Advertisement

Reflections on Glycobiology*1

  • Saul Roseman
    Affiliations
    From the Department of Biology and the McCollum-Pratt Institute, The Johns Hopkins University, Baltimore, Maryland 21218
    Search for articles by this author
  • Author Footnotes
    * This work was supported by Grant GM51215 from the National Institutes of Health.
    1 The original title of this article was: “Glycobiology: Past, Present, and a Very Bright Future.” It was intended to show something of the development of major concepts and to recognize the excellent contributions of pioneers in the field. I had no problem until the modern era, when I realized that it would take volumes to adequately describe the diversity of glycans and to cover, however briefly, current work and investigators. So, with regret, I am limited to the past, where I tried to capture some of the flavor of the field, the origins of some contemporary ideas, and how they may tie to the future. Insofar as the chemistry is concerned, I have chosen to emphasize the cell surface and extracellular matrix because these are where most of the glycoconjugates are found. The focus is on two examples, the cartilage aggrecan aggregate because it illustrates the enormous complexity that is possible with glycans and the erythrocyte surface and the ABO blood groups, which in some respects, at least, may be a model for other cell surfaces.
      GAG
      glycosaminoglycan
      PAPS
      3′-phosphoadenosine-5′-phosphosulfate
      MGT
      multiglycosyltransferase
      GT
      glycosyltransferase
      ST
      sialyltransferase
      Glycobiology has become a “hot” subject,
      A recent issue of Science features glycobiology (
      ), and the April, 2001 meeting of the Carbohydrate Division of the American Chemical Society emphasizes glycobiology as a major subject; their prestigious C. S. Hudson Award was presented to a well known glycobiologist, Y. C. Lee. How times have changed! In the 1950s, glycobiology was not a popular subject. There were a few interested biochemists at the meetings, and we had an annual lunch (Karl Meyer, Al Dorfman, Dick Winzler, Roger Jeanloz, Ward Pigman and a few others). After lunch, one might as well go home. My papers (glucosamine metabolism) were invariably scheduled as either last or next to last on Friday afternoon at the Federation Meetings in Atlantic City. The most hilarious incident was when my paper (next to last) was announced at one of these sessions. When I reached the platform, the chairman of the session apologized because he had to leave to make a train. My audience consisted of the next speaker and the slide projectionist. I stayed for the last paper, but unfortunately I never asked the projectionist how he liked the presentations. I had the same experience at the American Chemical Society meetings. Starch chemistry was a big thing for members of the Carbohydrate Division, and they walked out on papers devoted to the glycoconjugates or hexosamines. At one such meeting, my audience consisted of other members of the laboratory waiting to drive back to Ann Arbor with me. It was, however, a great time to do this kind of research. There was virtually no pressure. The handful of us in this country who worked in the field were supported by the National Institutes of Health. I can capture a little of the intellectual flavor of the times by my experience when I submitted my first independent application. It stated that I would work on the enzymatic synthesis of one of the monosaccharides in the glycoconjugates, but I did not know which to choose. I then listed about four monosaccharides (glucosamine, fucose, glucuronic acid, and galactosamine) and possible preliminary experiments for each; I would work on whichever problem appeared most fruitful. I was funded, and a short time later I met one of the members of the Study Section (Ef Racker) who told me that it was the best application he had read. What would happen today with an application that was so “unfocused” and with such nonspecific aims? Equally important to the National Institutes of Health support, we received unsparing help from a number of farsighted physicians such as Walter Bauer at Massachusetts General Hospital, who not only created a high caliber research unit (Roger Jeanloz, Jerome Gross, Karl Schmid, Morris Soodak, and others) but was also instrumental in the Helen Hay Whitney Foundation. In my case, it was William Robinson and Ivan Duff at the Rackham Arthritis Unit at the University of Michigan. Only once (when I was first interviewed) did I have to explain to Bill Robinson how work on Escherichia coli might relate to arthritis. Thus, we had the luxury of following our noses and serendipity wherever the work took us. We started with studies on the intermediary metabolism of glucosamine, which led in turn to the structure of the sialic acids and the discovery of N-acetylmannosamine, to the intermediary metabolism of these compounds, to CMP-sialic acid and its enzymatic synthesis, to the glycosyltransferases, and finally to the phosphotransferase system for sugar uptake by bacteria (reviewed in Refs.
      • Roseman S.
      and
      • Roseman S.
      ). In recent years, the complex process of chitin catabolism by marine bacteria has become a major project (
      • Keyhani N.O.
      • Roseman S.
      ).
      2A recent issue of Science features glycobiology (
      ), and the April, 2001 meeting of the Carbohydrate Division of the American Chemical Society emphasizes glycobiology as a major subject; their prestigious C. S. Hudson Award was presented to a well known glycobiologist, Y. C. Lee. How times have changed! In the 1950s, glycobiology was not a popular subject. There were a few interested biochemists at the meetings, and we had an annual lunch (Karl Meyer, Al Dorfman, Dick Winzler, Roger Jeanloz, Ward Pigman and a few others). After lunch, one might as well go home. My papers (glucosamine metabolism) were invariably scheduled as either last or next to last on Friday afternoon at the Federation Meetings in Atlantic City. The most hilarious incident was when my paper (next to last) was announced at one of these sessions. When I reached the platform, the chairman of the session apologized because he had to leave to make a train. My audience consisted of the next speaker and the slide projectionist. I stayed for the last paper, but unfortunately I never asked the projectionist how he liked the presentations. I had the same experience at the American Chemical Society meetings. Starch chemistry was a big thing for members of the Carbohydrate Division, and they walked out on papers devoted to the glycoconjugates or hexosamines. At one such meeting, my audience consisted of other members of the laboratory waiting to drive back to Ann Arbor with me. It was, however, a great time to do this kind of research. There was virtually no pressure. The handful of us in this country who worked in the field were supported by the National Institutes of Health. I can capture a little of the intellectual flavor of the times by my experience when I submitted my first independent application. It stated that I would work on the enzymatic synthesis of one of the monosaccharides in the glycoconjugates, but I did not know which to choose. I then listed about four monosaccharides (glucosamine, fucose, glucuronic acid, and galactosamine) and possible preliminary experiments for each; I would work on whichever problem appeared most fruitful. I was funded, and a short time later I met one of the members of the Study Section (Ef Racker) who told me that it was the best application he had read. What would happen today with an application that was so “unfocused” and with such nonspecific aims? Equally important to the National Institutes of Health support, we received unsparing help from a number of farsighted physicians such as Walter Bauer at Massachusetts General Hospital, who not only created a high caliber research unit (Roger Jeanloz, Jerome Gross, Karl Schmid, Morris Soodak, and others) but was also instrumental in the Helen Hay Whitney Foundation. In my case, it was William Robinson and Ivan Duff at the Rackham Arthritis Unit at the University of Michigan. Only once (when I was first interviewed) did I have to explain to Bill Robinson how work on Escherichia coli might relate to arthritis. Thus, we had the luxury of following our noses and serendipity wherever the work took us. We started with studies on the intermediary metabolism of glucosamine, which led in turn to the structure of the sialic acids and the discovery of N-acetylmannosamine, to the intermediary metabolism of these compounds, to CMP-sialic acid and its enzymatic synthesis, to the glycosyltransferases, and finally to the phosphotransferase system for sugar uptake by bacteria (reviewed in Refs.
      • Roseman S.
      and
      • Roseman S.
      ). In recent years, the complex process of chitin catabolism by marine bacteria has become a major project (
      • Keyhani N.O.
      • Roseman S.
      ).
      a timely one for “Reflections.” The primary reason, I think, is illustrated in Fig.1, which shows the surface of an erythrocyte in cross-section. Just outside the plasma membrane of this and nearly all cells is a coat of fuzzy material called the glycocalyx, consisting of a myriad of carbohydrate-rich molecules, polysaccharides, proteoglycans, glycoproteins, and glycolipids. If the cell shown here was a fibroblast or an intestinal epithelial cell that secretes polysaccharides or mucins, it would be difficult to determine the location of the cell boundary; the polymers begin on the cytoplasmic face of the lipid bilayer, within it, or on its periphery, but it is not clear where they end. These extensive, complex structures must serve essential roles in cell surface phenomena, but we are only beginning to understand what some of these functions are. I believe that the glycoconjugates or glycans can serve as important informational macromolecules.
      Figure thumbnail gr1
      Figure 1The erythrocyte glycocalyx. This is revealed in electron microscopy by special staining methods. It is up to 1400 Å thick, and the oligosaccharide filaments are 12–25 Å in diameter. (Taken from Voet and Voet, Biochemistry, with permission of the publisher. Original was by courtesy of Harrison Latta, UCLA).
      In this remarkable age of genomics, proteomics, and functional proteomics, I am often asked by my colleagues why glycobiology has apparently lagged so far behind the other fields. The simple answer is that glycoconjugates are much more complex, variegated, and difficult to study than proteins or nucleic acids. To understand where we are and to appreciate what it has taken to get here requires some background, so this article will briefly survey the history of glycobiology from early studies on fermentation to the beginning of the contemporary era.

      The Past

      Although glycobiology antedates biochemistry by many millenniums, their histories are inextricably linked. The principal foundations of both fields lie in the development of organic chemistry during the 19th century and in studies on the process of fermentation of glucose and sucrose.

