apolarity). In the chromatograms included in the papers, the peaks are separated close to the baseline. The distance between the peaks is ok.

Specifically, what prompts me to write is this:

1. The sample will solidify in the refrigerator. Can we add on a device to heat the auto-sampler and the tube to the column (to keep the sample liquid without overheating the room)?

2. If you have experience with such molecules (lipophilic molecules found in vegetable oils), can you please tell me: are the Inertsil ODS-3, Kromasil C18, and Develosil ODS-MG columns optimal for this separation? If not, what columns do you recommend for this? If the analyte is lipophilic and apolar, then should we not also pick a column packing with these characteristics?

3. Will we get better separation (resolution) or faster chromatograms, if we use ethanol instead of methanol, since ethanol is less viscous? (Even if it costs more, it may be more available.) Why did they use methanol:water 90:10 and not methanol 100%?

4. To keep it simple (and quicker to validate), we would prefer to have no gradient and no buffers to alter the pH, even if this reduces the separation slightly. If the molecules have similar lipophilicity, apolarity, and structure, is there an advantage to a gradient? One technical rep of an instrument manufacturer informed me that gradients are to be used mostly for matrices that contain molecules with very different characteristics. Do you have any comments on this?

5. What is an outline of a validation protocol other than: precision, accuracy, LOD, LOQ, linearity & range, selectivity, robustness, and ruggedness? For example, if we change the instrument (once) and change the column (once) and re-run the analysis ten times on each combination of instrument and column, calculate the variance among each set of ten measurements on the same combination of instrument and solvent, will this (forty measurements) be enough to validate the method, assuming acceptable variance? If not, what if we changed the solvent from methanol to ethanol and re-ran the forty analyses for a total of eighty samples processed? Will this be enough? What are the most frequent errors that would make a method invalid? How can I prepare for the validation? How many samples do they usually run? Will it make for a simpler validation without a gradient nor a buffer?

I am discovering that two months in the lab can save three to four hours in the library... The molecule has been detected with UV at 205 nm, with FTIR at ~1750 and ~3430 cm-1, (and also with TLC). We would use SFC for this separation, but will have not have access to SFC equipment for some time.

Sincere thanks for your time and any help with this.

Regards,

Ben

[bisnettrj2]

Wow. Ok, I'm not qualified to answer all your questions, but I'm throwing my few cents into the discussion.

1. Methanol is less viscous than ethanol, but with a UV absorbance of 205nm, you should look into running acetonitrile:water as your mobile phase - lower UV-cutoff, less noise at your analyte's UV-absorbance maxima. And running 90% methanol is pretty odd - typically, the more methanol, the less retention you get. If you're running a one-analyte standard, you can run whatever percentage of organic solvent you want. If you're trying to separate a mixture of analytes in a complex matrix, you'll probably want to optimize the retention times and elution order of your analytes with a mixture of pH, temperature and organic:water proportions, and this might require a gradient in order to elute co-extracted interferences earlier in the run.

2. Some autosamplers are thermostatted, or can have an add-on that will refrigerate or heat the autosampler compartment. I have an Agilent G1329A autosampler module with the G1330B thermostat module, and I can hold samples in a 4 to 40 C range. I would worry, though, about injecting something into my instrument that was a solid at room temperature. Perhaps you should use a mobile phase in which your sample will dissolve?

3. A quick search using my friend Google lead me to a site describing lipid HPLC using silver ion columns. http://lipidlibrary.aocs.org/topics/silver/index.htm. You may want to check out this site and possibly contact them to see if they can point you in the right direction in terms of analysis. Also, http://lipidlibrary.aocs.org/topics/fa_hplc/index.htm and http://www.varianinc.com/cgi-bin/nav?pr ... LKIQPLNPFM.

