Potential Band Broadening issues

[Eggstract]

Hi all - A little relevant back-information - we're a new startup trying to look at detection and quantification of cannabinoids. Running an Agilent 1100 with degasser, autosampler, column oven and DAD.

We have been doing a lot of runs recently trying to develop our method but keep coming into a certain problem with some samples but not others. The only relationship we can think of between the samples is the presence of acidic cannabinoids.

When ever we test products with THCa we get this giant broad band, see attached file Fig.1. At first, we thought this was due to overloading, so we diluted 100x instead of just 10x, and made a smaller injection of 2ul instead of 5ul, but we still get this broadband.

(Fig. 1)




We tried to vary the pH of our buffered water between 4.2-4.8 because we thought the broadening is happening with the acidic analytes but not the neutral ones, but we still get the 'hump back'.

We also tried to experiment with gradient to see what happens with a slow and fast gradient ( Fig 2,3).

(Fig. 2)


(Fig. 3)




We have a feeling the hump-back may also be 2 compounds co-eluting. There is a clear bump at one wave length, but not at another (see Fig 4.)

(Fig. 4)




When looking at the spectra (fig 5) although im sure its not pure, it appears to be one of our acidic analytes. We tried to run some standards we purchased to see what happens when we inject a known acidic cannabinoid compound and again we got issues with very broad bands.

(Fig. 5)




Fig 6 shows what we call the Batman.... we actually prepared a sample with CBD and CBD acid standards, same concentration (50uL aliquot of CBDa and 50uL aliquot of CBD). The large peak has a spectra that matches the known spectra for CBD, however our acidic analyte seems to have gone, or become batman!

(Fig. 6)




We currently use 25mM ammoniun Formate buffer in our water, pH to 4.3 with Formic Acid. We also tried putting a drop or formic acid in our Methanol to make 0.1% V/V and tried to see if we could resolve our issues, but no luck.

Our next step was to try 50mM buffer, and also start all over again with Acetonitrile as our B instead of methanol.

Now we're really quite lost as to where to go next. We feel like there is something happening to our acidic analytes, causing this broad hump back peak, but we're unsure what else to do. We are at our limit with flowrate, any higher and we are over pressure ( small bore colum 2.1mm x100mm, with 2.7um).

If anyone has any tips or suggestions, we would greatly appreciate some input.

Thanks in advance!

[mattmullaney]

Hi Eggstract,

This paper may help:

https://www.sciencedirect.com/science/a ... 8517311895

Journal of Pharmaceutical and Biomedical Analysis, Volume 147, 5 January 2018, Pages 565-579. "Pharmaceutical and biomedical analysis of cannabinoids: A critical review."

Open access!

Best Wishes!

[Multidimensional]

You probably do not want to hear this, but the reasons for your problems all relate to a complete lack of training in both the technique of HPLC and in the use of the Agilent 1100 series HPLC system. I suspect you are brand new at this and thought you could just "set it up", and it would work (we see this a lot. People have no idea how much training is needed to do this correctly). Your comments and interpretation of what you are seeing make this very clear (sorry to be so blunt, but you need to hear this from someone with experience to save you time and money now). It takes many years of professional training to develop (or even to simply run a pre-developed) HPLC method. Too many things to list here as to why, but you will not solve your problems by posting on a website like this. You need to hire a professional chromatographer to help you.

Some quick examples of obvious problems with your "method" and technique can be seen in what you have presented so far. First, you did not provide any information about the exact column used. No chromatography suggestions or advice could be provided without such very basic information being described. Next, you have the UV/VIS detector's 'Reference Wavelength' feature turned ON (indicating you have no training in how to use the DAD). Never, turn that on as it will delete your raw data rendering any output invalid [https://hplctips.blogspot.com/2011/03/reference-wavelengths-as-used-in-hplc.html]. Also, your UV signal and scale are so low. You are looking at a lot of noise and inadequate equilibration. Your integration settings have not been setup properly. Your attempt to compare spectra are also invalid because the chosen settings are inappropriate and the signals were not integrated properly. Lastly, where did you get this gradient from? It makes no sense at all. You have composition changes occurring when peaks are still eluting. You have changes when nothing elutes. The gradient is going back to initial conditions before the peaks are out???? The operator has no idea what they are doing and you are focusing on the wrong causes rather than realizing that several years of professional training plus lots of mentoring is needed before you attempt this. A website is not going to help you, nor is someone sharing articles with you.Please seek out some professional help with this before you proceed.

