Analysis of fungal high-mannose structures using CAZymes

Abstract

Glycoengineering ultimately allows control over glycosylation patterns to generate new glycoprotein variants with desired properties. A common challenge is glycan heterogeneity, which may affect protein function and limit the use of key techniques such as mass spectrometry. Moreover, heterologous protein expression can introduce nonnative glycan chains that may not fulfill the requirement for therapeutic proteins. One strategy to address these challenges is partial trimming or complete removal of glycan chains, which can be obtained through selective application of exoglycosidases. Here, we demonstrate an enzymatic O-deglycosylation toolbox of a GH92 α-1,2-mannosidase from Neobacillus novalis, a GH2 β-galactofuranosidase from Amesia atrobrunnea and the jack bean α-mannosidase. The extent of enzymatic O-deglycosylation was mapped against a full glycosyl linkage analysis of the O-glycosylated linker of cellobiohydrolase I from Trichoderma reesei (TrCel7A). Furthermore, the influence of deglycosylation on TrCel7A functionality was evaluated by kinetic characterization of native and O-deglycosylated forms of TrCel7A. This study expands structural knowledge on fungal O-glycosylation and presents a ready-to-use enzymatic approach for controlled O-glycan engineering in glycoproteins expressed in filamentous fungi.

Introduction

Filamentous fungi are known to be an attractive platform for heterogenous gene expression due to their ability to secrete large amounts of proteins in the growth media (Cherry and Fidantsef 2003). One common attribute of proteins produced by fungi is glycosylation. This post-translational modification occurs by attachment of a glycan chain either to the nitrogen atom of an asparagine residue in the consensus glycosylation motif Asn-X-Ser/Thr (N-glycosylation) or to the oxygen atom of a serine or a threonine residue (O-glycosylation) (Reily et al. 2019). In humans and higher eukaryotes, O-linked glycosylation is typically of the mucin- or α-dystroglycan-type (Hopkins et al. 2015), whereas in filamentous fungi, the O-glycosylation core is composed of mannose (Man) chains, optionally capped with a terminal galactofuranose (t-Galf) residues as demonstrated for some glycoproteins produced in Aspergillus spp. (Goto 2007).

Fungal expression hosts have been widely applied in the production of protein-based therapeutics for use in humans due to the overall ease of scale-up, high titers and the ability to preserve glycosylation. However, nonnative glycosylation of a heterologously produced protein can impact its half-life, cellular targeting, immunogenicity or therapeutic activity (Gerngross 2004; Gupta and Shukla 2018). Glycosylation also serves important roles in the function of industrial enzymes, including modification of activity (Adney et al. 2009; Rubio et al. 2019), proteolytic and thermal stability (Amore et al. 2017) and substrate affinity (Kołaczkowski et al. 2020). In these cases, precise identification of the purified glycoprotein and the composition of its glycoforms are required. Mass spectrometry (MS) is well suited for this purpose; however, result interpretation may be challenging due to unresolved signal in mass spectra imposed by glycan heterogeneity. This heterogeneity is especially challenging for electrospray ionization-mass spectrometry (ESI-MS), therefore the removal of glycans is often a prerequisite to obtain a lower number of possible glycoforms and higher ionization efficiency (Balaguer and Neusüss 2006).

Multiple strategies for glycoengineering of heterologous proteins are available, such as site-directed mutagenesis of consensus glycosylation motifs, manipulation of genes encoding glycosylation pathways or changing the growth medium (Gupta and Shukla 2018). After glycoprotein production, the undesired glycans can be removed using endo or exoglycosidases, without disruption of the protein backbone. Enzymatic removal of N-linked glycans is well established with enzymes like endoglycosidase H (endo H) or peptide-N-glycosidase F that hydrolyze the glycosidic bond between two N-acetylglucosamine (GlcNAc) units or the innermost GlcNAc and the asparagine residue, respectively (Freeze and Kranz 2010). Modification of the fungal O-glycosylation pattern is more challenging as there is no universal endotype enzyme. Rather, this must be approached with glycan “trimming” by application of exoglycosidases. The most commonly applied enzyme is a jack bean α-mannosidase (JBM) with broad specificity toward α-1,2/α-1,3/α-1,6 glycosidic bonds between Man residues. This enzyme is commercially available as a crude cellular extract, which often exhibits low efficiency or unwanted proteolytic side activity (Hopkins et al. 2015). Recently, α-mannosidase from Bacteroides spp. have been demonstrated to be active on α-linked mannosyl residues present in high-Man N-glycans that are often structurally similar to O-glycans from filamentous fungi (Cuskin et al. 2015; Briliūtė et al. 2019). To expand the variety of enzymes targeting glycosidic bonds in O-glycan chains, an array of exoglycosidases was screened on a well-described model glycoprotein produced in the filamentous fungus Aspergillus oryzae. Specifically, we used the cellobiohydrolase from Trichoderma reesei Cel7A (TrCel7A). TrCel7A harbors a peptide linker with a high number of serine and threonine residues for which the presence and structure of O-glycans have been characterized (Harrison et al. 1998; Amore et al. 2017). We identified an enzymatic toolbox composed of three enzymes capable of reducing O-glycosylation in fungal expressed glycoproteins: a GH92 α-1,2-mannosidase from Neobacillus novalis (NnGH92), a GH2 β-galactofuranosidase from Amesia atrobrunnea (AaGH2) and a JBM.

