Redesigning N-glycosylation sites in a GH3 β-xylosidase improves the enzymatic efficiency

Background β-Xylosidases are glycoside hydrolases (GHs) that cleave xylooligosaccharides and/or xylobiose into shorter oligosaccharides and xylose. Aspergillus nidulans is an established genetic model and good source of carbohydrate-active enzymes (CAZymes). Most fungal enzymes are N-glycosylated, which influences their secretion, stability, activity, signalization, and protease protection. A greater understanding of the N-glycosylation process would contribute to better address the current bottlenecks in obtaining high secretion yields of fungal proteins for industrial applications. Results In this study, BxlB—a highly secreted GH3 β-xylosidase from A. nidulans, presenting high activity and several N-glycosylation sites—was selected for N-glycosylation engineering. Several glycomutants were designed to investigate the influence of N-glycans on BxlB secretion and function. The non-glycosylated mutant (BxlBnon-glyc) showed similar levels of enzyme secretion and activity compared to the wild-type (BxlBwt), while a partially glycosylated mutant (BxlBN1;5;7) exhibited increased activity. Additionally, there was no enzyme secretion in the mutant in which the N-glycosylation context was changed by the introduction of four new N-glycosylation sites (BxlBCC), despite the high transcript levels. BxlBwt, BxlBnon-glyc, and BxlBN1;5;7 formed similar secondary structures, though the mutants had lower melting temperatures compared to the wild type. Six additional glycomutants were designed based on BxlBN1;5;7, to better understand its increased activity. Among them, the two glycomutants which maintained only two N-glycosylation sites each (BxlBN1;5 and BxlBN5;7) showed improved catalytic efficiency, whereas the other four mutants’ catalytic efficiencies were reduced. The N-glycosylation site N5 is important for improved BxlB catalytic efficiency, but needs to be complemented by N1 and/or N7. Molecular dynamics simulations of BxlBnon-glyc and BxlBN1;5 reveals that the mobility pattern of structural elements in the vicinity of the catalytic pocket changes upon N1 and N5 N-glycosylation sites, enhancing substrate binding properties which may underlie the observed differences in catalytic efficiency between BxlBnon-glyc and BxlBN1;5. Conclusions This study demonstrates the influence of N-glycosylation on A. nidulans BxlB production and function, reinforcing that protein glycoengineering is a promising tool for enhancing thermal stability, secretion, and enzymatic activity. Our report may also support biotechnological applications for N-glycosylation modification of other CAZymes.

enzymes were highly secreted, making them relevant targets for investigating the influence of N-glycosylation on enzymatic properties and secretion. Here, a GH3 (BxlB wt ) secreted by A. nidulans A773 with high activity toward ρ-nitrophenyl-β-d-xylopyranoside (ρNP-X), was selected as a model for N-glycosylation engineering. BxlB wt harbors seven putative N-glycosylation sites validated by LC-MS/MS. Site-directed mutagenesis was used to produce BxlB glycomutants exhibiting different N-glycosylation profiles, which were then used to investigate whether the absence or addition of specific N-glycosylation sites would influence enzyme production, secretion, kinetics, and stability. We observed two N-glycosylation sites that were important for BxlB catalytic efficiency. In addition, we observed that BxlB synthesis may be completely compromised by changing the N-glycosylation context.

BxlB wt has seven predicted N-glycosylation sites
BxlB wt is one of the five most secreted enzymes by A. nidulans when cultivated on different polymeric substrates, and is the most secreted hemicellulase during cultivation on beechwood xylan [4]. BxlB wt is a highly N-glycosylated enzyme belonging to the GH3 family, presenting seven predicted N-glycosylation sites (NetNGlyc 1.0 Server)all of them were validated by mass spectrometry: N63, N340, N408, N458, N419, N621 and N760.
N-glycosylation sequons were mutated by replacing asparagine (N) with glutamine (Q) and three BxlB glycomutants were designed: BxlB non-glyc , a non-glycosylated variant in which all N-glycosylation sites were mutated; BxlB N1;5;7 , a partially glycosylated variant in which N340, N408, N419, and N621 were mutated; and BxlB CC , a variant in which four new N-glycosylation sites were added using BxlB non-glyc as template. The design of the new sites was based on the homology with 33 β-xylosidases sequences from aspergilli (Additional file 1: Figure S1). In addition, the accessible surface area (ASA) for each new N-glycosylation site was calculated ( Fig. 1 and Additional file 1: Table S1). The design of BxlB CC enabled the explicit verification of the importance of N-glycosylation position for β-xylosidase production, secretion, and function.

