Developing a LC-MS method for carbohydrate derivatives

[haagiboy]

Hi, first time poster here.

I have just started on a PhD in chemical engineering, and my group are working on catalytic conversion of biomass to biofuels. One of our catalysts converts cellulose to ethylene glycol/polyethylene glycol and other -diols, both straight and ring formations. We also have some HMF derivates such as tetrahydrofurfuryl alcohol.

Setup:
Agilent 1260 HPLC outfitted with a HiPlex Ca-duo column
Agilent 6120 Quadrupole LCMS System (ESI ionization)

As of now, we have just used distilled water as our mobile phase. No buffer or anything added to it.

Questions: Is it possibe to remove the +Na adduct all together? We tried using 0.1% Formic acid, but to no avail. The baseline noise just increased and some products where not found using this mobile phase. The ONLY ion we see in our standard solutions is the [M+Na]+. No real fragments occuring. The peaks are also pretty broad.

However, when analyzing our products, we get several peaks that can be described by different fragmentations such as M+Na-H2O+H, M+Na-OH+H etc..

Is this ok to use for characterization alongside results from GC-MS? The problem is that we have peaks in the GC-MS that are not easily characterized, and we have the same problem in the HPLC.

So, should we try to remove the Na+ or just run with the ions and fragments we get at this point? Will changing dest.water to HPLC grade water or installing a desalter column between mobile phase and HPLC or between the HPLC and MS help at all?

[JMB]

In my experience, you can never eliminate [M + Na]+ for the simple reason that the Na+ comes from the glassware; don't waste your time.

If it cannot be eliminated, then take advantage of it; in my lab. where we also used ammonium formate as a modifier on occasion for RP-HPLC, we would typically see [M+H]+, [M+NH4]+, [M+Na]+ and [M+K]+ i.e. signals at +17, +5 and +16 apart above the molecular ion. Residual NH4+ in the system would be present.

You may not always see all four of these signals, but even two (better three) will enable the MW to be determined. This is critical, because your EI/GC/MS spectrum may well be dominated by successive losses of H2O.

The MW is the single most important piece of info to determine !!!

Regards

[haagiboy]

Yeah we've figured out that we should just run with the sodium adduct.

I have a question regarding a spectra I obtained with several peaks. I have analyzed it and found two compounds, and every peak can be explained by adding sodium, and then fragmenting of oxygen, water and some carbon chains.

I have read that ESI is a very mild ionization, so you should really only see M+Na peak, and not so many fragments.

The problem I face, is that it's hard to determine the first peak at around 8 minutes if it contains just 2 species, or several species. The product we obtain from the reaction can have all of the deoxygenates/dehydrogenates.

I can try to upload the specific spectra tomorrow and see if you agree with my interpretation of it. The interpretation is important to decide which "pure" compounds we should buy to test for retention times.

[JMB]

With regards to GC/MS, you might think about analyzing your carbohydrate mixture as the trimethylsilyl derivatives (TMS ethers).

These will generally give a lowish relative abundance of M+. and (M - Me.)+ as a strong signal; again, very useful for MW detmn.

All of the instrument libraries contain the carbohydrate-O-TMS spectra, which will help your identification.

Regards

[haagiboy]

All right! I'll see if that will help analyzing the GC-MS results. We use the NIST08 database, and I havent really used the GC-MS myself yet. Literally only 3 weeks into my PhD. I just don't want to waste time on wrong/bad methods.

I love that GC-MS has a database, but unfortunately this is not the case for LC-MS using ESI.

[haagiboy]

With regards to GC/MS, you might think about analyzing your carbohydrate mixture as the trimethylsilyl derivatives (TMS ethers).

These will generally give a lowish relative abundance of M+. and (M - Me.)+ as a strong signal; again, very useful for MW detmn.

All of the instrument libraries contain the carbohydrate-O-TMS spectra, which will help your identification.

Regards

thank you for guiding us in this direction. It definitely looks to be the route to go.
have you worked with silylation of carbohydrates before?

I try to find a simple method for the silylation, but they all seem to be rather labour intensive. Is this correct? Our sample that we want to analyze contains a large amount of water.

Is it possible to do something as simple as add x amount of TMSI or other silylating agent to the product containing vial directly?

