© 2004 National Academy of Sciences
Three bacterial and viral glycan-binding proteins were tested for their sugar-binding properties using a 200-member carbohydrate array. By looking at what the bound sugars have in common, scientists can infer a protein's particular binding motif. The influenza hemagglutinin shown at middle, for example, binds exclusively to glycans terminating with a Neu5Aca2-3Gal grouping. (Reprinted with permission from O. Blixt et al.,
Cells are coated with carbohydrates, as are many of the proteins and lipids contained within cells. Scientists increasingly are coating microarrays with carbs, too. More than just frosting, they hope that these carbohydrate arrays can advance research into the vital roles sugars play in cancer, infection, and cell interactions such as fertilization.
"Carbohydrates play these very important roles in biology, but our understanding of [them] ... lags very far behind the understanding of proteins and nucleic acids. There's a big opportunity here, now that the tools to characterize carbohydrates are becoming available," says Milan Mrksich of the University of Chicago.
Arrays of sugars bound to a solid support, called carbohydrate chips, screen a given sample – antibodies, lectins, nucleic acids, viruses, cells, and so on – for its carbohydrate-binding specificities, replacing labor-intensive binding assays performed in test tubes. What they don't do is identify specific sugars in a sample; that's the job of lectin arrays.
As with other areas of carbohydrate research, array development is in its infancy when compared with DNA and even protein arrays. The first carbohydrate arrays appeared just three years ago, and at the moment, they are not available commercially. But the technology and the expertise to develop custom arrays is growing, and with it, the list of possible applications.
"The excitement over carbohydrate arrays is due to the many, many findings of diseases caused by a lack or overabundance of certain carbohydrates displayed on cells, for instance, as hooks for viruses and bacteria to bind to," Mrksich says. "We're on the verge in the next 18 months or so to see a dozen or so more papers that start to make real advances with carbohydrate arrays, to discover unanticipated activities, not just validate things we already know."
AN EXCITING YEAR
The past year saw several research groups demonstrating potential applications for carbohydrate arrays. Peter Seeberger of the Swiss Federal Institute of Technology in Zurich and colleagues used arrays to analyze the binding profiles between five HIV surface proteins and gp120 glycans.1 "We identified four lead structures for HIV drug vaccines now in animal tests," he says. Denong Wang of Stanford University and Jiahai Lu of Sun Yatsen University in China developed an array of 51 carbohydrate antigens to analyze SARS coronavirus for autoimmunogenicity.2
Seeberger's group has also used carbohydrate arrays containing aminoglycosides to identify sugars that can bind RNA. Such glycans could potentially inhibit protein synthesis, making for potent antibiotics, if they can be found to specifically target bacterial, rather than human RNA, says Seeberger.
Pathogen detection is yet another potential application.3 In theory, specific bacterial strains can be typed based on their unique carbohydrate interactions. Using as a proof-of-principle device an array with five monosaccharide derivatives (glucose, galactose, mannose, N-acetyl-D-glucosamine, and fucose), Seeberger's group was able to observe specific binding of
"Their array was able to identify strains of bacteria very quickly and specifically," says Lara Mahal of the University of Texas in Austin. Aside from obvious biodefense and food-safety applications, these arrays also have potential clinical applications. Bacteria can be typed in minutes, as opposed to the hours required with conventional methods.
Carbohydrate arrays could play a role in cancer diagnostics. At last November's joint meeting of the Society for Glycobiology and the Japanese Society of Carbohydrate Research, Ola Blixt of the Scripps Research Institute in La Jolla, Calif., reported that an initial carbohydrate-array survey of 10 patients with cancer from Houston's M.D. Anderson Cancer Center "revealed sugars that lit up with cancer sera that did not light up with normal sera," says Jim Paulson, head of the Consortium for Functional Glycomics and a professor at Scripps.
