Heparin Dodecasaccharide Containing Two Antithrombin-binding Pentasaccharides

Background: Heparin is a linear sulfated polysaccharide used clinically as an anticoagulant. Results: A heparin dodecasaccharide, containing two contiguous antithrombin-binding sequences, has been described and characterized for the first time. Conclusion: The dodecasaccharide binds antithrombin in two different molecular assemblies enhancing the probability of the binding and the affinity. Significance: The discovery of this dodecasaccharide improves the knowledge of heparin structure. The antithrombin (AT) binding properties of heparin and low molecular weight heparins are strongly associated to the presence of the pentasaccharide sequence AGA*IA (ANAc,6S-GlcUA-ANS,3,6S-I2S-ANS,6S). By using the highly chemoselective depolymerization to prepare new ultra low molecular weight heparin and coupling it with the original separation techniques, it was possible to isolate a polysaccharide with a biosynthetically unexpected structure and excellent antithrombotic properties. It consisted of a dodecasaccharide containing an unsaturated uronate unit at the nonreducing end and two contiguous AT-binding sequences separated by a nonsulfated iduronate residue. This novel oligosaccharide was characterized by NMR spectroscopy, and its binding with AT was determined by fluorescence titration, NMR, and LC-MS. The dodecasaccharide displayed a significantly increased anti-FXa activity compared with those of the pentasaccharide, fondaparinux, and low molecular weight heparin enoxaparin.

The antithrombin (AT) binding properties of heparin and low molecular weight heparins are strongly associated to the presence of the pentasaccharide sequence AGA*IA (A NAc,6S -GlcUA-A NS,3,6S -I 2S -A NS,6S ). By using the highly chemoselective depolymerization to prepare new ultra low molecular weight heparin and coupling it with the original separation techniques, it was possible to isolate a polysaccharide with a biosynthetically unexpected structure and excellent antithrombotic properties. It consisted of a dodecasaccharide containing an unsaturated uronate unit at the nonreducing end and two contiguous AT-binding sequences separated by a nonsulfated iduronate residue. This novel oligosaccharide was characterized by NMR spectroscopy, and its binding with AT was determined by fluorescence titration, NMR, and LC-MS. The dodecasaccharide displayed a significantly increased anti-FXa activity compared with those of the pentasaccharide, fondaparinux, and low molecular weight heparin enoxaparin.
Heparin is a linear sulfated polysaccharide belonging to the glycosaminoglycan family and has been used clinically as an anticoagulant for many years, and its worldwide consumption has increased to 100 tons per year (1). Over the last few decades, heparin was shown to be involved in many other biological processes through interaction with a large number of proteins (2). One of the main molecular bases for its anticoagulant activity was elucidated about 30 years ago, when a specific pentasaccharide sequence within heparin chains was identified as responsible for binding and activating antithrombin (3). Other mechanisms such as tissue factor pathway inhibitor release also contribute to the anticoagulant properties of heparin (4). Heparin is a complex mixture containing linear chains of repeating disaccharide units composed of glucosamine and uronic acid residues. Although the main repeating disaccharide unit in heparin consists of (134) ␣-D-6-O-N-sulfoglucosamine (134)-linked to a ␣-L-2-O-sulfated iduronic acid, short undersulfated domains as well as more complex sequences are also present in the chains. Both the 2-O position of iduronic acid and the 6-O position of glucosamine can be desulfated; glucosamine can be N-acetylated instead of N-sulfated, and iduronic acid can be replaced by glucuronic acid (5). The rare 3-O-sulfated glucosamine (A*) 3 is present, almost exclusively, in the unique pentasaccharide domain (AGA*IA), corresponding to the antithrombin-binding sequence (6). The biosynthesis of heparin and heparin sulfate takes place in the Golgi network and is initiated by glycosylation reactions that generate saccharide sequences composed of alternating ␤-D-glucuronic acid and A NAc units. In close association with chain elongation, a series of reactions involving N-deacetylation/N-sulfation of A NAc residues, epimerization of ␤-D-glucuronic acid to IdoUA, and O-sulfation at different positions occurred to generate the final heparin polymer (7). Although earlier studies proposed that incorporation of the 3-O-sulfate group concludes the biosynthetic process, further research has demonstrated that several isoforms of 3-O-sulfotransferase may transfer a sulfo group to different positions of the chain depending on the particular tissues and species (8,9). Although some studies supported the presence of more than one AGA*IA sequence in a single hepa-rin chain, no oligosaccharides bearing more than one AGA*IA have been isolated from heparins until now (10,11).