[lmh]

sugars by GC-MS: yes, they work, but it's not a totally smooth ride. For a start, they all tend to fragment very similarly, so although you have the NIST library, you may find that you have sugars that are hard to distinguish on spectrum alone. Also you may find that there is an optimum length of time to derivatise, with samples gradually going off afterwards, meaning that it can be difficult to get really good reproducible results. It's also worth making sure that however you dry the samples, you don't end up frying them into a brown mess that doesn't redissolve or derivatise properly.

the lc-ms method you have: you are quite right, it can be difficult to tell if peaks are single compounds. Oligosaccharides fragment differently according to whether they are sodium or hydrogen adducts. The hydrogen adducts fragment much more easily than the sodium adducts, and usually fragment at the glycosidic bonds between the individual monosaccharides, with the neutral leaving group being dehydrated, so the fragments just look like normal, shorter oligosaccharides. You may find that you were actually forming hydrogen adducts in ESI, but they were fragmenting in the source, and unless you can get separation by LC, it's hard to tell. The sodium adducts are tougher, and the sodium intercalates between the sugar rings in oligosaccharides (or so I've been told) providing an additional bonding between the monosaccarides, and therefore meaning it's more likely you'll get other fragmentations not at the sugar-sugar bond. In methods using ammonium buffers you may form ammonium adducts but they tend to fragment very easily too, losing ammonia and then behaving like a hydrogen adduct. By the way, I notice your column can also be used with dilute sulphuric acid; do not even think about using this with MS. You will get beautiful sulphate adducts in negative mode for the first run or two, before the spray chamber dissolves. Don't ask me how I know this.

[JMB]

There are several standard handbooks on derivatization for GC/MS.
1) Daniel Knapp et al
2) K. Blau & King

If your products are in aqueous soln it needs to be dried down. I suggest freeze-drying (lyophilization) to avoid the brown mess mentioned by lmh.

TMS ether formation is or used to be the gold standard for carbohydrate dervn for GC/MS., hence the huge number of NIST library entries.

I know there is/was a TMS mixture that is suppose to work even in the presence of water (see Pierce & Co or Sigma-Aldrich catalogues) the problem being that TMS-Cl is very readily hydrolyzed.

By the way, do you know whether the catalyst produces only the analyses named in your OP ? Are mono- & oligosaccharides also produced ?

The TMS dervn. Is not itself lab our-intensive, just a dilute & shoot; however, getting the reaction to be anhydrous is, but freeze-dryers run overnight unattended.
What scale are you working on 1g, 10 g, 100 g ....??

[haagiboy]

So, here is an image of the LC-MS spectra obtained at 80degc column temperature with a fractionation voltage of 70.



I have added the numbering myself.

After analysing this spectra by looking at MS spectra of each peak, and finding the LCMS peak for each fragment/compund, I have found that several peaks overlap and are even slightly hidden.

Some peaks only have one MS ion peak corresponding to M+Na. And some mass fragments are found at several different retention times, e.g m/z=155.1. This I believe could either be fragments, or products. IF it is a fragment, then surely the fragment piece of 155.1 should show the same retention time as its parent molecule? E.g I find a m/z peak of 205.1 at RT=24.704 min, and a peak of 157.1 at 24.655 min. This could mean that the peak with a m/z value of 157.1 could be a fragment of the main compound at that retention time?

The problem occurs when I find a m/z value of 157.1 at other retention times... This could perhaps be explained that the 157.1 compound could be in our product solution, with different isomers?

Also, any ideas on how to perhaps increase the separability of the LCMS spectra? We have a RI detector coupled to this as well, and this shows good resolution. However, if I understand correctly the RID spectra doesn't really tell us much other then possible quantification of a peak? And the order of peaks in the RID = the order of peaks in HPLC? As in first peak in RID = first peak in HPLC?

EDIT:

Regarding possible products. We have different sugar alcohols. Mainly ethylene glycol, propylene glycol and glycerol. We also have several different isomers of butanediol, pentanediol, hexanediol and possibly butanetriols.

Since we start with cellulose, we expect to find glucose and fructose as well, but we only see the sugar alcohols of these (sorbitol, mannitol etc).

I am also wondering if we might see some xylitol/arabitol along with Arabinose/Xylose. Perhaps some Fucitol as well.

The problems with the products between the sugar cellulose and ethylene glycol is the number of possible isomers. If we remove an oxygen from glucose it could either be a compound, or a fragment right? And depending on where it is removed we will have different retention times. So all I can really say at this point is that we have deoxyglycose, dideoxyglycose, anhydro-dideoxy etc etc....

But that isn't really a problem, as long as we can understand the reaction pathway by doing more tests with both shorter and longer residence time.

Ideally I would be able to identify each of the main peaks in the HPLC spectra and mark it with "2,5-anhydroglucitol" etc.