Production of BxlB glycomutants in A. nidulans
All bxlB gene variants were cloned into the pEXPYR vector controlled by the glaA promoter and the glucoamylase signal peptide (SP) from A. niger, and then transformed into the A. nidulans A773 parental strain, before being cultivated with the inducer maltose [11]. Analyses of crude supernatants revealed that the non-glycosylated mutant (BxlB non-glyc ) presented similar levels of activity to the wild type (Fig. 2), while the BxlB N1;5;7 mutant showed 1.5-fold higher enzymatic activity. Although BxlB CC activity could not be detected, the identification of some peptides was verified by mass spectrometry (Additional file 1: Table S2).
In addition, very low BxlB CC intracellular activity was detected (similar to the parental strain) (Additional file 1: Figure S2), and intermediate activity levels were found for BxlB wt and BxlB non-glyc . BxlB N1;5;7 , in turn, presented fourfold higher intracellular activity in relation to BxlB wt . Real-time PCR revealed that all BxlB glycomutants were properly transcribed, and no direct relationship between gene expression and enzyme production could be observed (Additional file 1: Figure S3).
To investigate how the N-glycosylation might affect the overall flexibility of the enzyme, we performed molecular dynamics (MD) atomistic simulations of both BxlB N1;5 and BxlB non-glyc and computed the RMSF for each residue relative to the average structures (Fig. 3a). The presence of glycans at N1 and N5 increases the overall flexibility of the protein, especially in major segments that surround the catalytic site, but promotes stabilization of some structural elements of the enzyme (Fig. 3b).

The BxlB N1;5 catalytic site is more stable than in BxlB non-glyc
The stability of the catalytic site in each system was examined by computing the distances between the α-carbons of the two catalytic residues-Asp288 and Glu491along the molecular trajectories of both simulations. The average distances were 16.1 ± 1.6 Å and 18.4 ± 2.4 Å for the BxlB N1;5 and BxlB non-glyc models, respectively, with a slightly narrower distribution for BxlB N1;5 (Fig. 4). The average distance between the corresponding Asp/Glu residues in other GH3 members, as measured from the crystal structures in PDB: 1IEX [16], 2X41 [17], 3ZYZ [18], 4I3G [19], 5AE6 [to be published], and 6Q7I [20]), is approximately 13 Å.   The N-glycosylation at N1 and N5 also contributes to enhance substrate binding, as the xylobiose ligand showed higher residence times inside the catalytic pocket of BxlB N1;5 than in BxlB non-glyc . In three out of four auxiliary MD runs, xylobiose remained bound to the catalytic pocket at least twice as long in BxlB N1;5 (Additional file 1: Figure S4). This is attributed to the glycan chains that surround the catalytic pocket and hinder the release of the substrate.

N1 and N5 glycosylation affects the pattern of large-scale motions of the enzyme
The large-amplitude, low-frequency collective motions for both BxlB N1;5 and BxlB non-glyc were investigated via principal components analysis of the α-carbons Pearson cross-correlations using the MD trajectories [21]. Positive correlations denote pairs of residues that move synchronously along the same direction, while negative correlations indicate pairs of residues that move synchronously along opposite directions. If two residues move independently, then the cross-correlation between them is null. BxlB N1;5 shows a very distinct pattern of collective motions when compared to BxlB non-glyc . The glycosylation at position N1 and N5 promotes a significant increase in the overall amount of positive correlations within the structural elements of the enzyme, but also gives rise to negative correlations between spatially distant parts of the structure, which were absent in BxlB non-glyc (Fig. 5).