Results

TrCel7A wild-type (WT) and a variant TrCel7A ΔN-glyc, with the three known N-glycosylation sites removed by the point mutations (N45Q, N270Q, N384Q), were successfully expressed in A. oryzae and subsequently purified and characterized as described elsewhere (Kołaczkowski et al. 2020). To determine the extent and composition of O-glycans in the linker region of TrCel7A, O-glycomics analysis was performed. O-linked glycans were released from TrCel7A ΔN-glyc by reductive β-elimination. In this method, O-glycan release was catalyzed by addition of NaOH. In the resulting alkaline environment, the oligosaccharides are prone to degradation by “peeling” reactions targeting the reducing end of the released O-glycans. To prevent unwanted degradation, a reducing agent, NaBH₄, was employed to bind and stabilize the reducing end (Shajahan et al. 2017a; Shajahan et al. 2017b; Shajahan et al. 2020). Among the released O-glycans, only hexose (Hex) residues were identified, and these were comprised of Hex_1–14. The dominant species comprised 95% of the total O-glycans which consisted of Hex₁–Hex₅ with Hex₂ as the most prevalent O-glycan chain (Supplementary Figure S1). Monosaccharide composition analysis indicated the presence of Man and galactose (Gal) in a ratio of 1.3:1. Based on glycosyl linkage analysis, the main O-glycans components were terminal mannopyranose (t-Man, 45%), 2-linked mannopyranose at the reducing terminus (red-2-Man, 31%), 4-linked galactopyranose (4-Galp, 12%) and t-Galf (5%) residues (Supplementary Figure S2, Supplementary Table SI). The reduction (by NaBH₄) of the released O-glycans in the β-elimination reaction resulted in conversion of their proximal (“reducing end”) residues to alditols. As a result, since alditols are not in ring form, the subsequent permethylation reaction (here using red-2-Man as an example) installed methyl groups on O-1 and O-5, as well as on O-3, O-4 and O-6, but not on O-2 of this 2-linked Man residue. This is in contrast to any residues in the pyranose ring form, which all end up with acetyl groups on O-1 and O-5, regardless of whether they come from the reducing end or not. The fact that the partially methylated alditol acetate (PMAA) of red-2-Man contained five methyl groups and only one acetyl group caused it to elute much earlier than the other PMAAs in this study which had at least two acetyl groups. It also led to characteristic changes in the mass spectrum (Supplementary Figure S2, Supplementary Table SI). The structure of the individual PMAAs giving rise to the proposed linkage positions can be found in Supplementary Table SI.

Enzymes with putative activity toward the identified monosaccharide residues and their glycosyl linkages (Supplementary Table SI) were screened for their ability to trim O-glycans from the linker region of TrCel7A. These enzymes were selected based on their sequence homology to enzymes with activity toward α-linked mannopyranosyl and galactosyl residues already reported in literature (Goto 2007; Cuskin et al. 2015; Matsunaga et al. 2015). This led to the identification of a putative α-1,2-mannosidase (NnGH92) and a putative β-galactofuranosidase (AaGH2). Purity of the studied enzymes was evaluated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis (Supplementary Figure S3).

According to the CAZy database classification (Lombard et al. 2014), the GH92 family is comprised of exo-acting α-mannosidases with specificity toward α-1,2-, α-1,3-, α-1,6-linked mannooligosaccharides. These linkages are found in α-mannan, a polysaccharide attached as an N-glycan to the yeast cell wall proteins (Orlean 2012). Multiple genes encoding GH92 α-mannosidases have been identified in the human gut bacterium Bacteroides thetaiotaomicron (Zhu et al. 2010; Cuskin et al. 2015), supporting its ability to grow on yeast Saccharomyces cerevisiae α-mannan (Cuskin et al. 2015). In synergy with other α-mannan-degrading enzymes, this bacterium deconstructs yeast α-mannan to use it as a carbon source. Other bacterial GH92 α-mannosidases have been characterized which are involved in N-glycan processing (Robb et al. 2017; Li et al. 2020). N. novalis is a common soil bacterium having a gene encoding a GH92 α-1,2-mannosidase, suggesting that this bacterium might be also involved in processing of high-Man type glycans. Characterization of NnGH92 by high-performance anion-exchange chromatography (HPAEC) (see details in the SI) confirmed its α-1,2-mannosidase activity, as well as some degree of α-1,3-mannosidase activity if incubation time was prolonged significantly (17 h). No activity was seen toward α-1,6-mannobiose (Supplementary Figure S4).

GH2 is a much more diverse family with multiple activities ascribed to it, including β-galactofuranosidases and β-galactopyranosidases. The only characterized β-d-galactofuranosidase is a bacterial GH2 from Streptomyces spp., although homologs from filamentous fungi have been retrieved by blast searches (Matsunaga et al. 2015). β-galactofuranosidase activity has been reported for the filamentous fungi Aspergillus spp. (Wallis et al. 2001) and Penicillium fellutanum (Miletti et al. 1999), although no protein or gene responsible for this activity has been identified. Galf is a component of galactomannoproteins side chains (galactofurans), found in the filamentous fungal cell wall, and N- and O-glycans of extracellular enzymes (Komachi et al. 2013). The GH2 used in this study is from the filamentous fungus A. atrobrunnea, and although its natural function has not been elucidated, it is likely involved in reducing the Galf-containing glycoconjugates in certain stress-induced or unfavorable cultivation conditions.