A major breakthrough in the study of structural motifs responsible for AT interactions was in the separation and purification of distinct heparin fractions, differing markedly in their affinity for AT. AT-binding oligosaccharides are usually obtained after partial chemical or enzymatic depolymerization of heparin and purification using affinity chromatography on an immobilized Sepharose-AT column (12). In other studies, the AT-binding oligosaccharides were isolated from LMWHs, such as enoxaparin, using a combination of orthogonal separation techniques, including gel permeation chromatography, HPLC, and AT affinity chromatography (13).
These studies demonstrated that the AGA*IA sequence can be flanked by both sulfated and undersulfated disaccharides and that these elongating sequences can have an active role in binding AT (14,15). Notably, the presence of glucuronic acid instead of iduronic acid before the pentasaccharide moiety increases the affinity to AT by 1 order of magnitude, whereas the presence of 1,6-anhydrohexosamine at the reducing side of the pentasaccharide significantly reduces the affinity (15,16). Recently, an octasaccharide having a pentasaccharide sequence at the nonreducing end with two 3-O-sulfated glucosamine residues has been characterized (17). Fluorescence titration experiments indicated that the AT affinity of this octasaccharide was characterized by an equilibrium dissociation constant 20-fold lower than that measured for the AGA*IA-AT complex.
The type of depolymerization process applied to complex heparin chains may thus generate many structural variants, including cleavage within the pentasaccharide sequence, that strongly influence the AT binding properties. A depolymerization procedure able to preserve the AT-binding sequences is the main condition to achieve an ideal and cost-effective LMWH. Because of their lower molecular size compared with their parent heparins, LMWHs have improved bioavailability and higher anti-Factor-Xa (anti-FXa) activity than AT-mediated anti-Factor-IIa (anti-FIIa), with decreased hemorrhagic risk during prolonged treatment (18).
Particularly, the highly selective depolymerization of heparin by the phosphazene base used for preparation of the ULMWH semuloparin protects the native antithrombin-binding sequence from cleavage. This was demonstrated by the anti-FXa activity of semuloparin, which, in contrast to other LMWHs, remains substantially comparable with that of unfractionated heparin despite a 6-fold size reduction of the polysaccharide chain (19,20).
In this work, a novel dodecasaccharide containing two contiguous AT-binding sequences was isolated from semuloparin. In agreement with the method of depolymerization used, the dodecasaccharide terminates at the nonreducing end with a 4,5-unsaturated uronic acid residue (⌬U). The two AGA*IA moieties, both with the first glucosamine N-acetylated and 6-O-sulfated, are separated by a nonsulfated iduronic acid residue. The structure of the dodecasaccharide was unambiguously defined by controlled enzymatic digestion and NMR spectroscopy.
Mapping of the binding region involved in the interaction with AT was determined by saturation transfer difference (STD) experiments. The conformational and AT binding properties of this dodecasaccharide were also investigated by NMR spectroscopy (NOESY and transferred-NOESY experiments). The presence of two different 1:1 AT-dodecasaccharide complexes was supported by NMR and modeling results and finally confirmed by LC-MS spectrometry. The dodecasaccharide showed significantly higher affinity to AT as well as an increased anti-FXa activity when compared with the pentasaccharide fondaparinux and LMWH enoxaparin.
This procedure yields ULMWH products with an average molecular mass of ϳ2400 Da, anti-FXa of ϳ160 units/mg, and anti-Factor IIa of ϳ2 units/mg (21). All experimental details regarding the ULMWH preparation are reported in example 2 of United States Patent 8,003,623 B2 (22).
Procedure for Dodecasaccharide Isolation and Purification-Semuloparin is fractionated by gel permeation chromatography on a column packed with BioGel P30 (200 ϫ 5 cm) (2 g per injection). The dodecasaccharide fraction was collected and then desalted on a column packed with Sephadex G-10 (100 ϫ 7 cm). In a second step, 200 mg of the original fraction (5 g) was injected onto an AT affinity column (30 ϫ 5 cm) using a step gradient of NaCl concentration (0.7, 1.15, 1.6, 2.05, 2.65, and 3 M NaCl in 1 mM Tris-HCl, pH 7.4). The fractions eluted for conductivities between 80 and 130 millisiemens/cm, which are enriched in dodecasaccharide, are pooled (Fig. 2). These fractions were desalted and purified on semi-preparative CTA-SAX columns (250 ϫ 22 mm), and each desalted fraction was tested by analytical AS11 or CTA-SAX. After pH neutralization, the collected fractions were passed through a Mega Bond Elute C18 cartridge (Varian) before the final purification on a semi-preparative AS11 column (Dionex). The fractions containing the pure dodecasaccharide were pooled, desalted on Sephadex G-10, and freeze-dried.
