Synthesis of bioengineered heparin chemically and biologically similar to porcine-derived products and convertible to low MW heparin

Significance

Heparin products correspond to a critical class of anticoagulant drugs. However, heparin extraction from animal tissues presents multiple challenges, ranging from insecure supply chains to contamination and significant variability. The 2008 contamination crisis and the continued growth of the heparin market emphasize the need for an alternative, reliable supply of this essential medicine. We report a robust and scalable method to synthesize animal-free bioengineered heparin with structural and functional properties similar to that of USP heparin, including the ability to be converted into a low molecular weight heparin similar to enoxaparin.

Abstract

Heparins have been invaluable therapeutic anticoagulant polysaccharides for over a century, whether used as unfractionated heparin or as low molecular weight heparin (LMWH) derivatives. However, heparin production by extraction from animal tissues presents multiple challenges, including the risk of adulteration, contamination, prion and viral impurities, limited supply, insecure supply chain, and significant batch-to-batch variability. The use of animal-derived heparin also raises ethical and religious concerns, as well as carries the risk of transmitting zoonotic diseases. Chemoenzymatic synthesis of animal-free heparin products would offer several advantages, including reliable and scalable production processes, improved purity and consistency, and the ability to produce heparin polysaccharides with molecular weight, structural, and functional properties equivalent to those of the United States Pharmacopeia (USP) heparin, currently only sourced from porcine intestinal mucosa. We report a scalable process for the production of bioengineered heparin that is biologically and compositionally similar to USP heparin. This process relies on enzymes from the heparin biosynthetic pathway, immobilized on an inert support and requires a tailored N-sulfoheparosan with N-sulfo levels similar to those of porcine heparins. We also report the conversion of our bioengineered heparin into a LMWH that is biologically and compositionally similar to USP enoxaparin. Ultimately, we demonstrate major advances to a process to provide a potential clinical and sustainable alternative to porcine-derived heparin products.

Heparin, the most sulfated glycosaminoglycan (GAG), was discovered in 1916 (1) and remains a widely used intravenous anticoagulant drug. Initially isolated from various animal tissues (e.g., lung, liver, intestine), porcine intestinal mucosa has become the predominant tissue source worldwide for heparin active pharmaceutical ingredient (API) and is currently the sole source of heparin accepted by the US Food and Drug Administration (FDA). Moreover, porcine intestinal unfractionated heparin (UFH) is the only source material currently used in manufacturing low molecular weight heparins (LMWHs), the preferred class of subcutaneous anticoagulant drugs.

UFH is a critical drug for a wide range of clinical uses, including post-operative prophylaxis (2), treatment of deep vein thrombosis (3), and pulmonary embolism (4), among others. UFH is also routinely used as an anticoagulant in extracorporeal blood circulation devices (5), such as in open-heart surgery and hemodialysis (6), and it is the drug of choice when rapid anticoagulation is needed. As a result, heparin is on the list of the World Health Organization’s most essential medicines (7). Nearly 100 tonnes of heparin are produced annually worldwide, from close to one billion hogs (8), and the global heparin market was approximately $7.4 billion in 2022 (9).

LMWHs, a key class of heparin derivatives, are prepared through the controlled chemical or enzymatic depolymerization of UFH. Due to their shorter length, LMWHs exhibit less non-specific interactions with proteins, other than antithrombin (AT), affording more predictable pharmacokinetic/dynamic (PK/PD) profiles, better subcutaneous bioavailability, and lower incidence of heparin-induced thrombocytopenia (HIT) when compared with UFH (10). Among commercially available LMWHs (11), enoxaparin is the most commonly used worldwide. Since 2010, generic forms of enoxaparin have been approved by the US FDA for drugs demonstrating fulfillment of five criteria of equivalence to the original innovator product (12).

Heparin is naturally biosynthesized in mast cells, beginning in the endoplasmic reticulum and concluding in the Golgi apparatus (13). The enzymatic process (SI Appendix, Fig. S1) is highly conserved (14). Heparin biosynthesis starts in vivo with the formation of a tetrasaccharide linkage, or primer, βGlcA-(1→3)-βGal-(1→3)-βGal-(1→4)-βXyl-O-Ser (where GlcA, Gal, and Xyl are d-glucuronic acid, d-galactose, and d-xylose, respectively), which is attached to serine residues of the core protein serglycin. Subsequently, the GAG chain is extended through alternating addition of GlcA and N-acetyl-d-glucosamine residues (GlcNAc) by the glycosyltransferases exostosin 1 (EXT1) and exostosin 2 (EXT2) (15). As the polysaccharide chain extends, N-deacetylase/N-sulfotransferase (NDST) removes acetyl groups from some of the GlcNAc residues and replaces these with sulfo groups to afford N-sulfo-d-glucosamine residues (GlcNS). This is followed by conversion of GlcA to l-iduronic acid (IdoA) catalyzed by heparan sulfate C5-epimerase (C5-epi) and the action of 2-O-, 6-O- and 3-O- sulfotransferases (2OST, 6OST, and 3OST, respectively) in the presence of 3′-phosphoadenosine-5′-phosphosulfate (PAPS), leading to the final heparin structure. The epimerization of GlcA is a reversible reaction that can be locked in place by 2-O-sulfation, giving rise to 2-O-sulfo-l-iduronic acid (IdoA2S) (16). This suggests that C5-epimerase and 2OST are likely coupled in vivo (17, 18). The resulting heparin chains, having a molecular weight of 60 to 100 kDa, (19) are covalently bound to their core protein, serglycin, as a proteoglycan. Located in secretory granules of mast cells, they are cleaved by heparanase into GAG chains of 10 to 20 kDa fragments (20–22).

