Heparan sulfate 3-O-sulfation: A rare modification in search of a function

Abstract

2. Introduction

Heparan sulfate is a type of sulfated glycosaminoglycan found covalently attached to a small set of extracellular matrix and plasma membrane proteoglycans. Its origin is ancient, having emerged during metazoan evolution, and its utility is evident, as only minor changes in composition have occurred over more than 500 million years of evolution (Medeiros et al., 2000). Many so-called heparin-binding proteins are known, many of which bind to heparan sulfate under physiological conditions and modulate cell division and differentiation, tissue morphogenesis and architecture, and organismal physiology (Bishop et al., 2007; Ori et al., 2008; Xu and Esko, 2014). Binding to heparan sulfate can have many effects on the protein ligand, ranging from simple presentation and/or stabilization to induction of conformational change, receptor-ligand interactions and protein oligomerization as a prelude to signaling. Thus, much interest exists in understanding the rules that guide the selective engagement of proteins with heparan sulfate chains.

Heparan sulfate is a linear polysaccharide composed of alternating glucosamine and uronic acid residues (Fig. 1) (Esko and Selleck, 2002). During polymerization of the chains, several classes of sulfotransferases install sulfate groups at various positions, including C2 of the uronic acid and N-, C6 and C3 of the glucosamine units. These reactions occur substoichiometrically in segments of variable size along the chain resulting in highly heterogeneous products with variable sulfation. The addition of the 3-O-sulfate group to glucosamine units is a relatively rare modification, present in only a limited number of chains or absent entirely (de Agostini et al., 2008; Marcum et al., 1986a; Pejler et al., 1987a). It is also one of the last modifications in biosynthesis (Zhang et al., 2001a; Zhang et al., 2001b), meaning that the substrates for the 3-O-sulfotransferases (Hs3sts) are sulfated oligosaccharides that have already been modified at other positions by the N-, 2-O- and 6-O-sulfotransferases and by the C5 epimerase (Kusche et al., 1988).

Open in a separate window

Figure 1

Structure of the antithrombin-binding pentasaccharide found in heparin

The individual residues of the pentasaccharide are numbered relative to the N-sulfo-glucosamine-3-sulfate residue at position 0. The central 3-O-sulfate group generated by Hs3st-1 (shown in red) and the 6-O-sulfate group at residue −2 (shown in blue) account for the majority of the binding energy of antithrombin to heparin. Note that the 6-O-sulfate group on the 3-O-sulfated glucosamine residue is dispensable and that residue −2 can be N-sulfated.

Binding to heparan sulfate depends on complementarity between positively charged amino acids in the protein ligands and the negatively charged sulfate groups and uronic acid epimers in the polysaccharide chain (Lindahl and Li, 2009; Xu and Esko, 2014). In general, two types of interactions occur, those that depend on overall sulfation and those based on specific types or arrangements of sulfated residues and uronic acid epimers. The former group consists of ligands that can accommodate different arrangements of sulfated sugars, presumably because of conformational flexibility in the binding site that can fit different orientations of sulfate and carboxyl groups in the chains. The latter group consists of ligands that rely on specific subsets of sulfate groups or the spacing of sulfated domains, and ligands that depend on a specific sequence of sulfated sugar residues (some containing a 3-O-sulfated glucosamine) for optimal binding. The paucity of known proteins that are influenced by 3-O-sulfation is rather surprising given the large family of Hs3sts involved in heparan sulfate biosynthesis. This review focuses on 3-O-sulfation, in particular the origin, substrate specificity, protein structure, and expression pattern of the Hs3st family, and the small family of proteins described to date whose activity is influenced by 3-O-sulfation.

3. Prevalence of 3-O-sulfation

The prevalence of 3-O-sulfation in natural heparan sulfates is largely unknown. Reasons for this lack of information include difficulty in obtaining large quantities of heparan sulfate for structural analyses and lack of technology to quantitate the 3-O-sulfate group. Historically, the 3-O-sulfate group was discovered by while searching for enzymes that remove sulfate groups from heparin. An enzyme in the urine was discovered that released the 3-O-sulfate group from synthetic radioactive N-sulfoglucosamine-3-sulfate (Leder, 1980) and from the non-reducing end of heparin fragments (Lindahl et al., 1980). We now know this enzyme as arylsulfatase G and its deficiency results in a lysosomal storage disease in mice in which heparan sulfate accumulates (Kowalewski et al., 2012). Since Leder’s initial study, chemical analyses, NMR and mass spectrometry have confirmed the presence of 3-O-sulfate groups in heparin (Lindahl et al., 1980; Meyer et al., 1981; Yamada et al., 1993) and in some preparations of heparan sulfate (Edge and Spiro, 1990; Marcum et al., 1986a; Pejler et al., 1987a). These latter studies indicate that the prevalence of 3-O-sulfation varies based on the source of heparan sulfate. Heparan sulfate from endothelial cells contains about one 3-O-sulfate group per 100 disaccharides (Marcum et al., 1986a). Five to ten percent of disaccharides from heparan sulfate derived from Reichert’s basement membrane contains 3-O-sulfate (Pejler et al., 1987a), whereas basement membrane heparan sulfate from Engelbreth-Holm-Swarm mouse tumor does not contain any 3-O-sulfate (Pejler et al., 1987a). Heparan sulfate from follicular fluid contains about 6 percent 3-O-sulfated glucosamine units (de Agostini et al., 2008). Certain animal species produce copious amounts of 3-O-sulfated heparin. For example, the clam, Anomalocardia brasiliana, produces heparin with one 3-O-sulfate for every 5 disaccharides (Pejler et al., 1987b). Many tissues express one or more isozyme of Hs3st, suggesting the presence of 3-O-sulfate groups, but very few of these tissues have been analyzed structurally to date.

