UHPLC and MS Techniques Aid in Pharmaceutical Research

LCGC recently spoke with Pauline Rudd of University College, Dublin, and The National Institute for Bioprocessing Research and Training (NIBRT) in Dublin, Ireland, about her work using ultrahigh-pressure liquid chromatography (UHPLC) for the characterization of protein glycosylation. Rudd’s group has done extensive work in this area and has partnered with several companies to build databases of their findings to share with the research community.

How does your research group use high performance liquid chromatography (HPLC) or UHPLC for the characterization of protein glycosylation?
The analysis of pools of glycans depends on the application of orthogonal separations technologies. UHPLC using ethylene bridged hybrid columns provides high sensitivity, quantitative, high resolution separations of 2AB-labeled sugars. It is relatively inexpensive and straightforward to operate. In our lab anyone is able to carry out this technique, but mass spectrometry (MS) requires more training and we have fewer instruments.

How did your group develop high-throughput liquid chromatography with exoglycosidase digestion and MS technologies for this research? Was it a long process or did you adapt previously known techniques for your specific needs?
This technology was based on gel filtration (P4) technology developed in Akira Kobata's lab in Japan and Raymond Dwek's lab in Oxford. After the San Francisco earthquake, we were unable to obtain P4 gel and in frustration turned to columns we were using for protein separations. To our surprise, the IgG glycan pools we were analyzing were separated into 16 peaks compared with just three on the P4 columns. Geoffrey Guile, a research assistant, then tested every HPLC column we could buy and we selected amide columns as being the most robust and the most highly resolving. We used these columns to make preliminary structural assignments from the database we built and coupled these with information from exoglycosidase array digestions and MS to confirm the structures.

There were two other important breakthroughs. The first was that in the Oxford Glycobiology Institute (OGBI) and in Oxford GlycoSciences (Abingdon Oxfordshire, UK) conditions were established in which all the enzymes could be used in arrays at pH 5.5. The second was the realization that there was no need to isolate individual peaks as we had before. Since each sugar has a particular monosaccharide sequence and linkage each will take a different route through the array digestions. This reduced the time for a glycan analysis from 2 years to 2 days. I will always remember analyzing the glycans attached to calreticulum (admittedly not very complicated!) over a weekend!

What kind of information can you get with these technologies that cannot be obtained with other methods?
The most important information we can obtain with this technology is relative quantitation, both from the undigested glycan pool and on the products of the enzyme arrays. This, together with our automated glycan release robot, has enabled us to determine glycosylation changes in disease. More recently, we found associations between the serum glycome and the IgG glycome by analyzing the glycans released from all the serum proteins in blood and from IgG isolated from a few microliters of blood from thousands of individuals. This is the first time that genome-wide association studies (GWAS)/glycan studies have been possible. We have already identified an extremely good marker for MODY (maturity onset diabetes of the young) type diabetes and found five genes not previously known to be linked to glycan processing. Another important piece of information from these technologies is that we can distinguish between some arm-specific isomers and epitopes, such as Gal alpha (1,3)Gal, which is antigenic.

Can you detect differences in the attachment sites of glycans as well as differences in the glycan structures themselves?

No, we cannot determine the attachment sites using these technologies. For this we carry out glycopeptide analysis using MS. However, for a site analysis we analyze the released glycans before analyzing the glycopeptides to limit the search parameters in the MS step.

Are these methods useful for both the characterization of biopharmaceuticals as well as the potential identification of cancer biomarkers?

They are. UHPLC is useful for making clonal selection, quantitatively determining antigenic epitopes, at-line monitoring of glycosylation, determining, for example, the effect of media changes, pH, hypoxia, and the addition of metal ions. It is also a useful technique for obtaining a rapid profile of a biosimilar to compare with an innovator product. Within the National Institute for BioProcessing Research and Training in Ireland we use these techniques all the time in our research collaborations and contract services when we engage with pharma.

