Principal Investigator: Brian Haab, Ph.D.
Institution: Van Andel Research Institute, Grand Rapids, MI
Principal Investigator: Randy Brand, M.D.
Institution: University of Pittsburgh
The goal of our project is to develop new biomarkers for pancreatic cancer based on glycosylation variants of specific proteins. The team includes the co-Principal Investigators, Dr. Brian Haab (Van Andel Research Institute) and Dr. David Smith (Emory University); Dr. Randall Brand (University of Pittsburgh Medical Center); Drs. Marshall Bern and Doron Kletter (Palo Alto Research Center); Dr. Kelley Moremen (University of Georgia); and Dr. Ying Huang (Fred Hutchinson Cancer Research Center).
One of the most common features of pancreatic cancers is the increased abundance of a carbohydrate structure called the CA 19-9 antigen. This carbohydrate structure (glycan) is attached to many different proteins, many of which are secreted from the tumor into the blood circulation, making it available for detection as a biomarker. The detection of CA 19-9 from blood samples is widely used for confirmation of a diagnosis of pancreatic cancer and to get information about disease progression. Unfortunately, the test has limited usefulness for early detection or diagnosis because about 20-30% of incipient tumors produce low levels of CA 19-9. The low levels are usually due to inherited genetic mutations in the genes responsible for the synthesis of CA 19-9. Our preliminary work has shown that the patients who produce low CA 19-9 produce alternate carbohydrate structures that are abnormally elevated in cancer. Characterizing and identifying those structures could lead to an improved ability to detect cancer in the CA 19-9-low patients.
Our strategy is to identify carbohydrate structures that have elevated abundances specifically in patients with low CA 19-9. We are using novel glycobiology tools to analyze the carbohydrates from pancreatic tumors and control pancreas specimens. In an approach called 'shotgun glycomics', all the glycans from the tissue specimens will be isolated, fractionated into their individual components, and analyzed to determine structures and abundances. The detection and measurement of the glycans will be enabled by new protein reagents that bind specific glycan structures. These tools will allow us to determine which glycan structures are most elevated in the patients with low CA 19-9. We then will use the new glycan-binding proteins to develop assays for measuring the glycans and the associated proteins in blood samples.
The new detection assays developed in the above approaches will be applied to the testing of new biomarkers in blood samples from pancreatic cancer patients and controls. We have a valuable tool for efficiently measuring glycan levels of specific proteins in many different biological samples, the antibody-lectin sandwich array. Microarrays of antibodies can be used to capture multiple, different proteins out of biological samples, and lectins can be used to probe specific glycan levels on the captured proteins. We can then test the hypotheses that specific glycoforms of proteins secreted from pancreatic tumors are elevated particularly in the patients that are low in CA 19-9 and that the use of these assays in combination with the CA 19-9 provides highly accurate detection of pancreatic cancer. We anticipate these new approaches advancing pancreatic cancer diagnostics as well as benefitting other glycobiology research in cancer.
Synopsis of Research and Network Resources
Introduction: A Brief Summary of the State of the Science and Research Need
Nearly 80% of pancreatic cancers overexpress a glycan called sialyl-Lewis A (also known as CA19-9 antigen), which can be detected by CA 19-9 monoclonal antibodies, and forms the basis for the FDA-approved CA 19-9 assay. The rate of elevation of CA 19-9 makes it the current best biomarker for pancreatic cancer detection and diagnosis. But CA 19-9 is still not good enough because a useful biomarker would need to identify much more than 80% of the cancers. Searches for biomarkers that perform better than CA 19-9 so far have proven fruitless.
The basis of our Alliance project is to investigate whether the cancers that do not elevate CA 19-9 overexpress other glycan biomarkers that have not been defined. Given that most pancreatic cancers overexpress specific glycans, the possibility exists that the CA 19-9-low tumors make and secrete glycans that are different from the CA 19-9 antigen and also different from those of normal secretions. If we focus our discovery efforts on the CA 19-9-low subgroup, we may find markers that are complementary to CA 19-9 and useful in combination with CA 19-9 to form a highly effective biomarker panel for early detection and diagnosis of pancreatic cancer. We are testing this hypothesis in a variety of ways: mass spectrometry (MS) discovery in cell lines and tissue; glycogene profiling of cell lines; and antibody-based discovery and validation in human plasma samples.
Adding Blood Test for Pancreatic Cancer May Aid Early Detection
Using a new blood test for pancreatic cancer alongside the current blood test may improve early detection and help screen people at high risk for the deadly disease, researchers say.