4. Sites of modification by Hs3sts

The Hs3sts represent the largest gene family among all heparan sulfate sulfotransferases (Liu and Pedersen, 2007). Vertebrates generally have seven isozymes of Hs3st divided into two subgroups based on the homology of the sulfotransferase domain (Liu and Pedersen, 2007). Zebrafish has one additional Hs3st. Hs3st−2, −3a, −3b, −4 and −6 form one subgroup, sharing greater than 80 percent sequence identity in the sulfotransferase domain (Lawrence et al., 2007). This group is often referred to as “gD-type” Hs3sts because all members of the subfamily can generate binding sites for glycoprotein gD of Type I Herpes simplex virus (O'Donnell C et al., 2006; Shukla et al., 1999; Tiwari et al., 2005; Xia et al., 2002; Xu et al., 2005). Hs3st-1 and −5 form the other subgroup, sharing 71 percent identity in the sulfotransferase domain (Xia et al., 2002). These two sulfotransferases have in common the capacity to generate a binding site for antithrombin and thus are designated “AT-type” sulfotransferases. In vertebrates, the AT-type and gD-type subgroups share about 60 percent amino acid identity in the sulfotransferase domain (Lawrence et al., 2007).

Based on the large number of Hs3sts and the observation that they probably act after other sulfotransferases, one might assume that they show selectivity for substrates. Indeed, Hs3st-1 preferentially modifies sites in which a glucuronic acid devoid of 2-O-sulfate resides to the non-reducing side of the target glucosamine unit (Position −1, Fig 1) (Table 1) (Liu et al., 1996; Mochizuki et al., 2008; Shworak et al., 1997; Zhang et al., 1999). The enzyme will tolerate iduronic acid in this position, but 2-O-sulfation specifically prevents its action (Nguyen et al., 2012; Zhang et al., 2001b). In contrast, Hs3st−2, −3, −4 and −6 (gD-types) preferentially modify sites in which position −1 is 2-sulfo-iduronic acid (Lawrence et al., 2007; Liu et al., 1999a; Meissen et al., 2009; Mochizuki et al., 2008; Wu et al., 2004; Xu et al., 2005). Hs3st-5 modifies sites irrespective of 2-O-sulfation and consequently can produce both AT and gD type modifications (Chen et al., 2003; Chen and Liu, 2005; Duncan et al., 2004; Liu et al., 1999a; Mochizuki et al., 2003; Mochizuki et al., 2008; Xia et al., 2002; Xu et al., 2008).

Heparin lyases, which depolymerize heparin and heparan sulfate into disaccharides by beta-eliminative cleavage of the hexosaminic linkages, have the interesting property of generating “resistant” tetrasaccharide products bearing 3-O-sulfated glucosamine at the reducing end (Shriver et al., 2000; Yamada et al., 1993). A recent study of heparin lyase II showed that the resistance to digestion of the bond between positions −2 and −1 appears to arise from failure to bind the substrate caused by steric hindrance due to the proximity of Asn405 in the active site of the enzyme and the 3-OH group of the bound glucosamine residue (Zhao et al., 2011). Three tetrasaccharides produced by lyase treatment of heparan sulfate modified by Hs3st-1 consist of D0A6-G0S3, D0A6-G0S9 and D0A0-G0M3 (Zhang et al., 1999)¹. The tetrasaccharide structures produced by Hs3st-3 have been identified as D2S0-I2H3 and D2S0-I2H9, suggesting that Hs3st-3 modifies N-unsubstituted glucosamines (Liu et al., 1999b). This conclusion is not without controversy. The evidence suggesting an N-unsubstituted glucosamine derives from studies using nitrous acid to depolymerize the chains. At pH 4, nitrous acid treatment results in cleavage only at N-unsubstituted glucosamine residues (Shively and Conrad, 1976). Thus, the sensitivity of the tetrasaccharide to pH 4 nitrous acid suggested that the 3-O-sulfate group was located on an N-unsubstituted glucosamine (Liu et al., 1999b). A later study of Hs3st-3-modified octasaccharides enriched for gD binding came to a similar conclusion (Liu et al., 2002). However, in the original study the tetrasaccharide was also ar sensitive to low pH nitrous acid, which only reacts with N- sulfoglucosamine (Liu, Shriver et al. 1999), suggesting that the pH selectivity of nitrous acid might be altered by 3-O-sulfation (Liu et al., 1999b). Recent NMR experiments identified a persistent hydrogen bond within the AT-pentasaccharide between the internal glucosamine sulfamate NH and the adjacent 3-O-sulfate group, suggesting that the 3-O sulfation may confer certain structural constraints that could affect chemical and enzymatic sensitivity (Langeslay et al., 2012). Recently, we observed that resistant tetrasaccharides derived from CHO cells expressing Hs3st-3 consisted of structures containing GlcNS3S without evidence of N-unsubstituted 3-O-sulfated glucosamine (R.L. and J.D.E., unpublished results). Resistant tetrasaccharides have also been observed in heparan sulfate modified in vitro by Hs3st −2, −4 and −5 (Lawrence et al., 2007; Mochizuki et al., 2008). Limited analytical work on hexasaccharides derived from heparan sulfate modified by Hs3st-4 following partial digest with heparin lyases suggested two structures, D0A0-G0S0-I2S3 and D0A0-G0S0-I2S9 (Wu et al., 2004). Thus, the idea that 3-O-sulfation can occur on unsubstituted glucosamine residues may be incorrect.