[benrock]

Sincere thanks for your response. I will study these websites, but in concept if the analyte is lipophilic and apolar, in a matrix with similar molecules, then will a lipophilic and apolar column packing not separate the molecules optimally? I notice that Inertsil ODS-3, Kromasil C18, and Develosil ODS-MG are described by Mac-Mod as both the most lipophlic ("hydrophobic") and the most apolar.

If I understand the theory, then a column with similar characteristics to the analyte will yield better separation, is this not true? Like attracts like. This would point to one of these three columns, if the Mac-Mod study and review is correct.

We hesitate to put in a column packed with bits of silver, because I have read that metal ions in the column can somehow degrade due to the metal rubbing off over time. Also, from a cursory initial reading of the website, I am not convinced that the bond in the molecule that would bind to silver ions is the determining characteristic for the retention, compared with the terminal lactone moeity and the long-carbon chain. Also, for certain availability, we would prefer to pick more commonly used columns. I will study the websites that you found though.

Regarding the solvent, perhaps we are naive, but we would prefer to do green chemistry, using solvents that are ideally non-toxic, such as ethanol or even possibly vinegar (dilueted acetic acid) to lower the pH. The published method worked with methanol:water 90:10, so we would prefer to use this or ideally ethanol, or even the purest Russian vodka that we could find. Also, the cost and disposal of acetonitrile is a factor for us.

This puzzle could also be expressed as follows:

In the first analysis, what if we fixed the column to one of the three listed (Inertsil ODS-3, Kromasil C18, or Develosil ODS-MG), and what if we also fixed the solvent to 100% methanol or ethanol, then how could we optimize the method for separation and speed:

a. using columns smaller particle sizes, column lengths, and internal diameters?

b. perhaps adding 10% vinegar (diluted acetic acid) to the solvent to lower the pH?

c. increasing or decreasing the flow?

We would like to keep this as simple (robust) and rugged as possible, so that the method could be applied reliably in field laboratories in tropical countries.

Thanks for your time and attention.

Regards,

Ben

[danko]

Hi Ben,

I’m afraid you didn’t take Bisnettrj2’s remarks, about UV cutoff, too seriously.
If you’re going to detect your analyte at 205 nm, ethanol is not a viable mobile phase (if memory serves me the cutoff of ethanol is roundabout 210 nm).
And it’s getting much worse considering acetate as an additive.
So, maybe it’s a classic case of lack of experience (you mentioned it yourself) – overseeing the need for detection.
Actually my personal experience tells me that the method you’re considering is kind of strange – using 90 % organic modifier. As a matter of fact I believe the analysis might be better off utilizing the HILIC technique.
If you insist on using reversed phase however, then a C4 or a phenyl chemistry kind of stationary phase, combined with much less organic solvent, might yield better results compared to the C18 option.

Best Regards

[AA]

You have set youself a challenging task for your first try and I think it is unlikely that you will be able to seperate 10-15 structurally similar molecules using 100% organic on any C18 column. A gradient approach using ACN is pobably you best choice at this point. I applaud your desire to go green but if you really want to resolve these components you may have to consider alternate approaches. You should consider the book "Practical HPLC method development" by Snyder, Kirkland and Glajch which is one of the best guides out there for LC method development.

AA

[yangz00g]

I fully agree with AA's comments on this matter: you put yourself into a very challenging situtaion and the chance to success is relatively low. Remeber, determining 10 to 15 similar molecules by LC-UV in pure standards is a totally different story from doing the same thing in a real-world natural product extract, no matter what column you use, or how good the column is demonstrated in a well-controlled model system.

You have set youself a challenging task for your first try and I think it is unlikely that you will be able to seperate 10-15 structurally similar molecules using 100% organic on any C18 column. A gradient approach using ACN is pobably you best choice at this point. I applaud your desire to go green but if you really want to resolve these components you may have to consider alternate approaches. You should consider the book "Practical HPLC method development" by Snyder, Kirkland and Glajch which is one of the best guides out there for LC method development.