[lmh]

I think you're aiming for something unachievable. Looking at your chromatogram (note: if the x-axis is time, it's a chromatogram; if it's mass, wavelength etc., it's a spectrum), there are just too many peaks all combining. Just thinking statistically, even if you manage to identify them, what is the point of labelling, say, peak 7? It is probably quite a substantial component, but it's so close to 6 that it could easily become an unintegrated bump on the side, even while it's still quite abundant. And statistically, what is the likelihood of there being other peaks underneath the ones that actually show up in the TIC? (very high indeed).
Your sample is too complex for full separation by the LC method, and coelution is happening. The TIC is going to look bad, and integrating it won't measure anything.
You have two choices: if you want an attractive and convincing total chromatogram containing all compounds, you need to switch to a much higher-resolution method (GC might help; the wild-kid on the block would be capillary electrophoresis, which needs derivatisation for neutral sugars, but it doesn't seem very popular). Alternatively, you can accept that your separation isn't good, and look exclusively at extracted ion chromatograms for components of the mix that are interesting to you. If you do this, be aware of cosuppression: compounds affect each other's ionisation, so if A and B coelute (even partially), then if B goes up, the signal for A may go down.

[haagiboy]

Thanks for correcting my terminology. I am fairly new to this as you see

Yeah I definitely think I am asking for a bit too much here. At least we have identified some of the peaks as being pure. And luckily enough, these pure peaks with just one M+Na are the interesting products!

I think I will just use the LC-MS to evaluate different product composition in regards to amount of e.g Ethylene glycol. And maybe see a shift of some peaks to increase, and some to decrease in the RID.

Thanks for the help! I might return to discuss some GC-MS results at a later point if I need some help there as well. This forum has been very friendly and helpful!

Just one more question... Between 10-40 minutes in the chromatogram, we see a ion with a m/z value of 82.1. I am trying to look into it, and I believe this might be a propanal radical ion + Na? Or Sodium acetate?

[JMB]

haagiboy,

You have jumped, or been pushed, head-first into a challenging analytical chemistry problem; you should take a step back and do some quiet-time reading to establish some of the basics. Given that I don't know your level of background knowledge, I may overstep the mark here so apologies in advance.

Some basic issues are,

[1] Q: what is basis of separation by GC ?
A: differences in volatility (monosacch > disacch > trisacch > oligosacch) so monos elute before disacch before trisacch etc

[2] Q: why derivatize (e.g. as TMS ethers) ?
A: to increase volatility

[3] Q: what is basis of separation by your Calcium HPLC column ?
A: I don't know----do you ?

[4] Q: What is essential difference between the molecular ion formed in LC-MS & EI GC-MS ?? How does this drive subsequent fragmentation ?
A: this is VERY important

[5] Q:what is the difference between a mass chromatogram, an ion chromatogram and an extracted ion chromatogram ?

[6] Q: do you know how to use the MS data system to generate an extracted ion chromatogram ? Apply this to the total ion chromatogram (TIC) shown in one of your posts.

At this point, I would focus my efforts on the GC-MS analysis of the trimethylsilylated reaction mixture. A capillary GC column has about n-thousand theoretical plates (n~ 30-50 or more), whereas your Ca column has maybe only 30-50. [You can determine this from the chromatogram you have shown]. You will resolve many components cleanly using GC, AND you have the NIST library of TMS-ethers to help in identification.

If you don't have freeze-drying eqpt. the biochemistry dept will have it. While you're waiting to run GC-MS, apply [6] to your LS-MS data.

I would initially use the GC-MS data to drive the interpretation of the LC-MS data.

[7] Can you post the RID chromatogram that corresponds to your TIC ?

Regards,

JMB

[MSCHemist]

I personally like GC-FID or MS of PAAN derivatives peracetylated aldonitrile.

I take 0.5 ml of sample and 0.2ml ribitiol ITSD and blow it down to dryness at 45 deg C with nitrogen

add 200ul of 2% hydroxylamine pyridine incubate 1 hour at 90 deg C

cool

add 200 ul of acetic anhydride and incubate 1 hour at 90 deg C

cool and add 1ml water and 500 ul of methylene chloride (I am investigating substituting toluene in the future). 

Inject the methlene chloride layer. I use a db-5 column. monosaccharides and sugar alcohols come off ~ 200-260 deg C and disacchardies 315 deg C. I am planning to eventually get a high temp DB5 to do maltotriose. I am also eventually looking to replace the methylene chloride with either toluene or ethyl acetate.

Fructose seems to work well enough to quantitative. I get linearity of .996 vs .999-.9999 for the rest. The LOD is higher for it as it is split into two smaller peaks for the cis and trans oxime peracetate.

What is the best way to quantify sugars in gc-ms? - ResearchGate. Available from: https://www.researchgate.net/post/What_ ... s_in_gc-ms [accessed Mar 7, 2016].