There has been a chronic scarcity of crude heparin (23), which can affect both UFH and LMWH supplies. Over 70% of heparin is from a single country of origin (China) and attempts to diversify the supply chain have been challenging. A widespread outbreak of porcine reproductive and respiratory syndrome (PRRS, or Blue Ear disease) in Asia in 2006 to 2007 necessitated mass culling (24), resulting in a dramatic drop in crude heparin supply. In 2008, after the death of heparin recipients in the US and worldwide, several batches of heparin API from China were found to be adulterated with oversulfated chondroitin sulfate, an inexpensive, semi-synthetic GAG, that was difficult to distinguish from heparin based on the testing standards put in place at the time by the USP (25, 26). This triggered an overhaul of analytical quality control procedures for API validation and revision of the United States Pharmacopeia (USP) monograph (27). The need for diversification of the heparin supply is clear, from both alternative animal sources and animal-free manufacturing processes.

Multiple alternative sources of heparin have been sought in the past two decades. These include reintroduction of bovine (or ovine) heparin (28) as an animal-derived alternative. This has been hindered by the relatively low anticoagulant activity of these products (29, 30) compared to porcine heparin, as well as concern over prion-mediated diseases (31, 32). Attempts at animal-free synthesis include de novo chemical synthesis, especially of LMWHs (33), biosynthetic processes, and metabolic engineering for in vivo production of heparin from prokaryotic and eukaryotic hosts (34). Chemical synthesis has resulted in the ultra-LMWH fondaparinux (Arixtra), a pentasaccharide heparin analog (33), but not in a UFH-equivalent product. While fondaparinux enables unmatched control over product structure, it is obtained in very low yields and is costly. Similarly, the metabolic engineering of UFH in yeast (35) and in mammalian cells (36–38) results in very low yields, is poorly scalable, and gives a heparin compositionally distinct from USP heparin. The generation of a USP heparin-similar using in vitro chemoenzymatic methods has been attempted, such as the early heparin analog, neoheparin (39), which used heterologously expressed C5-epi and chemical per-O-sulfation/de-O-sulfation to convert N-sulfoheparosan (NSH) at milligram scale to an anticoagulant material. Other attempts included enzymatic synthesis of LMWH (40) and deuterated heparin (41), also synthesized at small scale and with minimal control over final product composition. Thus far, no chemical or biosynthetic routes have produced a synthetic heparin equivalent or similar to porcine-derived heparin.

Herein, we describe the gram-scale chemoenzymatic synthesis of bioengineered heparin (BEH) that mimics porcine UFH. The multienzymatic process uses NSH as starting material, generated chemically from Escherichia coli heparosan. As a result, we report the gram-scale synthesis of animal-free heparin-like product with chemical and biological properties similar to porcine USP heparin and with the potential to be converted into an enoxaparin-like LMWH.

Results

Chemoenzymatic Biosynthesis of BEH.

BEH synthesis (SI Appendix, Fig. S1) starts with heparosan, a polysaccharide isolated from the capsule of E. coli K5 featuring the same disaccharide backbone structure as heparin. As a first step, heparosan is chemically N-deacetylated, depolymerized, and N-sulfonated to generate NSH with a Mw range in line with that of porcine intestinal heparin. The level and distribution of N-sulfation/N-acetylation in NSH is critical in generating a BEH with similar composition to that of porcine-derived USP heparin, including degree of epimerization (42) and distinguishes this approach from previous attempts to generate a non-animal sourced heparin (39, 41, 43). The distribution of GlcNAc residues prepared chemically in NSH was compared to NSH prepared enzymatically using NDST-2 (44), which introduces N-sulfation during the in vivo biosynthesis of heparin/heparan sulfate. Using LC-MS oligosaccharide mapping, only subtle differences were observed between the two methods of NSH preparation (SI Appendix, Fig. S2).