Inspired by the published characterization of the Streptomyces spp. β-d-galactofuranosidase (Matsunaga et al. 2015), activity of AaGH2 toward 4-nitrophenyl β-d-galacto-furanoside, 4-nitrophenyl β-d-galactopyranoside and 4-nitrophenyl α-l-arabinofuranoside was determined (Supplementary Figure S5). No activity was seen toward β-d-galactopyranoside, suggesting that AaGH2 is not a β-d-galactopyranosidase, an otherwise common hydrolase activity in the GH2 family. In contrast to the GH51 and GH54 α-arabinofuranosidases from Aspergillus niger (Tefsen et al. 2012), AaGH2 activity was demonstrated only on β-d-galactofuranoside and AaGH2 did not show dual specificity for the structural similar α-l-arabinofuranoside (Supplementary Figure S5), indicating that AaGH2 may be Galf specific, like its counterpart from Streptomyces spp. A previous study using site-directed mutagenesis identified several aspartic acid and glutamic acid residues in the catalytic domain of Streptomyces spp. GH2 as having a critical function for its β-d-galactofuranosidase activity (Matsunaga et al. 2015). These residues are conserved in AaGH2, which further supports its annotation as a β-d-galactofuranosidase (Supplementary Figure S6).

To demonstrate the ability of NnGH92 and AaGH2 to hydrolyze O-linked glycans in fungal glycoproteins, TrCel7A WT was selected as a model system. N-linked glycans were removed from TrCel7A by treatment with endo H (Roche), either as a separate reaction or in combination with NnGH92 and AaGH2. The reaction mixtures were incubated for 19 h at 37°C and then analyzed by SDS-PAGE (Figure 1A) and LC–MS (Figure 1B–F).

Fig. 1.

Evaluation of NnGH92, AaGH2 and JBM GKX-5010 deglycosylation effectiveness of TrCel7A WT. (A) SDS-PAGE analysis: Coomassie blue staining (left) and periodic acid-Schiff staining method (right). Lane 1: Marker LMW (GE Healthcare) molecular weight standard; lane 2: TrCel7A; lane 3: +endo H; lane 4: +NnGH92 and AaGH2; lane 5: +endo H, NnGH92 and AaGH2. (B–F) LC–MS intact protein mass spectra of TrCel7A; (B) untreated; (C) + endo H; (D) + NnGH92 and AaGH2; (E) + NnGH92, AaGH2 and endo H; (F) JBM GKX-5010; (G) + endo H, NnGH92, AaGH2 and JBM GKX-5010. The peaks indicated with an arrow match the mass of TrCel7A + 3 × HexNAc and the corresponding number of hex. In the mass spectra (B–G), the adjacent peaks correspond to the mass difference of one hexose (162 Da). Theoretical molecular weight of TrCel7A was calculated based on the amino acid sequence without signal peptide from P62694 (UniProt entry). This figure is available in black and white in print and in color at Glycobiology online.

Each enzymatic deglycosylation step with endo H, NnGH92 and AaGH2, or a combination of both enzymatic treatments resulted in the reduction of the glycan content in TrCel7A. This was observed as a decrease in apparent molecular weight (Coomassie blue staining), with lower band intensity (acid-Schiff staining) by SDS-PAGE analysis (Figure 1A), and by changes in the mass profiles of intact protein LC–MS (Figure 1B–E). In this analysis, the mass profile of TrCel7A corresponded to the mass of the protein backbone and total number of sugar residues attached to it. However, intact protein LC–MS does not provide any information about the length or complexity of the O-glycan chains attached to TrCel7A. In TrCel7A, the deconvolution of the mass spectrum was only possible after the removal of N-glycans (Figure 1B and C), as the heterogenous nature of N-glycosylation gives a large multitude of glycoforms. Endo H-treated TrCel7A demonstrated a range of O-linked glycoforms with the highest intensity peak corresponding to the protein with the expected 3 × N-acetylhexosamine (HexNAc) positioned at the N-glycosylation sites and 19 × Hex (Figure 1C). Because endo H is active only on N-linked glycans, the remaining 19 × Hex most likely correspond to O-glycans (Freeze and Kranz 2010). Further refinement could be achieved by the additional combined activity of NnGH92 and AaGH2, which resulted in a mass spectrum with a major peak corresponding to a glycoform with 3 × HexNAc and 9 × Hex (Figure 1E). NnGH92 and AaGH2 also lowered the glycan heterogeneity of TrCel7A, leading to the presence of one dominant glycoform (Figure 1E). The enzymatic treatment with either NnGH92 or AaGH2 alone did not reach the same level of deglycosylation as when both enzymes acted simultaneously (Supplementary Figure S7). Other conditions were tested and neither pH 7.0 nor higher enzyme dose improved the overall O-deglycosylation activity of NnGH92 and AaGH2 (Supplementary Figure S8).