[AA]

I for one applaud your efforts. Most companies in MMJ industry don't do nearly enough quality control. Everybody starts at the beginning and learns as they go. Keep trying.

First thing that jumps out to is that the 1100 is really not a good choice to run smallish particles with 2.1 mm i.d. columns and the very low flow rate you are using (the system volume is way to high for that). Try something that is known to work like the HPLC method in the American Herbal Pharmacopoeia.

Column XBridge C18 4.6 X 150mm 3.5 µm Column Temp 30 C Mobile Phase A 9:1 50 mMol ammonium Formate (pH 3.75)/Acetonitrile Mobile Phase B 9:1 Acetonitrile/Water Detection UV @ 228 nm Flow Rate 1 mL/min

Gradient Table

Time % A %B
0 30 70
15 10 90
30 10 90
31 30 70
40 30 70



===========================================================

[mattmullaney]

Hi again, Eggstrsct,

If your firm can commit to the funds, you can modify the Agilent 1100 a bit to improve its performance with mac-mod's guide: (unfortunately you'll still be pressure-limited to 400 bar)

http://mac-mod.com/pdf/technical-report ... Report.pdf

I don't know that this is an option for you. Was easy enough to modify the 1260 system I used at one job I had using similar parts. It is important to replace the detector cell.

Anyhow, how certain are you that the broad peak is CBDA? Peak shapes such as these may indicate some sort of chemical conversion that proceeds swiftly compared to the speed of the separation.

[Multidimensional]

The Agilent 1100-series is well suited to 2.1 mm ID columns at linear flow rates of ~210 ul/min. Dwell volume for gradient runs must always be addressed by the USER of the HPLC system and many easy changes in both shorter and narrower ID tubing to the flow path, a smaller volume flow cell (as Matt suggested) and even a reduced mixer change (if it is a Quat pump) to optimize it for most any low volume application. We have run them with columns and flow rates in this range since the products were introduced around 1995. No problem at all, but should be done by someone with an intermediate level of experience.

Regarding backpressure limits of 400 bars, that should be of no concern at all as the backpressure would be the same at ~ 210 ul/min on a 2.1 mm column packed with the same particles as a 4.6 ID column running at 1.000 ml/min (linear flow rates). Most of the 2.5u and greater particles do not get you anywhere near the limits of the system as shown in this table: "Pressure Drop Across an HPLC / UHPLC Column "; https://hplctips.blogspot.com/2011/08/p ... olumn.html *A 2.1 x 100 mm column (~ 240 ul volume column) with 2.7u particles would present with very reasonable pressures.

We should not be worrying about the use of an 1100-series system or 2.1mm ID columns in the post. The problem areas are not related to the equipment or column choice.

[mattmullaney]

Hi Again, Eggstract,

Assuming that you're going to actually carry on with the work in the near future, and have no plans to modify the LC a bit, here are more links if you like:

https://www.agilent.com/cs/library/user ... _ebook.pdf

https://www.agilent.com/cs/library/appl ... 8647EN.pdf

These may help a bit with the "Peak Purity" aspect of things. If you have a service contract with Agilent, their engineers may be able to help you with using the DAD...never hurts to ask. Reviewing the documents prior to consulting Agilent folks will help enormously.

I encourage you to read the review article--you may find something within the review that will help your effort. You'll never know if you don't try. Someone may have written about a way or ways to suppress whatever dynamic behavior that seems to be going on with CBDa/CBD (?).

One thing to keep in mind is: (from G2180-90024)

"The most accurate and sensitive purity analysis is for peaks between about 250 and 800 mAU in height."