Analysis of resistant tetrasaccharides undoubtedly oversimplifies the complexity of domains bearing 3-O-sulfate. Because of the cleavage pattern of the lyases, all Hs3st-modified oligosaccharides studied thus far have the 3-O-sulfate group at the reducing end of the fragment. Thus, there is little information about the oligosaccharide sequence towards the reducing side of the 3-O-sulfation site (i.e. at the +1, +2 and +3 positions). To investigate this question, Hs3st-3 activity has been measured against libraries of defined oligosaccharides (Nguyen et al., 2012). These studies showed that active substrates contained an iduronic acid residue to the reducing side of the acceptor, whereas substrates containing glucuronic acid and 2-sulfo-iduronic acid at this position were inactive. In this respect, the specificity of Hs3st-3 differs from that of Hs3st-1, which permits a 2-sulfo-iduronic acid at the +1 position (Fig. 1). More extensive oligosaccharide libraries are needed to better define the acceptor specificity of the Hs3sts. Recently, Kowalewski et al. reported that arylsulfatase G catalyzes the removal of the 3-O-sulfate groups during lysosomal degradation (Kowalewski et al., 2012). Analysis of the non-reducing ends of the chains that accumulate in the liver showed that heparinase digestion released a significant amount of acetylated trisaccharide with the structure S3-U0A0. Thus the chains that accumulate in different tissues from the mutant could be used to obtain additional information about the structure on the reducing side of the 3-O-sulfated unit.

The possibility that the Hs3sts might also exhibit some promiscuity should also be considered. Thus, under some conditions, a gD-type 3-O-sulfotransferase might create AT-type heparan sulfate and vice versa. Glomerular epithelial cells produce AT-type heparan sulfate and express Hs3st-1 and-3, but none of the other Hs3sts, including Hs3st-5. Nevertheless, when the cells were derived from Hs3st-1^−/− mice, low levels of antithrombin-binding heparan sulfate were detected (Girardin et al., 2005). Similarly, small amounts of AT-binding heparan sulfate were produced using recombinant Hs3st-3 in vitro (Girardin et al., 2005).

5. Insights into substrate specificity from structural studies of Hs3sts

X-ray crystal structures have been solved for the sulfotransferase domains of Hs3st-1, −3 and −5 (Edavettal et al., 2004; Moon et al., 2004; Xu et al., 2008). These structures, accompanied by extensive mutational studies, form the basis of our understanding of the working mechanism of this enzyme family and provide insights into the substrate specificity of the enzymes. The overall fold of the Hs3st sulfotransferase domain is very similar to the sulfotransferase domain of N-deacetylase-N-sulfotransferases (Ndst), consistent with ~30 percent sequence identity of these enzymes (Edavettal et al., 2004). All of the sulfotransferases utilize the high-energy sulfate donor, 3'-phosphoadenosine-5'-phosphosulfate (PAPS). Thus, not surprisingly Hs3st and Ndst1 share almost identical structure in the binding site for PAPS, and they both use a conserved glutamate (E184 in Hs3st-3, Fig. 2A) as the catalytic base (Edavettal et al., 2004; Kakuta et al., 2003; Kakuta et al., 1999). The major difference between Hs3st and Ndst is in the heparan sulfate-binding cleft, where significantly more positively charged amino acid residues are present in Hs3st. The difference likely reflects the distinct substrate selectivity of Hs3st and Ndst; Ndsts act on N-acetylated domains of the chains and initiate polymer modification, whereas Hs3sts usually act late requiring a pattern of sulfate groups in the target site (Edavettal et al., 2004).