The remaining 9 × Hex (Figure 1E) may be distributed as single Man residues directly attached to the 2 × Ser and 7 × Thr residues in the TrCel7A linker region by Man-α-O-Ser/Thr linkages (Harrison et al. 1998; Amore et al. 2017), making them inaccessible to the NnGH92. In literature, the GH38 (either jack bean or lysosomal) α-mannosidases have been described as being able of hydrolyzing this linkage (Gomathinayagam and Hamilton 2014; Hopkins et al. 2015). To test this hypothesis, two commercially available JBMs were applied to the N-deglycosylated TrCel7A, either alone or in combination with NnGH92 and AaGH2. On its own, JBM GKX-5010 (Agilent) was able to further deglycosylate the target protein, as indicated by the appearance of a predominant peak corresponding to the TrCel7A backbone with 3 × HexNAc residues and 2 × Man residues (Figure 1F). Interestingly, the combined action of JBM GKX-5010 with NnGH92 and AaGH2 resulted in an even more uniform glycan population, harboring fewer glycoforms. This is clearly illustrated by a reduction in higher mass glycoforms following the predominant peak of TrCel7A with 3 × HexNAc +2 × Hex (Figure 1G). The addition of NnGH92 and AaGH2 in combination with endo H and JBM GKX-5010 (Figure 1G) lowered the number of higher mass glycoforms when compared with the treatment solely with endo H and JBM GKX-5010 (Figure 1F). Moreover, in Figure 1G, the intensity of peaks corresponding to the TrCel7A backbone with 3 × Hex to 5 × Hex was increased, most likely pointing at an accumulation of a subset of glycan structures, resistant to hydrolysis by the enzymatic mixture presented in this study.

The other JBM product (M6944, Sigma) showed inferior O-deglycosylation effectiveness compared with the JBM GKX-5010 (Supplementary Figure S7). After prolonged contact with JBM, the occurrence of multiple bands in SDS-PAGE suggested truncation of TrCel7A, maybe reflecting the presence of proteolytic enzymes (Gomathinayagam and Hamilton 2014) acting on the flexible, deglycosylated linker.

After the enzymatic deglycosylation of TrCel7A, the hydrolysates were analyzed using PA1/Dionex. The products released from TrCel7A by NnGH92 and AaGH2 were Man and Gal (Figure 2). The combined use of AaGH2 and NnGH92 provided a higher total release of Man compared with the sole use of the NnGH92 (Figure 2). This indicated that a subset of the O-glycan chains was shielded from the NnGH92, most likely by a steric hindrance from the t-Galf. This prevented NnGH92 from acting on these O-glycan chains, even during prolonged incubation. The observed release of Man by NnGH92 alone points at a population of O-linked glycans chains with terminal Man residues.

Fig. 2.

Product profile after enzymatic hydrolysis of endo H treated TrCel7A linker O-glycans with NnGH92 and AaGH2 measured by HPAEC-PAD Dionex PA1. (A) HPAEC-PAD traces of 3 h long incubations with NnGH92 and/or AaGH2. The top panel presents monosaccharide standards, including Gal, Glc and Man. TrCel7A panel corresponds to a negative control. The subsequent panels represent 3-h enzymatic treatments of TrCel7A by the mixture of NnGH92 and AaGH2, and either NnGH92 or AaGH2. (B) TrCel7A was incubated with either NnGH92 or AaGH2 for 3 h and 20 h and the released Man (green) and Gal (yellow) was quantified. The combined effect of NnGH92 and AaGH2 on Man and Gal release from TrCel7A was evaluated by sequential addition of the enzymes. This was done by initially running a 3-h incubation with AaGH2, followed by an extended incubation (17 h) in presence of both AaGH2 and NnGH92. This figure is available in black and white in print and in color at Glycobiology online.

The impact of reduced O-glycosylation on TrCel7A was investigated by steady-state kinetics using the model cellulosic substrate Avicel (Figure 3). This demonstrated that enzymatic deglycosylation by NnGH92 and AaGH2 did not change the kinetic profile of TrCel7A as compared with the native glycosylated TrCel7A. In particular, the changes in the Michaelis constant (K_M) were not significantly different and only a minor difference in turnover frequency (expressed as V_max/E₀) was observed for the different glycoforms.

Fig. 3.

Steady-state kinetic analysis of TrCel7A with two different glycosylation patterns: native (Fig 1B) and O-deglycosylated (NnGH92 and AaGH2). The analysis was done at 25°C on the cellulosic substrate Avicel. TrCel7A was enzymatically deglycosylated by NnGH92 and AaGH2. Error bars represent standard deviations from triplicate measurements. This figure is available in black and white in print and in color at Glycobiology online.

