An ATIII–Sepharose column (30 × 7 cm) prepared by coupling 2 g of human ATIII to CNBr-activated Sepharose 4B (Sigma) as described by Höök and coworkers (8) was used. A step gradient of NaCl concentration was applied. Low-affinity fractions were eluted using a 0.25 M NaCl solution buffered at pH 7.4 with 1 mM Tris–HCl at 11 ml/min; high-affinity fractions were eluted by a five steps gradient of NaCl, typically (0.74, 1.23, 1.71, 2.2, and 3.5 M NaCl and 1 mM Tris–HCl, pH 7.4). The NaCl gradient was monitored by conductivity measurements, and the heparin fractions were detected in the UV (Ultraviolet) at 219 nm. This wavelength, with a limited influence of the NaCl present in the mobile phase, is sensitive to N-acetyl functions present all over the heparin chain. The chromatograms of the separation were obtained after subtraction of the signal obtained on a blank run. Injected quantities could vary between 250 mg and 500 mg of heparin depending on the capacity of the column and the type of heparin injected.

Multiple desalting steps were mandatory to eliminate NaCl, especially for fractions of highest affinity. First, high-affinity fractions (1 L) were diluted 1/5 in water and passed through a 20 × 1.6 cm column filled with Q-Sepharose Fast Flow (Sigma Aldrich, Saint-Quentin-Fallavier, France). The column was then washed with water to eliminate free sodium chloride. The heparin was then flash eluted by NaClO₄ 2.5 N. UV detection at 215 nm was used to monitor the elution. In a second step, the heparin solution was desalted on a 100 × 7 cm column filled with Sephadex G10 monitored with UV detection at 215 nm and conductimetry.

Two chromatographic AS11 columns (25 × 0.21 cm) (Thermo Scientific Dionex, Montigny-le-Bretonneux, France) were connected in series. The column temperature was set at 40°C. Mobile phase A was 2.5 mM NaH₂PO₄ at pH 3.2, and mobile phase B was an aqueous solution of 2.5 mM NaH₂PO₄ with 1 M NaClO₄ adjusted to pH 3.0. A linear gradient (t0 min B% 0; t 80 min B% 60) was applied for elution at a flow rate of 0.22 mL/min. Diode array detection was used. Double UV detection was performed at 265 nm and 232 nm. An NRE building block-specific signal was obtained by the reconstruction of 265 nm - 2.21 × 232 nm (7).

Antithrombin III affinity chromatography was used to fractionate heparins into an ATIII low-affinity fraction (LA) and five fractions of increasing affinity for ATIII (HA1 to HA5) (Figure 1) as previously described (3), using 475 mg injections of PMH. The applied five-step concentration gradient (0.97, 1.55, 2.14, 2.79, and 3.5 M NaCl) was different to that previously used for BMH (3) because it was implemented on a recently bonded affinity column. The gradient was adapted to keep the HA fraction structurally constant. The capacity and retention of ATIII affinity columns continuously decrease during their lifecycle, and care must be taken to avoid column overloading, otherwise the content of the affinity fraction would be affected. The steps were adapted to the evolution of the column by an analytical control (building block analysis) of the affinity fractions. The object of the step gradient is to split the affinity fractions of heparin in a repeatable way. However, it is difficult to ensure a complete equivalence of the ATIII affinity of respectively collected fractions with the three heparins at different chromatographic states of the affinity column. A comparison of the percentages of the affinity fractions (measured w/w on 1.5 g of heparin through the weight of desalted fractions) under equivalent experimental conditions obtained values of 42.7, 34.7, and 29.9% for PMH, BMH, and BLH, respectively. The properties of the three heparins used for this study are shown in Table 2, and those from the collected fractions in Table 3. The AXa and AIIa activities of the fractions were measured for PMH and showed a continuous increase of activity with increasing ATIII affinity associated with a decreasing mean molecular weight.