MORE SUGARS, PLEASE
At the moment, researchers interested in doing their own glycomics research must be self-reliant, as commercial carbohydrate arrays do not yet exist. Some companies, such as Glycominds in Israel, develop carb arrays for in-house use, but researchers interested in the technology generally must make their own arrays. Recently, Blixt, Paulson, and colleagues at the Consortium of Functional Genomics developed an alternative and freely available array.4
Created using standard robotic printing technology, the chip bears some 200 different natural and synthetic sugars. Blixt's team used it to profile the binding specificities of several different plant and human lectins, antibodies, and pathogen proteins. Yet despite its utility, this particular array contains but a tiny fraction of the body's carbohydrate complement.
"The glycome is all the sugars made by an organism, and we know that's a limited set, but the human glycome alone over the course of a lifetime is certainly greater than 20,000, though probably less than 100,000. We're at 200 right now," Paulson says.
This limitation stems from sugars being considerably harder to synthesize and characterize than proteins or nucleic acids. No reliable and general methods exist for determining the composition or ordering of carbohydrates, and only recently have reliable methods to generate glycans of defined sequence and structure emerged. Moreover, synthesis of certain glycan classes is not always possible due to limitations in chemical and enzymatic synthesis technology.
The Consortium for Functional Glycomics doesn't even deal with one particular class of sugars. Abandoning proteoglycans (polysaccharides such as heparin, which are hundreds of sugars in length), the consortium focuses instead on glycoproteins and glycolipids, the carbohydrate forms likely recognized by most glycan-binding proteins. These typically involve branched structures of up to 17 or so sugars, which have an estimated 500 different terminal structures.
Paulson explains the challenge facing carb chemists: If you take a sugar with, say, four branches, "and you take one terminal structure and put it on one branch, that's a different sugar than if you had a sugar with that terminal structure on a different branch. With permutations you multiply up to many thousands of sugars." The answer, he says, is to populate arrays with increasing numbers of complex branched structures.
"We can probably expand to 1,000 or even 2,000 sugars in a few years," Paulson says. An attractive option is to match automated carbohydrate synthesis with array printing, "so arrays can rapidly expand. That hasn't been thoroughly explored yet," he adds.
Another challenge involves taking into account the close packing of carbohydrates on cell surfaces. "It's rare in cells to find a single protein binding to a single carbohydrate. The binding constant can be very poor, but if you put 500 of a carbohydrate on a surface, it could be extremely high. Binding is almost always multivalent," says Ajit Varki of the University of California, San Diego.
"If a person makes an array and the ligand density is high enough, that person might always get tight binding, but if another person has the same content but the density is lower, and you're buying arrays from both, you could find different results," Mrksich says. It will prove important to collect data for binding to carbohydrate arrays with different glycan densities and match up the data with cell biology, Paulson says.
Mrksich says carbohydrate arrays might eventually be read using mass spectrometry instead of via labels. "When you use labels, whether fluorescent tags or radioisotopes or antibodies, you add a lot of steps to the assay, making it a lot more complicated, and you bias the assay towards what the labels look for and might miss a lot of other interactions," Mrksich says.
Mrksich's lab has developed self-assembled alkanethiolate monolayers on gold as attachment surfaces compatible with mass spectrometry and is collaborating with Seeberger to develop advanced carbohydrate arrays. "I do see mass spectrometry becoming the standard for carbohydrate arrays," Mrksich says. "To do the actual reading of the chip, mass spectrometry is slower by up to a factor of 10. But if you look at the time for the entire experiment, the throughput is faster."
Given, say, an array of 100 sugars and a desire to know what specific modifications an enzyme makes on a given sugar, such as adding a sulfate or another sugar to build a longer oligosaccharide, with a label-based array, "I only know that some sugars are modified and others aren't," Mrksich explains. Using his arrays, however, "I can take a mass spectrum of each spot, and based on mass change have a pretty good idea of what the modification was within an hour. With fluorescent labels you can see the same DNA-binding activity you always see with DNA, but mass spectrometry can be relevant for the variety of activities you see with carbohydrates."
For the general researcher, however, these developments will be of little value unless commercial arrays become available. "I think in another two years there will be commercial sources for both carbohydrate content and chip assembly," Mrksich says. "Then it will take a good three to five years in the marketplace to evaluate and improve carbohydrate chips for them to become broadly available and reliable, the same as what happened with gene chips." In the meantime, researchers unable to make their own arrays can use the Consortium's Glycan Array Screening service