Enzyme immobilization offers a robust tool to scale up the production of longer, diverse, natural glycosaminoglycan oligosaccharides

Abstract

Although structurally diverse, longer glycosaminoglycan (GAG) oligosaccharides are critical to understand human biology, few are available. The major bottleneck has been the predominant production of oligosaccharides, primarily disaccharides, upon enzymatic depolymerization of GAGs. In this work, we employ enzyme immobilization to prepare hexasaccharide and longer sequences of chondroitin sulfate in good yields with reasonable homogeneity. Immobilized chondroitinase ABC displayed good efficiency, robust operational pH range, broad thermal stability, high recycle ability and excellent distribution of products in comparison to the free enzyme. Diverse sequences could be chromatographically resolved into well-defined peaks and characterized using LC-MS. Enzyme immobilization technology could enable easier access to diverse longer GAG sequences.

Introduction

Glycosaminoglycans (GAGs), especially heparin/heparan sulate (Hp/HS) and chondroitin sulfate (CS), are natural biopolymers that impact human health and disease (Rabenstein 2002; Trowbridge and Gallow 2002; Gandhi and Mancera 2008). Growing evidence points to the role of GAGs in modulating several distinct processes such as hemostasis (Huntington 2003; Desai 2004), growth (Mikami and Kitagawa 2017; Morla 2019; Vitale et al. 2019), immune response (Casu et al. 2010; Monneau et al. 2016; Collins and Troeberg 2019), tauopathy (Sugahara and Mikami 2007; Iannuzzi et al. 2015; Zhao et al. 2020) and others. Yet, of the multitude of GAG structures present in nature, only a few have been identified as very selective modulators of proteins and cells (Maimone and Tollefsen 1990; Jin et al. 1997; Patel et al. 2016; Pomin 2016; Sankaranarayanan et al. 2017; Zhao et al. 2020). The primary reason for this state of the art is the lack of availability of a library of diverse GAG sequences that can be expected to bind to proteins, e.g., tetra- (dp4), hexa- (dp6) and octa-(dp8) saccharides, with high affinity and high selectivity. Majority of the longer sequences available commercially today are repeating units of the common building blocks, which offer a reasonable starting point, but may not bind in a selective manner.

Three major approaches are available for preparation of GAG oligosaccharides including total synthesis, chemoenzymatic synthesis and enzymatic depolymerization of natural polymers (Xiao et al. 2011; Bedini and Parrilli 2012; Pomin et al. 2012; Sugiura et al. 2012; Mizumoto et al. 2013; Liu and Linhardt 2014; Brown and Kuberan 2015; Carnachan et al. 2016; Li et al. 2016; Mende et al. 2016; Kang et al. 2018). Although the synthetic approaches have many advantages, they are difficult to implement for longer oligosaccharides. Enzymatic degradation of GAGs is a relatively easier approach for preparing sequences with different chain lengths (Xiao et al. 2011; Pomin et al. 2012; Mizumoto et al. 2013; Brown and Kuberan 2015; Carnachan et al. 2016; Li et al. 2016; Kang et al. 2018). Yet, the major products observed in free solution degradation of GAGs are dp2s, which may account for as much as 90% and small amounts of dp4 and dp6. Longer sequences are typically isolated by implementing partial digestion of GAGs, which has to be monitored and controlled in a rigorous manner, thereby introducing its own challenges. Thus, despite the availability of this technology for decades, major breakthroughs have not been achieved in the rapid and readily implementable generation of libraries of GAG oligosaccharides of longer chain lengths (>dp6).

Recently, we introduced immobilization of heparinase I (Hep1), an enzyme that cleaves Hp/HS, as an approach to produce longer oligosaccharides (Bhushan et al. 2017). Immobilized Hep1 displayed “altered pH and temperature optima and a higher propensity for generation of longer chains (hexa- and octa-) with variable sulfation” in comparison to the free enzyme. This raised a simple question whether immobilization of GAG lyases could be a general approach for preparing structurally diverse, longer oligosaccharides. Herein, we show that immobilized chondroitinase ABC (chABC) depolymerizes a range of substrates to yield higher proportions of dp6, dp8 and other longer sequences. ChABC immobilization also enables reusability for more than 10 depolymerization cycles, which presents exciting possibilities of automated production for the first time.

Results and discussion

ChABC immobilization onto sepharose is quantitative and easy

Following the protocol developed for heparinase I (Bhushan et al. 2017), we immobilized commercial chABC from Proteus vulgaris (EC 4.2.2.4, Sigma) onto CNBr-activated Sepharose-4B (see detailed experimentation in Supplementary Information). The immobilization is quick under mild conditions and essentially quantitative. Briefly, the Sepharose beads were allowed to swell, washed to remove excess acid and incubated in the coupling buffer (pH 8.3) at room temperature. ChABC, dissolved in the coupling buffer, was added to the beads such that approximately 5 mL gel is covalently modified with the enzyme (~1–5 IU). The gel was shaken at 4°C for 12–16 h, which yielded nearly 100% covalent coupling as suggested by subsequent washings that displayed no chABC activity. Following blocking of remaining active CNBr groups, the gel was ready for digestion studies.