NSH, with controlled NS content, is then enzymatically modified by C5-epimerization (catalyzed by C5-epi) of GlcA and 2-O-sulfation of the resulting IdoA catalyzed by 2-O-sulfotransferase (2OST), followed by 6-O- and 3-O-sulfation of GlcN with 6-O-sulfotransfase isoforms 1 and 3 (6OST-1/6OST-3) and 3-O-sulfotransferase isoform 1 (3OST-1), respectively. The sequential, as opposed to simultaneous sulfotransferases reactions is essential for generating a heparin with similar composition and anticoagulant activity to porcine-derived USP heparin (45). Indeed, it is known that the 6-O-sulfation of GlcNy residues (y = Ac or S) sterically hinders access of 2OST to its target (46, 47), and 3-O-sulfation of GlcNy residues by 3OST-1 is blocked if located at the reducing end of IdoA2S residues (48).

Analysis of USP Heparins to Determine NSH’s GlcNS/GlcNAc Content.

The strongly anionic charge of heparin chains is central to their anticoagulant properties. Key sulfation patterns are the 3-O-sulfated GlcNS at the center of the AT-binding pentasaccharide and stretches of tri-sulfated (TriS) disaccharides (IdoA2S-GlcNS6S), which are essential to the binding of thrombin/factor IIa (49). The distribution of N- and O-sulfation on the disaccharide repeats forming the heparin backbone varies depending on the species of origin (50, 51). Thus, prior to generating BEH, the compositions of the major structural features in porcine intestinal heparin were established. Specifically, 32 lots of USP heparin API from various commercial manufacturers (SI Appendix, Table S1) were analyzed to determine disaccharide and 3-O-sulfo-containing tetrasaccharide compositions (dp2 & dp4 analysis) (SI Appendix, Table S2 and Dataset S1). Since the structures of the main tetrasaccharides (T1-5) have been established (52), their constitutive disaccharides could be quantified, thereby allowing calculation of the total sum of all GlcNS- and GlcNAc-containing disaccharides in USP heparin. As a result, we set the NS content to 88% in NSH. This is consistent with the 12.8 ± 1.2% GlcNAc obtained using NMR analysis of porcine mucosal heparin (50) and, since there are only a few GlcNAc residues in the average USP heparin or BEH chain, it is very unlikely that extended clustering of these residues takes place.

Biocatalyst Engineering for Improved Expression and Activity.

While highly efficient in their native Golgi environment, heparin biosynthetic enzymes are difficult to produce in soluble and active forms using heterologous bacterial expression. Starting from enzymes with their transmembrane domains truncated and substituted with an MBP tag (53), the constructs were further modified, notably by performing computational optimization to reduce surface hydrophobicity. An additional N-terminal His₆ tag was added for purification and the eukaryotic sequence was codon-optimized to match the E. coli expression system. As a result, the soluble expression yields were improved from four-fold (6OST-3) to 54-fold (2OST) (SI Appendix, Table S3). We also observed an apparent increase in the specific activities of the enzymes, up to 14-fold for 6OST-3. We hypothesize that the reduced propensity to aggregate, due to better solubility, improved the enzymes folding into its catalytically active conformation (54).

The enzymes were then immobilized onto a suitable support to enhance process performance and simplify separation of the intermediates and BEH product from the biocatalysts. CNBr-activated Sepharose was chosen as a commonly used immobilization resin (55). In all cases, except for C5-epi, immobilized engineered enzymes showed higher activity, approximately 4- to 5-fold, compared to immobilized control enzymes (SI Appendix, Table S4).

Multi-Gram Scale BEH Biosynthesis.

The enzymatic synthesis of BEH was performed using immobilized enzymes and NSH with NS content consistent with that of porcine heparin. Initially, reproducibility was assessed by performing three parallel syntheses starting with 0.5 g of NSH, each reaction using the same biocatalyst preparations but three independently prepared batches of NSH (SI Appendix, Fig. S3 A–D and Table S5). The resulting BEH products had disaccharide and tetrasaccharide compositions similar to that of USP heparin and with average anti-IIa and anti-Xa activities of 239 ± 15 IU/mg and 233 ± 9 IU/mg, respectively (anti-IIa/anti-Xa ratio of 1.03), which conforms to USP heparin specifications (SI Appendix, Table S6).