Either increase the concentration of analytes (if you can) or the injection volume, or perhaps both. A rule of thumb; if the sample solvent is of the same composition as the initial eluent composition of the gradient program, you might be able to inject as much as 10% of the column volume, a bit more than ca. 20 microliters or so. In any case, doesn't seem that you're injecting too much to me.

My Best Wishes for You and the Successful Completion of this Project!

[Eggstract]

Some quick examples of obvious problems with your "method" and technique can be seen in what you have presented so far. First, you did not provide any information about the exact column used. No chromatography suggestions or advice could be provided without such very basic information being described.

I'm sorry. The post wasn't formatted for this forum. It was copy-pasted from correspondence we have with a friend whose job is method development and who knows our system.

Column is a Poroshell 120 EC C-18 2.1x100mm 2.7um.

Next, you have the UV/VIS detector's 'Reference Wavelength' feature turned ON (indicating you have no training in how to use the DAD). Never, turn that on as it will delete your raw data rendering any output invalid [https://hplctips.blogspot.com/2011/03/reference-wavelengths-as-used-in-hplc.html].

The method is a gradient run, not isocratic, hence the use of the reference wavelength. That's a useful link though, thank you.

Also, your UV signal and scale are so low. You are looking at a lot of noise and inadequate equilibration. Your integration settings have not been setup properly. Your attempt to compare spectra are also invalid because the chosen settings are inappropriate and the signals were not integrated properly. Lastly, where did you get this gradient from? It makes no sense at all. You have composition changes occurring when peaks are still eluting. You have changes when nothing elutes. The gradient is going back to initial conditions before the peaks are out???? The operator has no idea what they are doing and you are focusing on the wrong causes rather than realizing that several years of professional training plus lots of mentoring is needed before you attempt this. A website is not going to help you, nor is someone sharing articles with you.Please seek out some professional help with this before you proceed.

The gradient is based on what's in the literature. You may think we're cluelessly fumbling around, but we've been head deep in the fundamentals for months now. Thanks for your input all the same, even if it seems like you'd rather criticize than help

I for one applaud your efforts. Most companies in MMJ industry don't do nearly enough quality control. Everybody starts at the beginning and learns as they go. Keep trying.

First thing that jumps out to is that the 1100 is really not a good choice to run smallish particles with 2.1 mm i.d. columns and the very low flow rate you are using (the system volume is way to high for that). Try something that is known to work like the HPLC method in the American Herbal Pharmacopoeia.

Column XBridge C18 4.6 X 150mm 3.5 µm Column Temp 30 C Mobile Phase A 9:1 50 mMol ammonium Formate (pH 3.75)/Acetonitrile Mobile Phase B 9:1 Acetonitrile/Water Detection UV @ 228 nm Flow Rate 1 mL/min

Gradient Table

Time % A %B
0 30 70
15 10 90
30 10 90
31 30 70
40 30 70



===========================================================

AHP uses ACN as a mobile phase, and they based it off Swift 2013. We instead opted to look to De Backer (2009) and the improvements to the method made by Patel (2017).

However, interestingly, Giese et al (2015) have published in AOAC their method and validation which also uses ACN, but without buffers. Perhaps most interestingly of all for us is their column is almost identical (150mm instead of 100mm being the only difference).

This week we're going to be looking at ACN as our B, and go from there.

We should not be worrying about the use of an 1100-series system or 2.1mm ID columns in the post. The problem areas are not related to the equipment or column choice.

On this much we strongly agree.

The Agilent 1100 is used widely in cannabinoid analysis.

Hi Again, Eggstract,

Assuming that you're going to actually carry on with the work in the near future, and have no plans to modify the LC a bit, here are more links if you like:

https://www.agilent.com/cs/library/user ... _ebook.pdf

https://www.agilent.com/cs/library/appl ... 8647EN.pdf

These may help a bit with the "Peak Purity" aspect of things. If you have a service contract with Agilent, their engineers may be able to help you with using the DAD...never hurts to ask. Reviewing the documents prior to consulting Agilent folks will help enormously.