Open in a separate window

Figure 2

Structural interactions of 3-O-sulfotransferases with oligosaccharide substrates

(A) Stereo diagram of the interaction between Hs3st-3 and the tetrasaccharide D2S6-I2S6 (PDB code 1T8U). Side chains of critical amino acid residues that mediate substrate binding are shown as thick grey sticks. Possible hydrogen bonds between side chains and tetrasaccharide are pictured as black broken lines. The C3-OH group of the glucosamine residue to be sulfated is shown as a red sphere. The sulfate donor mimic, 3’-phosphoadenosine 5’-phosphate (PAP) is shown, and the protein backbone is shown as a cartoon in the background. (B) Superposition of Hs3st-3 bound tetrasaccharide (D2S6-I2S6; green, PDB code 1T8U) and heptasaccharide (A6-G0S6-I2S6-G0M0; white, PDB code 3UAN). The surface of Hs3st-1 is rendered in gray. Residues in tetrasaccharides are labeled as T-1 to T-4. Residues in the heptasaccharide are labeled as H-1 to H-6 (the anhydromannose residue is not shown). (C) Role of K259 in Hs3st-3 substrate recognition. The tetrasaccharide and heptasaccharide are colored and numbered as in (B). Putative hydrogen bonds between K259 and the 2-O-sulfate group of T-1 and the carboxylate of T-3 are shown in black dashed lines. The distance between the 2-O-sulfate of T-1 and carboxylate of T-3 are shown as a yellow broken line. The distance between the 2-O-sulfate of T-1 and the carboxylate of H-4 are shown as a red broken line. (D-F) Role of gate residues in substrate recognition. The gate residues are shown as large spheres and labeled in white, and the tetrasaccharide is shown with the C3-OH group shown as a red sphere. The molecular surface of each enzyme is shown in grey. The distance between gates residues is pictured as black broken lines. The tetrasaccharide is modeled into Hs3st-1 (PDB code 1VKJ) and Hs3st-5 (PDB code 3BD9) by superimposing their structures on that of the Hs3st-3/tetrasaccharide complex (PDB code 1T8U) using PyMol (Schrödinger).

Perhaps the most interesting information to glean from structural studies of the Hs3sts is how the isozymes distinguish subtle differences in the target substrates, which in turn allows the enzymes to generate distinct products. Co-crystallization of Hs3st-3, 3’-phosphoadenosine-5’-phosphate (PAP) and a tetrasaccharide substrate has revealed how enzyme structure can influence substrate specificity (Moon et al., 2004). An extensive hydrogen-bonding network was observed in the substrate-binding cleft involving K161, R166, K215, Q255, K368 and R370, the acceptor glucosamine (sulfation-site, sugar residue T-2 in the tetrasaccharide substrate, Fig. 2A) and neighboring uronic acids (T-1 and T-3, Fig. 2A). All six residues are completely conserved in all Hs3st family members and across species. Ndst1 lacks three of the six residues, K161, R166 and Q255, suggesting that they contribute to the recognition of substrates unique to Hs3sts. However, the conservation of these residues across the Hs3st isozymes suggests that they do not dictate substrate specificity across the Hs3sts.

A comparison of Hs3st-1 and Hs3st-3 co-crystal structures with oligosaccharide substrates revealed that the substrates orient differently in the two enzymes (Moon et al., 2012). The orientation of the acceptor glucosamine (position 0, T-2 and H-3 in the hexasaccharide substrate, Fig. 2B) and non-reducing end uronic acid (position −1, T-1 and H-2) were nearly identical in the two crystal structures, but the 2-sulfo-iduronic acid unit at the reducing side of the acceptor glucosamine (position +1, T-3 and H-4) was found in a chair conformation in Hs3st-1 (H-4, Fig 2B), whereas in Hs3st-3 it was found in a skew boat conformation (T-3, Fig 2B). The chair conformation preferred by Hs3st-1 approximates the carboxyl group of 2-sulfo-iduronic acid unit (H-4) and the C2-OH of the uronic acid on the non-reducing side of the acceptor glucosamine (H-2). If the iduronic acid H-2 in the heptasaccharide were 2-O-sulfated, charge repulsion would result between the sulfate and carboxylate groups due to their close proximity (Fig. 2C, red dashed line indicates a distance of only 2.8 Å). In contrast, Hs3st-3 induces a skew boat conformation of the 2-sulfo-iduronic acid T-3 (Fig. 2C), which provides more distance between the carboxylate and the sulfate groups (Fig. 2C, yellow dashed line indicates a distance of 3.2 Å). Importantly, Hs3st-3 also places a lysine residue (K259) between the 2-O-sulfate group of residue T-1 and the carboxylate group of residue T-3, which would help neutralize the negative charges and thus prevent charge repulsion (Fig 3C). \K259 is conserved among all gD-type Hs3sts, suggesting that it might also participate in other gD-type Hs3sts in the selection of substrates containing 2-sulfo-iduronic acid (Lawrence et al., 2007; Liu et al., 1999a; Mochizuki et al., 2008; Wu et al., 2004; Xu et al., 2005). Consistent with this idea, Hs3st-1 contains an uncharged asparagine residue in place of K259 (N167, Fig. 2C), which might explain why Hs3st-1 will not act on substrates bearing 2-sulfo-iduronic acid at the −1 position, but will tolerate either glucuronic or iduronic acid (Mochizuki et al., 2008; Zhang et al., 2001b; Zhang et al., 1999).

Open in a separate window

Figure 3

Phylogenetic analysis of Hs3sts

The number of Hs3st isozymes identified in each species was deduced from existing publications, NCBI protein database searches and comparison of key amino acids that define substrate and catalytic sites. The phylogenetic tree depicts their evolutionary relationship (adapted from (Medeiros et al., 2000)). The proposed origin of Hs3st is indicated with an arrow.