Having successfully generated BEH with compositional properties similar to USP heparin, we proceeded to scale up the three sequential enzymatic steps from NSH to BEH at the multi-gram scale. Two sequential sub-batches of C5-epi/2OST and 6OST-1/6OST-3 reactions were used, which also provided an opportunity to evaluate immobilized enzyme reusability. The C5-epi/2OST reactions (600 mL per sub-batch) were performed with 10 mg/mL NSH (89.3% NS) and 1 mg/mL C5-epi and 1.7 mg/mL 2OST in MES buffer containing 25 mM PAPS supplemented with 10 mM MgCl₂. After the first sub-batch reaction, the slurry was thoroughly washed with MES/NaCl buffer to recover the NS2S-Hep intermediate and the biocatalyst was then used in the second sub-batch reaction. The reaction mixtures from each sub-batch were then desalted and concentrated by tangential-flow filtration. The purified NS2S-Hep sub-batches were then pooled and used in two sequential 6OST-1/6OST-3 sub-batch reactions. The substrate concentration was 12 mg/mL (500 mL per sub-batch) in MES buffer containing 25 mM PAPS and 0.25 mg/mL each of 6OST-1 and 6OST-3. The reaction mixtures from each sub-batch were desalted, concentrated, and pooled as described above to provide the final TriS-Hep intermediate for use as substrate for the 3OST-1 reaction. Finally, BEH was obtained from a single 3OST-1 reaction, consisting of 40 mg/mL TriS-Hep (300 mL) and 7.5 mM PAPS in MES buffer containing 0.5 mg/mL 3OST-1.

C5-epi/2OST and 6OST-1/6OST-3 reactions were monitored by disaccharide analysis, while the 3OST-1 reaction was monitored by measuring anti-IIa activity (Fig. 1). The C5-epi/2OST and 6OST-1/6OST-3 enzymatic reactions followed two-phase time courses with an initial rapid sulfation, corresponding to abundant availability and accessibility of target reaction sites on the polysaccharide chain, up to approximately 60% conversion. This was followed by a marked reaction rate reduction, leading to a plateau presumably corresponding to the loss of reactive sites on the polysaccharide chains.

Fig. 1.

Time courses of multi-gram synthetic reactions leading to a 10 g BEH batch. (A and B) Time course of two sequential sub-batches of the C5-epi/2OST reactions (#1 and #2), with magnification (B) to visualize 2S, IdoA, and GlcA2S disaccharides. (C) Time course of two sequential sub-batches of the 6OST-1/6OST-3 reactions (#1 and #2). (D) Time course of the single batch 3OST-1 reaction. Curves fitted using two-phase association model.

The first and most complex reaction is the combined C5-epimerization and 2-O-sulfation, as the interplay of the two enzymes is required to reach high levels of epimerization (16). Reaction time courses for the two C5-epi/2OST sub-batches were nearly identical in dp2 analysis with NS2S levels plateauing within 18 to 24 h (Fig. 1A). The NS2S-Hep produced from the two reactions were similar with NS2S disaccharide (IdoA/GlcA2S-GlcNS) contents of 72.2 and 73.5% and GlcA2S contents of 6.9 and 5.6%, respectively. The minor 2S disaccharide (IdoA/GlcA2S-GlcNAc) reached 1.0 and 0.8% (SI Appendix, Table S7). The bienzymic reaction time course revealed highly dynamic reaction kinetics. During the first hour of reaction, a rapid transient increase in non-sulfated IdoA was observed, approximately up to the equilibrium concentration (56), before diminishing slowly while the IdoA2S level rose. Simultaneously, GlcA2S accumulated, albeit at a much lower rate (Fig. 1B). While 2OST shows a strong preference for IdoA residues, it has low, but not negligible, activity on GlcA (57). The accumulation of IdoA was likely due, in part, to the higher rate of C5-epi than that of 2OST (SI Appendix, Table S4). Moreover, unlike in the native mammalian cell where C5-epi and 2OST appear to form a complex (18), the in vitro reaction has the enzymes separately immobilized and unable to form a bienzymic complex. This may explain why there is a build-up of IdoA and a lag between the generation of IdoA and its 2-O-sulfation, which may also have increased the likelihood of 2OST acting on GlcA residues.

The apparent Mw of the NS2S-Hep intermediate was consistently lower than that of the NSH precursor (~500 Da reduction, P < 0.005, paired t test) instead of the expected higher Mw due to addition of multiple sulfate groups per GAG chain (SI Appendix, Fig. S4A). Molecular weight analysis of NSH treated only with C5-epi and using the same GPC method and USP heparin for calibration revealed that epimerization caused a statistically relevant drop in the apparent Mw of the GAG chains (SI Appendix, Fig. S4B). These results are in line with previous work showing that IdoA residues introduce flexibility through conformation changes (58) and that the length of a NSH oligomer decreased after C5-epimerization and 2-O-sulfation (59). Altogether, this suggests that epimerization of GlcA to IdoA induces a conformation change of the NSH chains, from a linear conformation for the GlcA-containing NSH to a more compact and helicoidal configuration for the IdoA-containing epimerized NSH.