Heparin fractions were characterized by analysis of the building blocks generated by exhaustive heparinase digestion. Reductive amination with sulfanilic acid was performed before chromatographic analysis (7), and sulfanilic tagged derivatives were indicated with the superscript ^sulf. This new assay method is more selective than the classical method and includes non-reducing end (NRE) building blocks that are not accessible with the classical method (13) since they do not contain any Δ^4–5 unsaturated acids. The major derivatives identified were the four glucosamines GlcNS^sulf, GlcN(NS,6S)^sulf, GlcN(NS,3S)^sulf, GlcN(NS,3S,6S)^sulf and the three NRE disaccharides Is_id^sulf [IdoA(2S)-GlcN(NS,6S)^sulf], IIs_glu^sulf ([GlcA-GlcN(NS,3S,6S)^sulf] and U(2,2,0)^sulf. Detection was performed at 265 nm and a signal specific of NRE building blocks (265 nm – 2.21 × 232 nm) was implemented (7). Tables 4–6 show the analytical results obtained respectively with PMH, BMH and BLH. Figures 2–4 shows the corresponding chromatograms obtained respectively for the three heparin sources. Each figure shows chromatograms of separation of building blocks on AS11 columns for a single heparin source and compared results when the affinity for ATIII of the analyzed heparin fraction is modified, that is for LA, HA1, HA3, and HA5 fraction. Another view of the same chromatograms is given in the Supplementary Material, where the influence of the animal origin of the three heparins (PMH, BMH, and BML) is shown for each type of affinity fraction (LA, HA1, HA3, and HA5) (Supplementary Figures 1–4). The quantified building blocks were those described previously (7), including the eight classical disaccharides (Table 1), the two galacturonic acid disaccharides (ΔIVs_gal and ΔIIs_gal), the 3S disaccharides (ΔIs, ΔIIs, and ΔIIIs) and the five main 3S tetrasaccharides (ΔIIa-IVs_glu, ΔIIa-IIs_glu, ΔIIs-IIs_glu, ΔIa-IIs_glu, and ΔIs-IIs_glu). The 3S tetrasaccharides evidenced in ion-pair liquid chromatography–mass spectrometry (LC/MS) of BMH fractions exhaustive digests (3) [ΔU(4,4,0)-2, ΔU(4,4,0)-3, ΔIIIs-IIs_glu, ΔIs-IVs_glu, ΔU(4,5,0)-4 to 6, ΔU(4,6,0)-7 to 9] were not quantified here. As a matter fact, if their presence is, on the structural point of view, interesting in relation to BMH specific 6-OH ATIII binding sequences, their content, estimated from LC/MS data in the first part (3) for BMH, is rather low (< 0.5%). Moreover, they were not all identified in the AS11 chromatographic system, due to its incompatibility with MS detection and its chromatographic specificity being different to ion pair chromatography. Thus, even though in some cases [ΔIIIs-IIs_glu, ΔIs-IVs_glu, ΔU(4,5,0)-4, and ΔU(4,6,0)-7 to 9], chromatographic specificities allowed their probable recognition, as shown in chromatograms of heparinases I+II+III digests of fractions LA, HA1, HA3, and HA5 (Figures 2–4), it is obvious that the resolving power of AS11 columns is too limited to enable their full quantification.

Sulfate to carboxylate ratios in Tables 4–6 were determined using cumulative contributions for each building block of carboxyl and sulfate groups in the heparin. A specific determination of sulfate to carboxylate ratio and NAc% based on the same calculation with only 3S building blocks [Sulfates/Carboxylates (sites) and NAc (sites)] was also performed, as it has good specificity to the heparin source. This specificity clearly appears in Figure 5 where the variations of these three parameters were plotted against the affinity for ATIII of the fractions. In PMH, NAc (sites) had an average value of 42 (42% of disaccharides in 3S building blocks are acetylated) which means that 84% of ATIII pentasaccharide binding sites were acetylated, if the major tetrasaccharide structure of 3S building blocks is considered. In the case of BMH, NAc (sites) is lower than in PMH, usually at about 19–23, in part due to the high content of 3S disaccharides (all N-sulfated and included in the calculation). If only 3S tetrasaccharides were considered, values at about 30 would be obtained, to be compared with the 42 obtained for PMH. With Sulfates/Carboxylates (sites), these two parameters reflect the differences of sulfate distribution in the ATIII binding sites for each heparin source.