Enzymatic Sequencing of the Dodecasaccharide-The sequencing of the dodecasaccharide was performed by a combination of controlled and exhaustive cleavage with heparin lyases and HPLC analysis of fragments. To obtain a unique form of the glucosamine at the reducing end, a preliminary reducing step was performed before sequencing experiments. Briefly, 15 l of a 150 g/liter NaBH 4 solution in 100 mM sodium acetate was added to 1 ml of a 1-2 g/liter aqueous solution of the dode-casaccharide. After 2 h, the remaining NaBH 4 was neutralized by perchloric acid to obtain a pH between 5 and 6. The solution was then desalted on Sephadex G-10. The structure was determined using a controlled digestion with heparin lyase I (heparinase I, EC 4.2.2.7) followed by HPLC analysis according to the procedure described previously (13). In the reduced dodecasaccharide, only one site was cleaved by heparinase I, and a mixture of the tetrasaccharide ⌬U-A NAc,6S -GlcUA-A NS,3S,6S and the octasaccharide ⌬U 2S -A NS,6S -I-A NAc,6S -GlcUA-A NS,3S,6S -I 2S -A NS,6S-ol was predominantly obtained as shown in Fig. 3. Additionally, the hexasaccharide ⌬U 2S -A NS,6S -I-A NAc,6S -GlcUA-A NS,3S,6S was also observed in low quantity due to the partial cleavage of the reduced octasaccharide. When added to the mixture of the three heparinases, the reduced dodecasaccharide was transformed into a mixture of the tetrasaccharide ⌬U-A NAc,6S -GlcUA-A NS,3S,6S and two disaccharides, ⌬U 2S -A NS,6S and the ⌬U 2S -A NS,6S-ol providing the proposed structure.
NMR-NMR spectra were recorded using a Bruker Avance III spectrometer (600 MHz), Bruker Avance III spectrometer (800 MHz), and Avance III spectrometer (900 MHz), all equipped with a high sensitivity 5-mm TCI cryoprobe at 33°C. Samples were dissolved in 0.2 ml of 10 mM phosphate buffer, 0.5 M NaCl, pH 7.4, and 3 mM EDTA in 2 H 2 O (99.996%) and placed in 3-mm NMR tubes. For the transferred NOE experiments, samples were prepared by dissolving 0.47 mg of AT and 300 g of the dodecasaccharide in phosphate buffer reaching a molar ratio of 10:1 AT/dodecasaccharide. Proton spectra were recorded with presaturation of the residual water signal with a recycle delay of 12 s and 64 scans. Bidimensional double quantum filter-COSY and two-dimensional TOCSY spectra were acquired using 32 scans per series of 2048 ϫ 512 data points with zero filling in F1 (4096 ϫ 2048), and a shifted (/3) squared cosine function was applied prior to Fourier transformation. Heteronuclear single quantum coherence (HSQC) spectra were obtained in phase sensitivity-enhanced pure absorption mode with decoupling in the acquisition period. The matrix size of 1024 ϫ 320 data points was zero-filled to 4096 ϫ 2048 by application of a squared cosine function prior to Fourier transformation. All two-dimensional NOESY and two-dimensional transferred NOESY experiments were performed in a similar way. A total of 32 scans was collected for each free-induction decay (matrix 2048 ϫ 512 points), and the data were zero-filled to 4096 ϫ 2048 points before Fourier transformation, and mixing time values of 150, 200, 300, and 400 ms were used.
For STD experiments dodecasaccharide-AT samples were prepared with a ligand/AT molar ratio of 50:1. The pulse sequence used for the monodimensional STD NMR experiments includes a 15-ms spin-lock pulse to eliminate the broad resonances of the protein. A train of 20, 40, and 60 Gaussianshaped pulses of 50 ms each were applied to produce selective saturations. The on-resonance irradiation was performed at 7.2 ppm, whereas the off-resonance control irradiation was performed at 24 ppm. The STD spectrum was obtained by phase cycling subtraction of the on-resonance and off-resonance data acquired in interleaved mode. The number of scans and dummy scans were 2048 and 16, respectively.
Antithrombin Affinity-The interaction with antithrombin of the dodecasaccharide and of fondaparinux, as a reference compound, was studied by the fluorimetric titration method, according to the method previously described (15).