BxlB glycomutants showed lower secretion
BxlB secretion in A. nidulans was evaluated by Western blot (WB) using a polyclonal BxlB antibody. All the glycomutants were secreted at lower levels than that in the wild type (Fig. 6). In silico analyses indicated that N-glycosylation assisted BxlB wt folding by encouraging a more specific route with a lower free energy barrier compared to the non-glycosylated enzyme. The curve of specific heat capacity at constant volume-C v (Additional file 1: Figure S6) showed that, for the wild type, there was a higher peak and a wider width that can be understood as a less cooperative protein folding. Following the wild-type analysis in particular, the free energy profile (Additional file 1: Figure S7) indicated a small shoulder in the curve around Q = 0.7, meaning that no intermediate states were being populated in the transition state, but that perturbations along the folding process may occur, similar to what happens for the other cases. Complementing both results, the φ-value analyses (Additional file 1: Figure S8) show for all cases that residues from the N-terminal portion (1 to 400) were involved in interactions during the transition state (TS), while the C-terminal portion (from 400 to the end of the chain) interactions for the wild type in the TS were insignificant, indicating a more specific route where the N-terminal portion was formed earlier. The reason is the limitation imposed upon the residue angles due to the N-glycosylation that restricts the conformational space, entropically favoring the process. Analyzing the other cases, it is possible to see an increase of residues from the C-terminal portion involved in contacts in the TS proportional to the removal of N-glycosylation. Thus, during folding, the protein is conducted to intermediate states resulting in traps by the competition of contacts from both sides, increasing the free energy profiles and providing an explanation for the C v analysis that suggests a more cooperative process (narrowest barrier).
In silico data revealed the fundamental role of N-glycans during the protein folding and highlighted how N-glycosylation engineering can be a challenge. The evaluation was mainly focused on BxlB wt , BxlB non-glyc , BxlB N1;5;7 , BxlB N5;7 , and BxlB N1 forms. The constant volume specific heat analysis (Additional file 1: Figure  S6) showed a higher thermal stability for BxlB wt and BxlB N1;5;7 when compared to BxlB non-glyc , BxlB N5;7 and BxlB N1 , corroborating to the results obtained experimentally (Additional file 1: Table S3). The inset in Additional file 1: Figure S6 shows the superimposition of the curves by critical temperature (T c ) normalization. The width of the distribution indicates a slightly less cooperative folding process for BxlB wt , which in this case is favorable as previously presented. This result is complemented by the analysis of the free energy profiles (Additional file 1: Figure S7) and φ-values (Additional file 1: Figure S8).
BxlB wt presents a lower free energy barrier in the transition state (3.4 kT) when compared to BxlB non-glyc (3.8 kT), correlating to faster folding according to the Arrhenius equation. Even so, the BxlB non-glyc free energy profile presents an intermediate state due a less harmonious folding process by the removal of restraints (higher conformational space) imposed by the glycans (Additional file 1: Figure S7). BxlB wt has more specific folding, first forming C-terminal, for then finalize the process. Furthermore, the N-glycosylations in BxlB wt seem to stabilize the folded and unfolded states. Since simulations are in reduced units the results are qualitative.

Monitoring BxlB CC secretion
The absence of detectable BxlB CC secretion cannot be attributed to transcriptional impairment, as previously shown. Moreover, real-time PCR showed a non-significant difference in the transcription of bipA (ER chaperone) and pdiA (protein disulfide isomerase), genes normally associated with ER stress (data not shown) [23][24][25]. To further investigate the low secretion of BxlB CC , we asked if the enzyme was trapped inside the cell by WB analysis. Again, BxlB CC was detected at a very weak intensity (Additional file 1: Figure S9). Furthermore, BxlB wt and BxlB CC were detected via immunohistochemistry. The polyclonal BxlB wt antibody was identified with a red fluorescence antibody by confocal microscopy (Additional file 1: Figure S10 and Additional files 2, 3, 4, 5: Videos S1-4). Strong BxlB wt signals were detected in A. nidulans hyphae, especially in the tips, which also had a very weak BxlB CC signal, corroborating the WB.

N-glycosylation enhances BxlB folding
The BxlB wt gene from A. nidulans encoding β-xylosidase was selected from a set of 265 N-glycoproteins. The high levels of β-xylosidase secretion found when A. nidulans was cultivated on beechwood xylan provided evidence that this enzyme was highly important for biomass degradation [4]. β-Xylosidases have been detected in various prokaryotes and eukaryotes, but, according to the mycoCLAP database [26,27], only 11 eukaryotic GH3 β-xylosidases have been characterized. Among them, four β-xylosidases have been expressed in hosts such as Aspergillus spp., Pichia pastoris, and Saccharomyces cerevisiae (Additional file 1: Table S5). Here, a homologous expression system was used to prevent concerns related to accurate gene expression and protein production. In addition to the high secretion levels in the presence of lignocellulosic substrates, BxlB wt can be considered a good model for studies regarding the influence of N-glycosylation on enzyme secretion and functional parameters, due to its considerable levels of N-glycosylation and enzymatic activity.
As observed in the free energy profiles (Additional file 1: Figure S7), the wild-type protein lacks an intermediate state after the transition state peak, thus showing a more specific and favorable folding process. This result is complemented by the φ-values analysis (Additional file 1: Figure S8) that indicates the formation of the N-terminal side of the protein preceding the folding of the C-terminal portion. The presence of an intermediate state after the transition state peak in the non-glycosylated protein's free energy profile (red line in Additional file 1: Figure  S7) indicates difficulties in achieving the folded state. The φ-value analysis (Additional file 1: Figure S8) also indicates simultaneous folding of the N-and C-terminal portions, allowing for a less specific folding process that traps the protein in an intermediate state. The nonglycosylated state enables a higher number of accessible conformations, suggesting that for some mutations folding may not occur. N-glycans play an important role by interacting with polypeptide chains, favoring promising folding pathways and the stabilization of the native and unfolded states. The secretion of glycoproteins is generally facilitated by their higher stability and the presence of calnexin/calreticulin quality control [28]. Furthermore, low thermodynamic stability is known to decrease protein export efficiency-i.e., very unstable proteins are poorly exported [29]. The influence of N-glycan composition on protein folding was also studied in the human immune cell receptor adhesion domain (CD2ad), and the presence of complete N-glycans was reported to provide a fourfold acceleration in folding, in addition to stabilizing the protein structure [30].