When the crystal structures of Hs3st-1, −3 and −5 were superimposed, it was observed that two residues in Hs3st-1, H271 and E88, form a narrow opening that would contact the polysaccharide substrate at multiple sites including the two sugar residues to the non-reducing side of the sulfation site (+2 and +3 positions) (Fig. 2D) (Moon et al., 2012; Xu et al., 2008). The corresponding residues in Hs3st-3 (G182 and G365) and Hs3st-5 (S120 and A306) are much smaller and form a significantly wider opening (Fig. 2E, F). The nature of these “gate” residues seems to regulate the substrate specificities of Hs3st-1 and Hs3st-5 in a substantial way. Mutation of these residues to glycine in Hs3st-1, or to histidine and glutamic acid in Hs3st-5, had little or no effect on catalytic activity, but had a profound effect on substrate specificity. Remarkably, interchanging these two residues in Hs3st-1 and Hs3st-5 was sufficient to convert the substrate specificity of one enzyme into another.

Based on these studies, it is clear that substrate specificity is determined by amino acids distal to the catalytic site. Also, the evidence strongly suggests that Hs3sts recognize at least a pentasaccharide motif spanning from one sugar unit to the reducing end of the sulfation site, to three sugar units to the non-reducing side. Larger oligosaccharides may confer additional specificity. The diversification of the Hs3st family suggests that there may be distinct substrate specificities even within a subgroup of Hs3sts.

6. Evolutionary origin of the Hs3sts

As mentioned above, Homo and other mammals express seven Hs3sts, whereas Danio (zebrafish) expresses eight. In comparison to vertebrates, Hs3st isozymes in invertebrates are much less numerous. Drosophila (fruit fly) and Caenorhabditis (nematode) both have one gD and one AT-type transferase (Kamimura et al., 2004; Tecle et al., 2013), whereas Strongylocentrotus (sea urchin) and Planaria (flatworm) have only a single Hs3st related most closely to an AT-type transferase (Fig. 3). The Nematosetella (sea anemone), a member of the Cnidarian phylum, has two Hs3sts whereas Hydra has only one, but all three share ~50 percent amino acid sequence identity to human Hs3sts. In contrast, Trichoplax, a placozoan generally considered as a sister clade to bilaterians and Cnidarians, do not have any Hs3st homologs, but homologs of all other heparan sulfate sulfotransferases appear to be present. Porifera (sponges) do not express any homologs of the sulfotransferases, suggesting that sponges lack heparan sulfate. Based on this information, we propose that the primordial Hs3st originated early in eumetazoan evolution (excluding Porifera), probably first emerging in a common ancestor of bilaterians and cnidarians (Fig. 3).

Although Hs3sts appear to be widely distributed in nature, demonstration of 3-O-sulfate groups in heparan sulfate derived from non-vertebrate species other than Anomalocardia brasiliana (clam) is lacking. Attreed et al. (Attreed et al., 2012) showed that the single chain antibody (HS4C3), which reacts with a 3-O-sulfated determinant containing 2-O- and 6-O-sulfate groups in heparan sulfate and heparin (Ten Dam et al., 2006), binds to specific cellular targets in C. elegans, suggesting the presence of 3-O-sulfated motifs. We have detected 3-O-sulfated disaccharides and tetrasaccharides in heparan sulfate from zebrafish embryos. These structures were diminished by specific Hs3st morpholinos. Furthermore, expression of cDNAs encoding the different zebrafish Hs3sts in CHO cells results in expression of both AT- and gD-type heparan sulfate. (R.L., H. Joseph Yost, Adam Cadwallader and J.D.E, unpublished findings). The context of the 3-O-sulfate groups in these organisms remains unknown.

7. Expression patterns of the Hs3sts

The expression of Hs3st genes is exquisitely controlled in a spatiotemporal manner in vertebrates, befitting a family of seven isozymes (Table 2). Human HS3ST-1, −3a and −3b transcripts are widely expressed in many organs, whereas expression of HS3ST-2 and −4 has been primarily detected in the brain (Lawrence et al., 2007; Mochizuki et al., 2008; Shworak et al., 1999). HS3ST-5 is expressed in skeletal muscle (Xia et al., 2002). In the mouse, Hs3st-6 is expressed predominantly in the liver and kidney with lower expression in the heart, brain, lung and testis (Xu et al., 2005). At least one Hs3st gene is expressed in nearly every cell line examined thus far and many express multiple Hs3st genes simultaneously (Deligny et al., 2010; Girardin et al., 2005; Vanpouille et al., 2007). Notable exceptions are Chinese hamster ovary cell line and Engelbreth-Holm-Swarm mouse tumor, which synthesize heparan sulfate devoid of 3-O-sulfation (Pejler et al., 1987a; Zhang et al., 2001b).