Anti-FXa Activity in Plasma or in a Buffer Medium-The anti-FXa activity of the dodecasaccharide was determined with the 10 ACL 7000 automated instrument for measuring coagulation (Instrumentation Laboratory), using the Heparin kit (Instrumentation Laboratory), which contains antithrombin, Factor Xa, and the chromogenic substrate S-2765. The measurements were carried out according to the manufacturer's recommendations. The second low molecular weight heparin international standard (National Institute for Biological Standards and Control, London, UK, code 01/608) was used to construct the calibration curve. Enoxaparin (Clexane) and fondaparinux (Arixtra) were used as internal standard compounds.
The test samples or the second LMWH international standard were first diluted in standard human plasma (Instrumentation Laboratory) or in a buffer (0.05 M Tris-HCl, 0.154 M NaCl, pH 7.4). Test solutions containing samples in plasma or in buffered medium were then diluted 1:20 with a buffered solution containing antithrombin and placed in duplicate in dedicated sites of the apparatus. Factor Xa reagent and chromogenic substrate were introduced into reservoirs provided for this purpose in the ACL 7000 instrument. Anti-FXa activity test was carried out according to the "Heparin" program integrated into the user interface (software) of ACL 7000. During the assay, 50 l of the sample diluted in buffer were mixed with 50 l of Factor Xa reagent. After an incubation time of 60 s at 37°C, 50 l of chromogenic substrate (1.1 mM) was added, and changes in 405 nM absorption were measured as a function of time. Anti-FXa activity of tested samples was determined using a calibration curve constructed with the second LMWH international standard (code 01/608).
LC-MS-Size exclusion chromatographic separations were performed with a liquid chromatography system (Dionex Ultimate 3000) equipped with a quaternary pump, an eluent degasser, an auto-sampler, and a variable wavelength detector. A sample volume of 25 l was injected on to the Superdex 75 10/300 GL column (300 ϫ 10 mm; 13 m average particle size). For sample concentrations, an AT solution at 2.0 mg/ml was run to verify separation profiles and retention times. Protein/ oligosaccharide solutions were prepared at 1:1 molar ratio containing AT at 2 mg/ml and an equimolar amount of the added oligosaccharide in CH 3 COONH 4 (100 mM). After 30 min of incubation at room temperature, the separation was run using an isocratically 100 mM CH 3 COONH 4 /CH 3 CN 97:3 (v/v) at 0.1 ml/min and monitored at 280 nm.
Mass spectrometric analyses were made on an ESI-TOF instrument (microTOFQ Bruker Daltonics), operating both in positive mode, in the expanded mass range from 1000 to 6000 m/z (normal conditions), and in negative mode, in the mass range from 200 to 2000 m/z (denaturing conditions). In normal conditions, a fraction (33 l/min) of total eluent flow was delivered by a splitter in the ionization source of the ESI-TOF mass spectrometer and the acquisition in positive polarity over the higher mass range allowing detection of high molecular weight proteins and binding studies. (Instrumental parameters were as follows: capillary Ϫ4500 V; nebulizer 1.0 bar; dry gas 7 liters/ min; dry temperature 200°C; mass range 1000 -6000 m/z).
In the first experiment, denaturing conditions used a solution of 5% HCOOH in 50% CH 3 CN introduced at 10 l/min in the ESI source from a syringe pump (located between the UV detector and ESI ionization), with the aim of reducing noncovalent interactions. In the next experiments, denaturing conditions were increased using 25% HCOOH in 100% CH 3 CN introduced at 10 l/min in the ESI source. In that case, the acquisition in negative mode allows detection of the previously interacting oligosaccharide, and it was then removed from the noncovalent protein-oligosaccharide complex by applying denaturing conditions. (Instrumental parameters were as follows: capillary ϩ3200 V; nebulizer 1.0 bar; Dry gas 7 liters/min; dry temperature 200°C; mass range 200 -2000 m/z).