I encourage you to read the review article--you may find something within the review that will help your effort. You'll never know if you don't try. Someone may have written about a way or ways to suppress whatever dynamic behavior that seems to be going on with CBDa/CBD (?).

One thing to keep in mind is: (from G2180-90024)

"The most accurate and sensitive purity analysis is for peaks between about 250 and 800 mAU in height."

Either increase the concentration of analytes (if you can) or the injection volume, or perhaps both. A rule of thumb; if the sample solvent is of the same composition as the initial eluent composition of the gradient program, you might be able to inject as much as 10% of the column volume, a bit more than ca. 20 microliters or so. In any case, doesn't seem that you're injecting too much to me.

My Best Wishes for You and the Successful Completion of this Project!

Thanks for the links mattmullaney. The review article in particular was interesting. It's not often we stumble across relevant papers we're not aware of, especially given cannabinoid analysis is still an industry in its infancy with no real commercial regulations or guidelines as of yet.

Interesting you mention increasing concentration or injection volume. We read somewhere (it escapes me where right now) of a method where a 6x dilution was used for quantification of trace cannabinoids, and a 96x dilution used for the major cannabinoids. There's certainly a lot which isn't detected in some of our samples when going from 10x dilution to 20x.

At first we suspected our humpback might be due to overloading, but given it appears in 100x dilutions as well, I think we can be pretty confident that's not the case, and a few hints lead us to believe co-elution might be more responsible.

[Multidimensional]

"The gradient is based on what's in the literature. You may think we're cluelessly fumbling around, but we've been head deep in the fundamentals for months now."

If you copied that method from any literature source, then you may not realize that many of the examples provided are invalid. Never assume that because you have attempted to copy a method from someone else that it must be OK to use (It is a sad truth that much of the information provided in application notes and articles is so poorly done and of no scientific value).

As previously noted, your use of the reference wavelength ('ON') alone in your gradient will result in the method failing (even if someone else made the same mistake and you copied it). If your goal is to learn HPLC fundamentals, then please turn it off. Copying methods and other examples does not make something right. As you gain some experience, you will soon discover how many of them are inaccurate and poor in quality. Your gradient profile, as noted before, also does not follow any fundamental guidelines of analysis and should not be used in any form for analysis. The sooner you realize this and seek professional help with your methods, the better off you will be.

*BTW: Yes, I am being very critical of what you have presented and that is because your efforts and information are misdirected. You provided a lot of information in your post. In doing so, you made it very easy for an experienced chromatographer to see all of the flaws present and provide you with some good suggestions. I want you to be successful, but by ignoring my free advice, you will not progress forward and learn from your mistakes at all. You will waste time, solvents and of course money.

[Eggstract]

by ignoring my free advice, you will not progress forward and learn from your mistakes at all. You will waste time, solvents and of course money.

That was the issue though. Up until this post, you hadn't really given any practical advice, you simply said 'Stop and hire a professional'.

We will turn the reference gradient off and proceed from there. I appreciate the criticism - there's nothing less helpful than praise where it's certainly not due. However, if criticism can be constructive as well, this is always preferred.

[mattmullaney]

Hi again, Eggstract,

Probably should have given you this link above as well:

https://www.agilent.com/cs/library/user ... USR_EN.pdf

While Chpt 5 discusses ordinary 2-D data acquisition, it does explain what the reference wavelength and bandwidth are and what they do. From this, you can infer why it is not a good idea to disable this feature while using 3-D data acquisition.

[Multidimensional]

"Up until this post, you hadn't really given any practical advice, you simply said 'Stop and hire a professional'."

That is not true. Please review my earlier posts to you.

If your company hires someone with demonstrated practical experience, then you will move forward much faster than trying to learn as you go. Right now, you are back-peddling. In the interest of time/money, hiring a professional chromatographer is still the best advice there is.