AA

[Uwe Neue]

1. The fact that your sample solidifies means nothing, since you will inject a dilute solution of your sample in mobile phase or related solvent.
2. Any C18 column will do the job.
3. Ethanol is more viscous than methanol. You also need to stay away from alkohols if you want your UV detector to see the analytes. You should use acetonitrile, with about 20% water instead of methanol with 10% water.
4. No need for buffers unless your analytes are ionic. Whether you need a gradient or not will depend on the hydrophobicity range of your analyte set. Since you know nothing about this yet, I suggest you start with a gradient.

[benrock]

Sincere thanks to each of you who replied. I will be the first to admit that I have no practical experience with LC. This is why I ask for your help.

This method has already been developed & published. It worked with an RP-18 column and methanol:water solvent. The retention time was about 45 minutes. I am simply trying to repeat what has been published, but modify it for optimal selectivity/resolution and speed.

Regarding the analyte, it is both very lipophilic and very apolar. Regarding the analyte set (matrix), it is in fact relatively uniform, unlike most extracts, because we intend to use SFE for the extraction. This technique makes it possible to fractionate the oil to select either heavy oil or light oil, filtering out other molecules, such as alkaloids, which are also found in these plants.

"Any C18 column will do"? Forgive me, but I don't believe this, because everybody else informs me that the specific C18 column will affect the results, the retention times, and the separation of the peaks on the chromatogram. In the Mac-Mod comparison of RP18 columns (www.mac-mod.com), they categorize columns by both hydrophobicity (lipophilicity) and polarity (and apolarity). They list Inertsil ODS-3, Kromasil C18, and Develosil ODS-MG as both very hydrophobic and very apolar. This matches our analyte set, so for retention, I assume that we should start with one or another of these columns, is this not logically correct? I see no reason to use another column packing material, such as HILIC (hydrophilic, unlike our analyte), C4, phenyl, or any other column packing material other than the one that already worked, RP18.

Regarding the solvent, this method already worked with methanol. In fact, various studies report UV detection at 205, 210, 215, and 220 nm. If I understand correctly, because ethanol is more viscous, then the selectivity will be less than methanol, if all else is equal, so we should use methanol. I would prefer (naively perhaps) to use a solvent that I could drink, but barring this, if the analyte is oily, then perhaps hexane would be the optimal solvent, no? Hexane poisoning has been reported recently, so I hesitate to consider this option, but logically, this would be the optimal solvent, no? Hexane is already widely used for food oil extraction.

We would like to avoid acetonitrile not only because of the toxicity, but also because of the earlier and ongoing shortage and consequent higher price of this solvent. What is the theory behind using acetonitrile? If the analyte is apolar, should we not use an apolar solvent, methanol?

Assume that we start with a stationary phase that matches the analyte (lipophilic and apolar).

Assume that we start with methanol:water 90:10 or methanol 100 as the mobile phase.

Assume that we start with an isocratic solvent (to keep the method simple and robust). (One instrument manufacturer informs me that the advantage of a gradient has to do with sample with a matrix that contain molecules with different polarities, and our matrix will have molecules with similar polarities.)

How can we optimize this for resolution and run-time?

Smaller particle size, shorter column, and smaller internal diameter?

Decreasing the flow?

Increasing the injection volume?

Using a co-solvent (vinegar - ?) to lower the pH?

Other?

Regards,

Ben

[bisnettrj2]

*turns around, walks away from discussion with a low whistle and a shake of the head...*

[Uwe Neue]

If you have a method - bad as it may be - then it should specify a column, and you should not even think about departing from the specified method and column. If you choose another column than the one specified, you will get randomized results.