       Fermentation

      Fermentation was known to the cave man and has been the subject of intense study ever since. The Old Testament has many references to wine and libations, the first being to Noah (Genesis, 9:20–21): “Noah, the husbandman, began and planted a vineyard. And he drank of the wine and was drunk.”
      This reference was kindly called to my attention by Dr. Michael Edidin.
      Treatises and philosophical discourses were published on the process during and after the Middle Ages.
      One that struck a chord was a 74-page treatise by John Richardson (1790) entitled “Theoretic Hints on an Improved Practice of Brewing Malt-liquors … ”. He defines fermentation as: “A spontaneous internal motion of constituent parts, which occasions a spontaneous separation and removal from their former order of combination, and a remarkable alteration in the subject, by a new arrangement and re-union.” Not a bad definition of intermediary metabolism and the thermodynamics of glycolysis.
      Fermentation was not confined to making alcohol but has been used for thousands of years to make cheese, soy sauce, etc.
      The first chapter of the biochemical classic Alcoholic Fermentation by Arthur Harden (1st edition, 1911) reviews the history. (a) The most important early study was that of Lavoisier (1789) who quantitatively established the stoichiometry of the process and concluded that the sugar was split into two parts, one of which was oxidized (carbonic acid) at the expense of the other (alcohol). Furthermore, “if it were possible to recombine these two substances, sugar would result.” The methodology was insufficient to permit him to see an increase in the weight of the yeast or of other products that were formed. (b) Yeast at the time was regarded as a catalyst, much like alumina or diatomaceous earth, and during the first 60 years of the 19th century, this was the prevailing view of the leading chemists and journal editors of the time (Liebig, Berzelius, Wohler). This was despite the fact that in 1837 three independent investigators, Cagniard-Latour, Schwann, and Kutzing, presented evidence that yeast was a living organism, an idea that was ridiculed by the establishment. (c) In 1860, the pivotal experiments of Louis Pasteur finally laid this ghost to rest (
      • Conant J.B.
      ), and he showed unequivocally that yeast was a living organism. He also did careful stoichiometry. The balance sheet showed that about 95% of the C,H,O of the sugar was converted to CO2 and ethanol. The remainder, from 1.6 to 5%, consisted of substances that the “yeast had taken from the sugar.” The result of this and subsequent work by Pasteur led to his famous dictum, “no fermentation without life.” In an extension of this work, he came to the conclusion (1875) that fermentation was the result of life without oxygen, the cells being able under anaerobic conditions to avail themselves of the energy liberated by the decomposition of substances containing combined oxygen (i.e. anaerobic glycolysis). (d) Enzymes (called ferments), generally hydrolases, were known during the 19th century; indeed, invertase (i.e. sucrase) had been extracted from yeast. In 1858, Traube proposed that fermentation resulted from the action of ferments secreted by cells on sugar. Many attempts were made to extract yeast cells and obtain cell-free fermentation of sugar but without success. Finally, while attempting to preserve yeast extracts for therapeutic purposes, Eduard Buchner succeeded in 1897. The preservative was sugar, and he noted that carbon dioxide was formed. This fortunate and serendipitous result marks the beginning of biochemistry as we know it today. It is interesting to note that the Journal of Biological Chemistry was founded only a few years later by Christian Herter.
      The story would not be complete without summarizing what was learned between Buchner’s landmark result in 1897 and the publication of Harden’s monograph in 1911. Kinetic experiments were conducted using yeast extracts and glucose, and the rate of fermentation was followed by measuring the rate of CO2 evolution. The following results, especially by Harden and Young, were obtained. (i) Fructose and mannose were fermented as well as glucose, but the yeast had to be “trained” (i.e. adapted) for the extract to ferment galactose. They speculated that different ferments were required for galactose utilization. (ii) Inorganic phosphate was required. (iii) A hexose diphosphate was isolated, characterized as fructose-di-P, and was shown to be an intermediate in the process. (iv) The extract was pressure filtered through a gelatin film, giving a dialysate and a “residue.” Neither alone supported fermentation, but it was restored by mixing the two. The residue contained the heat-labile zymase, and the dialysate contained the heat-stable coenzyme(s) or cozymase. Soon after, it was shown that yeast anaerobic glycolysis was closely connected to anaerobic glycolysis by muscle and muscle extracts. The cozymase, of course, was the source of ATP, NAD+, etc.

       Development of Organic and Carbohydrate Chemistry

      The close connection between the development of organic chemistry and biochemistry in the 19th century is summarized in an exemplary, early textbook (
      • Fruton J.S.
      • Simmonds S.
      ). However, carbohydrate history goes back many centuries earlier. Cellulose in the form of cotton, for instance, was known from ancient times, and sucrose was one of the first organic substances to be crystallized (300 A.D., from the juice of sugar cane in India). Because the climate in Europe was not favorable for growing sugar cane, alternative sweetening agents were sought early in the 19th century, leading to the discovery of new sugars (glucose, fructose, mannose, galactose, etc.), all with the same elementary composition (CH2O)n. Clarifying the structural relationships between these compounds occupied carbohydrate chemists for most of the century. Finally, the structure of d-glucose was established by Emil Fischer in 1891, which marks the beginning of modern carbohydrate chemistry. Fischer's multitudinous and brilliant contributions were likewise in the fields of amino acid and purine/pyrimidine chemistry. It is worth reminding the reader that chromatography and electrophoresis were unknown at the time, and substances were purified by fractional crystallization and characterized by elemental analyses and their physical properties (melting point, optical rotation, solubility, etc.).
      My interest in carbohydrate chemistry began as a graduate student working in the laboratory of Karl Paul Link at the University of Wisconsin. He was both a carbohydrate and natural products chemist, with very high standards and an ability to inspire the best in us. The laboratory had isolated and characterized dicumarol as the hemorrhagic factor in spoiled sweet clover hay prior to my arrival (warfarin is a synthetic analogue). My project was to study the metabolism of its parent compound, 4-OH-coumarin, which was not toxic. Four large dogs used as subjects were fed the drug and maintained in very large metabolic cages so that their urine could be collected. (In those days, graduate students took complete and very good care of their animals, including feeding, exercising them, and cleaning their cages.) The metabolic product turned out to be 4-OH-coumarin β-d-glucuronide. However, this had to be established by synthesis and also by elemental analysis. I spent a very muggy, frustrating summer in Madison recording the swings on a microbalance and learning how to do microanalyses before the standards finally came out right. Somewhat later, I developed considerable experience with fractional crystallization, particularly of anomeric glycosides. They were being synthesized for Joshua Lederberg, a young faculty member in a neighboring department (genetics), who was using them for assaying the expression of glycosidases, such as β-galactosidase in E. coli. Fractional crystallization, like elemental analysis, is tedious work, but above all it is a real art and when it works, it is most gratifying. In doing this kind of work, we even invoked the help of the Lord. To this day my children remember that my wife Martha (who is not a scientist) concluded the evening prayers over the Sabbath candles with the following phrase: “and may Daddy have crystals.” It worked!
      In this age of electronics and the internet, one always thinks that science moves forward too slowly, but it is mind boggling to realize how far we have come since the 1890s (Fischer, Buchner).

      Glycobiology in the 20th Century: Chemistry

       Structural Studies on Glycosaminoglycans (Mucopolysaccharides)

      Although mucins from various sources were studied by organic chemists as early as 1846 (see reviews by Blix, Gottschalk, and Morgan (
      • Gottschalk A.
      )) and were thought to contain sugars, there was always an unresolved question of purity. In 1925, the distinguished chemist, P. A. Levene, who had made fundamental contributions to the structures of the nucleic acids, published a monograph entitled “Hexosamines and Mucoproteins.” Chondroitin sulfate had been isolated in 1884 from cartilage, but the nature of its monosaccharides and structure were controversial until Levene showed conclusively that the constituents were d-glucuronic acid, chondrosamine (d-galactosamine), acetic acid, and sulfuric acid in equimolar ratios. He depicted the structure as GalNAc linked to GlcUA and sulfated at C-6 on the GalNAc. As might be expected from the available methodology and misinformation on sugar ring structures, there were major errors in the structural assignment, including the fact that it was a tetrasaccharide. Similarly, mucoitin sulfate (i.e. hyaluronic acid) was depicted as a tetrasaccharide containing GlcNAc but also sulfate. He also questioned whether substances such as ovalbumin were “glucosidoproteins” or whether such substances even existed.
      In 1934 (
      • Meyer K.
      • Palmer J.W.
      ), hyaluronic acid was isolated in pure form from vitreous humor, and its correct composition was determined. This groundbreaking paper was the first of many from Karl Meyer's laboratory, creating a science from chaos. His laboratory subsequently isolated and characterized the chondroitin sulfates, keratan sulfate, and various hyaluronidases.
      Karl Meyer was a delightful person with a keen sense of humor. His exchanges with Albert Dorfman at the meetings were the highlight for many of us. For instance, at one meeting Al gave a talk, and in the questioning period Karl asked Al, “How did you quantitate the keratosulfate?” Al responded that he had not. In a stage whisper, Karl said: “I thought as much.” Al, with whom I did my postdoctoral work, was a principal figure in the field. He held both M.D. and Ph.D. degrees but what made him really unusual was his expertise in both fundamental biochemical research and in clinical practice (pediatrics). He was a leader in the University of Chicago Medical School and later became Chair of Pediatrics. Al came around to see me every day, and we would get into the most vigorous discussions on how to interpret results, the next experiments, etc. He had to be the most tolerant person, considering that I was fresh out of graduate school and was convinced that I knew everything there was to know (it has been downhill ever since). My paying job was to direct the pediatric blood chemistry laboratory, which was actually very interesting because one had to develop ultramicroanalytical methods, especially for samples from the newborn, which were often obtained by heel puncture. Most of my research was conducted late in the afternoon and evening. Al lived across the Midway and could see the laboratory window (top floor of Bobs Roberts Hospital) from his bedroom. I always left the lights on when I went home.
      Establishing the structures of heteropolysaccharides can be exceptionally difficult, and the problems can be summarized as follows: (i) identification and quantitation of the monosaccharides; (ii)d- or l-configurations; (iii) branched or unbranched; (iv) sequence; (v) α or β anomers; (vi) pyranose or furanose rings; (vii) positions of the linkages; (viii) many of these polymers are derivatized (e.g. phosphate, sulfate, acetate, etc.), and polymers with different biological and chemical properties are formed, depending on the position of the linkage in the derivative; and (ix) to complicate matters even further, some of the polymers and oligosaccharides are covalently linked to proteins or lipids.
      One of the major problems confronting workers in this field was protein and how to get rid of it because it was regarded as a contaminant of the “mucopolysaccharides,” now called glycosaminoglycans or GAGs.7Protein was not easily removed.
      At the University of Chicago we were fortunate to have the large meat packing houses close by, which were sources of necessary tissues, such as bovine eyes (for vitreous humor), testis (for hyaluronidase), etc. The isolation of chondroitin sulfate started with bovine nasal septa, which were obtained by working on the line and cutting them out of the skulls as they came by on a belt (very hard on the hands). The cartilage was ground and extracted with about 0.1n NaOH for several days in the cold with constant stirring. The alkaline extract was then deproteinized and the polysaccharide isolated. By hindsight we know now that the alkaline extraction procedure split the polysaccharide from its O-serine (or threonine) linkage in the protein by β-elimination.
      Meyer, for instance, thought that the protein formed ionic bonds with the polysaccharides. In the 1950s, Maxwell Schubert's laboratory showed that cartilage chondroitin sulfate was linked to protein, thus opening a new chapter in the chemistry of these polymers, now called proteoglycans. The next essential step was to characterize the linkage region between the GAG and the protein. Work on different polymers around the same time (late 1950s) by Pigman (mucins), Kabat (blood group substances), and Muir (chondroitin sulfate) suggested that the sugars were linked to serine.
      In the alkaline β-elimination step, the oligo- or polysaccharides glycosidically linked to serine or threonine are first released from the protein and then degraded by the alkali at the reducing end of the chain, a reaction called “peeling.” An important advance in the field was Carlson’s alkaline borohydride procedure, which reduced the aldehyde group as the glycosidic bond was cleaved and protected the oligomer from alkaline degradation (
      • Carlson D.M.
      ).
      In 1964, Lindahl and Roden found that the “linkage fragment” in heparin was O-β-d-xylopyranosyl-l-serine (reviewed in Ref.
      • Roden L.
      ). They later showed that the sequence at the linkage region in these polymers (chondroitin sulfates, dermatan sulfate, and heparan sulfate) to which the polysaccharide is attached is GlcUA-Gal-Gal-Xyl-Ser. In skeletal keratan sulfate, the O-linkage is to α-GalNAc in place of the Xyl.
      At the same time, a different class of complex carbohydrates, now call glycoproteins, was the subject of intensive study. Neuberger’s laboratory in England showed by isolation and synthesis that the linkage region in ovalbumin is β-GlcNAc→Asn,i.e. to the amide N of asparagine. There are, of course, a wide variety of N-linked glycoproteins, particularly the glycoproteins in serum. Since the overriding question in these early studies was purity, the isolation and characterization of the major serum glycoprotein, α1-acid glycoprotein (orosomucoid), by Karl Schmid was a key breakthrough. The protein (44 kDa) contained 17% hexose and 12% hexosamine.
      A characteristic of carbohydrate polymers is that they are polydisperse or microheterogeneous. The template mechanisms of protein and nucleic acid synthesis do not apply to the carbohydrate polymers, thereby resulting in polydispersity. Human orosomucoid, for instance, contains 6 oligosaccharide chains per molecule, but the chains are different from each other. In the collection of molecules called orosomucoid, at least 8 oligosaccharides have been identified (
      • Yoshima H.
      • Matsumoto A.
      • Mizuochi T.
      • Kawasaki T.
      • Kobata A.
      ). Each oligosaccharide can contain up to 5 different kinds of sugars, a given sugar can occur several times in the chain, and the number of possible combinations is overwhelming (see below).