Laboratory-specific Studies to Meet the Research Needs
MS-based discovery using cell lines
Cell lines are useful in experiments that require purity or highly controlled conditions. To develop materials suitable for thorough glycan profiling by mass spectrometry (MS), we identified three high-CA 19-9 and five low-CA 19-9 cell lines that give a good cross-section of the phenotypes and genotypes observed in pancreatic cancers. Matrix-assisted laser desorption/ionization (MALDI)-MS profiling of both the N-linked and O-linked glycans of the cell lines, performed by the Smith lab at Emory University, revealed notable differences between the cell lines, including increased high-mannose structures in the CA 19-9-high cells and loss of O-glycans in the CA 19-9-low cells. At present the Smith group is confirming these findings and identifying specific candidate structures that are present uniquely in the CA 19-9-low cells.
An additional discovery approach currently underway is shotgun glycan array analysis. To produce material for shotgun glycan arrays, the glycans from a biological sample are isolated, tagged with a fluorophore, and fractionated by multidimensional chromatography to near purity of each glycan. The resulting Tagged Glycan Library (TGL) is then used as a glycan source to produce a printed glycan microarray. The Smith group is making arrays from both CA 19-9 expressing and non-expressing cell lines. We will probe the arrays with lectins and glycan-binding antibodies to identify differences between the cells and then further characterize the relevant glycans by Metadata Assisted Glycan Sequencing.
MS-based discovery using tissue
Primary tumor material allows a direct look at what is happening in the tumor, without potential alterations induced by culturing. To identify pancreatic cancer cases for in-depth study, we made a tissue microarray containing several cores each from tissue blocks of 43 different tumors, obtained from the Van Andel Research Institute Biorepository. We performed several analyses: H+E stains to determine histology; immunohistochemistry to determine CA 19-9 levels; and genotyping of DNA isolated from each patient sample to determine the genotypes of several glycosyltransferases involved in the biosynthesis of Lewis antigens and other glycans. We selected five cases covering a range of characteristics, and the Smith group is performing glycomics analyses on cell lysates from the selected blocks. In addition, Dr. Richard Drake at the Medical University of South Carolina is applying a novel glycan imaging analysis to mounted sections from the blocks. The initial results from Dr. Smith appear to corroborate the findings from the cell lines where the CA 19-9-high cells have increased N-glycans and that the CA 19-9-low cells have decreased O-glycan expression. We still need to confirm the results and to identify the structures, but we already are seeing the usefulness of the multi-pronged approach in this project.
Glycan discovery using gene expression profiling
The glycosyltransferases involved in synthesizing sialyl Lewis A—the antigen recognized by CA 19-9 antibodies—generally are known, but researchers have not been able to specify the gene expression changes that lead to upregulation of the structure. We are taking a new approach that might help answer this question and predict the glycans made in the absence of sialyl Lewis A. We discovered that CA 19-9-expressing cells greatly upregulate sialyl Lewis A, without major changes in protein expression, when grown in 3D culture relative to 2D culture. Now we can compare between CA 19-9-high and CA 19-9-low in a single cell line, giving us a common background between the conditions. Lectin blots of the lysates showed increases in GlcNAc and GalNAc in specific presentations, but we do not yet know the details of those structures.
Dr. Kelley Moremen at the University of Georgia is collaborating on the gene expression profiling of 65 glycogenes that are involved in the biosynthesis of Lewis antigens and glycans containing GalNAc and GlcNAc. We anticipate uncovering features that are specifically involved in sialyl Lewis A production and in the production of alternate glycans. Because we have MS and antibody/lectin methods to use on the same cell lines, we can immediately test the presence of glycans predicted by these analyses.
Antibody-based discovery and validation in human plasma specimens
Ultimately we need to test our biomarker candidates using antibody and lectin assays, because such assays provide the specificity, reproducibility, and throughput needed for biomarker validation. We are investigating the possibility that the CA 19-9-low patients make structures that are related to sialyl Lewis A but are not detected by CA 19-9 antibodies. Previous studies provided evidence that modifications to sialyl Lewis A such as di-sialylation, sulfation, and lack of sialylation are present in some disease settings. Using queries of our GlycanBinder database (see below), we identified a panel of antibodies to probe the various candidate glycans. We then used antibody array technology to obtain measurements from every combination of capture and detection antibody in a series of plasma samples. A few of the capture-detection combinations indeed showed elevations in the CA 19-9-low cancer cases, and an analysis of the specificities of these antibodies (using glycan array technology) gave more details about the nature of the elevated glycans. Subsequent experiments using additional antibodies confirmed the predicted structures. We are now in the process of bringing these results together into a single, validated test for the detection of early-stage pancreatic cancer.
Resources and Reagents for Sharing
We are hopeful that the GlycanBinder database and GlycoSearch software program will be useful for many projects in glycobiology. We developed these tools to facilitate the mining and using of glycan array data. Developers of glycan arrays are continually increasing the diversity and scope of the arrays, bringing the potential to better determine the rules that govern lectin binding. However, this opportunity has also presented new challenges with understanding the data . The GlycanBinder database contains analyzed data, raw data, and metadata from thousands of glycan array experiments, and the GlycoSearch program enables mining of the experiments and the database. Researchers can use this resource in several ways: to delve into individual data sets; to analyze the relationships between various lectins and antibodies; to find lectins or antibodies that bind to particular glycans; and to use the output of the analyses in the interpretation of lectin experiments. The package is available upon request, and we are working to enable web access.