Model Complex Building-The molecular editing was done with the Maestro 9.1 graphical interface. Two models of the 1:1 AT-dodecasaccharide complex, with the nonreducing and reducing AGA*IA framed by a nonsulfated IdoUA residue and terminating at the reducing end with a 4,5-unsaturated (⌬U) residue (structure b of Fig. 1) were built fitting the corresponding AT-binding site. The basic element of the complex between AT and AGA*IA pentasaccharide was isolated by molecular editing from the ternary complex AT-T-Hep (antithrombinthrombin-heparin) included in the Protein Data Bank code 1TB6. Both AT-dodecasaccharide complexes were created with the conformation of unsaturated uronate and IdoUA residues in 2 H 1 and 1 C 4, respectively, in accord with a previous study (15). The conformation of the I 2S residue of the interacting AGA*IA sequence was in 2 S 0 conformation as found in previous studies (14 -17), whereas that of the noninteracting AGA*IA I 2S was maintained in 1 C 4 , according to the prevalent conformation assumed when AGA*IA is in free solution. The torsional angles between ⌬U-A NAc,6S 1 and I-A NAc,6S 2 of the AT-interacting pentasaccharide are in agreement with the previous conformational analysis of the AT-octasaccharide complexes (octasaccharide 2 and 3 of Ref. 15). The structure of the two complexes, AT/⌬U-AGA*IA 1 -I-AGA*IA 2 and ⌬U-AGA*IA 1 -I-AGA*IA 2 /AT, was optimized by energy minimization applying the bmin procedure included in Macromodel/Maestro 9.1. Amber* force field (all atom force field) was used to represent the potential energy for both the AT and the sugar. The nonbonded interactions (electrostatic and dispersive) were treated using a continuum with a dielectric constant value of 80.0, whereas their cutoff was set to 20.0 and 8.0 Å, respectively. The hydrogen bond cutoff was defined separately as 4.0 Å. The minimization procedure stopped until the default gradient threshold of 0.05 kJ Å Ϫ1 mol Ϫ1 was reached. The hypothetical AT 2 -dodecasaccharide complex geometry was built starting from two complex structures, AT/⌬U-AGA*IA 1 and AT/AGA*IA 2 , that are linked together by an IdoUA residue in 1 C 4 conformation, forming the two glycosidic bonds A NS,6S 1 -I, and I-A NAc,6S 2 . To avoid molecular interpenetration of the AT molecules, the two AGA*IA-binding sites and the corresponding sulfate groups were oppositely oriented through proper rotation of A NS,6S 1 -I and I-A NAc,6S 2 torsional angle pairs /. After that, energy minimization, using the setting parameters previously defined, was performed.

Isolation and Characterization of the Dodecasaccharide-
The dodecasaccharide (Fig. 1b) was isolated from semuloparin, using the procedure described under "Experimental Procedures." The isolated dodecasaccharide fraction, representing ϳ6 -11% of the product, was injected onto an AT affinity column. The high affinity fraction was analyzed by HPLC (column UPLC Waters Acquity BEH C18) (Fig. 2). The fraction, eluted at conductivities between 80 and 130 millisiemens/cm (enriched in the dodecasaccharide), was collected and desalted on a column packed with Sephadex G-10 (100 ϫ 7 cm). The dodecasaccharide was then purified by CTA-SAX chromatography. By applying enzymatic sequencing methodology to the dodecasaccharide analysis, its key structural determinants were obtained such as building block composition and their respective order within the polysaccharide (Fig. 3). This first structural analysis was confirmed subsequently with full assignment of NMR spectra. Proton and carbon resonances of each residue were assigned using standard COSY, TOCSY, and HSQC pulse sequences (Tables 1  and 2). The sequence-specific resonance assignments have been made by identifying inter-residue H1-H4 NOE connectivity in NOESY spectra. Briefly, the A NAc,6S 1 was assigned through H4-H1 correlation with ⌬U, whereas A NS 1 and A NS, red were assigned for the H1-H4 correlation with IdoUA and lack of the interglycosidic correlation, respectively (Fig. 4). The H1-H4 correlation between I 2S 2 and A NS, red allowed assignment of A* 2 and, subsequently, A* 1 . It was more difficult to  assign G 1 and G 2 , whose resonances are superimposed; only the small difference of H4 chemical shifts, being observable from the NOEs with H1 of A NAc,6S 1 and A NAc,6S 2 residues, allowed these residues to be assigned.
Antithrombin Affinity-The equilibrium dissociation constant, K d , for the interaction between AT and dodecasaccharide or fondaparinux was measured by fluorescence titration (Table 3).    The K d value was obtained by monitoring the enhancement of intrinsic fluorescence of the serpin upon its reaction with increasing concentrations of each products, using the procedure previously described (15). Binding of the dodecasaccharide to AT was investigated in the presence of 0.1, 0.25, and 0.5 M NaCl yielding K d values of 0.0099, 0.086, and 1.23 M, respectively. The K d values obtained for fondaparinux were very similar to those previously published (14). Depending on the salt concentration, the dodecasaccharide had 2.1 to 3.2 times higher affinity for AT than fondaparinux. These results are surprising in view of current knowledge. Indeed, only a single affinity site would have been expected to bind with the antithrombin, and therefore the dodecasaccharide should have exhibited the same affinity as fondaparinux. However, our data imply that either the two affinity sites bind to AT or that just one site binds, and the second site reinforces the interaction.