Discussion

In this study, the target glycoprotein, TrCel7A, was enzymatically deglycosylated through the removal of N-linked glycans by endo H and O-linked glycans by NnGH92, AaGH2 and JBM GKX-5010. In addition, O-glycans were investigated by O-glycomics, including reductive β-elimination, methylation and Gas chromatography-mass spectrometry (GC-MS) analysis. This lead to the identification of the component glycan residues and their linkage positions, and elucidation of glycosidic linkages in the O-glycan chains (Shajahan et al. 2017a). Based on these results, the predominant O-glycan structures in TrCel7A are summarized in Figure 4. From the glycosyl linkage analysis (Shajahan et al. 2017a) of reduced glycans, three types of residues are commonly identified: residues at the reducing terminus, internal residues and residues at the nonreducing terminus. Residues at the reducing terminus are directly linked to the protein backbone, whereas residues at the nonreducing terminus are positioned as the end-capping monosaccharides in the glycan chain. If a residue is not identified at a terminal position, then this residue is located in the internal part of the glycan chain between nonreducing and reducing ends. The combination of endo H, NnGH92 and AaGH2 resulted in a major homogenous population of TrCel7A with a glycoform comprised of 3 × HexNAc residues and 9 × Hex residues (Figure 1E). The α-1,2-mannosidase activity of NnGH92 (Supplementary Figure S4) suggests that Man residues in O-glycans are connected mainly by α-1,2-glycosidic bonds. This is well reflected in the glycosyl linkage analysis that indicates red-2-Man (directly attached to the protein backbone) is linked to t-Man through an α-1,2-glycosidic bond, in the predominant O-glycan chain with degrees of polymerization (DP) of 2 (structure A, Figure 4). Moreover, NnGH92 can most likely target glycosidic bonds of the internal 2,3-linked Man (structure C, Figure 4) based on its partial activity on α-1,3-mannobiose during prolonged reactions (17 h, Supplementary Figure S4).

Fig. 4.

Possible O-glycan structures from TrCel7A expressed in A. oryzae. The presented structures are based on the linkage analysis (Supplementary Fig. S1, Supplementary Table SI) and the exoglycosidase treatment. Only glycan structures having DP of 1–4 are demonstrated. The percentage values above the residues indicate their relative abundance (Supplementary Table SI). The glycans were represented by following the Symbol Nomenclature for Glycans (SNFG) guideline (Varki et al. 2015; Neelamegham et al. 2019). This figure is available in black and white in print and in color at Glycobiology online.

The activity of AaGH2 indicated the presence of t-Galf residues, which was also confirmed by the glycosyl linkage analysis (Supplementary Table SI). The t-Galf residues are presumably linked by β-1,6-glycosidic bonds as shown for other O-linked glycans from glycoproteins expressed in Aspergillus spp. (Goto 2007). t-Galf residues are unlikely to exist as DP2 glycans as it would require t-Galf to be linked through an α/β-1,2-glycosidic bond to red-2-Man, a linkage not previously reported in O-glycosylated proteins from Aspergillus spp. (Goto 2007; Komachi et al. 2013). Therefore, the t-Galf residues appear to be primarily present in longer chains (DP > 2), linked to internal residues as presented in structures D and F in Figure 4. As none of the internal residues were 5-linked (Supplementary Table SI), this indicates that t-Galf were linked through β-1,6-glycosidic bond.

The presence of the O-glycan linkage 4-linked galactopyranosyl (4-Galp, Supplementary Table SI) was not found by enzymatic hydrolysis using well-characterized β-galactopyranosidases (Supplementary Table SII). This strongly suggests that O-glycans containing 4-Galp are extended or capped with a β-1,4- or α-1,4-linked Man (structure B, Figure 4), presenting a glycosidic bond which could not be hydrolyzed by the NnGH92 or the JBM GKX-5010. The presence of a capping residue attached to 4-Galp glycan would create a glycan conformation that is not likely to be accommodated in the active site of the β-galactopyranosidase and therefore no β-galactopyranosidase activity was observed. Despite the significant presence of 4-Galp (12%), all the internal residues can be present only in longer glycan chains with (DP > 2) (12% relative abundance, Supplementary Figure S1). Therefore, 4-Galp and the other internal residues may have a rather low impact on the overall O-glycan heterogeneity.

The enzymatic hydrolysis of O-glycans with NnGH92 and AaGH2 led to the presence of a TrCel7A linker with the remaining Man residues directly linked to the Ser and Thr residues, which can be further removed by the action of JBM GKX-5010 (Figure 1G). This treatment also exposed low abundant O-glycan structures, like B, E and F (Figure 4), which were not hydrolyzed by the NnGH92, AaGH2 and JBM GKX-5010. The appearance of high intensity peaks in Figure 1G may represent an enrichment of these nonhydrolyzed O-glycan structures including DP3 chains containing 4-Galp (structure B, Figure 4).

The enzymatic toolbox defined by this study provides the necessary deglycosylation activities to reduce O-glycan heterogeneity and examine different profiles of O-glycosylation in other glycoproteins expressed in filamentous fungal hosts such as Aspergillus spp. (Figure 5). Most studies on the relationship between activity and glycosylation rely on variants with point mutations that prevent the addition of selected glycans (Beckham et al. 2012; Kołaczkowski et al. 2020). A new toolbox of enzymes that trim O-linked glycans under nondenaturing conditions allows for the characterization of proteins in their native form rather than using an “all or nothing approach” as previously discussed by Beckham et al. (Beckham et al. 2012). Individual application of either NnGH92, AaGH2 or JBM may allow better control over the type and length of O-glycan structures. Having a homogenous glycan population, as demonstrated in Figure 1E, would help to investigate the functional changes between a protein with one major glycoform and a protein with a complex mixture of glycoforms. Modification of the amount of Man and GAl present in the TrCel7A linker could simplify the functional studies of linker glycans in TrCel7A and potentially verify the computational work on this topic (Payne et al. 2013; Amore et al. 2017).