Immobilized chABC displays broader pH and temperature profile against different CS substrates

CS from different sources presents different structures and compositions. CS from bovine cartilage (CS_bc) can carry sulfation at 6-, 4-, 4,6- or no sulfation, whereas that from shark cartilage (CSsc) is predominantly 6-sulfated. In contrast, CS from bovine trachea is mostly at 4-sulfated on the GalNAc (Pomin et al. 2012; Prabhakar et al. 2005). To compare the activity of chABC from Proteus vulgaris in free and immobilized forms, we studied the digestion of the different CS substrates, as evidenced by the increase in the formation of unsaturated uronic acid chains (ΔA₂₃₂), under a broad range of pH and temperature conditions.

Figure 1A (and Figure S1) shows the effect of pH on the activity of chABC in free and immobilized forms between pH 5 and 9, wherein the temperature was maintained constant at 37°C. Immobilization uniformly displayed better activity at pHs and appeared to shift the optimum pH toward alkaline conditions (ΔpH_OPT = ~ 0.5–1.0 units) in comparison to the free chABC for all four substrates. More strikingly, the free chABC was essentially inactive at pH 5, whereas the immobilized chABC was 5–15-fold more active. On a finer point, the different biopolymers appear to be differentially cleaved at different pHs. Although both the free and immobilized chABC robustly digested CS_bc, their digestion of CS_bt and CS_sc was moderate and much lower of DS (Figure S1).

Fig. 1.

Effect of pH (A) and temperature (B) on the catalytic activity of free and immobilized chABC using chondroitin sulfate from bovine cartilage (CS_bc). Results with CS substrates from other sources are shown in Supplementary Figures S1 and S2. Average A₂₃₂ (n = 3) for CS_bc (5 mg/mL) incubated with either 0.1 IU free or immobilized chABC at 37°C is being reported. (C) Lineweaver–Burk plot and linear fit (solid line) for CS_bc digestion with free and immobilized chABC. (D) Catalytic activity of immobilized chABC as a function of repeat cycles of reaction in batch and flow processes. Average A₂₃₂ values for CS_bc (5 mg/mL) were measured at the end of incubation time (24 h). Percent residual activity was calculated by setting the measured activity in first run as 100% (n = 3). The red dotted line shows 80% activity. (E) Preparative size-exclusion chromatography of CS_bc digested chABC under either immobilized (flow) or free enzyme conditions at pH 6.5 and 37°C. The numbers on peaks correspond to degree of polymerization (dp).

One of the problems with chABC is its thermal instability (Tester et al. 2007 ; Shahaboddin et al. 2017). We studied the activity of free and immobilized chABC against the four substrates between 5 and 50°C at either pH 6.5 (CS_bc, CS_bt and CS_bc) or pH 5.5 (DS). Uniformly, immobilized chABC exhibited a higher relative activity profile at all temperatures studied compared to the free enzyme, which displays much narrower range of activity (25–37°C) (Figure 1B). The optimal temperature for both free and immobilized chABC was between 30 and 37°C for CS substrates and about 25°C for DS (Figure S2). The excellent thermal stability probably originates from the physical dampening of unfolding transitions due to covalent immobilization onto Sepharose, although this hypothesis remains to be tested.

Immobilization induces better kinetics of substrate cleavage

To study whether immobilization impacts kinetics, we measured rate of cleavage at pH 6.5 and 37°C (Figures 1C and S3). The kinetic parameters of free and immobilized chABC digesting CS_bc, CS_bt CS_sc and DS were compared using Lineweaver–Burk double reciprocal analysis (see Table SI). Immobilization lowered the Michaelis constants (K_m) of the four substrates by 0.3- to 0.6-fold, whereas it increased the maximal velocity (V_max) by 1.3- to 2.6-fold. Alternatively, both the affinity and eliminase potential of chABC for each substrate increase following immobilization. This profile of immobilized chABC was superior to immobilization by entrapment on nanoparticles (Daneshiou et al. 2017) and indicates the possible role of the Sepharose in the elimination reaction. It is likely that the hydrated environment provided by Sepharose plays a role in recognition as well as catalytic conversion of the substrates. Considering that large-scale unfolding transitions are likely to be dampened by immobilization (as above), immobilization appears to be engineering conformational catalytic movements of the right type to enhance catalytic efficiency. Thus, the nature of immobilization support is also very important.