Particular attention has been focused on Hs3st expression in the nervous system. Several Hs3st genes are spatially regulated in the mouse cerebrum and cerebellum throughout development (Yabe et al., 2005). Whole mount in situ hybridization demonstrated unique expression patterns of each of the Hs3st genes in the Zebrafish brain (Cadwallader and Yost, 2006). Studies of a transgenic mouse, in which the human placental alkaline phosphatase (hPLAP) was inserted next to the start codon of Hs3st-2, showed expression in trigeminal and dorsal root ganglia during development (Hasegawa and Wang, 2008). Expression of Hs3st-2 and −4 appeared associated with a subset of neuronal cells within the trigeminal ganglia (Lawrence et al., 2007). In D. melanogaster and C. elegans, the two Hs3st genes are expressed in specific cells, including small subsets of neurons (Kamimura et al., 2004; Tecle et al., 2013). The spatially and temporally restricted pattern of expression of Hs3st genes in the nervous system suggests a neurological role for 3-O-sulfation. A mild neuronal patterning defect in C. elegans was seen after loss of function of the enzymes (discussed in section 8.6), but to date no such role has been demonstrated in other organisms.

There is a notable lack of information about the factors that regulate the expression of the Hs3st genes (e.g. transcription factors, promoter and enhancer sequences, etc.). This should be a fruitful area of investigation based on the differential expression of the isozymes across different tissues and during development and the observation that external stimuli can affect the expression of the Hs3st genes. For example, Hs3st-2 expression and 3-O-sulfation in the pineal gland varies with Circadian rhythm and is inducible by light (Borjigin et al., 2003; Kuberan et al., 2004). In intestinal epithelium, Hs3st-1 expression increases in response to IL-4 and IL-13 (Takeda et al., 2010). In contrast, stimulation of human microvascular endothelial cells with TNF-a or LPS diminishes expression of HS3ST-1 mRNA (Krenn et al., 2008). Finally, Hs3st-3 was upregulated in urothelium in response to IL-6 (Wood et al., 2011). In most of these examples, it remains unknown if sulfotransferase protein levels, enzyme activity or heparan sulfate structure were also affected. When F9 mouse embryonal carcinoma cells were differentiated into parietal endoderm using retinoic acid, cAMP and theophylline, the expression of Hs3st1 increased more than 100-fold with a corresponding increase in production of heparan sulfate that binds to antithrombin (Zhang et al., 1998).

8. Biochemical and physiological implications of 3-O-sulfation

Hundreds of heparan sulfate-binding proteins have been identified and much effort has been expended to determine the structural features in heparan sulfate that mediate binding (Xu and Esko, 2014). Most studies have overlooked the effect of 3-O-sulfation on binding, generally because of the lack of available material for study. For this reason only six proteins have been demonstrated biochemically to be affected by 3-O-sulfation.

8.1 Antithrombin

Antithrombin is the best characterized protein in which 3-O sulfate groups play a crucial role in binding to heparin/heparan sulfate and it serves as the paradigm for how one can unravel the structural context of relevant, biologically active 3-O-sulfated glucosamine residues (Lindahl et al., 1980). Determining the location of these rare groups in a sea of differentially sulfated domains is a formidable problem, compounded by the fact that obtaining multimilligram amounts of heparan sulfate from any source is quite difficult and costly. The analysis of the antithrombin binding site was greatly simplified because of its enrichment in heparin, which is available in kilogram quantities (Atha et al., 1985; Höök et al., 1976; Hopwood et al., 1976; Lindahl et al., 1979; Lindahl et al., 1983; Lindahl et al., 1980; Riesenfeld et al., 1981). Thus, it was possible to partially cleave the chains into smaller oligosaccharides and to obtain sufficient material for separating fragments into high and low affinity fractions, disaccharide analysis, chemical degradation and NMR studies (Yamada et al., 1993). The antithrombin binding motif was subsequently identified in heparan sulfate expressed by endothelial cells, fibroblasts and in other tissues (De Agostini et al., 1990; Marcum et al., 1986a; Marcum et al., 1986b; Marcum et al., 1983; Marcum and Rosenberg, 1985). It should be pointed out that all of these studies focused on sequences that bind to antithrombin, but in fact heparan sulfate chains contain 3-O-sulfate groups at sites that do not bind with high affinity to antithrombin. In the absence of a ligand like antithrombin, the nature of many of these sites remains uncharacterized.

Antithrombin is an inhibitor of the proteases thrombin, factor IXa and factor Xa and plays a central role in hemostasis. It also plays less appreciated roles as an inhibitor of inflammation and angiogenesis (O'Reilly et al., 1999; Wiedermann Ch and Romisch, 2002). Heparin binding to antithrombin induces a conformational change that increases its catalytic activity by several orders of magnitude (Huntington et al., 1996; Rosenberg and Damus, 1973). Indeed, heparin is routinely used as an anticoagulant in the clinic based on its ability to activate antithrombin.

Much effort went into identifying the minimum active heparin structure, which turns out to be a pentasaccharide with the structure A6-G0S9-I2S6 or A6-G0S3-I2S6 bearing a critical 3-O-sulfate on the middle N-sulfo-glucosamine residue (Fig. 1) (Choay et al., 1983; Thunberg et al., 1982). Based on genetic studies in mice, Hs3st-1 is responsible for the overwhelming majority of antithrombin-binding structures (HajMohammadi et al., 2003). Hs3st-1 is generally thought to be a limiting factor in generating antithrombin binding sites (Zhang et al., 1998). However, in the presence of excess Hs3st-1, the precursor structures become limiting. Evidence also suggests that non-precursor saccharides can inhibit Hs3st-1, which presumably compete with the substrate for binding to the enzyme (Razi and Lindahl, 1995).