Fig. 5.

Scheme of the proposed model of enzymatic O-deglycosylation of TrCel7A WT expressed in A. oryzae. The enzyme architecture consists of the catalytic domain (CD) linked to CBM through heavily O-glycosylated linker. The O-glycosylation profile demonstrates a simplified “snapshot” of glycoforms present in the whole O-glycan population. O-linked glycans present on the TrCel7A are trimmed by NnGH92 α-1,2-mannosidase and AaGH2 β-galactofuranosidase. N-linked glycans present in the CD of TrCel7A can be removed by the action of endo H. The removal of the Man residues directly attached to the Ser/Thr residues was achieved by the application of JBM (GKX-5010). The lines between the circles colored in black, purple, yellow and gray correspond to the glycosidic bonds with configuration of α1, α2, β6 and α/β4. The glycans were represented by following the SNFG guideline (Varki et al. 2015; Neelamegham et al. 2019). This figure is available in black and white in print and in color at Glycobiology online.

Molecular dynamics (MD) simulations have shown that O-glycosylation of the linker region of TrCel7A may modulate linker binding affinity to cellulose (Amore et al. 2017). In another study (Badino et al. 2017), site-directed mutagenesis of a Thr-rich region (TTTT to PPGP) of the linker region limited the extent of O-glycosylation. This resulted in a higher turnover, but reduced substrate affinity as compared with TrCel7A with its native O-glycosylation pattern. It has been speculated that interaction between linker glycans and cellulose appears to depend on the location of O-glycosylation sites and that changes in location may change structure and dynamics of the linker (Badino et al. 2017). The functional impact of posttranslational O-deglycosylation of TrCel7A by NnGH92 and AaGH2, was investigated by steady-state kinetics using the model cellulosic substrate Avicel (Figure 3) (Kołaczkowski et al. 2020). The derived K_m and V_max of the native and deglycosylated TrCel7A (Figure 3) did not reveal any major differences at the tested conditions. This suggests that the reduced DP and high homogeneity of O-glycans in the O-deglycosylated TrCel7A (Figure 1E) do not influence kinetics. Since the majority of O-glycans are located in the linker region of TrCel7A (Harrison et al. 1998; Amore et al. 2017) and the activity of NnGH92 and AaGH2 reduce the extent of O-glycans to DP1, one could speculate that the presence of a single Man residue at each glycosylation site is sufficient to maintain the proposed linker substrate interactions as suggested by Amore et al. (2017) and Badino et al. (2017). This may indicate that the extent of O-glycosylation plays a lesser role in TrCel7A kinetics than the impact of site occupancy at every glycosylation site in the linker region.

Complete removal of O-glycosylation may not be beneficial as O-glycosylation has been shown to protect the protein structure against proteolytic attack. This has been demonstrated (Amore et al. 2017) by an attempt to express TrCel7A with a nonglycosylated linker (Thr/Ser clusters in linker mutated to Ala residues). The deglycosylated enzyme was found to lose the carbohydrate-binding module (CBM) module, indicating a proteolytic susceptibility of the linker during enzyme expression. It has been proposed that glycosylation provides a steric hindrance around the peptide backbone (Solá and Griebenow 2009). Therefore, it is likely that at least one monosaccharide residue attached to each glycosylation site in the exposed peptide, like TrCel7A linker, can provide a sufficient level of protein stability. Thus, enzymatic O-deglycosylation with NnGH92 and AaGH2 can be advantageous, as it simplifies the overall glycosylation profile while retaining some degree of protein stability. Most commercial exoglycosidases are suited for mammalian-type O-glycan modification and the selection of enzymes for glycoproteins produced in filamentous fungi remains limited. The enzymes described in this study provide a way to trim O-glycan structures attached to glycoproteins produced by filamentous fungi. This will be relevant in studies on the influence of glycosylation on biochemical properties. Moreover, a low glycan heterogeneity will enable an analysis of glycoproteins in top-down MS and improve deconvolution of mass spectra. In proteomic studies with bottom-up MS, protein coverage can also be enhanced by increasing the population of nonglycosylated peptides.

Materials and methods

Cloning, expression and purification of AaGH2 and NnGH92

The data for AaGH2 were deposited in the GenBank at National Center for Biotechnology Information (NCBI) under accession number MW316479. The AaGH2 was expressed as an extracellular enzyme in A. oryzae in a similar setup as described elsewhere (Borch et al. 2014). The fermentation broth was sterile filtrated and then 1.8 M ammonium sulphate was added. After filtration on 0.22 μm Polyethersulfone (PES) filter (Nalge Nunc International, Nalgene labware Cat. No. 595-4520), the filtrate was loaded onto a Phenyl Sepharose™ 6 Fast Flow column (high sub) (GE Healthcare, Piscataway, NJ) equilibrated in 25 mM HEPES pH 7.0 (buffer A) with 1.8 M ammonium sulphate. After loading, unbound material was washed with five column volume (CV) of equilibration buffer, and bound proteins were eluted in buffer A. The fractions were pooled and applied to a Sephadex™ G-25 (medium) (GE Healthcare, Piscataway, NJ) column equilibrated in buffer A. The fractions were applied to a SOURCE™ 15Q (GE Healthcare, Piscataway, NJ) column equilibrated in buffer A, and bound proteins were eluted with a linear gradient from 0 to 1000 mM sodium chloride over 10 CV. Fractions were analyzed by SDS-PAGE, and fractions containing the enzyme were combined.