Immobilization offers a reproducible, recyclable digestion process for scaled-up production of GAG oligosaccharides

Immobilization offers a major advantage over the free enzyme in terms of repeated catalysis, which is especially useful for scaling up product formation. We studied recyclability of immobilized chABC under two engineering formats—batch process and continuous flow process. The A₂₃₂ of CS_bc cleavage was monitored at 37°C in repeated cycles over ~20 days under both formats (Figure 1D). Briefly, an aqueous solution of CS_bc was either incubated or circulated through immobilized chABC beads or gel, respectively, monitored at the end of a 24 h period and then washed exhaustively before introducing new substrate solution for continued digestion. The results showed that chABC remained essentially fully active (~80% activity) over at least 10 cycles. Interestingly, in the flow process, chABC remained fully active over some 20 catalytic cycles, whereas no activity was observed in the batch process by then. Although both processes were fundamentally very similar, mechanical shaking (100 RPM) in the batch process may have led to degradation of beads with concomitant loss in activity.

Aliquots of reaction mixture drawn after each cycle were analyzed using reversed-phase ion-pairing (RPIP), ultrahigh performance liquid chromatography (UPLC), electrospray ionization (ESI) and mass spectrometry (MS), which is now routinely used for characterization of GAG sequences (Kuberan et al. 2002; Yang et al. 2011; Langeslay et al. 2013; Mangrum et al. 2017). The results showed that the digestion profiles of CS_bc were very similar up to at least the 10th cycle (Figure S5), which is extremely promising for the development of a reproducible engineering process for scaled-up production of natural GAG oligosaccharides. It is important to note that such a recycling process is simply not feasible for digestion under free enzyme conditions.

Immobilized chABC generates higher proportion of longer CS oligosaccharides

Our earlier work on immobilized heparinase I showed that higher proportions of longer HS sequences could be produced (Bhushan et al. 2017). To assess whether this observation is generalizable for GAG lyases, we compared CS_bc and CS_sc depolymerization with the free enzyme under 100% (1 h) and 30% digestion (15 min) and with the immobilized enzyme (24 h). The chromatographic profiles of the two CS samples showed a series of size-fractionated oligosaccharides (dp2 → dp18), whereas complete depolymerizaton using free chABC yielded ~ 86% and ~97% disaccharides for CS_bc and CS_sc, respectively, and not much of chains longer than dp6 as also shown earlier (Pomin et al. 2012). Partial depolymerization of both biopolymers yielded much improvements in dp4, dp6 and dp8 yield; however, yields for longer sequences, especially for CS_sc, were miniscule (Figure S5). In contrast, immobilization showed distinct reduction of dp2 proportions for both CS_bc and CS_sc (33% and 25%, respectively), while more than 65% were oligomeric sequences. In fact, immobilization appeared to yield significant proportions of dp4–dp18 sequences under flow conditions, especially for CS_bc (Figure S5). Alternatively, the conditions (pH, temperature, Sepharose matrix, enzyme kinetics, etc.) employed in immobilized depolymerization of CS intrinsically bias the product distribution to longer oligosaccharides.

Preparatory size-exclusion chromatography yields reasonable amounts of longer oligosaccharides in a single run

The possibility of translating analytical results to preparatory levels was evaluated by preparing a column of immobilized chABC gel (0.4 IU; 1 × 5 cm) through which CS_bc (25 mg) was circulated for 24 h at pH 6.5 and 37°C. The digested mixture was then resolved by size-exclusion matrix (Bio-Gel P-10; 120 × 1.5 cm) with continuous monitoring (A₂₃₂). Appropriate fractions were collected, desalted, lyophilized, weighed and analyzed by RPIP-UPLC ESI-MS. Figure S6 shows the RPIP-UPLC ESI-MS analysis, which shows good separation of chains of different lengths. Based on the ESI-MS spectra of different peaks present in the dp4 through dp18 fractions, we identified nearly three dozen different sequences (Figure S7). This raises the possibility of isolating these unique sequences in homogeneous form from a single preparative digestion of CS_bc.

Table I lists the isolated yields of dp2 through dp18 fractions. The yields compare favorably with the results from analytical, small-scale studies described above. More than 60% of the isolated digested mass arose from oligosaccharides (>dp4), of which dp6 through dp12 constituted nearly 46%. In comparison, free enzyme yielded miniscule amounts of sequences longer than dp6 (Table I). Further, immobilization helped isolated sequences longer than dp12 in small amounts.

Table I.