If you want a better method, you should listen to the advise given, and not go back to silly prejudices such as "acetonitrile is toxic". Acetonitrile has no UV absorption at the wavelengths you want to use. You will have better sensitivity and you can run gradients without difficulties. If you want to develop your own method, most C18 columns will do. If you want a column that is reproducible for future work, you should use Symmetry C18 (I work for the manufacturer of this column, but I also have the data that proves this reproducibility). You should choose a particle size and a column dimension that fits your instrument. The standard is a 15 cm 4.6 mm column with 5 micron packing. For more resolution (and somewhat more backpressure) you can use a 10 cm 4.6 mm 3.5 micron column. Stay away from smaller column diameters, since they create headaches with extra-column bandspreading of the instrument, and since you are new to the game, you want to be in a safe spot.

You do not need "vinegar" if nobody else has used vinegar for this method, and the type of your analytes does not appear to require such a thing. Plus, your detector will now really go blind if you throw vinegar into the mobile phase and you will not see anything anymore.

You may need to figure out what solvent composition your analytes are soluble in. But remember, the concentrations needed in LC are tiny (unless you use methanol in your mobile phase).

[lmh]

... and just a comment on acetonitrile. Yes, we would all like to keep our solvent consumption low, particularly of "nasty" solvents, but all solvents carry at least some environmental impact, even water! Where do the purifying cartridges in RO water purifers come from (or go to)??.

The best approach is probably to optimise your method for minimum solvent consumption (while balancing carefully Uwe Neue's concerns about columns that may be difficult to handle). You could consider 2mm diameter columns run at 20% of the flow-rate of a 4.6mm column, if your hplc is good for this.

[benrock]

Thanks again for your replies, but snide comments such as "turns around, walks away from discussion with a low whistle and a shake of the head..." are not helpful. If this poster were really as intelligent as he appears to think he is, perhaps he could just reply to the specific questions...

I am open to your advice. I informed you that I am a newbie, but what I have seen and heard over the last year of studying this analyte set is that many technicians have different opinions about what conditions are optimal, naturally usually based on their own experience and direct observations. N.B. Even in these replies, some say "use C4, use phenyl, use silver ion in the column". So I have to make sense of this and try to reconcile the various opinions and pieces of advice with the facts.

The facts are still:

1. The method worked, using RP18 and MeOH:H20 90:10.
2. They did not specify the brand nor packing material of RP18.
3. The analyte and analyte set is very lipophilic and very apolar and relatively uniform (unlike most extracts of natural products).
4. The retention time of the key analyte (base peak) was ~ 45 minutes.

The questions remain:

1. If the analyte (and analyte set) is both very lipophilic and very apolar, then among all of the RP18 columns available, should we not choose the column or columns with the packing material that is the most lipophilic (hydrophobic) and apolar? Mac-Mod independently compares various columns by both of these characteristics. They rate Inertsil ODS-3, Kromasil C18, and Develosil ODS-MG as both the most lipophilic and most apolar, among all of the RP18 columns available, so logically and rationally should we not start with one or another of these columns?

2. If methanol worked as the solvent, what is the theory behind using acetonitrile to improve the resolution or run-time? If acetonitrile is less polar than methanol, then why not use the ultimate in apolarity, hexane? The UV detection was ok at 205-220 nm, even using methanol as the solvent, according to the published papers. For the detection, we are not constrained to using UV. We could use MS. The molecule has also been detected, using FTIR and even TLC.

3. How can we optimize this separation for resolution and run-time? Heat? Acidity? Particle size?

Would you please reply, only if you have practical experience with this type of analyte and specific advice to offer, ideally answering the specific questions, or asking your own questions, if this is not clear to you.

You could say: just try it. This is what we intend to do, but we would like to plan and focus on the specific variables that will optimize resolution and run-time for this analyte and matrix. I still think that two months in the lab can save three to four hours in the library, but I have already been to the library many times looking for these answers. Most, if not all, of the books about HPLC method development are not specific to lipids, but are more oriented to Rx.

What I am really looking for is advice, not from a technician, who knows how it has always been done but not why, but from a scientist, who is capable of both thinking outside the box and explaining his or her thinking clearly. Is this too much to ask for?