Anti-Xa Activity-Results indicate a comparable anti-FXa activity on weight basis between the dodecasaccharide and the pentasaccharide fondaparinux (Table 4). However, on a molar basis, which is more representative of the phenomenon taking place at the molecular level, the dodecasaccharide is about twice as active as the fondaparinux reference. Interestingly, this is in agreement with the dodecasaccharide/fondaparinux molar mass ratio (respective molecular masses of fondaparinux and of dodecasaccharide in the form of sodium salts, 1728 and 3666 g/mol).
Interaction Study between AT and Dodecasaccharide-The comparison between free dodecasaccharide and the dodecasaccharide-AT complex proton spectra is shown in Fig. 5. Both high and low field protein-induced shifts were observed in the NMR spectra. Together with the increased linewidth, arising from the higher correlation time induced by the protein binding, the observed shifts indicate interaction between the dodecasaccharide and AT. Notably, the anomeric signal of A* 1 and H5 of I 2S 1 residues belonging to the nonreducing pentasaccharide show the larger shifts and line width increase, indicating stronger involvement in the binding of the corresponding residues. Evidence of the intermolecular interaction was also supported by the increased NOE magnitudes induced by the protein (Table 5). Particular attention is drawn to the H5-H2 NOE of the iduronate residues. This NOE contact is the marker of the 2 S 0 conformation, since it cannot take place in the pure 1 C 4 conformer (23). The ratio between H5-H2/H5-H4 NOE magnitudes in the absence of AT is around 0.4 for all iduronate moieties, compatible with an equilibrium between 1 C 4 and 2 S 0 conformation. In the presence of AT, the H5-H2/H5-H4 NOE ratio never reaches a value close to 1.0, typical of the pure 2 S 0 conformation. This is in contrast to what was observed for the pentasaccharide (24) and other AGA*IA-containing oligosaccharides so far described, where the AT-binding surface forces I 2S to assume the 2 S 0 conformation (14 -17). Assuming that the dodecasaccharide interacts with AT with a single pentasaccha-  ride sequence, the interacting iduronate residue should adopt the 2 S 0 conformation, whereas the iduronate residue belonging to the pentasaccharide not involved in the binding should remain in the classical equilibrium between 1 C 4 and 2 S 0 conformations. The higher H5-H2/H5-H4 NOE ratio (Ϸ0.7) found for I 2S 1 with respect to I 2S 2 (Ϸ0.5) suggests that the nonreducing pentasaccharide moiety preferentially interacts with AT (Table  5). However, the simultaneous binding of dodecasaccharide with two AT molecules cannot, at least in principle, be excluded. The inter-glycosidic NOE magnitude ratios between H1 A NS 1 -H3 I/H1 A NS 1 -H4 I (ratio 1) and H1 I-H4 A NAc,6S 2 /H1 I-H6 A NAc,6S 2 (ratio 2) in the absence and presence of AT does not change significantly, indicating that these geometries are not influenced by the protein binding (Table 5). Similar results have been obtained in previous studies on octasaccharides having sequences comparable with the reducing and nonreducing parts of the dodecasaccharide (Fig. 1, a and c) and showing A NS 1 -I and I-A NAc,6S 2 in the bound state geometries similar to those found in the dodecasaccharide (Table 6) (15).
In this conformation two AT molecules might interact with both pentasaccharide moieties, entailing a small torsional change of one or both A NS 1 -I and I-A NAc,6S 2 glycosidic linkages. Furthermore, small changes of and dihedral angles of the I-A NAc,6S 2 linkage allow the dodecasaccharide to bind two AT molecules (Table 6), spatially orienting sulfate groups in the way that can fit the two AT binding regions.