The data for NnGH92 were deposited in the European Nucleotide Archive at EMBL-EBI under accession number LR963497.1. The NnGH92 expressed as an extracellular enzyme in Bacillus subtilis in a similar setup as described elsewhere (Jensen et al. 2010) with the following modifications. The native signal peptide was replaced by the Alcalase signal peptide followed by a histidine tag (6×His + PR) resulting in an N-terminal sequence of MKKPLGKIVASTALLISVAFSSSIASAHHHHHHPR. The fermentation broth was sterile filtrated and then 500 mM NaCl was added and adjusted to pH 7.5/NaOH. The sample was loaded onto a Ni-Sepharose™ 6 Fast Flow column (GE Healthcare, Piscataway, NJ) equilibrated in 50 mM HEPES, pH 7.5 with 500 mM NaCl (buffer A). After loading, the column was washed with 10 CV of buffer A, and bound proteins were eluted with 500 mM imidazole in buffer A. The fractions containing the enzyme were pooled and applied to a Sephadex™ G-25 (medium) (GE Healthcare, Piscataway, NJ) column equilibrated and eluted in 50 mM HEPES pH 7.5. Fractions were analyzed by SDS-PAGE, and fractions containing the enzyme were combined.

Cloning, expression and purification of TrCel7A WT and variant TrCel7A ΔN-glyc

TrCel7A enzymes were cloned and expressed in A. oryzae as described earlier (Borch et al. 2014). The changes in N-glycosylation were introduced by point mutations of asparagine to glutamine in the consensus N-glycosylation motifs. This resulted in a variant TrCel7A ΔN-glyc with the following mutations (N45Q, N270Q, N384Q). Numbering corresponds to the protein sequence of P62694 (UniProt entry). Enzyme purification was performed according to the protocol described elsewhere (Sørensen et al. 2015).

Analysis of released O-glycans from TrCel7A ΔN-glyc

O-glycans were released by reductive β-elimination reaction. The solution containing the peptides and glycopeptides was made alkaline with 250 μL 50 mM NaOH and 250 μL of a 76 mg/mL solution of sodium borohydride in 50 mM NaOH. The resulting mixture was incubated at 45°C for 18 h. The samples were cooled, neutralized by 10% acetic acid, passed through Dowex H⁺ resin column, lyophilized and borates were removed under a stream of nitrogen. The released, reduced glycans were permethylated for structural characterization by MS as reported previously (Shajahan et al. 2017a; Shajahan et al. 2017b; Shajahan et al. 2020). Briefly, the dried eluate was dissolved with dimethyl sulfoxide and methylated by using methyl iodide on Dimethyl sulfoxide (DMSO)/NaOH mixture. The reaction was quenched with water and the reaction mixture was extracted with methylene chloride and dried. The permethylated glycans were dissolved in methanol and analyzed using ESI-MS.

Glycosyl linkage analysis

For determination of sugar linkages, PMMAs were prepared from the released, reduced and permethylated O-glycans. Briefly, permethylated glycans were hydrolyzed with 2 M Trifluoroacetic acid (TFA) at 121°C for 2 h, followed by reduction with NaBH₄ in 50 mM NaOH (400 μL, 10 mg/mL) and acetylation with acetic anhydride/TFA (1:1, v/v) at 45°C for 25 min. The PMMAs thus obtained were analyzed by GC–MS.

Enzyme assay conditions

To confirm the theoretical activity of NnGH92, 66 nM of the enzyme was incubated with 1 mM mannobiose solubilized in the assay buffer (50 mM MES pH 6.0, 50 mM NaCl, 2 mM CaCl₂, 0.01% Triton X-100) for 1 h. Then the enzymes were inactivated with 0.15 M NaOH and the reaction hydrolysates were analyzed on a Dionex HPAEC (as described later).

AaGH2 activity was tested on three different 4-nitrophenyl-linked substrates: α-l-arabinofuranoside β-d-galactopyranoside, β-d-galactofuranoside. Reaction mixture was composed of 20 μL of diluted enzymes and 120 μL 1 mM substrate dissolved in the assay buffer. The reaction mixture was incubated for 15 min at 37°C and then quenched with 100 μL 1 M glycine pH 10 buffer. The absorption of the released 4-nitrophenol was measured at 405 nm using a plate reader (Spectra Max 3; Molecular Devices, Sunnyvale, CA).

Standard deglycosylation mixture contained 300 μg of TrCel7A dissolved in 50 mM sodium acetate pH 5.0 buffer with addition of 200 mM NaCl and 2 mM CaCl₂ incubated with 7.5 μg of NnGH92 and/or AaGH2. The reaction mixture had a final volume of 250 μL. Commercially available JBMs (M6944, Sigma and GKX-5010, Agilent) were applied as recommended by the manufacturer.