       Aggrecan Aggregate

      The major components of cartilage are collagen and a huge macromolecular complex called the aggrecan aggregate. An electron micrograph of one such aggregate is shown in Fig. 2 A, and Fig.2 B presents a schematic view of 6 aggrecan monomers bound to hyaluronan. Determining the details of these structures is an extraordinary achievement in this field, equivalent (at least) to delineating the structure of collagen. The structure was developed through work in the laboratories of Hascall, Muir, and Heinegard and has recently been reviewed (
      • Hascall V.C.
      ,
      • Wight T.N.
      • Heinegard D.K.
      • Hascall V.C.
      ). This unusually complex “molecule” can have an apparent mass of >6 × 109Da and is a composite of all of the structural units described above. The relationship between the structure of the aggregate and its function is briefly discussed below.
      Figure thumbnail gr2
      Figure 2Electron micrograph of a proteoglycan aggregate purified from calf epiphyseal cartilage. A, the aggregate was spread as a monolayer in a cytochrome cfilm. Under these conditions the chondroitin sulfate chains of the proteoglycan collapse onto the core protein so that each monomer (aggrecan) of the aggregate is distinct. The nearly uniform length of the monomers is characteristic of proteoglycans from young cartilages, with each ∼3.5 million Da in molecular mass. This aggregate contains ∼180 aggrecan monomers, and the overall molecular mass of the aggregate is ∼6.5 billion Da. The central strand of hyaluronan is ∼5500 nm in length. The boxed area encloses 6 monomers, the number depicted in the model in B. (Micrograph kindly provided by Joseph Buckwalter, University of Iowa.)B, model of a portion of the proteoglycan aggregate showing 6 aggrecan monomers (see box in A). Each of the six monomers is depicted with a central core protein strand to which the glycosaminoglycans are covalently linked, giving the appearance of bristles. The core protein (∼250,000 Da) contains a midregion with ∼100 chondroitin sulfate chains (blue) of ∼30,000 average molecular weight and a nearly equal number of keratan sulfate chains (red) of ∼10,000 average molecular weight. The monomers are anchored (non-covalently) to the central hyaluronan strand (orange) by: (a) a hyaluronan-binding site in the N-terminal globular-1 (Gl) domain (pink sphere) of the core protein, and (b) a link protein (crystal sphere) that binds to both hyaluronan and to the G1 domain of aggrecan. The core protein of aggrecan contains two other globular domains, G2 and the C-terminal G3, shown as green and lavender spheres, respectively. The functions of G2 and G3 are not known. The globular domains of aggrecan also contain N-linked oligosaccharides (red Y symbols). The model is fully extended as it would appear in dilute solution. In the tissue the structure would be compressed to approximately one-tenth its extended size. See Ref.
      • Wight T.N.
      • Heinegard D.K.
      • Hascall V.C.
      for additional details. (Model and supplemental three-dimensional, rotational view kindly provided by Mark Sabo (Art Department) and Vincent Hascall, Cleveland Clinic Foundation.)

       The Erythrocyte Surface, Human Blood Group Activity, and Erythroglycan (Poly-N-acetyllactosamine)

      The frequent incompatibility of the blood of a donor and recipient was recognized in the 17th century. Starting with the work of Landsteiner (1900), who defined the ABO group, we now know that there are at least 27 such families of human blood group substances expressed on the surfaces of erythroid cells and often other cells as well. The general characteristic of these antigens is that they comprise integral membrane glycoproteins, both O- and N-linked, and in some cases, glycolipid. Thus far, it has been shown that the glycan units are the epitopes in four of the systems, ABO, Lewis, P, and H/h.
      I am very grateful to Dr. Olga Blumenfeld (Department of Biochemistry, Albert Einstein Medical School) for helpful discussions on the blood group substances.
      Some aspects of the ABO system will be discussed here.
      Work on the ABO family was greatly aided by finding these activities in water-soluble form in various secretions and mucins, such as ovarian cysts. The major antigenic determinants were established by Morgan and his co-workers (particularly Watkins and Aminoff) and by Kabat and his co-workers (reviewed in Ref.
      • Kabat E.A.
      ). These determinants were sugars at the non-reducing termini of oligosaccharide chains linked via Ser and Thr to polypeptides, similar to the mucins. Blood group O chains were terminated by a trisaccharide Gal(β,1–4)[Fuc-(α,1–2)]GlcNAc–X. Blood group A activity was expressed by linking an α-GalNAc to C-3 of the Gal, whereas in B activity a Gal is substituted for the GalNAc.
      The erythrocyte membrane was quite another problem. Although Yamakawa showed that red blood cell glycolipids exhibited such activity (1953), this conclusion was disputed as late as 1956 (
      • Kabat E.A.
      ). It is now clear that the antigens are carried on the erythroid surface by both lipids and polypeptides (see review by Hakomori (
      • Hakomori S.-I.
      )).
      These structures are closely related to the glycosaminoglycan keratan (desulfated keratan sulfate). The repeating unit in this GAG is N-acetyllactosamine: Gal-(β,1–4)-GlcNAc-(β,1–3) linked to the next Gal in the chain. The same structural unit but in shorter chains than the polysaccharide, called poly-N-acetyllactosamine or polylactosamine, is found both O- and N-linked to integral membrane proteins on many cell surfaces and is also found linked to ceramide. Polylactosamine can be straight chain or branched and can be “decorated” with Fuc or sialic acid residues. Apparently the first references to Gal-, GlcNAc-, and Fuc-rich glycopeptides in cell membranes came from work by the eminent geneticist/molecular biologist Francois Jacob and his group on the cell surface antigens found in early embryonic differentiation (reviewed in Ref.
      • Jacob F.
      ). At about the same time (
      • Järnefelt J.
      • Rush J.
      • Li Y.-T.
      • Laine R.A.
      ,
      • Laine R.A.
      • Rush J.S.
      ), Laine and co-workers isolated “erythroglycan” by extensive Pronase digestion of lipid-free red blood cell stroma and characterized the large branched oligosaccharides (7,000–10,000 Da) by methylation, etc. “Band 3,” the major red blood cell integral membrane protein and the anion transporter, is the source of the polylactosamine, and it accounts for more than 30% of the total Gal and GlcNAc in the red blood cell membrane. Further, at its non-reducing termini the polymer can carry Fuc and αGal or αGalNAc, thereby becoming an antigenic determinant for A, B, or O activity. The large quantity of polylactosamine peptide derived from the red blood cell membrane corresponds to most of the antigenic sites in the intact erythrocyte (about 2 × 106). There is now an extensive literature on polylactosamine, its enzymatic synthesis (
      • Renkonen O.
      ), and how branching occurs during development, tumorigenesis, etc.
      Blood group activity is also carried by glycolipids, which are present in small quantities in the red blood cell membrane (reviewed by Hakomori (
      • Hakomori S.-I.
      ). They consist of a large number of compounds derived from N-acetyllactosamine oligomers. This family comprises oligosaccharides, both straight and branched chain, linked to glucosylceramide and terminated by one of the antigenic determinant sugars. The glycolipids change, especially with respect to branching, during the development of erythroid cells.