To better characterize the mode of binding of the dodecasaccharide with AT, STD experiments were performed. All monosaccharide residues show high or medium intensity signals in the corresponding STD spectra, indicating that all residues are on average positioned in the proximity of the AT-binding site (Fig.  6). The most intense signals are those belonging to the nonreducing pentasaccharide moiety (-A NAc,6S 1 ; A* 1 ; A NS 1 ; I 2S 1 and I) indicating that this portion of the molecule preferentially interacts with the protein. However, assuming that the nonreducing part of the molecule exclusively binds the AT, according to the model shown in Fig. 7a, the reducing part of the dodecasaccharide should be placed far away from the protein surface. Such an hypothesis is in contrast with STD results, showing that the reducing pentasaccharide also interacts with the protein. However, the sole interaction of the reducing pentasaccharide with AT (Fig. 7b) is not in agreement with the experimental results, because the corresponding signals are weaker than those belonging to the nonreducing pentasaccharide (Fig.  6). A dynamic equilibrium between the population of the two AT-dodecasaccharide complexes (i.e. AT interacting with the more prevalent form at the nonreducing end and the less prevalent one at the reducing end) can better recapitulate the NMR experimental data. Such an explanation is supported by previous studies indicating that the octasaccharide having AGA*IA at the nonreducing end (Fig. 1c) showed a slightly higher affinity for AT compared with the octasaccharide with AGA*IA positioned at the reducing end (Fig. 1a) (14). Nevertheless, the simultaneous presence of a small amount of AT 2 -dodecasaccharide tertiary complex cannot be excluded by STD data.
To verify a possible contribution of the 2:1 AT/dodecasaccharide stoichiometry, the complex was also analyzed by mass spectrometry. The LC SEC/MS method was optimized using a solution of free AT and its mixture with AGA*IA pentasaccharide in a 1:1 molar ratio, as described under "Experimental Procedures." Signal deconvolution of the mass spectrum corre-

TABLE 5 Inter-residue and intra-residue experimental NOEs (in parentheses) and transferred NOEs of dodecasaccharide and dodecasaccharide-AT complexes, respectively
Spectra were measured at 600 MHz, 33°C, in phosphate buffer (0.5 M NaCl) at five mixing times (150, 200, 300, and 400 ms). ND means not determined.   A further confirmation of these results was established by running the same AT/AGA*IA 1:1 solution using denaturing experimental conditions, able to destabilize noncovalent interactions. More specifically, the ESI interface polarity was switched to negative mode, and a formic acid solution was introduced from a syringe pump (see "Experimental Procedures") after the SEC separation and before the ionization source, allowing the weakly bound oligosaccharide to be dissociated from the interacting protein before detection. The mass spectrum shows the pentasaccharide molecular weight (M r 1506) and the loss of sulfate groups as usually observed for this type of compound in acidic conditions (Fig. 8C). The optimized LC-SEC/ESI-TOF method was then applied to the complex between AT and the dodecasaccharide. Results for the AT-dodecasaccharide complex, run both in normal and denaturing conditions, are reported in Fig. 9. The lower peak at R t ϳ82 min, attributable to a protein, does not produce any mass information as observed for the AT-pentasaccharide complex. The evidence of a 1:1 noncovalent complex between the AT protein and the dodecasaccharide was proven showing the corresponding peak with M r 61,118. The mass difference between the complex and the free AT corresponds to the molecular weight of the interacting oligosaccharide (M r ϳ3300). The same mass was also measured by negative mass spectrum in denaturing conditions (Fig. 9C).

DISCUSSION
Whereas natural variants of the AT-binding pentasaccharide (i.e. N-sulfation instead of N-acetylation of the first amino sugar residue and 6-O-desulfation of the 3-O-sulfated residue) have been described (25), the location of the pentasaccharide along the heparin chain is still uncertain. Although an early study suggested that this domain was located prevalently toward the nonreducing end of the molecule (26), other studies suggested a more random distribution (27). In recent studies by our groups, the presence of the pentasaccharide in both sulfated and unsulfated sequences was shown, suggesting a more random distribution of the pentasaccharide along the heparin chain (14). The different order of elution on the AT affinity column of an octasaccharide mixture suggested the active role of the flanking residue in the binding of the pentasaccharide sequence with AT (15). Moreover, structural modifications within the AGA*IA sequence, originating with the process used to depolymerize the polysaccharide, such as the presence of the 1,6-anhydrohexosamine residue at the reducing end or naturally present in the parent heparin, such as a pentasaccharide with two A* residues, have a strong influence in the AT binding properties of these oligosaccharides (16,17). The possibility of heparin chains with high AT-binding site density was suggested for high molecular weight fractions of porcine mucosal heparins (10), clam heparins (27), "macromolecular" rat skin heparins (28), and high affinity low molecular weight heparins (11). However, two active pentasaccharides in a row were never reported. It should be noted that the difficulties of isolating and purifying long heparin oligosaccharides and the fact that both enzymatic and chemical depolymerization procedures were not sufficiently selective to produce short chains containing this partic- ular dodecasaccharide containing two active sites could explain this. For example, the enzymatic depolymerization procedure with heparinase I depolymerizes the pentasaccharide sequence at the level of A* and generates truncated and inactive sequences (29), which is likely the reason why such sequences were not isolated by this methodology (21). The possible presence of more than one pentasaccharide sequence in a given heparin chain was supported by data obtained by fluorescence spectroscopy on the stoichiometry of the interaction of AT with active high molecular weight heparin fractions (10). This finding was confirmed by quantification, by NMR spectroscopy, of the glucuronic acid residue linked to A* (G-A*). This disaccharide, contained only in the pentasaccharide sequence and considered the real marker of the AT-binding sequence, was found to be present in high affinity fractions of LMWHs (i.e. dalteparin and enoxaparin), with contents slightly higher than those corresponding to one disaccharide per chain (11).