Increased affinity of the fraction was associated with both an increase in the number of ATIII binding sites per chain and a structural transformation of these binding sites. The number of sites per chain was estimated from the content of the five major 3S tetrasaccharides (ΔIIa-IVs_glu, ΔIIa-IIs_glu, ΔIIs-IIs_glu, ΔIa-IIs_glu, and ΔIs-IIs_glu), based on the assumption that theses tetrasaccharides were all digested from true ATIII binding pentasaccharides. Values greater than two were obtained from fractions of highest affinity, even in BMH, although the calculation did not consider the influence of 3S disaccharides. The degree of 3-O-sulfation, which follows a similar increasing trend, offers another way to detect the same phenomenon. The structure-activity relationships governing the binding of pentasaccharides sites to ATIII are common to all heparins, so that the structural transformation of binding sites observed in fractions of increasing affinity to ATIII had similarities between the three heparins, but there were however significant specific differences, particularly for BMH, that highly impacted this transformation. Figure 6 shows the variations for PMH, BMH and BLH of the percentages of the main 3-O sulfated building blocks when the affinity of the fraction for ATIII increases. For PMH and BLH, NAc (sites) decreased between HA1 and HA5 (Figure 5B), reflecting the smaller growth of acetylated building blocks ΔIIa-IVs_glu and ΔIIa-IIs_glu compared with the N-sulfated blocks, ΔIs-IIs_glu and ΔIIs-IIs_glu. Doubling of AXa activity induced by the substitution of NAc for NS in the non-reducing glucosamine of ATIII binding pentasaccharides has previously been observed (14, 15). ΔIIs-IIs_glu appears to generate pentasaccharides with highest affinity to ATIII since it is the building block where the concentration increase between fractions HA1 to HA5, reaches highest values, to a much greater extent than for ΔIs-IIs_glu. We have already observed in high-affinity hexa- and octasaccharides from Semuloparin (10) that the 2-O-sulfate of ΔIs-IIs_glu- induced a marked decrease of affinity for ATIII binding sites, illustrated by the stronger binding to the ATIII affinity chromatography stationary phase of the hexasaccharide ΔIIs-IIs_glu-Is_id than the Δ2-O-sulfated equivalent ΔIs-IIs_glu-Is_id. The strong increases of ΔIs, observed for bovine heparins, will be discussed later with other 3S disaccharides.

Some other general trends were detected in Figure 7 where the variation of the parameters reflecting the sulfation pattern of the heparin fractions were plotted against the ATIII affinity of the fraction: with increasing affinity for ATIII, the percentage of 6-O-sulfation increased (Figure 7C), and 2-O-sulfation decreased (Figure 7B), which can largely be explained by the increase in ATIII binding sites (one 3-O-sulfate and one 2-OH GlcA), and the higher affinity of their 6-O-sulfated derivatives.

Other specific patterns were detected, such as the presence of the NRE PMH marker IIs_glu^sulf in fractions of PMH (7) (Table 4 and Supplementary Figures 1–8). Another specificity observed for BMH (Table 5) was the increase of the sulfate to carboxylate ratio of fractions with their affinity for ATIII (Figure 5A). This small but significant difference, only observed with BMH, was probably the reason why fractions with higher anticoagulant activity could be obtained from BMH by anion exchange chromatography (16).

However, despite all differences observed in Figures 5–7 between the 3 heparin origins, the similar trend of the percentages of 3-O sulfated glucosamines with the affinity for ATIII, observed in Figure 7D, is an indicator of a good comparability of the affinity chromatography experiments performed on the three heparins.

The distribution of 3S disaccharides in the affinity fractions LA and HA1–HA5 was central to ascertaining their role in binding to ATIII. ΔIIs, which was only present in the HA fractions of the three heparin sources and at a content increasing with the fraction affinity (Figure 6), must be classified as ATIII-binding. It strengthens the option chosen in the first part (3) of IIs_glu as precursor of ΔIIs within structural sequences with consecutive 3S disaccharides. Figure 6 and the data from Tables 4–6 also show that the content of ΔIs was clearly associated with the affinity of the fraction for ATIII, which reaches significant levels mainly in bovine heparins (Figures 6B,C) and, to a lesser extent, in ovine mucosa heparin (7).