      Glycobiology in the 20th Century: Biosynthesis

       Isotope Experiments

      The complex carbohydrates contain up to 8 different monosaccharides, including d-xylose, hexoses, hexosamines, and hexuronic acids, in addition to various sialic acids, such N-acetylneuraminic acid (NAN or NeuAc). Until 1950, we did not know how most of these monosaccharides were biosynthesized. For the 6-carbon sugars, the theories ranged from (a) direct conversion of the 6-carbon skeleton ofd-glucose to the sugar to (b) fragmentation of glucose through glycolysis and other catabolic cycles and recombination of suitable fragments. It was suggested, for example, that the GlcNH2-6-P carbon skeleton was formed by condensing glyceraldehyde-3-P (G3P) and serine, with subsequent reduction of the carboxyl to the aldehyde. And how could l-fucose possibly arise directly from d-glucose without inversion of the carbon skeleton by 180o, which would give thel- from a d-sugar?
      These problems were addressed by treating an appropriate biological system with specifically labeled glucose, such as 1-[14C]- and 6-[14C]glucose in companion experiments. The pure polymer was isolated and hydrolyzed, and the monosaccharides were isolated and dissected carbon by carbon to determine the specific activity at each C-atom in the skeleton.
      My major postdoctoral project was to determine the modes of synthesis of the glucosamine and glucuronic acid moieties in hyaluronic acid. The biological system was a strain of Group A streptococcus that secreted the polysaccharide, and the organism was grown (in a rich medium) on 1-[14C]- and 6-[14C]glucose. One of the many problems was the cost of the labeled sugars (far too expensive for these experiments). [14C]NaCN was more reasonable, and the labeled sugars were synthesized from this starting material. In the experiments, because of the rich medium, the labeled glucose, acetate, and lactate were isolated from the medium, as well as the polysaccharide, and were dissected as well. Konrad Bloch, who was a Professor in the Department of Biochemistry, was of enormous help to me during this phase of the work.
      If the origin of the 6-carbon hexoses was via fragmentation of the Glc 6-carbon chain, followed by recombination, isotope scrambling would result.
      For instance, at the triose-P level, because of triose-P isomerase 1-[14C]glucose would become 1,6-[14C]hexose, and the specific activity at C-1 would be half that of the 1-[14C]glucose used for the experiment.
      The results were conclusive, showing that the 6-carbon skeleton of Glc was converted intact to GlcNH2, glucuronic acid, galactose, and mannose. Surprisingly, d-Glc was converted to l-Fuc without inversion (
      • Heath E.C.
      • Roseman S.
      ).

       Enzyme Experiments

      The next step, of course, was to determine the pathways of synthesis and degradation using appropriate “ferments.” A review (
      • Roseman S.
      ) published in 1959 shows how rapidly the field grew in 10 years. Many of the anabolic/catabolic pathways were established, and although they have since been added to and somewhat modified, the essential elements remain the same today.
      Sialic acid was a separate problem in that enzymatic studies could not proceed until after its correct structure was established (
      • Comb D.G.
      • Roseman S.
      ), and N-acetyl-d-mannosamine was found to occur naturally and to be a precursor of NeuAc. Glucosamine-6-P is the precursor of all nitrogen-containing sugars and is formed from Fru-6-P and glutamine (
      • Ghosh S.
      • Blumenthal H.J.
      • Davidson E.
      • Roseman S.
      ), although the catabolic enzyme, GlcN-6-P deaminase (
      • Comb D.G.
      • Roseman S.
      ), which gives Fru-6-P and NH3 as products, is reversible and can be utilized anabolically when the synthase is mutated in bacteria.

       Sugar Nucleotides, Dolichol, and PAPS

      Aside from establishing the intermediary metabolism of the monosaccharides, there were five major developments in the field over the course of the next 20 years (listed in order of the discussion): (a) isolation of “active sulfate” or PAPS (1958); (b) recognition of lectins (sugar-binding proteins) in animal tissues; (c) identification of dolichol-linked oligosaccharides as intermediates in the synthesis of the Man-rich core oligosaccharides of N-linked glycoproteins (1976); (d) isolation of the sugar nucleotides (1950); (e) elucidation of the pathways of synthesis and degradation of the complex carbohydrates and of the number and specificities of the glycosyltransferases.
      The precursor of the sulfated glycoconjugates, such as chondroitin sulfate, is 3′-phosphoadenosine-5′-phosphosulfate (or PAPS) characterized and enzymatically synthesized by Robbins and Lipmann (
      • Robbins P.W.
      • Lipmann F.
      ). PAPS is, of course, “high energy” or “active” sulfate.
      Ricin was apparently the first lectin (proteins that bind carbohydrates) recognized more than a century ago. Early in their history, lectins were found to agglutinate erythrocytes depending on blood type. Lectins by now have become a field unto themselves, and the work of I. J. Goldstein, who developed quantitative methods for accurately defining specificity, as well as in isolating new lectins, is especially significant. The plant lectins are not only powerful tools for analyzing macromolecules and cell surfaces, but the field became particularly interesting to cell biologists when it was realized that animal cells express lectins.
      In 1968, Ashwell and co-workers (
      • Morell A.G.
      • Irvine R.A.
      • Sternlieb I.
      • Scheinberg I.H.
      • Ashwell G.
      ) discovered that liver hepatocytes bind and take up asialoglycoproteins (the asialoglycoprotein endocytosis receptor). This receptor is a Gal-specific lectin in mammals and GlcNAc-specific in birds. It is called the “Ashwell protein” in what follows. Animal lectins, such as the Siglecs (bind to sialic acids), Ig superfamily lectins, selectins, the integrins, CAMs (cell adhesion molecules), collectins, CD44, and others, have now become major areas of research.
      In the 2-year period 1999–2000, SciFinder lists 1400 papers on selectins and 5900 on the integrins. Early in my service on National Institutes of Health Study Sections, our section, which comprised a distinguished group of biochemists, reviewed what I think may have been the first National Institutes of Health application for funds to study a plant lectin (concanavalin A). A vigorous debate ensued with those opposed asking why a plant protein that binds carbohydrates should be of any interest to the National Institutes of Health. It should be funded by the National Science Foundation! Fortunately, the application was funded. In this connection it was this same group that reviewed applications by Fritz Lipmann and by Luis Leloir, which were of course funded; these applications basically consisted of describing what the applicants planned to do with very little detail or particular focus. How would they fare today?
      Lipid-linked intermediates were discovered around 1964–1965 by three groups (reviewed by Osborn (
      • Osborn M.J.
      )). These studies were conducted in the laboratories of Horecker, Robbins, and Strominger, who were working on the enzymatic synthesis of bacterial lipopolysaccharides and the peptidoglycan cell wall. This work led to similar studies in a number of laboratories on lipid-linked intermediates in the biosynthesis of complex carbohydrates in eukaryotic organisms, including yeast, plants, and higher animals (for review, see Ref.
      • Waechter C.J.
      • Lennarz W.J.
      ). A polyisoprenoid, dolichol, was known to occur in animal tissues and was identified as the lipid carrier of the carbohydrate groups. This early work led to the well established dolichol pathway for the synthesis of the N-linked glycoproteins (
      • Kornfeld R.
      • Kornfeld S.
      ). The dolichol pathway does not apply to the O-linked glycoproteins or to the glycolipids.
      The isolation and characterization of sugar nucleotides is one of the most important achievements in the field of carbohydrate metabolism in the 20th century. They were discovered in two laboratories at about the same time, those of Luis Leloir in Argentina and of James Park in this country. Leloir’s group (Caputto, Cardini, Paladini, and Cabib) was working on the enzymatic synthesis of Glc-6-P from Gal-1-P using yeast extracts and found that a heat-stable cofactor (“cozymase”) was required. One of the factors was isolated and fully characterized as UDP-Glc (
      • Caputto R.
      • Leloir L.F.
      • Cardini C.E.
      • Paladini A.C.
      ). This was followed by isolation of UDP-Gal and recognition that the “galactowaldenase” reaction (epimerization at C-4) occurred at the level of the sugar nucleotides. In an independent discovery, Park found that Staphylococcus aureus treated with penicillin (which inhibits cell wall synthesis) accumulated considerable quantities of UDP derivatives and showed that they contained the cell wall sugar muramic acid and amino acids (
      • Park J.T.
      ). The Park compounds are intermediates in cell wall biosynthesis.
      The list of sugar nucleotides is huge (
      • Gabriel O.
      ). It includes virtually every naturally occurring monosaccharide, purines, pyrimidines, and in some cases, 2-deoxyribose in place of ribose. They are most frequently formed by the action of pyrophosphorylases, which catalyze the reaction (N indicates nucleoside): NTP + glycose-1-P → PPi + N-P-P-glycose. As usual, the sialic acids are exceptions in that the nucleotide is CMP (not the diphosphate). CMP-sialic acid, originally isolated from E. coli (
      • Comb D.
      • Watson D.
      • Roseman S.
      ), is formed by condensation of NeuAc or N-glycolylneuraminic acid and CTP (
      • Roseman S.
      ,
      • Kean E.L.
      • Roseman S.
      ). A similar nucleotide (
      • Ghalambor M.A.
      • Heath E.C.
      ) was obtained with 2-keto-3-deoxyoctanoic acid (KDO), a component of bacterial lipopolysaccharides.

       Functions of the Sugar Nucleotides

      Some sugars in glycoconjugates (Man, GlcNAc, NeuAc) are synthesized from Fru-6-P, whereas others (Gal, GlcUA, Xyl) are synthesized as nucleotide sugars from UDP-Glc or, in the case of l-Fuc, from GDP-Man. Aside from their participation in the biosynthesis of monosaccharides, the sugar nucleotides serve as glycose donors in the biosynthesis of oligo- and polysaccharides. Enzymes that utilize sugar nucleotides as donors are designated glycosyltransferases and are the major catalysts for generating the glycosidic bond (also formed by transglycosidases and phosphorylases).
      Glycosyltransferases were first reported by the Leloir group (
      • Cabib E.
      • Leloir L.F.
      ,
      • Leloir L.F.
      • Olavarria J.M.
      • Goldemberg S.H.
      • Carminatti H.
      ). The enzymes catalyzed the synthesis of disaccharides (trehalose, sucrose) and of the α,1→4 linkage in glycogen. At the time, it was generally believed that the α,1→4 linkage in glycogen was synthesized from Glc-1-P by reversing the glycogen phosphorylase-catalyzed reaction.
      The following general glycosyltransferase catalyzed reaction occurs in animal tissues.
      GlycosePP(U or G)+HO­acceptor(U or G)DP+GlycosideO­acceptor
      Reaction 1


      In the case of the sialic acids, such as NeuAc, the donor is CMP-NeuAc.