Isolated yields of oligosaccharides from preparative chABC digestion of CS_bc followed by size-exclusion chromatography^a

	Free chABC		Immobilized chABC
Fraction	Isolated yields mg (%)		Isolated yields mg (%)
Δdp2	~20 mgb (~80%c)		~8 mgb (~35%c)
Δdp4	~2.5 mg (~10%)		~3.5 mg (~15%)
Δdp6	≤1 mg		~4 mg (~16%)
Δdp8	≤ 1 mg		~4.5 mg (~18%)
Δdp10	n.d.d		~2 mg (~7%)
Δdp12	n.d.		~1 mg (~5%)
Δdp14	n.d.		~0.5 mg
Δdp16	n.d.		~0.7 mg
Δdp18	n.d.		~0.6 mg
Δdp20	n.d.		d.nq.e

^aExperimental conditions are described in the Supplementary Information section. ^bPercent yields were calculated based on total isolated amounts. ^cYields in mgs are average weights from two separate experiments. ^dNot detected. ^eDetected but not quantified.

To assess whether the sequences isolated from a single digestion run exhibit diversity of structures, we utilized a second round of SEC for some of the more promising fractions. This is in contrast to strong anion-exchange chromatography used in the literature, which is well-established as an orthogonal method for resolving GAGs (Fasciano and Danielson 2016). Using a second SEC column, we isolated three each of dp6, dp8 and dp18 sequences (Figures 2, S8 and S9) in reasonable homogeneity (>80%). These sequences carried varying levels of sulfation supporting the prediction on diversity of structures. For example, a dp18 sequence was found to have only five sulfate groups, which is significantly less than the expected average of one group per repeating unit of CS (Figure 2). Likewise, another dp18 sequence was identified as carrying 10 sulfate groups, which is one more than the average. Similar features, i.e., more or less than average sulfation, were identified for dp6 and dp8 sequences (Figures S8 and S9). Thus, in principle, enzyme immobilization and preparative digestion of CS biopolymers could result in dozens of oligosaccharide sequences that present the diversity found in nature.

Fig. 2.

ESI-MS analysis of dp18 fractions obtained from preparative size-exclusion chromatography of CS_bc digested under flow conditions with immobilized chABC (see Methods for details and Figures S8 and S9 for dp6 and dp8 results). Three relatively homogeneous (>80%) fractions obtained from a single digestion experiment correspond to different octadecasaccharide sequences. Δ refers to unsaturation at the nonreducing end. Numbers following dp (degree of polymerization) refer to length of chain (octasaccharide), number of sulfate groups and number of acetyl groups, respectively, in the sequence. # refers to the mass peak corresponding to the primary oligosaccharide identified in the fraction.

Significance

The primary innovation and value of this work originates from chABC immobilization. Although free solution enzyme digestion (full or partial) of CS biopolymer has been known for decades and has been implemented for isolation of smaller sequences, this work shows for the first time that it is feasible to isolate much longer and diverse sequences in relatively higher proportions. In this context, immobilization of GAG lyases has also been known for decades. For example, HepI was first immobilized to develop technologies for clearing heparin from blood (Yang et al. 1988). More recently, chABC was immobilized on nanoparticles of different types to enhance its thermal stability (Daneshjou et al. 2017). Yet, this work establishes a new use and feasibility of GAG lyase immobilization.

We show that the nature of immobilization matrix is important in the generation of product distribution. Sepharose appears to play a role in altering substrate affinity and catalytic activity. Alternatively, the matrix appears to be playing a role in altering substrate selectivity. Alternatively, a different matrix may yield a different product distribution. Finally, the observation that different CS substrates are cleaved to different extents at different pHs (see Figure S1) adds to the value of chABC immobilization on Sepharose. In fact, this raises the possibility that a reasonably sized library of CS oligosaccharides may be possible to prepare for the first time. More importantly, this work enables access to a library of diverse GAG oligosaccharides for many biochemistry-oriented labs, which have no means of resorting to synthetic and/or chemoenzymatic methods.

The second major value of this work is the demonstration of recyclability of immobilized enzyme. Under the flow conditions, immobilized chABC was essentially as active in the 10th cycle as in the first (Figure 1D). Even the product distribution was very similar (Figure S4). This results points to the possibility of automated generation of CS oligosaccharides. One could envisage the possibility of using several molecular membranes to enrich chains of defined lengths in an automated manner. Engineering such a process, not a trivial task by any means, would further ease the generation of a library of diverse sequences. To achieve this it would be important to identify more stable covalent linkages. One possibility is N-hydroxysuccinimide (NHS) linkage, which is likely to be more stable than the cyanate linker of CNBr-activated Sepharose.

Overall, we put forward a simple technology, which can be implemented in most biochemistry labs, for preparing longer GAG sequences with diversity in mimicking nature. We expect this technology to greatly help with the discovery of GAG structure–function relationships, which are key to developing GAG-based therapeutic agents.

Funding

This work was supported by the grants HL107152, HL090586 and HL128639 from the National Institutes of Health to URD.

Conflict of interest statement

Authors declare no conflict of interest.