During the second enzymatic step, the 6OSTs catalyze 6-O-sulfation onto GlcN residues generating the tri-sulfated disaccharide (IdoA2S-GlcNS6S), which is important to the binding of Factor IIa (thrombin) to heparin (49), as well as the NS6S disaccharide (IdoA/GlcA-GlcNS6S) often at the center of the AT binding site (60). Reaction time courses for the sequential 6OST-1 and -3 sub-batches were nearly identical (Fig. 1C) with TriS levels leveling off in less than 12 h. Analysis of the recovered and desalted TriS-Hep intermediate yielded TriS contents of 68.5 and 68.2%, and 10.3 and 9.4% NS6S, respectively, indicating that the reactions were highly reproducible and generated very similar products. The minor 6S disaccharide (corresponding to GlcA/IdoA-GlcNAc,6S) reached 6.1 and 5.8% (SI Appendix, Table S7).

In the final enzymatic step, 3OST-1 catalyzes the 3-O-sulfation of TriS onto GlcNS or GlcNS6S residues to generate the AT binding pentasaccharide. The presence of a 2-O-sulfated uronic acid on the non-reducing end of GlcN restricts 3OST-1 activity (48), greatly limiting the availability of target sites, and hence, overall conversion and reaction rate. The 3OST-1 reactions (Fig. 1D) surpassed the desired anti-IIa activity threshold of 220 IU/mg in ~72 h. After desalting, the final product was freeze-dried and amounted to 11.34 g of BEH, corresponding to 71% isolated yield.

Compositional analysis of the NSH, NS2S-Hep, and TriS-Hep intermediates is summarized in SI Appendix, Table S7. The only significant deviation from USP heparin composition is the GlcA2S content, which at 6.9 and 5.6% is higher than USP heparin in both C5/2OST sub-batches. The minimal deviation observed between sub-batches (time courses and NS2S-Hep and TriS-Hep intermediates composition) indicated that reuse of the C5-epi, 2OST, 6OST-1, and 6OST-3 biocatalysts was successful without significant loss of activity.

Complete Characterization of BEH.

BEH isolated from the gram-scale multi-enzymatic sequential reactions was purified to remove endotoxins, residual protein, and nucleic acids, and then compared to USP heparin. The BEH met USP heparin specifications, with anti-IIa and anti-Xa activities of 246 and 241 IU/mg, respectively (anti-IIa/anti-Xa ratio of ~1.0) and a Mw of 17,600 Da and PDI of 1.2. Additionally, BEH composition, as reflected in disaccharide and tetrasaccharide analysis (Table 1 and Fig. 2A and SI Appendix, Fig. S4), was very similar to that of USP heparin. LC-MS “bottom-up” heparinase 2-catalyzed oligosaccharide mapping was performed on BEH and USP-porcine heparin from three different vendors to further examine the similarity of their building blocks. The distribution of the sequences observed was very similar for BEH and the USP-porcine heparins (SI Appendix, Fig. S6).

Table 1.

The dp2 & dp4 composition of BEH and comparison to porcine mucosal heparin (USP heparin)

	USP heparin	BEH 10 g batch
dp2 & dp4 analysis
0S	3.9 ± 0.4	4.0
NS	3.0 ± 0.3	4.9
6S	3.1 ± 0.2	3.3
2S	1.7 ± 0.2	0.6
NS6S	11.1 ± 0.4	7.5
NS2S	7.3 ± 0.4	5.4
2S6S	1.4 ± 0.1	0.3
TriS	65.9 ± 1.0	71.7
T1 (6S-NS3S)	0.4 ± 0.0	0.3
T2 (6S-NS3S6S)	1.7 ± 0.2	1.5
T3 (NS6S-NS3S6S)	0.2 ± 0.0	0.2
T4 (2S6S-NS3S6S)	0.1 ± 0.0	0.1
T5 (TriS-NS3S6S)	0.2 ± 0.0	0.2
Total T1–5	2.6 ± 0.3	2.3
Total	100.0 ± 0.0	100.0
USP heparin specifications
Criteria	USP specification	10 g Batch
Anti-Xa (U/mg)		246
Anti-IIa (U/mg)	NLT 180	241
Anti Xa/IIa ratio	0.9–1.1	1.0
Mw (Da)	15,000–19,000	17,600
Additional Mw parameter
PDI (not a USP specification)	1.09–1.32	1.16

Fig. 2.

Characterization of BEH. (A) Di- and tetrasaccharide composition of final, desalted BEH compared to USP heparin (mean ± SD, USP-H n = 32, BEH 0.5 g n = 3, BEH 10 g n = 1). (B) 1D and 2D NMR analysis of USP heparin vs. BEH. BEH (10 g batch) and USP heparin samples were dissolved in ²H2O and lyophilized repeatedly to achieve full deuterium exchange. The samples were then analyzed by 1D ¹H and 2D ¹H/¹³C HSQC NMR on 600 MHz spectrometer. Acquisition parameters and peak assignments were set according to Mauri et al. (50, 61). (C) 1D ¹H-NMR analysis of LMWHs, enoxaparin sodium. Samples were dissolved in ²H₂O and lyophilized repeatedly to achieve full deuterium exchange and analyzed by 1D ¹ H NMR on 600 MHz spectrometer. LMWH products were prepared using current depolymerization process from Celsus USP UFH (purple) and BEH UFH (green). Commercially available samples (controls) were a generic enoxaparin sodium (red) and Lovenox® (blue).