Another resource is the shotgun glycan arrays made from pancreatic cancer cells. Many of the glycans contained in these arrays are likely to be different from those in any other available glycan array, given the uniqueness of the starting material. Therefore, researchers may find the arrays to be a rich resource for discovery.
We also are developing recombinantly-produced lectins for probing glycans. Lectins exist throughout biology, and their glycan-binding specificities cover an immense range of motifs, yet only a limited set of lectins is routinely used for probing glycans in biological samples. We are identifying additional lectins that may be useful as probes and expressing them recombinantly. Recombinant expression gives a higher level of purity than possible from purification from natural sources, and it allows tagging of the lectin for detection using a secondary reagent. In addition, recombinant production allows optimization or modification of binding properties though various sequence manipulations.
Public Health Implications/Advancing the Field of Glycobiology
The most direct impact of this project will be through the new biomarkers developed to improve outcomes for pancreatic cancer patients (see below). Another area of impact could be through the methods and strategies developed in the project. The usefulness of the methods and strategies for research in pancreatic cancer indicates potential value for research on other diseases, thus having an impact beyond pancreatic cancer. In a similar way, the resources and reagents developed in this project could advance the glycobiology of various other research areas besides pancreatic cancer.
Cancer-Specific Relevance: Detection, Prevention, and Treatment
Many patients with pancreatic cancer receive late or inappropriate treatment due to lack of accurate information about their disease. A blood test that distinguishes early-stage pancreatic cancer from other, benign diseases of the pancreas could have a significant impact on patient care. It could allow patients with early-stage disease to get the appropriate treatment as rapidly as possible, leading potentially to better outcomes; it could be portable to locations distant from major hospitals, thus benefitting a broader population than possible with the current diagnostic procedures; it could be applied to high-risk individuals as part of a surveillance program to guide the use of more expensive and invasive methods; and it could spare some patients from unnecessary procedures or surgery.
Opportunities for Collaboration
Several opportunities for collaboration are available. For example, researchers may be interested in probing specific glycans using lectins, but are not aware of lectins for that purpose or would like a better way to interpret data from lectins. The GlycanBinder and GlycoSearch software tools assist with those tasks. We can use the database to search for lectins that bind particular motifs and to precisely characterize the specificities of particular lectins, and when researchers use lectins or glycan-binding antibodies to probe biological samples, we use that information from GlycoSearch to quantitatively interpret the data to give an assessment of the motifs that are present in the sample.
Researchers also may be interested in using the pancreatic cancer glycan arrays. These arrays could prove valuable for defining the glycans that are present in well-defined subtypes of pancreatic cancer. Follow-up studies can be performed on primary tissue, which we can provide on a collaborative basis.
Tools available from the Haab Laboratory
MotifFinder is a software program for the automated analysis of glycan array data. The program is designed both for researchers with no expertise in glycobiology as well as experts in glycobiology. The important features of the program are:
- Full automation; accessible to non-experts
- Accurate output on basic specificity; insights into complex, fine specificity
- Powerful additional features:
- Ability to integrate information from replicate datasets, giving high-quality output
- Platform-independent analysis; supports all types and platforms
- Concentration-dependent analysis; calculation of binding
The CarboGrove database is a collection of glycan-binding results produced by MotifFinder. This database of readily accessible reports makes the correct understanding and interpretation of lectin binding possible for non-glycobiologists. Used in conjunction with MotifFinder, CarboGrove enables non-experts to identify the relative levels of glycan epitopes from profiles of lectin binding.
SignalFinder-IF is an image analysis software package which can be used to quantitatively evaluate tissues stained with immunofluorescent markers. The algorithm rapidly and objectively sets thresholds specific to the image being analyzed without relying on specific signal distributions or shapes. This tool can assist researchers and clinicians in making informed decisions by providing unbiased, quantified signal information.
The software comes with two utilities: one for determining the locations and amount of patterns of markers, using multimarker staining data as the input, and another for producing images of the signal locations overlaid on brightfield images of the tissue. These utilities are called ColocFinder and Overlay, respectively. This package is thoroughly presented in Barnett, Hall, and Haab, Am J Path 2019. Information on protocols for multimarker immunofluorescence also are available from the Haab Laboratory. We use multi-round staining combined with whole-slide fluorescence scanning.
Another version of the software is available for the analysis of microarray images, called SignalFinder-MA. This program provides rapid and automated analysis of microarray images and offers many data processing tools.
To view the Haab Laboratory Tools, go to: https://haablab.vai.org/?tabUrl=tab-4.