The molar ratio within 3S building blocks between tetrasaccharides and disaccharides is given in Tables 4–6. This ratio increased with increasing affinity to reach a maximum for HA2 or HA3 for each type of heparin. High values were obtained with PMH due to the low content of 3S disaccharides. Much lower values were obtained for bovine heparins but in BLH, the 3S disaccharides were mainly ΔIs and ΔIIs, found in the high-affinity fractions, while in BMH, ΔIs and ΔIIIs were both present, in line with the low 6-O-sulfation of this heparin.

The variations of ΔIs in Figure 6 show for the three heparin origins, a more pronounced increase of the ΔIs content was observed for fractions of highest ATIII affinity (HA3–HA5), due to the presence in these fractions of ATIII binding sites with a supplementary 3-O-sulfate such as -IIa_id-IIs_glu-Is_id-, which have a higher affinity for ATIII than conventional binding pentasaccharides (3, 17, 18). ΔIIa-IIs_glu-Is_id is heparinase II-resistant but was digested into ΔIIa-IIs_glu and ΔIs by the heparinase mixture. These sequences were obtained by a secondary reaction of 3-O-sulfotransferases on the last glucosamine of the pentasaccharide. However, ΔIs was also detected in HA1, HA2, HA3 and partial LA fractions where the double 3-O-sulfated ATIII binding pentasaccharides were not present.

The main precursors of ΔIs, found in the corresponding heparinase II digests, were the two tetrasaccharides ΔIIs-Is_id and ΔIs-Is_id. The sequences -IIs_glu-Is_id-Is_id [-GlcA-GlcN(NS,6S)-IdoA(2S)-GlcN(NS,3S,6S)-IdoA(2S)-GlcN(NS,6S)-) and -Is_id-Is_id-Is_id (-IdoA(2S)-GlcN(NS,6S)-IdoA(2S)-GlcN(NS,3S,6S)-IdoA(2S)-GlcN(NS,6S)-] with potential affinity for ATIII can be hypothesized since significant affinity for ATIII and AXa activity were measured for the biosynthetic octasaccharide GlcN(NS,6S)-GlcA-GlcN(NS,6S)-IdoA(2S)-GlcN(NS,3S,6S)-IdoA(2S)-GlcN(NS,6S)-GlcA (5).

The presence of ΔIs in the LA fraction could be used as an argument against the affinity of sequences including Is_id. However, the LA fraction of PMH also contained some ΔIIa-IIs_glu and had a residual activity (Table 3). Indeed, the selectivity of ATIII affinity chromatography between HA and LA fractions had an inverse relationship to molecular weight. The BLH batch used for this analysis is much smaller than the PMH and BMH batches (Table 2), so that the LA fraction of BLH contains very little residual 3S tetrasaccharides (Table 6), with the ratio of the percentages of ΔIs between HA1 and LA almost twice that obtained for BMH (Table 5). If we consider a LMWH synthesized from BMH using enoxaparin process (unpublished data), the content of ΔIs in the high affinity fraction is more than twice that in the starting heparin when a residual 0.5% is detected in the LA fraction. When the length of the chain is decreased, the structural specificity of fractions increases so that, for hexasaccharides, the percentages of ΔIs in HA and LA are 8.1% and 0.4%, respectively. As shown in Figure 6, particularly in the case of BMH (Figure 6B), the case of ΔIIIs is less clear, with desulfation in position 6-O probably responsible for the lower affinity of sequences. Unlike ΔIs, there was no indication of any double 3S ATIII pentasaccharides ending with IIIs_id. Moreover, the same shared presence of ΔIIIs as for ΔIs was observed with LA and HA, but with a higher part in LA, particularly in the case of BMH. In bovine LMWH, as for ΔIs, ΔIIIs is still more concentrated in HA fractions (2.1% vs. 0.5% for LA). But finally, it seems that the affinity of sequences containing IIIs for ATIII should be at least low if not totally negligible.