BEH composition was also analyzed by 1D ¹H-NMR and 2D ¹H/¹³C HSQC NMR and compared to USP-porcine heparin following the method of Mauri et al. (50, 61) (Fig. 2B). An overlay of the USP-porcine heparin and BEH structures revealed that 1D peaks in the ring proton region (3 ppm to 6 ppm) showed excellent match; however, there was one clear difference near the acetyl (-CH₃) peak at around 2 ppm. A small set of peaks, both to the left and right of the major peak, was present only in the BEH and corresponded to 3-deoxy-d-manno-oct-2-ulosonic acid (Kdo), an eight-carbon sugar coming from the bacterial synthesis of heparosan (62).

The 2D NMR spectra was divided into anomeric (C1) region that stretches from 4.2 ppm to 5.8 ppm in the ¹H dimension and 90 to 110 ppm in the ¹³C dimension, and an aliphatic (C2 through C6) region, from 3.2 ppm to 4.6 ppm in the ¹H dimension and 50 ppm to 80 ppm in the ¹³C dimension. In the anomeric region, the peaks from GlcNS, GlcNy (y = S or Ac), IdoA, and GlcA for USP heparin and BEH nicely overlaid one another, suggesting very high similarity. Nevertheless, two major differences between USP heparin and BEH were observed: the presence of GlcA2S in the BEH sample and the presence of xylose (Xyl) or galactose (Gal) in the USP heparin sample. The presence of the former in some porcine heparin samples, as well as bovine and ovine heparin, has recently been demonstrated (63), and the latter is known to be part of the linker tethering heparin to the anchor protein, which does not exist in BEH. In the aliphatic region, overlap of the C3 through C5 peaks prevented full assignment of the spectral peaks; however, this region for USP heparin and BEH nicely overlaid one another, suggesting very high similarity. Small differences in this region were attributable to the earlier noted presence of GlcA2S in BEH and the presence of Xyl or Gal in USP heparin. The C2 peaks (for GlcNS6X, GlcNS3S6X, and GlcNAc) all showed good overlap between USP-porcine heparin and BEH, as did the C6 peaks (for GlcNy3X and GlcNy3X6S). The major difference was the presence of a peak near the GlcNy3X peak that was observed in USP heparin but not in BEH and could be assigned to the C6 of a galactose linker.

Lastly, we tested the affinity of BEH for AT and platelet factor 4 (PF4) in a competitive binding study using SPR. The interaction of PF4 with heparin unveils epitopes recognized by the immune system, leading to the generation of antibodies to the complex. The SPR competition studies performed with AT (SI Appendix, Fig. S7) and with PF4 (SI Appendix, Fig. S8) showed that BEH and USP-porcine heparins had comparable binding affinities to AT and PF4. Overall, the various analytical methods confirmed that BEH was highly similar to porcine intestinal heparin in terms of physicochemical properties, composition, and biochemical activities, beyond the minimal USP monograph specifications.

Conversion of BEH into Enoxaparin.

The similarity of BEH to porcine intestinal heparin was further assessed by converting it into a LMWH similar to enoxaparin. Chemical β-elimination (64) was used to convert both USP heparin and BEH into LMWHs, with characteristics consistent with USP enoxaparin acceptance criteria (Table 2). With respect to BEH, the Mw was reduced below 5,000 Da, and 1,6-anhydro derivatives, typical of the alkaline treatment (61), totaled 19 mol%. The anti-Xa/anti-IIa ratio increased to 3.9 (from approximately 1.0 in BEH), indicating lower anti-IIa activity. 1D ¹H-NMR analysis of BEH-derived LMWH was performed to compare its structure to commercial enoxaparin (Fig. 2C). The spectra of the innovator drug, Lovenox®, and a generic enoxaparin compared well with those of LMWHs obtained through chemical β-elimination of a commercial unfractionated USP-porcine heparin (Celsus) and of BEH using our depolymerization process. Additionally, we performed fragment mapping of these LMWHs by LC-MS “bottom-up” analysis. The sequence distribution was nearly identical to that of Lovenox (SI Appendix, Fig. S9). These results further confirm that chemoenzymatically synthesized BEH has properties comparable to that of porcine mucosal USP heparin.

Table 2.