       Specificity of the Glycosyltransferases, Biosynthetic Pathways, and Multiglycosyltransferase Systems

      At the time we began to study macromolecular glycans (around 1960), nothing was known about the biosynthetic pathways leading to glycoproteins, mucins, and glycolipids. To undertake this work, we required substrate quantities of the sugar nucleotides. Fortunately Moffatt and Khorana (
      • Moffatt J.G.
      • Khorana H.G.
      ) had just published a method for the synthesis of UDP-Glc. A very fruitful and enjoyable summer in Vancouver led to a modified, general method for the synthesis of sugar nucleotides (
      • Roseman S.
      • Distler J.J.
      • Moffatt J.G.
      • Khorana H.G.
      ),
      I arrived with the sugar 1-phosphates and was given space on John Moffatt's bench. He measured my waist and marked off the corresponding width on the bench top. Fortunately, my waist was substantially greater than his.
      giving us the tools for studying glycosyltransferases.
      Some 15–20 glycosyltransferases were characterized, and it soon became obvious that they formed families such as sialyl-, Gal-, GlcNAc-, GalNAc-, Glc-, and Fuc-transferases. The enzymes we studied are involved in the synthesis of the mucins, glycolipids, and terminal trisaccharides of N-linked glycoproteins, and the results can be summarized as follows (see Ref.
      • Roseman S.
      for review). 1) Each glycosyltransferase is specific for both the donor sugar nucleotide and the acceptor. 2) A different transferase usually catalyzes each step in a pathway. When a sugar occurs twice in a molecule, such as NeuAc in disialoganglioside, two different transferases are required.
      The glycosyltransferases that synthesize the GAGs have exceptional characteristics. (a) Elongation of the polysaccharide chains in chondroitin sulfate, heparan sulfate, and in one hyaluronan (produced by Pasteurella) takes place in a stepwise manner at the non-reducing ends of the polymers. In these cases, a single glycosyltransferase transfers first one and then the other glycose unit from their respective sugar nucleotides to the ends of the chain. (b) Single glycosyltransferases also catalyze hyaluronan synthesis by eukaryotic and Streptococcal cells, but in these cases elongation occurs at the reducing ends of the chain by mechanisms that are not entirely clear. One phenomenon that has always intrigued this reviewer is how the enzymes or cells “know” when to stop the process of polysaccharide elongation (see Ref.
      • Hascall V.C.
      for review).
      3) Chain elongation can be at the non-reducing end or can form branch points. 4) The product of one reaction is the substrate for the next, which leads to the concept of multiglycosyltransferase (MGT) systems,namely that the transferases required for synthesis of one glycoconjugate are associated. A different MGT system is required for mucins, glycoprotein trisaccharide termini, and glycolipids (e.g. gangliosides). 5) Polydispersity in glycoconjugates is explained by the MGT hypothesis. For instance, Svennerholm (
      • Svennerholm L.
      ) showed that there is a particular array of human brain gangliosides of different chain length and complexity. This array exactly fits the pathway that is predicted by the specificities of the enzymes in the corresponding MGT. One would expect to find only the final products of the pathway (e.g. tetrasialoganglioside) if all conditions were optimum for each enzyme and they are expressed at high enough levels. It should be noted that many of the transferases require Mn2+ for activity and not necessarily at the same concentrations,
      One mechanism for regulating glycosyltransferase activity could well be via local Mn2+concentrations. A brief literature search found references to analyses of tissues, mostly autopsy material, for trace metals including Mn2+ but little on its transport. The relevant analyses will require that they be conducted on actively metabolizing cells and organelles (such as the Golgi) to preserve the in vivo ion gradients.
      and this may be an important means of regulating these activities. 6) There is some evidence to support the idea that glycosyltransferases in an MGT complex bind to one another. In the original work, we found that all of the transferases in a given MGT were found in the same particulate fraction, ultimately identified as the Golgi apparatus (
      • Schachter H.
      • Jabbal I.
      • Hudgin R.
      • Pinteric L.
      • McGuire E.J.
      • Roseman S.
      ). The Gal-transferase involved in synthesizing the Gal-O-Xyl-O-Ser (protein) linkage region in chondroitin sulfate was purified by binding to the immobilized xylosyltransferase, and it coprecipitates with antibody directed at the xylosyltransferase (
      • Schwartz N.B.
      • Roden L.
      ). Recent papers present evidence for binding of a glycolipid GalNAc-transferase to a Gal-transferase (
      • van Meer G.
      ,
      • Giraudo C.G.
      • Daniotti J.L.
      • Maccioni H.J.F.
      ).

       The Structure of the Acceptor Can Determine Glycosyltransferase Activity

      The activity of a GT is not only determined by whether the acceptor is a glycolipid, mucin, or an N-linked glycoprotein but can also depend on the fine structure of the termini in these potential acceptors. One example will be given. Enzymatically synthesized, labeled CMP-NeuAc and CMP-N-glycolylneuraminic acid (
      • Roseman S.
      ) were used to detect and characterize sialyltransferases (STs), first from rat mammary gland and then in colostrum (goat, bovine, human), bacterial extracts (for synthesizing colominic or polysialic acid), submaxillary gland, and embryonic chicken brain (summarized in Ref.
      • Roseman S.
      ). Bovine colostrum and human milk contain mixtures of 3′-sialyllactose (NeuAc-α2,3-Galβ1,4Glc) and 6′-sialyllactose (NeuAc-α2,6-Galβ1,4Glc). The rat mammary gland ST synthesizes 3′-sialyllactose when incubated with CMP-NeuAc and lactose, whereas the colostrum enzyme yields the 6′-isomer. The colostrum enzyme shows great specificity for its acceptors (
      • Bartholomew B.A.
      • Jourdian G.W.
      • Roseman S.
      ). In quantitative terms, when the efficiency of the enzyme is expressed as Vmax/Km the following values were obtained (% relative to N-acetyllactosamine): Gal(β,1–4)GlcNAc or N-acetyllactosamine, 100; Gal(β,1–4)Glc or lactose, 2; Gal(β,1–3)GlcNAc, 13; Gal(β,1–6)GlcNAc, 0.4; and asialoorosomucoid, 1000.
      Thus, the nature of the penultimate sugar in an acceptor, the precise linkage of the terminal to the penultimate sugar, and the size of the acceptor can all play major roles in determining the activity of a given GT.
      The total number of GTs thus far identified exceeds many hundreds (reviewed in Ref.
      • Varki A.
      • Marth J.
      ). Many of the structural genes have been cloned, and the enzymes were overexpressed, purified to homogeneity, and characterized kinetically. At least two have been crystallized and their three-dimensional structures determined. Insofar as the topics covered here are concerned, the GlcNAc transferases that act on polylactosamine ((Gal-(β1,4)-GlcNAc-(β1,3))n), a constituent of many cell membranes, are of considerable interest (see review by Renkonen (
      • Renkonen O.
      )). One GlcNAc transferase is required for increasing the chain length at the non-reducing terminal Gal. Two others add GlcNAc to internal Gal residues, thereby starting the branching process. One of the branching enzymes works at the distal end of the chain, and the other acts “centrally.” Both are greatly influenced by the presence of Fuc residues on the chains. Thus, the combination and interplay of the GalT, the three GlcNAc transferases, the FucT, and possibly the sialyltransferases determine the final structure on the cell surface, but how these are regulated with respect to each other remains to be determined.

      Speculations on Cell-Cell Adhesion

      The human brain contains approximately a trillion neurons, and each averages around 103 connections with other cells or about 1015 specific connections. How can this happen given a total of about 40,000 genes in the human genome? The data banks list 72,000 publications on “cell adhesion,” and they report CAMs (cell adhesion molecules), cadherins, catenins, ephrins, Eph receptor tyrosine kinases, laminins, selectins, integrins, their relationships to the extracellular matrix and the cytoskeleton, to cytokines, and much, much more.
      In some instances, the role of carbohydrates is well documented. (a) Leukocyte extravasation (recruiting leukocytes from the blood to the site of infection, injury, or lymphatic circulation) involves a sequence of complicated interactions between the leukocytes and the blood capillary endothelium comprising selectins, other proteins, and carbohydrates (reviewed in Ref.
      • Ebnet K.
      • Vestweber D.
      ). (b) CD44, a cell surface receptor, binds to hyaluronan (
      • Hascall V.C.
      ). (c) Myelin-associated glycoprotein is a Siglec (sialic acid-specific lectin) that binds to complex gangliosides, an interaction essential for maintaining normal myelin structure (
      • Collins B.E.
      • Kiso M.
      • Hasegawa A.
      • Tropak M.B.
      • Roder J.C.
      • Crocker P.R.
      • Schnaar R.L.
      ,
      • Collins B.E.
      • Yang L.J.S.
      • Mukhopadhyay G.
      • Filbin M.T.
      • Kiso M.
      • Hasegawa A.
      • Schnaar R.L.
      ).
      As indicated below, there is now a rapidly developing interest in the role of glycans in development and in cell recognition. However, in surveying the literature, it appears that some old ideas bear repeating. The discussion will be limited to cell-cell recognition.