Your most recent research has been on the underlying molecular mechanisms of some types of cancer and the potential for biomarker discovery. What have your results indicated so far?

We learn a lot about glycan processing pathways from our fundamental research, which enlightens our understanding of the features that are important in bioprocessing, particularly on our genome glycome studies. We also published the first study of the effects of demethylation on glycosylation and looked at the effects of hypoxia, relevant to cancer as well as bioprocessing. The unfolded protein response involves glycosylation pathways and this becomes really important in maintaining the integrity of protein structure during bioprocessing. The advantage of working on glycosylation pathways in cancer is that there is already a lot of information in the literature, which enables us to test the general principles in a bioprocessing setting. Another important reason for working on cancer is that it enables us to understand more about the effector functions required by the drugs that target disease. We have recently expressed human Fc gamma receptors for an analysis of the binding of IgG, the first step in understanding which cellular receptors in an individual patient will respond to particular monoclonal antibodies.

We have identified a number of glyco-markers that outperform existing markers both in human and animal diseases. We are currently working on a new format for moving these into a clinical setting. We do have to contend with the conservatism of large diagnostic companies since most of their assays are based on enzyme-linked immunosorbent assays (ELISAs), which don't work well for sugars.

You also build customized glycan databases for HPLC, UHPLC, MS, and capillary electrophoresis (CE). How will these databases help other scientists and laboratories in the future?

Our aim has been to develop technologies that can readily be undertaken by any lab using UHPLC with fluorescence detection. We have published the details of our automated release platform as well as the use of the enzyme arrays and databases. These protocols are all in the public domain; they have been accomplished mainly through European funding and SFI where we built grant proposals through GlycoScience Ireland. This organization was founded by Lokesh Joshi at NUI Galway and me when we first came to Ireland.

Some of our databases and software have been commercialized and include reporting software that really helps the experimentalist to couple UHPLC and MS during data analysis.

The hydrophilic interaction liquid chromatography (HILIC), reversed-phase, and UHPLC databases are all experimental and are being added to all the time. By consulting the tables it is possible to obtain a preliminary assignment for sample peaks by comparing their glucose units with the gu values of structures in the databases. It is also possible to download the databases and personalize them. The Waters database includes lists of glycans we found attached to a range of biopharmaceuticals.

CE is a very high-resolution technology, but it is hard to know which structures are in each peak. Therefore we used a grant an outside company to build a CE database that can be used with enzyme array digestions in the same way as the LC databases. We hope this will be useful to the many groups who use CE.

What are the next plans for your research?

We are very excited by the possibility opened up by our automated technologies for us and others to carry out more genome and glycome studies, both in humans and in cell lines.

We also hope to promote the integration of different “omics” data to obtain a more holistic view of health, disease, and bioprocess pathways, as we now have sets of omics data from various patient groups. It is very important that glycomics is included in “big data” sets now because it will be much harder to do this later when the bioinformatics have been designed without including this major post-translational modification (PTM). In the glyco field, over the last decade, we have been developing bioinformatics platforms that links many glyco databases. The most recent version of this is UniCarbKB.

We are also very interested in understanding more about the glycosylation pathways and the control points in cell culture. There are a number of theoretical pathway modeling initiatives, including one in Gavin Davey's group at Trinity College Dublin.

Another interest area is in O-glycosylation, particularly in the structural changes that occur on mucins with the addition of sugars since this determines the types of bacteria that colonize the gut. We have an exciting program with Steve Carrington's group at University College, Dublin.

These are big ideas and we are very privileged to have so many internationally recognized collaborators who will be able to build on the areas we have opened up. I hope I will always remain in touch so I can always be excited by the new science that emerges.

Related content:

Ultrahigh-Performance Liquid Chromatography–Mass Spectrometry in Lipidomics

Application of Quality by Design Principles to a Pharmaceutical Sample Using UHPLC Method Development with Modeling Technologies