LMWH derived from both heparin sodium USP (3 lots tested) and BEH meet all key USP acceptance criteria

		Source material
Criteria	USP-Enoxaparin Specification	USP heparin	BEH
Mw (Da)	3,800–5,000	4,200–4,900	4,550
M>8,000	NMT 18%	9–14	12
M2,000-8,000	68.0–92.0%	68.5–75.0	68.0
M<2,000	12.0–20.0%	16.0–20.0	20.0
1,6-anhydro (mol%)	15–25	15–22	19
Anti-Xa activity (IU/mg)	90–125	95–119	90
Anti-Xa/Anti-IIa	3.3–5.3	3.5–4.6	3.9

Discussion

We have demonstrated the synthesis of a non-animal sourced unfractionated heparin that is biologically and compositionally similar to porcine-derived USP heparin. This bioengineered heparin was convertible into a LMWH similar to enoxaparin, further confirming its resemblance with USP unfractionated porcine heparin and expanding the commercial utility of the biosynthetic approach. The opportunity to have a highly controlled biosynthetic route to heparins is critical due to recurring spot shortages of USP heparin (23) and a series of contamination issues since 2005 (25–27), both of which present dire public health consequences. BEH was produced at the >10 g scale through a precise sequence of immobilized enzyme reactions with a tailored NSH possessing NS% comparable to porcine heparin. We envision that such a process can be further scaled up and streamlined by using various reactor designs, including large-scale batch and packed bed continuous reactions to enhance productivity.

In this work, we extensively evaluated BEH, including requirements described in the USP monograph for heparin sodium (e.g., Mw, anti-coagulant activity), as well as conversion into a LMWH that satisfy USP enoxaparin requirements (e.g., composition, sequence, immunogenicity) (12). The distribution of “major” NS containing disaccharides and tetrasaccharides, corresponding to NS (IdoA/GlcA-GlcNS), NS2S (IdoA2S/GlcA2S-GlcNS), NS6S (IdoA/GlcA-GlcNS,6S), TriS (IdoA/GlcA2S-GlcNS,6S), and T2 (IdoA/GlcA-GlcNAc,6S-GlcA-GlcNS,3S,6S) sequences, were nearly identical between BEH and USP-porcine heparin as determined by dp2 & dp4 analysis. Similar concurrence was obtained for the NAc-containing “minor” disaccharides. Longer oligosaccharide sequence information via fragment mapping was also performed and the distribution of sequences was very similar for BEH and USP porcine heparins. A somewhat greater deviation of BEH from USP heparin was a higher content of the GlcA2S residue in BEH, which did not adversely impact anti-IIa and anti-Xa activities nor did it prevent the successful generation of LMWH that met USP enoxaparin specifications. Indeed, GlcA2S has been identified in porcine intestinal heparins (65) and is present in relatively high amounts in enoxaparin. Since this monosaccharide residue solely depends on the C5-epi/2OST step, the interplay between these two enzymes remains important for future investigations.

In conclusion, we have used a chemoenzymatic, multi-step process to generate a BEH, at more than 10 g scale, with biological and chemical properties similar to that of USP heparin, and suitable for conversion to enoxaparin. The use of immobilized enzymes facilitated product recovery and will likely simplify larger-scale bioprocessing. We believe that this approach can lead to the large-scale production of a non-animal sourced heparin and its derived LMWHs. Moreover, this methodology can address other animal-derived GAG supplies, including chondroitin sulfate, dermatan, and keratan sulfates, and various heparan sulfates, which should be amenable to a similar animal-free process.

Materials and Methods

N-Sulfoheparosan (NSH) Synthesis from K5 Heparosan, Expression and Purification of Heparin Biosynthetic Enzymes, and Covalent Immobilization of Enzymes.

NSH was prepared from K5 heparosan, the capsular polysaccharide produced by E. coli K5 strain (ATCC #23506). Fermentation and purification from the culture supernatant were described previously (65). NSH was then generated by controlled, partial depolymerization, and de-N-acetylation of N-acetylglucosamine residues using NaOH, followed by complete N-sulfonation of free primary amino groups using sulfur trioxide triethylamine (Sigma, St. Louis, MO) (66). The resulting NSH contained 87 to 89% GlcNS and 11 to 13% GlcNAc, had a weight average molecular weight (M_w) ∼17,000 Da with a polydispersity index (PDI) of approximately 1.2, which is similar to that of USP heparin.

The source of parental protein sequences, N-terminal truncations, amino acid substitutions relative to the wild type (WT), fusion partners, and expression strains are summarized in SI Appendix, Table S8. Details about the protein engineering approach are available in SI Appendix. Sequences of the optimized enzymes are provided in SI Appendix, Table S9. E. coli fermentations were performed in 40 L cultures in a 100 L bioreactor (Biostat D-100) in a fed-batch mode according to published methods (67, 68). The biomass was harvested and stored at −80 °C. Enzyme purification from biomass was performed using Ni-NTA affinity chromatography. Details about the purification are provided in SI Appendix. Purified enzymes were stored at −80 °C until immobilization. Enzymes were covalently immobilized on CNBr-activated Sepharose 4B (Cytiva, Marlborough, MA) using a method adapted from the manufacturer’s instructions. Details can be found in SI Appendix.