       Specific Intercellular Adhesion

      The crucial importance of cell recognition in development was well established in the late 1800s. In normal embryos, cells exhibit exquisite adhesive specificity. They “know” where they are, and they “know” where they are going. Under in vitro conditions, cells adhere nonspecifically to many substances, including tissue culture plastic, glass, serum proteins, etc. Nevertheless, adhesive specificity can be demonstrated in vitro and was shown in 1907 in a classic case of serendipity. Wilson (
      • Wilson H.V.
      ) found that when single-cell suspensions from two species of marine sponges were mixed they first aggregated to form a heterologous chimera, but with time they sorted out to yield aggregates of homotypic cells. Holtfreter (1930s) obtained the same results with cell suspensions from different embryonic tissues. Although cadherins are thought to be involved in cell sorting, the underlying biochemical basis is very complex, and yet to be fully explained.
      Humphreys (
      • Humphreys T.
      ) showed that dissociation of the sponges to single cells released species-specific, heat-labile, large molecular weight “aggregation factors.” These observations were followed by a series of studies from many laboratories, particularly by Burger’s group. Polysaccharides, sulfated polysaccharides, proteoglycans, and lectins have been invoked as participants in a multistep process, some of which require Ca2+. In a recent paper (
      • Jarchow J.
      • Jurgen F.
      • Anselmetti D.
      • Calabro A.
      • Hascall V.C.
      • Gerosa D.
      • Burger M.M.
      ), a unique supramolecular circular proteoglycan complex is described as one of the components involved in the process. One of the N-linked glycans contains glucuronic acid, fucose, mannose, galactose, N-acetylglucosamine, and sulfate (
      • Misevic G.N.
      • Burger M.M.
      ).
      It was subsequently demonstrated that adult cells, such as hepatocytes and mycocytes (
      • Albanese J.
      • Kuhlenschmidt M.S.
      • Schmell E.
      • Slife C.W.
      • Roseman S.
      ,
      • Obrink B.
      • Kuhlenschmidt M.S.
      • Roseman S.
      ), exhibit adhesive specificity and that in liver homogenates, the specific factor was localized to the plasma membranes (
      • Obrink B.
      • Kuhlenschmidt M.S.
      • Roseman S.
      ). The active factor(s) in the chick membranes is a trace glycolipid (Fig. 3).
      Figure thumbnail gr3
      Figure 3Scanning electron micrographs of chicken hepatocytes and immobilized chicken plasma membrane glycolipid. A quantitative procedure was devised for assaying the effects of immobilized glycolipids and similar substances on the rate of adhesion of chicken hepatocytes (
      • Park L.
      • Kuhlenschmidt M.
      • Roseman S.
      ). A glycolipid, present in trace quantities in chicken liver, was the only substance of many pure and mixed lipids and glycolipids tested that stimulated the adhesion of these (but not rat) hepatocytes. The scanning electron micrographs are as follows.A, mixture of chicken hepatocytes and polystyrene beads (red balls) previously immersed in a dilute solution of the specific glycolipid. No such structures were seen with beads treated with the inactive lipids. B and C, cell and bead. The filopodia-like structures were observed in all cases of cell-bead adhesion. D, cell-cell adhesion. Filopodia-like structures are evident in the regions of contact. These experiments were conducted with purified but not homogenous preparations of the glycolipid. The specific glycolipid has recently been purified to apparent homogeneity by Dr. Ming Chuan Shao and Barbara Rauch. (The photographs were kindly prepared by Michael McCaffery of the Integrated Imaging Center, Department of Biology, Johns Hopkins University.)
      A quantitative assay (
      • Orr C.W.
      • Roseman S.
      ) was used to study the kinetics of homologous adhesion (
      • Umbreit J.
      • Roseman S.
      ) and showed that the process is multistep. The first step does not require metabolic energy; the cells form a loose association that dissociates even by simple dilution of the suspension. In the second energy-requiring step, the aggregate is stabilized and can only be dissociated by vigorous treatment, e.g. proteases. In the third step, the stable aggregates synthesize collagen and sulfated GAGs. All of this takes minutes at 37 °C.
      Insofar as the underlying biochemical mechanisms are concerned, there are two obvious questions. (a) What cell surface molecules participate in the process? (b) How is the information transmitted to the interior of the cell? Two hypotheses were suggested to answer these questions, as indicated in Figs. 4 and5.
      Figure thumbnail gr4
      Figure 4Hypothesis I: carbohydrates are involved in specific intercellular adhesion. The suggested mechanisms are as follows. (a) Hydrogen bonding between oligosaccharide chains on adjacent cell surfaces. The scheme is not meant to imply that hydrogen bonds can only form between identical monosaccharides. (b) Enzyme-substrate complex. The model is meant to suggest that cells can bind to and/or modify adjacent cells or extracellular matrix through the action of cell surface glycosyltransferases. See text for discussion.
      Figure thumbnail gr5
      Figure 5Hypothesis II: membrane messengers. An extrapolation of the Sutherland second messenger idea. A variety of extracellular signals are received by cell membrane receptors, which in turn send specific messages to the cytoplasm or nucleus. The membrane is a transducer, and the membrane messengers were suggested (
      • Roseman S.
      ,
      • Roseman S.
      Neuroscience, Study Program.
      ) to comprise both low and high molecular weight substances, such as proteins. Additionally, it was suggested that in some cases the messenger molecules would act stoichiometrically (e.g.repressors of operons), whereas in others, they could be enzymes.

       Hypothesis I: Carbohydrates Are Involved in Specific Intercellular Adhesion

      Two mechanisms were proposed (
      • Roseman S.
      ) for carbohydrate participation as indicated in Fig. 4. 1) Cell adhesion is mediated by hydrogen bonds between carbohydrates on neighboring cells. That hydrogen bonds can be important in maintaining carbohydrate structures is exemplified by polysaccharides such as cellulose and chitin. 2) Cell adhesion is mediated by the binding of carbohydrates to cell surface proteins and enzymes. There were two reasons for extrapolating from proteins to enzymes and in particular to the glycosyltransferases. (a) The glycosyltransferases as a class appeared to be much more specific than the lectins, a critical requirement for specific intercellular adhesion. (b) If glycosyltransferases are involved, then one cell could also modify the surface of its neighbor. However, extracellular modification requires an extracellular cell surface or soluble enzyme and a source of sugar nucleotides and/or PAPS, either from the cytoplasm or extracellularly. Is any of this possible? 1) Enzymatically active, soluble extracellular glycosyltransferases do occur in the fluid surrounding intact embryonic chicken brain and in embryonic and adult chicken serum, vitreous humor, and human spinal fluid (
      • Den H.
      • Kaufman B.
      • McGuire E.J.
      • Roseman S.
      ). 2)Cell surface glycosyltransferases may occur. Chick embryonic neural retina cells transferred Gal from UDP-Gal to soluble high molecular weight acceptors (
      • Roth S.
      • McGuire E.J.
      • Roseman S.
      ), suggesting that the reaction was catalyzed by a cell surface Gal-transferase. Evidence for and against this conclusion has been presented by other laboratories, and at this time, it remains controversial. However, Fig. 3 suggests that only a vanishingly small percent of the cell surface appears to be involved early in specific cell-cell interactions. If cell surface glycosyltransferases participate in these interactions, they may be present in traces and difficult to detect by any method, including immunological procedures. 3) Sugar nucleotides may be secreted. In a recent paper (
      • Chambers J.K.
      • Macdonald L.E.
      • Sarau H.M.
      • Ames R.S.
      • Freeman K.
      • Foley J.J.
      • Zhu Y.
      • McLaughlin M.M.
      • Murdock P.
      • McMillan L.
      • Trill J.
      • Swift A.
      • Aiyar N.
      • Taylor P.
      • Vawter L.
      • Naheed S.
      • Szekeres P.
      • Hervieu G.
      • Scott C.
      • Watson J.M.
      • Murphy A.J.
      • Duzic E.
      • Klein C.
      • Bergsma D.J.
      • Wilson S.
      • Livi G.P.
      ), a G protein-coupled plasma membrane receptor for UDP-Glc was identified in a wide variety of human tissues, including many regions of the brain. Thus, extracellular sugar nucleotides may indeed occur.

       Hypothesis II: Membrane Messengers

      In 1958–1962, a series of studies by Sutherland and co-workers (
      • Rall T.W.
      • Sutherland E.W.
      ,
      • Hardman J.G.
      • Robison G.A.
      • Sutherland E.W.
      ) led to the characterization of cAMP, adenylate cyclase, and the effects of certain hormones on this enzyme. Sutherland designated cAMP as the “second messenger” (hormones were the first messenger). This seminal work surely ranks as one of the most important biochemical findings of the past century. Somewhat later, Rasmussen invoked Ca2+ as another “second messenger.” These hypotheses were obviously correct, but it seemed to us that they were insufficient. Could the diverse stimuli received by a cell and the many responses that these signals elicited be explained by only two second messengers? A “membrane messenger” hypothesis was therefore devised in 1974 as illustrated in Fig. 5. The membrane acts as a “transducer” containing multiple receptors that respond to external signals by releasing specific intracellular messengers. Signal transduction by the plasma membrane is now well established, and a section of each issue of this Journal is devoted to papers in this field.

       Quantitative Changes in Carbohydrate Ligands Can Have Global Effects on Cellular Phenotypic Behavior

      Qualitative changes in carbohydrate composition of the cell surface or the substrata to which the cell adheres can have far reaching effects on cell behavior, but what of quantitative changes? Although the Ashwell protein catalyzes receptor-mediated endocytosis of glycoproteins in hepatocytes, it does not participate in intercellular adhesion. Nevertheless, it served as a useful model to answer this question.
      We have often tried to mimic cell surfaces by adsorbing or covalently linking potential carbohydrate ligands to solid matrices (e.g. Fig. 3). This approach was used to test hepatocytes (
      • Schnaar R.L.
      • Weigel P.H.
      • Kuhlenschmidt M.S.
      • Lee Y.C.
      • Roseman S.
      ,
      • Weigel P.H.
      • Schnaar R.L.
      • Kuhlenschmidt M.S.
      • Schmell E.
      • Lee R.T.
      • Lee Y.C.
      • Roseman S.
      ) with sugar derivatives covalently linked to polyacrylamide gels. Chicken hepatocytes specifically adhered to GlcNAc-derivatized gels and rat hepatocytes to Gal-derivatized gels, in accord with the known specificities of the Ashwell receptors in the two cell types. However, there was a remarkable threshold or critical concentration effect of the sugars as shown in Fig.6. Below this concentration of sugar in the gel, the cells did not bind to the gels. At the threshold, ∼15% increases in GlcNAc and Gal concentrations, respectively, in the gels resulted in 100% cell binding to the gels.
      An interaction between a protein and its monovalent ligand may be weak, but if the ligand is polyvalent such that many protein molecules can interact, the binding affinity for the polyvalent ligand greatly increases. An excellent example of this is CD44, a cell surface receptor that binds to hyaluronan (
      • Hascall V.C.
      ). Hyaluronan oligosaccharides with 6–10 sugars are sufficient to interact with CD44 monovalently, and relatively high concentrations of these oligosaccharides can prevent binding of macromolecular hyaluronan, which otherwise binds with high affinity. However, the interaction of the monovalent oligosaccharides with CD44 is sufficiently weak that they do not remain bound through a routine washing step (
      • Lesley J.
      • Hascall V.C.
      • Tammi M.
      • Hyman R.
      ). The cytoplasmic tail of CD44 interacts with anchorin in the cytoskeleton. Therefore, interaction of CD44 with its polyvalent, linear ligand can contribute to alignment and stabilization of the cytoskeleton and consequently influence cell behavior.
      Figure thumbnail gr6
      Figure 6Effect of quantitative changes in carbohydrate concentration on cell binding in a model system. A variety of carbohydrates were covalently linked to polyacrylamide gels and tested with chicken and rat hepatocytes. As expected, the cells bound to the gels in accord with the known specificities of the Ashwell receptor, chicken to GlcNAc, and rat to Gal. In this model system, the concentrations of the sugars in the gels were varied, as indicated. A threshold or critical concentration effect is observed. Below this concentration, there is no binding, and above it, all the cells bind to the gels.
      The physiological implications are plain if this model represents what can happen in cell adhesion. A non-adherent cell can become adherent by a slight change in the cell surface concentration of the appropriate ligand and/or its receptor or in the extracellular matrix. Even more likely, the “grouping” of receptors or ligands into microdomains in the plasma membrane results in binding, and the size of these domains is apparently affected (regulated?) by other factors, such as cholesterol.