The presence of N-acetylglucosamine-6-sulfate and glucuronic acid at the −2 and −1 positions of the pentasaccharide sequence suggests that the binding site occurs in the transition zones between highly sulfated “NS” domains and unmodified “NA” domains (Lindahl et al., 1983; Zhang et al., 2001a). Biochemical studies demonstrated that the 3-O-sulfate group at position 0 and 6-O-sulfate group on the N-acetylglucosamine residue at position −2 account for more than half of the binding energy released when antithrombin interacts with the pentasaccharide (Atha et al., 1985; Atha et al., 1987). The absence of the 3-O-sulfate group decreases affinity by several orders of magnitude (K_D = 30 nM versus 500 µM) and reduces both the conformation change induced in antithrombin and the inhibition of Factor Xa (Atha et al., 1985; Atha et al., 1987). The co-crystal structures of antithrombin and bound oligosaccharide suggest that while several basic and polar residues in antithrombin make salt bridges and hydrogen bonds with the pentasaccharide, the most pivotal interaction that stabilizes the complex is the one between residue K114 and the 3-O-sulfate group (Jin et al., 1997; Li et al., 2004). This interaction orients K114 in an optimal conformation to form hydrogen bonds and salt bridges with sugar residues at position +1 and +2 (Fig. 4) (Richard et al., 2009).

Open in a separate window

Figure 4

Stereo diagram of pentasaccharide-antithrombin interaction

(PDB code: 2GD4). Heparan sulfate binding residues of antithrombin (salmon) and the pentasaccharide (grey) are shown in sticks, and the antithrombin protein backbone is shown in cartoon (salmon). Putative salt bridges and hydrogen bonds stabilized by the 3- O-sulfate group and K114 are shown as broken lines. Sugar residues in the pentasaccharide are labeled as in Figure 1.

The literature frequently describes antithrombin-binding heparin as “the pentasaccharide structure,” suggesting that only a unique pentasaccharide can bind to antithrombin. This notion is overly simplistic in that other oligosaccharides can bind and activate antithrombin. Some of these structures have higher affinity and result in more robust activation of antithrombin than the minimal pentasaccharide. These include an identical pentasaccharide with a second 3-O-sulfate group on the N-sulfo-glucosamine at position +2 or 3-O-sulfated heparan sulfate devoid of iduronic acid or 2-sulfo-iduronic acid at the + 1 position (Chen et al., 2007; Guerrini et al., 2013; Guerrini et al., 2008; Zhang et al., 2001b).

Only about one-third of typical unfractionated heparin preparations have the ability to bind and activate antithrombin (Höök et al., 1976; Lam et al., 1976). In most tissues examined to date, the percentage of heparan sulfate chains bearing AT-type sequences is quite low. For example, only five percent of heparan sulfate chains from bovine aortic endothelial cells, human umbilical vein endothelial cells and rat microvascular endothelial cells binds with high affinity to antithrombin (Kojima et al., 1992; Marcum et al., 1986a; Mertens et al., 1992). In several tissues examined in the rat, approximately 5–10 percent of heparan sulfate chains contained a high affinity-binding site (Horner, 1990). The brain and spleen have antithrombin binding sites on ~20 percent of the chains (Horner, 1990). Surprisingly, over 50 percent of heparan sulfate chains found in follicular fluid bind with high affinity and activate antithrombin (de Agostini et al., 2008). The production of anticoagulant heparan sulfate proteoglycans in granulosa cells peaks prior to ovulation, linking AT-type heparan sulfate production to the menstrual cycle (Princivalle et al., 2001). Heparin/heparan sulfate with low affinity for antithrombin also contains 3-O-sulfation, apparently in a context that does not support antithrombin activation (Shworak et al., 1997). The purpose of 3-O-sulfation catalyzed by Hs3st-1 in contexts other than antithrombin-binding sites is currently unknown.

Mutations that affect heparin binding to antithrombin underscore the importance of heparan sulfate-induced activation of antithrombin in vivo. Several mutations that interrupt the heparin-binding site of antithrombin have been identified in human patients, some with thrombotic phenotypes (Gandrille et al., 1990; Koide et al., 1984; Lane et al., 1996). Accordingly, a knock-in mouse with a point mutation (R48C) corresponding to a human variant produces antithrombin with reduced affinity for heparin (Dewerchin et al., 2003). This mouse suffers from severe spontaneous thrombosis beginning as early as birth and is unresponsive to heparin. As described above, AT-type 3-O-sulfation unquestionably endows heparan sulfate with high inhibitory activity towards Factor Xa, yet 3-O-sulfation seems dispensable in maintaining hemostatic tone in vivo based on inactivation of Hs3st-1 in the mouse (HajMohammadi et al., 2003). Surprisingly, Hs3st-1^−/− mice do not exhibit a prothrombic phenotype despite having drastically reduced levels of AT-type heparan sulfate (HajMohammadi et al., 2003). Two explanations for this finding have been proposed. First, the other 3-O-sulfotransferases (in particular Hs3st-5) may create sufficient AT-type heparan sulfate to compensate for loss of Hs3st-1. Other gD-type 3-O-sulfotransferases can create AT-type heparan sulfate as well, albeit much less efficiently than Hs3st-1 (Girardin et al., 2005; HajMohammadi et al., 2003). Second, studies have shown that suboptimal heparan sulfate structures, including those without 3-O-sulfation, may activate antithrombin sufficiently to maintain hemostasis (Nordenman et al., 1978; Richard et al., 2009; Scully et al., 1988; Streusand et al., 1995).