Preparation of BEH by Enzymatic Transformation of NSH.

Multi-gram scale biocatalytic reactions were performed at 37 °C in 1-L bioreactors (Applikon MiniBio 1000 systems, Getinge, Sweden) stirred with a low-shear marine style impeller. Biocatalyst slurries (either a mixture of C5-epi and 2OST slurries, or a mixture of 6OST-1 and 6OST-3 slurries, or 3OST-1 slurry only) were washed in 50 mM MES buffer, pH 7, containing 125 mM NaCl and then added into the bioreactor vessel where they were equilibrated to 37 °C. Substrate mixtures containing the substrate (NSH for C5-epi/2OST reactions, N-sulfo, 2-O-sulfo heparosan (NS2S-Hep) for 6OST-1/6OST-3 reactions, or N-sulfo, 2-O-, 6-O-sulfo heparosan (trisulfated (TriS)-Hep) for 3OST-1 reactions) and PAPS were prepared in MES/NaCl buffer, sterilized through 0.22 µm PES filter, and brought to 37 °C. The substrate mixtures were added atop the biocatalyst slurries to start the reactions (t₀). In-process monitoring was performed by analyzing reaction samples for disaccharide composition (C5-epi/2OST and 6OST-1/6OST-3 reactions) or anti-IIa activity (3OST-1 reaction). Upon completion of the reactions, the mixtures were filtered to separate the resin from the given reaction product using EZ Flash Cartridges (Sorbtech, Norcross, GA). The resin was further washed with three column volumes of MES/NaCl buffer to collect the reaction product. The products were then desalted and concentrated using tangential flow filtration (TFF) with a 3 kDa molecular weight cut-off (MWCO) cartridge (KrosFlo® KR2i TFF System, Repligen, Waltham, MA). Desalted, concentrated products were analyzed for disaccharide composition, M_w and concentration before being prepared for the next process step.

For conversion of NSH to NS2S-Hep with immobilized C5-epi and 2OST biocatalysts, the reaction conditions were as follows: 10 g/L (~2.4 mM dp2) NSH, 25 mM PAPS, in 50 mM MES (pH 7), 125 mM NaCl, 10 mM MgCl₂, 0.05% NaN₃, and 1 mg/mL C5-epi and 1.7 mg/mL 2OST. For conversion of NS2S-Hep to TriS-Hep with immobilized 6OST-1 and 6OST-3 biocatalysts, the reaction conditions were as follows: 12 g/L (~2.4 mM dp2) NS2S-Hep, 25 mM PAPS, in 50 mM MES (pH 7), 150 mM NaCl, 0.05% NaN₃, and 0.25 mg/mL 6OST-1 and 0.25 mg/mL 6OST-3. For converting TriS-Hep to BEH with immobilized 3OST-1 biocatalyst, the reaction conditions were as follows: 40 g/L (~75 mM dp2) TriS-Hep, 7 mM PAPS, in 50 mM MES (pH 7), 150 mM NaCl, 0.05% NaN₃, and 0.5 mg/mL 3OST-1. Indicated volumes for each reaction excluded the resin volume.

BEH recovered from the 3OST-1 reaction was desalted by TFF and freeze-dried. For final purification, BEH (10.3 g) was dissolved in water for injection (WFI) then bleached using 1% hydrogen peroxide, then using 0.85% peracetic acid. Eventually, BEH was washed by ethanol precipitation and dried in a vacuum oven at 40 °C to obtain 9.0 g (90% yield) of purified BEH. Additional details about the procedure are provided in SI Appendix, as are disaccharide, tetrasaccharide compositional (dp2 and dp4 analysis), GlcA and IdoA analysis, molecular weight determination, anticoagulant activity, and NMR spectroscopy analysis.

Conversion of BEH to Enoxaparin.

LMWH was prepared through a chemical β-elimination method based on the enoxaparin process and has been previously described (65). Briefly, USP heparin or BEH was added to benzethonium chloride solution to form the quaternary ammonium salt. The heparin benzethonium salt precipitated from this aqueous solution, and the precipitate was collected and dried. Dichloromethane was used to re-dissolve the dry remodeled benzethonium salt. Benzyl chloride was added to this heparin salt, and the reaction was incubated at 35 °C for 25 h. The heparin benzyl ester was formed and recovered by filtration. Depolymerization was performed by incubating heparin benzyl ester with NaOH solution. Hydrochloric acid was used to neutralize the excess NaOH after the reaction was completed. The resulting LMWH products were precipitated using methanol, recovered by centrifugation, and dried before analysis.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.