      The Future: Glycans as Informational Macromolecules

      The particular advantage of carbohydrates is that they have enormous potential for serving as informational macromolecules, starting with their de novo biosynthesis. Laine (
      • Laine R.A.
      ), for instance, has calculated that a hexasaccharide has 1012isomeric permutations. Second, the glycans are readily modified after synthesis of the core structure. A few such modifications are sulfation (thought to be essential for leukocyte extravasation),O-acetylation of individual sugars such as sialic acid, addition of a few sugar residues that can convert blood group O to A or B, or initiating a branch point by the action of the branching GlcNAc transferases on polylactosamine. For instance, in the neural retinotectal system where neuronal pathfinding is essential, immunological methods have shown a dorsoventral gradient in a cell surface antigen of the rat embryonic neural retina (
      • Sparrow J.R.
      • Barnstable C.J.
      ). The antigen was identified as 9-O-acetyl-GD3. At the same time, there was no apparent gradient of the parent ganglioside GD3. Thus, relatively few enzymes can create a large number of molecular variants.
      There is no doubt that the molecular events underlying embryogenesis, especially of the nervous system, will be the major goal of biology well into the foreseeable future. Experiments in a number of laboratories are now in progress to elucidate the roles of glycans in these processes, and some of these are cited above. However, other examples can be given. (a) By constructing specific glycosyltransferase mutants in mice and other organisms, the synthesis of specific glycans or classes of glycans can be eliminated. This approach has shown that gangliosides and glycoproteins of the N-glycan type are essential for the survival of the embryo and/or its normal development in the mouse (
      • Kawai H.
      • Allende M.L.
      • Wada R.
      • Kono M.
      • Sango K.
      • Deng C.
      • Miyakawa T.
      • Crawley J.N.
      • Werth N.
      • Bierfreund U.
      • Sandhoff K.
      • Proia R.L.
      ) and in the nematode Caenorhabditis elegans(
      • Chen S.
      • Zhou S.
      • Sarkar M.
      • Spence A.M.
      • Schachter H.
      ).
      Unpublished data: on the N-type glycoproteins by Schachter et al; on the gangliosides by Sandhoff, Proia et al.
      (b) A number of papers have reported that proteoglycans and glycosaminoglycans, especially heparan, are essential for normal development of Drosophila and C. elegans. The affected genes include Wingless,tout-velu, sugarless,sulfateless, dally, and sqv 3,7,8 (
      • Bulk D.A.
      • Wei G.
      • Toyoda H.
      • Kinoshita-Toyoda A.
      • Waldrip W.R.
      • Esko J.D.
      • Robbins P.W.
      • Selleck S.B.
      ,
      • Toyoda H.
      • Kinoshita-Toyoda A.
      • Fox B.
      • Selleck S.B.
      ,
      • Toyoda H.
      • Kinoshita-Toyoda A.
      • Selleck S.B.
      ,
      • Nakato H.
      • Futch T.A.
      • Selleck S.B.
      ,
      • Perrimon N.
      • Bernfield M.
      ,
      • Baeg G.H.
      • Lin X.
      • Khare N.
      • Baumgartner S.
      • Perrimon N.
      ,
      • Toyoda H.
      • Kinoshita-Toyoda A.
      • Fox B.
      • Selleck S.B.
      ). (c) Notch receptors are highly conserved intercellular signaling pathways that direct embryonic cell-fate decisions. The activities of these receptors are regulated by Fringe proteins, and recent evidence (
      • Brückner K.
      • Perez L.
      • Clausen H.
      • Cohen S.
      ,
      • Moloney D.J.
      • Panin V.M.
      • Johnston S.H.
      • Chen J.
      • Shao L.
      • Wilson R.
      • Wang Y.
      • Stanley P.
      • Irvine K.D.
      • Haltiwanger R.S.
      • Vogt T.F.
      ) shows that Fringe is a fucose-specific GlcNAc-transferase.
      To summarize, the huge gap between the 1015 specific connections in the brain and the number of genes in the human genome can readily be filled by the glycans.
      It is presumptuous to try to predict the future. Who, in the 1960s, could have predicted what happened to the field we called genetics? At the moment, primary interest seems to be shifting from genomics to proteomics and functional proteomics. But as others have said, glycobiology is the field of the future. However, the problems are formidable, as I have tried to indicate in this brief overview.
      One “problem” is nothing more than a false perception. On several occasions I have heard structural biologist colleagues state that the glycan units in a glycoprotein, for instance, cannot be important because they are too flexible to be seen in an x-ray crystal structure or by NMR. In other words, if they don’t have structure, how can they have function? That this conclusion is gratuitous requires no more than a moment’s reflection.
      For instance, one important physiological property of cartilage is that it is reversibly compressible, acting like a spring to the application of a force. This feature emanates from the flexibility of the aggrecan aggregate, which can be compressed to one-tenth its volume. Hyaluronan provides another example. It is essential as a lubricant in joint fluids where it has high viscosity and an extended helical or possibly random coil structure. It is more restricted in the aggregcan aggregate but must still be flexible, and it forms a gel in cumulus cell-oocyte complexes (
      • Hascall V.C.
      ).
      These should be sufficient to make the point. Glycans aredifferent because frequently they are flexible and adjust to physiological need. In other words, in these substances, function defines structure, not vice versa. Certain glycans clearly have highly favored conformations, and lectins may have evolved to reflect those particular three-dimensional structures. Furthermore, whereas energy minimization methods can yield the thermodynamically favored conformers, the less favored conformers may be the biologically active structures that bind to their ligands.
      It would not be a big surprise if different conformers of a single oligosaccharide interacted with different ligands or receptors or enzymes or possibly even other carbohydrates under different physiological conditions. It is this interplay between proteins and different conformers that likely allows a single carbohydrate structure, such as hyaluronan, to be used in many different ways. In the excellent book by Cantor and Schimmel on the conformation of macromolecules (
      • Cantor C.R.
      • Schimmel P.R.
      ), they raise a number of questions about carbohydrate polymers similar to those discussed above and then say: “These are all interesting questions, but it will probably take much hard work to answer most of them.” Amen to that! What is lacking is adequate biophysical methodology.
      The problem is much more complicated when we deal with membranes. Trying to assign structure or even distribution (if it is not random) of a particular glycolipid on the surface seems impossible at this point because of fluidity of the external monolayer of the lipid bilayer. If glycolipids do exist primarily in “rafts” or domains, these domains are in a constant state of flux and motion within the monolayer, and their sizes, frequency of formation, etc., depend on the lipid composition of the remainder of the monolayer and whether they are or are not associated with membrane-bound signaling proteins, such as the Src family of kinases. The same problem exists with cell surface glycoproteins, except possibly for those tethered to cytoplasmic components, such as the cytoskeleton. Even in the latter case, publications suggest that perturbation of the cell can rapidly result in drastic reorganization of the cytoskeleton.
      Thus, it appears that present methods will permit us only to obtain “snapshots” of limited areas of the cell surface. There is no doubt that the task ahead of us is difficult, but if cells “talk” to other cells via cell surface substances such as the glycans, the problem cannot be avoided. I am optimistic. Breakthrough technological advances are produced at an astonishing rate these days.
      Who could have predicted the development of polymerase chain reaction and its consequences?

      Acknowledgments

      I am especially grateful to Drs. Ronald Schnaar and Mark Roseman for critical reading of this manuscript and for many helpful suggestions. The sections on aggrecan and the aggrecan aggregate and Fig. 2 could not have been written without the help of Dr. Vincent Hascall, who also provided numerous other insightful comments.