The isolation of a dodecasaccharide containing two contiguous AT-binding sequences was only possible due to the particular alkaline ␤-eliminative cleavage method used to prepare the ULMWH semuloparin (see under "Experimental Procedures"). The combined use of phosphazene base and aprotic solvent in the depolymerization process protected the AT binding region from cleavage, allowing isolation of longer AT-binding oligosaccharides (19). Moreover, as the two pentasaccharide sequences within the dodecasaccharide are resistant to phosphazene depolymerization, they accumulate in the dodecasaccharide fraction of semuloparin. As can be seen in the chromatogram (Fig. 2), dodecasaccharide is one of the most abundant compounds in this fraction, although it was not detected in other LMWHs obtained by using different ␤-elimination procedures. The higher affinity of the dodecasaccharide for AT compared with fondaparinux is combined with a higher anti-FXa activity, both in agreement with the presence of two ATbinding sequences. The possibility of the dodecasaccharide binding AT in two different molecular assemblies enhances the probability of the binding and consequently the affinity to the protein. STD experiments confirmed that both pentasaccharide sequences interact with AT, even though the binding with the nonreducing pentasaccharide (AGA*IA 1 ) is slightly preferred to that in which the reducing AGA*IA 2 binds to the protein. Notably, these data are in agreement with the slightly higher affinity found for the octasaccharide with a structure identical to the nonreducing side of dodecasaccharide (Fig. 1c), with respect to that of the octasaccharide corresponding to the reducing part (Fig. 1a) (15). In addition to these results, the unexpected conformational equilibrium in the bound state between 1 C 4 and 2 S 0 forms of both I 2S 1 and I 2S 2 residues also suggested the simultaneous presence of two dodecasaccharide/AT assemblies. However, a "sandwich" assembly in which the ligand is located between two protein molecules, similarly to what already observed in the complex between heparin and FGF1 (30), cannot be excluded. Whereas LC-MS data did not detect the 2:1 complex, STD results are compatible with a possible contribution of AT 2 -dodecasaccharide assembly. All these data support the idea that a second AT-binding sequence contributes to reinforcement of the binding with AT, through both the 1:1 interactions and the possible 2:1 tertiary complex, both of them being present at the equilibrium.
The discovery of this dodecasaccharide confirms the extreme heterogeneity and structural complexity of heparin, extending the action of enzymes involved in the last step of the biosynthetic pathway (8) to 3-O-sulfation of two contiguous AGAIA sequences. There are different hypotheses on the abundance and distribution of the AT-binding site along the heparin chains. Whereas some authors envisage a limited access of polysaccharide to membrane-bound enzymes (i.e. 3-O-sulfotransferase) in the Golgi apparatus, others suggest that the distribution of recognition sites for the 3-O-sulfotransferase may control the synthesis of the AT-binding site (28,31). Differences in structure and specificity of different 3-O-sulfotransferases, each of them exhibiting unique substrate recognition properties and distinct functional roles, may also influence the distribution of the AT-binding site in both physiological and pathophysiological conditions (9,32). At the present state of knowledge on the actual physiological function of heparin (26), the significance of biosynthesis of two contiguous active sites for AT remains elusive.
The occurrence of more than one AT-binding sequence in a specific oligosaccharide was described for the first time in this work, and this finding was made possible by the peculiar chemo-selective depolymerization procedure and original separation techniques used. Other depolymerization methods to prepare LMWHs and ULMWHs can affect the structure and the distribution of the active sequences along the originated fragments, without preserving the complex multiple AT-binding sequences. Knowledge of the detailed structure of heparin may help in designing new LMWHs and ULMWHs as potential drugs in different therapeutic fields, and the dodecasaccharide described in this study is a particularly attractive target for chemical and/or chemo-enzymatic synthesis of a novel antithrombotic drug.