PS Matt: Nothing in that chapter (1290 DAD) suggested turning OFF or ON the Reference wavelength feature. It only shows and discusses examples, where a sample has NO ABS at all within the Reference Wavelength range selected and never discuses what happens in real life when you run actual samples which may or may not absorb in the Ref range you select. They also state clearly that the examples follow the default settings provided (with an understanding that no one would ever use the settings that they put in, because those are for a hypothetical sample, not your real one. You would actually choose your own wavelengths, bandwidths and special features (e.g. Ref Wave) for your own samples, not use the ones which are just place holders in the system). They also failed to mention how the feature actually works and why you should never use it (probably why they left it out). Esp if you work in a regulated environment OR if you wish to provide accurate data to someone. The optional Reference Wavelength feature takes all of the raw data from the original analysis signal, then subtracts out all of the data gathered by the reference signal and creates a new signal file which is used for the final output (final data file). All of the original signal data is destroyed and is not recoverable. So you end up with a corrupted data file which can never be used or validated as real (since the original data was destroyed). Its use (or not use) is taught as part of most formal DAD classes for method development. Please read the link provided for more info and also, if desired how to do the same type of thing, without data corruption.

[mattmullaney]

Hello again,

Well-explained, multidimensional--my thanks!

[AA]

The Agilent 1100-series is well suited to 2.1 mm ID columns at linear flow rates of ~210 ul/min."

I gotta disagree with this one. A stock 1100 has a system volume of over 1 mL, so at 200 uL a minute you get a 5 min isocratic hold and some extra re-equil time, not helpful. I said it was not a good choice, I didn't say you couldn't or it wouldn't work (with extensive modifications) with 2.1 mm i.d columns a low flow rate.  That instrument was designed for traditional column i.d's (i.e. 4.6 mm) at traditional flow rates (1-2 ml/min).

I have run thousands of cannabinoid assays out of flower, trim and extracts and never had issues like the ones described.  I gave up using the AHP method fairly early on but I know it works.  Running something you know works help focus in on the problem.

[Multidimensional]

My boss just looked at your initial post (among many other things,he provides professional consulting services to the cannabinoid and e-juice labs and knows these topics well) and agreed with my initial post. He said that "Your data and presented method indicate that you do not have any chromatography experience and need to hire someone who does before proceeding. You also have a very large gradient delay, your run is too short, ended too early and the gradient program makes no sense. Why not use regular 4.6 mm ID columns which work so well for this application, esp when starting off (much easier)?".

Besides that he may have also caught one item which could explain your very bizarre gradient method. You made a later comment about different column sizes in one of the methods that you used. He noticed that your proposed gradient program looks like it was designed for a 2.1 x 150 mm column, not a 2.1 x100 mm column as you are using. If you change the column length (or size), then you must also change the gradient program. Perhaps you partially copied a method and then used a shorter column without changing the gradient? That would explain why the method is just getting started when you end it. It does not explain all of the problems with the method, but is consistent with the types of mistakes someone new to chromatography would make. He also suggested that the gradient delay would also require that you update the gradient composition program to account for it too.

As to the std 1100 for 200 ul flow rates. For isocratic analysis, no problem. Gradient analysis at 200 ul. Due to gradient delay (~900 ul dwell for a stock system), this is why we routinely modify them with smaller mixers and narrow ID tubing (an easy change for an experienced chromatographer) to accommodate gradient methods. (new dwell is ~ 330 ul after changes). They are very reliable at 200 ul/min and do not start to become less reliable until you run at microbore flows. The pumps are some of the best made, but they do have limits.

In my opinion, the initial poster should not worry about the equipment choice right now, that can be optimized later (or just use conventionally sized columns which work great right now plus no modifications). Without having any training in chromatography or an understanding of the basic fundamental concepts he is going at this blind. So while the list of problems is very large (large gradient delays, poor gradient programming, troubleshooting the wrong items, use of Reference Wavelength) the key problem which needs to be addressed first is that of training. Hire one or more professional chromatographers and get on that learning curve. On your own, this will take many years to gain only an introductory level of experience and you need more experience than that right now.