Intact protein mass spectrometry (LC–MS)

Enzymatically deglycosylated enzymes were analyzed for their intact molecular weight using a MAXIS II electrospray mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany). The samples were diluted to 0.1 mg/mL in 50 mM sodium acetate pH 5.0 buffer with addition of 200 mM NaCl and applied to an AdvanceBio Desalting-RP column (Agilent Technologies). Samples were eluted from the column with an acetonitrile linear gradient from 5 to 95% (v/v) and introduced to the electrospray source with a flow of 0.4 mL/min by an Ultimate 3000 LC system (Thermo Fisher Scientific). Data analysis was performed with DataAnalysis version 4.3 (Bruker Daltonik GmbH, Bremen, Germany). The molecular weight of the samples was calculated by deconvolution of the raw data in the range 30,000 to 70,000 Da.

Dionex HPAEC

Products of mannobiose and TrCel7A O-linked glycans hydrolysis were analyzed on a Dionex ICS-3000 HPAEC-PAD (Thermo Fisher Scientific, Waltham, MA) and separated on a CarboPac PA-1 column. Samples were eluted using the following multistep gradient program at a flow rate 0.8 mL min⁻¹: 15 mM NaOH (0–4.5 min), 35 mM NaOH (4.5–7 min), 25 mM NaOAc +75 mM NaOH (7–10 min), 175 mM NaOAc +75 mM NaOH (10–20 min), 15 mM NaOH (20–30 min). As standards, glucose (Glc), Man, Gal were used. The sugars were purchased from Sigma. Data were analyzed using the instrument software Chromeleon 7.2 (Thermo Fisher Scientific, Waltham, MA). Before the experiment, N-glycans were removed from TrCel7A by overnight endo H treatment and separated with 10 kDa Molecular weight cut-off (MWCO) VivaSpin columns.

Kinetic characterization of TrCel7A before and after enzymatic O-deglycosylation

Intact (Figure 1B) and O-deglycosylated (NnGH92 and AaGH2) TrCel7A was subjected to a steady-state kinetic characterization on Avicel PH101 (Sigma-Aldrich, St. Louis, MO). This characterization was done in an identical setup as described elsewhere (Kołaczkowski et al. 2020). The substrate was washed in MiliQ water and then washed in 50 mM sodium acetate buffer pH 5.0. After the washes, it was stored in the same buffer in the presence of 5 mM sodium azide. A single experimental setting was investigated in which intact and deglycosylated TrCel7A were saturated with Avicel (Michaelis–Menten kinetics). In this experiment, 230 μL substrate aliquots from the Avicel stock (vigorously stirred) with various concentrations were pipetted into 96-well plate (96F 26960 Thermo Scientific, Waltham, MA). The hydrolysis reaction was initiated by addition of 20 μL enzyme with a final concentration of 0.1 μM. The reactions were conducted for 1 h at 25°C mixed at 1100 rpm and quenched using a centrifuge for 3 min at 2500 × g. Then 60 μL supernatant was withdrawn from each reaction and mixed with 90 μL p-hydroxybenzoic acid (PAHBAH) to perform reducing sugar assay (Lever 1973; Sørensen et al. 2015). The obtained colored products were measured at 405 nm in plate reader (Spectra Max 3; Molecular Devices, Sunnyvale, CA). The absorbance readouts were recalculated to reducing sugar end concentration using a standard curve prepared with cellobiose (0–1 mM). The reaction rate curves were fitted with Michaelis–Menten equation: ν_ss/E₀ = V_maxS₀/(K_M + S₀); using Origin Pro (version 2019, OriginLab Corporation, Northampton, MA). All reactions were performed in triplicate.

Data availability statement

The data for a GH92 α-1,2-mannosidase from Neobacillus novalis (NnGH92) were deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number LR963497.1 (GenBank sequence ID). The data for GH2 β-galactofuranosidase from Amesia atrobrunnea (AaGH2) were deposited in the GenBank at NCBI under accession number MW316479. All other data are included in the main article and supplementary materials.

Funding

Roskilde University, Novozymes A/S, Innovation Fund Denmark [Grant number: 5150-00020B], the Novo Nordisk Foundation [Grant number: NNF15OC0016606 and NNFSA170028392] and the Carlsberg Foundation. This research was also supported in part by National Institute of General Medical Sciences R24 grant number R24GM137782 at the Complex Carbohydrate Research Center.

Conflict of interest statement

The authors declare the following conflicts of interest: Christian I. Jørgensen, Nikolaj Spodsberg, Mary A. Stringer, Kristian B. R. M. Krogh, Kenneth Jensen work for Novozymes A/S, a major manufacturer of industrial enzymes.

Abbreviations

CBM, carbohydrate-binding module; CD, catalytic domain; endo H, endoglycosidase H; ESI, electrospray ionization; HPAEC-PAD, high-performance anion-exchange chromatography with pulsed amperometric detection; LC, liquid chromatography; MS, mass spectrometry; GlcNAc, N-acetylglucosamine; PNGase F, peptide-N-glycosidase F; SDS-PAGE, sodium dodecyl sulphate–polyacrylamide gel electrophoresis