In the first part of these experiments (3), heparinase II digests of BMH fractions were studied to highlight new specific 3S tetrasaccharide and hexasaccharide building blocks cleaved into unsaturated disaccharides (ΔIs, ΔIIs, and ΔIIIs) in exhaustive digests, by heparinase I. Nine new tetrasaccharides [ΔU(4,5,0)-10 to 13, ΔU(4,6,0)-14 to 16, ΔIIs-Is_id, ΔIs-Is_id] were detected, and most were specific to BMH. ΔU(4,6,0)-16 was also present in BLH while ΔIIs-Is_id and ΔIs-Is_id were both present at least in PMH, BLH and ovine mucosa heparin (OMH). Hexasaccharides were mainly detected in fractions of highest affinity (HA4 and HA5), mainly ΔIIa-IVs_glu-Is_id, ΔIIa-IIs_glu-Is_id, ΔIa-IIs_glu-Is_id, ΔIIs-IIs_glu-Is_id, and ΔIs-IIs_glu-Is_id, with other hexasaccharides as yet unidentified.

Integration data obtained from AS11 chromatograms of the heparinase II digests from PMH, BMH and BLH (Figures 8–10) are shown in Tables 7–9. As done for the heparinase I+II+III digests, the influence of the heparin origin on the separation on AS11 column of the building blocks obtained from heparinase II digestion of the different affinity fractions (LA, HA1, HA3 and HA5) is shown is the Supplementary Figures 5–8. Additionally, the influence of the digestion type (Heparinases I+II+III vs. Heparinase II) for each type of affinity fraction and for one animal origin of heparin is also illustrated in the Supplementary Figures 9–21. As observed in the heparinases I+II+III digests, some of the heparinase II specific building blocks detected in the ion pair LC/MS could be identified in the AS11 chromatographic system. However, Tables 7–9 only include the w/w% of the main building blocks with the already ascertained heparinase II-specific ones, ΔIIs-Is_id, ΔIs-Is_id, ΔIIa-IVs_glu-Is_id, ΔIIa-IIs_glu-Is_id, ΔIa-IIs_glu-Is_id, ΔIIs-IIs_glu-Is_id, and ΔIs-IIs_glu-Is_id. Sulfate to carboxylate ratios and percentages (NAc, 2-OH, 6-OH, 3S)% were not given, because the N-acetyl and 3S building blocks could not be exhaustively identified due to their scattering into unidentified glycoserines and other minor heparinase II-specific building blocks. Overall, aside from ΔIVa, ΔIVs and 3S building blocks, integration data from heparinase II and heparinases I+II+III were similar. In the three heparins (Tables 7–9), ΔIIs-Is_id and ΔIs-Is_id were major fragments containing -Is_id and their presence, detected essentially in the affinity fractions, was a strong argument for their ATIII binding contribution within a non-conventional pentasaccharide where a 2-O-sulfated iduronic acid is substituted for the canonical central glucuronic acid. These data, compatible with an ATIII binding capacity of sequences including ΔIIs-Is_id and ΔIs-Is_id, are in line with those recently published (5, 6). Table 7 also shows that even in the fraction HA5, the phenomenon of double 3S ATIII pentasaccharides was limited for PMH (5–10% of ΔIIa-IVs_glu, and ΔIIa-IIs_glu is in ΔIIa-IVs_glu-Is_id, ΔIIa-IIs_glu-Is_id sequences) in connection with the low content of ΔIs. Much higher percentages (10–40%) were observed in BMH (Table 8) and in BLH. The sites present in double 3S sequences reflected similar sulfation patterns than the classical ATIII pentasaccharides detected in lower affinity fractions, supporting the idea that this effect is due to a secondary reaction of 3-O-sulfotransferases after the first 3-O-sulfation of the site, key element of the pentasaccharide anticoagulant activity. However, maximum percentages (40%) were observed in acetylated sites (ΔIIa-IVs_glu-Is_id, ΔIIa-IIs_glu-Is_id).