Antithrombin also has anti-inflammatory and anti-angiogenic properties that may be influenced by 3-O-sulfation (Wiedermann Ch and Romisch, 2002). Hs3st-1^−/− mice are hypersensitive to LPS-induced inflammation and antithrombin has a protective effect in the presence of AT-type heparan sulfate (Shworak et al., 2010). The details and mechanism of this effect have not yet been elucidated. Cleaved and latent forms of antithrombin lose their anticoagulant properties but have anti-angiogenic properties (O'Reilly et al., 1999). These forms of antithrombin retain their preference for 3-O-sulfated heparan sulfate and binding to heparan sulfate influences the anti-angiogenic activity (Schedin-Weiss et al., 2008; Zhang et al., 2005). Hs3st-1^−/− mice experience intrauterine growth retardation and spontaneous eye degeneration (HajMohammadi et al., 2003), suggesting that other undiscovered ligands may require 3-O-sulfated sequences created by Hs3st-1.

9. 3-O-Sulfation in development and disease

10. Conclusions

Any discussion of heparan sulfate-protein interactions must address the question of specificity (Kreuger et al., 2006; Lindahl and Li, 2009; Xu and Esko, 2014). Antithrombin is clearly unusual in that its biological activity critically depends on high affinity binding to a 3-O-sulfated pentasaccharide. The requirement for the 3-O-sulfate group is so great, that in its absence, the affinity of the interaction drops several orders of magnitude. Stated in another way, the affinity of the interaction is of sufficient magnitude to drive a conformational change adequate to activate antithrombin. Can we expect other ligands to show similar characteristics? Many, but not all, heparan sulfate binding proteins show an “analog” response with respect to charge, i.e. affinity and biological activity increases as the degree of sulfation increases. Thus, one could argue that the presence of a 3-O-sulfate group in the context of other modifications merely increases the local charge density providing another opportunity for non-specific interaction. However, it is difficult to explain the evolution of seven 3-O-sulfotransferases differing in substrate specificity and tissue expression to merely increase charge density. Clearly, the information emerging from genetic studies in model organisms indicates selective and crucial interactions occur between some ligands and 3-O-sulfated heparan sulfate. Future studies should focus on characterizing the physical interaction of these ligands with 3-O-sulfated oligosaccharides and expanding the repertoire of known 3-O-sulfate dependent proteins.

The large number of Hs3st genes and their complex spatial and temporal expression provides a system that has the potential to control ligand binding and signaling in multiple tissues and at different stages of development. To date very few proteins have been described whose activity is enhanced by or dependent upon 3-O sulfation. We also know little about the chemical context in which 3-O-sulfation takes place and how the biosynthetic process is orchestrated to generate the precursor structures required by the Hs3sts. To a large extent, technical issues have limited progress. First, the small amounts of heparan sulfate available from defined sources makes classical fractionation schemes, like that used to define the antithrombin-binding site in heparin, technically demanding. Second, most methods to characterize sites of 3-O-sulfation have been based on a “bottom-up” approach using degradative techniques. Third, structural studies have been limited by the lack of defined standards, which also limits characterization of the sulfotransferases. Fourth, few researchers have focused on 3-O-sulfation, probably due to its rarity and the various analytical and synthetic barriers described above.

Although this assessment of the field may appear gloomy, we believe that the field has turned a corner. New “top-down” analytical methods based on tandem mass spectrometry and ion-mobility methods hold great promise for providing much greater contextual information about the structures of 3-O-sulfation sites (Kailemia et al., 2013; Meissen et al., 2009). Chemical and chemoenzymatic methods have been recently developed that have the capacity to produce multimilligram quantities of defined oligosaccharides, providing much needed reagents to aid in the discovery of new ligands and to characterize the binding sites in the enzymes and in the protein ligands (Arungundram et al., 2009; Xu et al., 2011; Xu et al., 2012). Genetic studies of model organisms focused on 3-O-sulfation have begun to yield interesting insights into the biological function of the Hs3sts and new ligands to study. Inactivation of the genes in cells by silencing methods, TALEN technology and gene targeting in embryonic stem cells provide new tools to study the biological implications of altering 3-O-sulfation in vitro and in vivo. These advances should fuel a resurgence of interest in 3-O-sulfation and eventually lead to a more complete understanding of this relatively rare and mystifying modification.

Acknowledgements

Abbreviations

Footnotes

References