[Uwe Neue]

If the published method did not specify a brand of the packing material, you might as well start from scratch. The information that you have is not useful.
Otherwise:
ad1.: it has been stated several times that the hydrophobicity of the stationary phase is irrelevant based on the information that you have. Any C18 will do. I recommend Symmetry C18.
ad2.: methanol adsorbs strongly at the low wavelength that you want to use. You may or may not have trouble with the sensitivity of the method. Note that MeOH has the same detectable groups as your analyte. So you are looking for a particular fish in a pond full of fishes. Not a good thing to do. If you use acetonitrile as the solvent, this problem will go away, and detection is easier.
ad3.: for the type of analyte described, acidity will do nothing. You can manipulate the separation by gradient profile and temperature. You will get better results with a smaller particle size, but considering that you are an amateur and we do not know anything about your instrument, the best choice is a 4.6 mm x 10 cm 3.5 micron column. Smaller particles have the promise of higher resolution and or shorter run times. But going below the suggested particle size may create instrument headaches, and are not the best choice for a newbie.

[mbicking]

You have a fairly challenging separation problem and no/little experience with method development. This is not a good recipe for quick success.

You have been getting free method development advice on this forum, from people who have considerable experience in HPLC. If you have questions about how to implement their suggestions, then this may be the time to decide to contract with someone to do the method development for you. Perhaps you can find someone who will let you "look over their shoulder" while they work on it, so that you can learn.

I am tempted also to say you should get more training, but you have to realize that method development is art as well as science, and at some point you just have to dive in and try something, recognizing that failure will happen often during the early experiments. There really isn't any other way; that's how we all learned it.

[lmh]

a few further comments:

(1) you want good resolution, i.e. separation between two compounds. Therefore it's not really about optimising a strong interaction between column and analyte (which is what you seem to be pursuing), but getting a situation where the compounds of interest have different interactions with the column. Hence the statement that any C18 will (probably) do. Yes, they will give slightly different retention times, but how do you know they will give worse resolution? We don't actually have the structures or properties of the things you are trying to separate from your target chemical.

(2) The thinking behind suggesting a C8, C4, or even phenyl column is that your compound of interest clearly interacts very strongly with the column (which is why you need 90% MeOH to get it off). If your next analyte of interest interacts even more strongly, it might not come off at all. Choosing a less hydrophobic column will reduce the interaction, and allow everything to elute at lower organic content, which may improve the range of things you can separate, and gives you more gradient range with which to experiment in improving resolution. Incidentally, phenyl columns allow pi-pi bonding with aromatic and desaturated systems, so they're not so relevant to you, and in any case, this interaction would be cancelled by using acetonitrile as solvent, so a phenyl column would basically work like a low-hydrophobicity RP column.

(3) In addition to its UV properties, acetonitrile has very low viscosity which makes it easy to use with hplc columns without creating high back pressure. You can run things faster at higher flows, which is a form of optimisation if instrument time is an issue (and frankly 45min is a long run by modern standards).

(4) Hexane has miscibility issues with most of the solvents people want to have attached to their LC systems, amongst other practical problems.

(5) It is cheaper to vary the concentration of methanol (or acetonitrile) than buy large numbers of columns with different sorts of C18 and different particle sizes. Varying the solvent is a good way to improve resolution and shorten run time. If you know something of the properties of your various analytes, there is software available that will predict the effect of the changes (ACD labs etc.).

(6) If you want to use MS, go ahead. The downside is that you wanted a robust method easy to transfer to other people; LC-MS is more complicated than LC-UV, MS instruments are more variable between manufacturers, and ionisation efficiency isn't a fixed value like the extinction coefficient of a compound, so accurate and repeatable quantitation is less straightforward.

(7) I am a technician, and proud of it. I am not by training an analytical scientist and have only the shakiest knowledge of the basic principles of chromatography. I don't consider myself a lateral thinker, or particularly skilled in explaining myself. So please feel free to ignore all I've written!