      REFERENCES

      1. Science. 2001; 291: 2263-2502
        • Roseman S.
        Chem. Phys. Lipids. 1970; 5: 270-297
        • Roseman S.
        FEMS Microbiol. Rev. 1989; 63: 3-12
        • Keyhani N.O.
        • Roseman S.
        Biochim. Biophys. Acta. 1999; 1473: 108-122
        • Conant J.B.
        Pasteur's Study of Fermentation. Harvard University Press, Cambridge, MA1952
        • Fruton J.S.
        • Simmonds S.
        General Biochemistry. John Wiley & Sons, Inc., New York1953
        • Gottschalk A.
        Glycoproteins: Their Composition, Structure and Function. Elsevier Science Publishers B.V., Amsterdam1972
        • Meyer K.
        • Palmer J.W.
        J. Biol. Chem. 1934; 107: 629-634
        • Carlson D.M.
        J. Biol. Chem. 1966; 241: 2984-2986
        • Roden L.
        Lennarz W.J. The Biochemistry of Glycoproteins and Proteoglycans. Plenum Press, New York1980: 267-371
        • Yoshima H.
        • Matsumoto A.
        • Mizuochi T.
        • Kawasaki T.
        • Kobata A.
        J. Biol. Chem. 1981; 256: 8476-8484
        • Hascall V.C.
        Glycoconj. J. 2000; 17: 599-608
        • Wight T.N.
        • Heinegard D.K.
        • Hascall V.C.
        Hay E.D. Olson B. Cell Biology of Extracellular Matrix. Plenum Press, New York1991: 45-78
        • Kabat E.A.
        Blood Group Substances. Academic Press, New York1956
        • Hakomori S.-I.
        Biochim. Biophys. Acta. 1999; 1473: 247-266
        • Jacob F.
        Curr. Top. Dev. Biol. 1979; 13: 117-137
        • Järnefelt J.
        • Rush J.
        • Li Y.-T.
        • Laine R.A.
        J. Biol. Chem. 1978; 253: 8006-8009
        • Laine R.A.
        • Rush J.S.
        Adv. Exp. Med. Biol. 1988; 228: 331-347
        • Renkonen O.
        Cell. Mol. Life Sci. 2000; 57: 1423-1439
        • Heath E.C.
        • Roseman S.
        J. Biol. Chem. 1958; 230: 511-519
        • Roseman S.
        Annu. Rev. Biochem. 1959; 28: 545-578
        • Comb D.G.
        • Roseman S.
        J. Biol. Chem. 1960; 235: 2529-2537
        • Ghosh S.
        • Blumenthal H.J.
        • Davidson E.
        • Roseman S.
        J. Biol. Chem. 1960; 235: 1265-1273
        • Comb D.G.
        • Roseman S.
        J. Biol. Chem. 1958; 232: 807-827
        • Robbins P.W.
        • Lipmann F.
        J. Biol. Chem. 1957; 229: 837-851
        • Morell A.G.
        • Irvine R.A.
        • Sternlieb I.
        • Scheinberg I.H.
        • Ashwell G.
        J. Biol. Chem. 1968; 243: 155-159
        • Osborn M.J.
        Annu. Rev. Biochem. 1969; : 501-538
        • Waechter C.J.
        • Lennarz W.J.
        Annu. Rev. Biochem. 1976; 45: 95-112
        • Kornfeld R.
        • Kornfeld S.
        Annu. Rev. Biochem. 1985; 54: 631-664
        • Caputto R.
        • Leloir L.F.
        • Cardini C.E.
        • Paladini A.C.
        J. Biol. Chem. 1950; 184: 333-350
        • Park J.T.
        J. Biol. Chem. 1952; 194: 877-904
        • Gabriel O.
        Methods Enzymol. 1982; 83: 332-353
        • Comb D.
        • Watson D.
        • Roseman S.
        J. Biol. Chem. 1966; 241: 5637-5642
        • Roseman S.
        Proc. Natl. Acad. Sci. U. S. A. 1962; 48: 437-441
        • Kean E.L.
        • Roseman S.
        J. Biol. Chem. 1966; 241: 5643-5650
        • Ghalambor M.A.
        • Heath E.C.
        Biochem. Biophys. Res. Commun. 1963; 10: 346-351
        • Cabib E.
        • Leloir L.F.
        J. Biol. Chem. 1958; 231: 259-275
        • Leloir L.F.
        • Olavarria J.M.
        • Goldemberg S.H.
        • Carminatti H.
        Arch. Biochem. Biophys. 1959; 81: 508-520
        • Moffatt J.G.
        • Khorana H.G.
        J. Am. Chem. Soc. 1958; 80: 3756-3761
        • Roseman S.
        • Distler J.J.
        • Moffatt J.G.
        • Khorana H.G.
        J. Am. Chem. Soc. 1961; 83: 659-663
        • Svennerholm L.
        J. Neurochem. 1963; 10: 613-623
        • Schachter H.
        • Jabbal I.
        • Hudgin R.
        • Pinteric L.
        • McGuire E.J.
        • Roseman S.
        J. Biol. Chem. 1970; 245: 1090-1100
        • Schwartz N.B.
        • Roden L.
        J. Biol. Chem. 1975; 250: 5200-5207
        • van Meer G.
        Proc. Natl. Acad. Sci. U. S. A. 2001; 98: 1321-1323
        • Giraudo C.G.
        • Daniotti J.L.
        • Maccioni H.J.F.
        Proc. Natl. Acad. Sci. U. S. A. 2001; 98: 1625-1630
        • Bartholomew B.A.
        • Jourdian G.W.
        • Roseman S.
        J. Biol. Chem. 1973; 248: 5751-5762
        • Varki A.
        • Marth J.
        Semin. Dev. Biol. 1995; 6: 127-138
        • Ebnet K.
        • Vestweber D.
        Histochem. Cell Biol. 1999; 112: 1-23
        • Collins B.E.
        • Kiso M.
        • Hasegawa A.
        • Tropak M.B.
        • Roder J.C.
        • Crocker P.R.
        • Schnaar R.L.
        J. Biol. Chem. 1997; 272: 16889-16895
        • Collins B.E.
        • Yang L.J.S.
        • Mukhopadhyay G.
        • Filbin M.T.
        • Kiso M.
        • Hasegawa A.
        • Schnaar R.L.
        J. Biol. Chem. 1997; 272: 1248-1255
        • Wilson H.V.
        J. Exp. Zool. 1907; 5: 245-258
        • Humphreys T.
        Davis B.B. Warren L. The Cell Surface and Specific Cell Aggregation. Prentice-Hall, Englewood Cliffs, NJ1967: 195-210
        • Jarchow J.
        • Jurgen F.
        • Anselmetti D.
        • Calabro A.
        • Hascall V.C.
        • Gerosa D.
        • Burger M.M.
        J. Struct. Biol. 2000; 132: 95-105
        • Misevic G.N.
        • Burger M.M.
        J. Biol. Chem. 1990; 265: 20577-20584
        • Albanese J.
        • Kuhlenschmidt M.S.
        • Schmell E.
        • Slife C.W.
        • Roseman S.
        J. Biol. Chem. 1982; 257: 3165-3170
        • Obrink B.
        • Kuhlenschmidt M.S.
        • Roseman S.
        Proc. Natl. Acad. Sci. U. S. A. 1977; 74: 1077-1081
        • Orr C.W.
        • Roseman S.
        J. Membr. Biol. 1969; 1: 109-124
        • Umbreit J.
        • Roseman S.
        J. Biol. Chem. 1975; 250: 9360-9368
        • Den H.
        • Kaufman B.
        • McGuire E.J.
        • Roseman S.
        J. Biol. Chem. 1975; 250: 739-746
        • Roth S.
        • McGuire E.J.
        • Roseman S.
        J. Cell Biol. 1971; 51: 536-547
        • Chambers J.K.
        • Macdonald L.E.
        • Sarau H.M.
        • Ames R.S.
        • Freeman K.
        • Foley J.J.
        • Zhu Y.
        • McLaughlin M.M.
        • Murdock P.
        • McMillan L.
        • Trill J.
        • Swift A.
        • Aiyar N.
        • Taylor P.
        • Vawter L.
        • Naheed S.
        • Szekeres P.
        • Hervieu G.
        • Scott C.
        • Watson J.M.
        • Murphy A.J.
        • Duzic E.
        • Klein C.
        • Bergsma D.J.
        • Wilson S.
        • Livi G.P.
        J. Biol. Chem. 2000; 275: 10767-10771
        • Rall T.W.
        • Sutherland E.W.
        J. Biol. Chem. 1962; 237: 1228-1232
        • Hardman J.G.
        • Robison G.A.
        • Sutherland E.W.
        Annu. Rev. Physiol. 1971; 33: 311-336
        • Schnaar R.L.
        • Weigel P.H.
        • Kuhlenschmidt M.S.
        • Lee Y.C.
        • Roseman S.
        J. Biol. Chem. 1978; 253: 7940-7951
        • Weigel P.H.
        • Schnaar R.L.
        • Kuhlenschmidt M.S.
        • Schmell E.
        • Lee R.T.
        • Lee Y.C.
        • Roseman S.
        J. Biol. Chem. 1979; 254: 10830-10838
        • Lesley J.
        • Hascall V.C.
        • Tammi M.
        • Hyman R.
        J. Biol. Chem. 2000; 275: 26967-26975
        • Laine R.A.
        Glycobiology. 1994; 4: 759-767
        • Sparrow J.R.
        • Barnstable C.J.
        J. Neurosci. Res. 1998; 21: 398-409
        • Kawai H.
        • Allende M.L.
        • Wada R.
        • Kono M.
        • Sango K.
        • Deng C.
        • Miyakawa T.
        • Crawley J.N.
        • Werth N.
        • Bierfreund U.
        • Sandhoff K.
        • Proia R.L.
        J. Biol. Chem. 2001; 276: 6885-6888
        • Chen S.
        • Zhou S.
        • Sarkar M.
        • Spence A.M.
        • Schachter H.
        J. Biol. Chem. 1999; 274: 288-297
        • Bulk D.A.
        • Wei G.
        • Toyoda H.
        • Kinoshita-Toyoda A.
        • Waldrip W.R.
        • Esko J.D.
        • Robbins P.W.
        • Selleck S.B.
        Proc. Natl. Acad. Sci. U. S. A. 2000; 97: 10838-10843
        • Toyoda H.
        • Kinoshita-Toyoda A.
        • Fox B.
        • Selleck S.B.
        J. Biol. Chem. 2000; 275: 21856-21861
        • Toyoda H.
        • Kinoshita-Toyoda A.
        • Selleck S.B.
        J. Biol. Chem. 2000; 275: 2269-2275
        • Nakato H.
        • Futch T.A.
        • Selleck S.B.
        Development. 1995; 121: 3687-3702
        • Perrimon N.
        • Bernfield M.
        Nature. 2000; 404: 725-728
        • Baeg G.H.
        • Lin X.
        • Khare N.
        • Baumgartner S.
        • Perrimon N.
        Development. 2001; 128: 87-94
        • Toyoda H.
        • Kinoshita-Toyoda A.
        • Fox B.
        • Selleck S.B.
        J. Biol. Chem. 2000; 275: 21856-21861
        • Brückner K.
        • Perez L.
        • Clausen H.
        • Cohen S.
        Nature. 2000; 406: 411-415
        • Moloney D.J.
        • Panin V.M.
        • Johnston S.H.
        • Chen J.
        • Shao L.
        • Wilson R.
        • Wang Y.
        • Stanley P.
        • Irvine K.D.
        • Haltiwanger R.S.
        • Vogt T.F.
        Nature. 2000; 406: 369-375
        • Cantor C.R.
        • Schimmel P.R.
        Biophysical Chemistry: The Conformation of Macromolecules. W. H. Freeman, San Francisco1980 (First Edition)
        • Park L.
        • Kuhlenschmidt M.
        • Roseman S.
        Fed. Proc. 1983; 42 (abstr.): 2129
        • Roseman S.
        Moscona A.A. Cell Surface in Development. John Wiley & Sons, Inc., New York1974: 255-271
        • Roseman S.
        Neuroscience, Study Program.
        in: Schmitt F.O. 3rd Ed. MIT Press, Cambridge, MA1974: 795-804