N-glycans at positions 63, 458 and 760 are important for enzyme activity
BxlB wt , BxlB N1;5;7 , BxlB CC and BxlB non-glyc secretion levels were evaluated after their homologous expression in A. nidulans. Activity assays of the extracellular enzymes revealed that complete deglycosylation did not affect BxlB non-glyc activity. On the other hand, the improvement in BxlB N1;5;7 enzymatic activity compared to the wild type denotes that the N-glycosylation sites at N63, N458, and N760 are important for protein folding and export. Moreover, BxlB CC activity and secretion were not detected in the extracellular fraction, indicating an impairment in either gene expression, protein production, or secretion. Real-time PCR showed high transcription levels for wild type, BxlB N1;5;7 , BxlB CC , and BxlB non-glyc , suggesting that transcription is not a bottleneck for BxlB CC production. There are few studies dealing with the influence of N-glycans on protein secretion, and variable results have been found. For example, changing two N-glycosylation sites (N14Q and N48Q) of a heterologous lipase expressed in P. pastoris did not affect its secretion, while the N60Q mutation completely abolished secretion [12]. Hence, it is possible that N-glycans attached to specific positions in the protein are be essential to folding kinetics and to the secretion pathway quality control.

BxlB wt deglycosylation reduces thermal stability
Secondary structure analyses of BxlB wt and glycomutants revealed that all recombinant enzymes shared a classic profile, exhibiting α-helices as the predominant secondary structure (Additional file 1: Figure S5) [31]. Similarities in CD data were confirmed by deconvolution analysis, which found the same α-helix and β-strand rates (Additional file 1: Table S3). The deconvolution data were highly similar to what was found in A. niger β-xylosidase composed of 41% α-helix and 16% β-sheet; however, there was not as much similarity to the T. reesei β-xylosidase which presented 23% α-helix and 27% β-sheet composition [32,33].
Generally, glycoproteins have higher stability than their partially glycosylated or non-glycosylated counterparts, despite the absence of structural links associated with N-glycosylation [34]. This was also observed in our study where all BxlB mutants were less stable when compared with the wild type (Additional file 1: Table S3). Recently, thermal stability was evaluated in T. reesei cellobiohydrolase (TrCel7A) by differential scanning calorimetry, with lower stability reported for all 15 N-glycosylation mutants [35]. Interactions between N-glycan sugars and amino acid residues often stabilize the protein structure; consequently, the glycosylated forms of a protein are more stable than the non-glycosylated counterpart [34,36].

Glycosylation at N63, N458, and N760 enhances BxlB kinetic parameters
The optimal activity conditions for all BxlB glycomutants were pH 5.0 and 60 °C using ρNP-X as substrate. The same enzyme expressed in P. pastoris showed optimal activity at pH 4.4 and 48 °C using rye arabinoxylan and xylohexaose as substrates [37]. This difference in reaction conditions due to N-glycosylation was also reported for β-glucosidase from A. terreus expressed in P. pastoris and T. reesei [13]. Kinetic parameters showed that the removal of any N-glycosylation site decreased BxlB wt affinity for ρNP-X, indicating that N-glycans may facilitate substrate recognition or affect catalytic site flexibility (Km values in Table 1). Structural dynamics can influence positions and flexibility of catalytic residues, impacting kinetic parameters [38]. Wei et al. showed that removing some N-glycosylation sites decreased the specific activity of A. terreus β-glucosidase produced in filamentous fungi or yeast [13]. In accordance with this observation, we demonstrated that the removal of four N-glycosylation sites to form BxlB N1;5;7 increased V max and K m , suggesting important changes to structural dynamics.

N5 glycosylation site is essential to BxlB catalytic efficiency
Six additional mutants were designed to further understand the influence of each N-glycosylation site on folding, secretion, and BxlB wt functional parameters. BxlB N1 secretion and catalytic efficiency decreased in comparison to BxlB non-glyc . Similarly, BxlB N1;5;7 secretion and catalytic efficiency decreased in comparison to BxlB N5;7 . These data suggest that the N1 site has a slight negative influence on BxlB wt kinetic parameters and secretion levels. However, for the other glycomutants that contained a single N-glycosylation site, maintaining N5 or N7 sites, showed a drastic negative influence on enzyme secretion and function. These alterations in secretion and enzyme activity were probably impacted by the N-glycan position in the 3D structure, as well as protein dynamics and stability [38,39]. MD analysis of the catalytic site suggests that the N-glycosylation at positions N1 and N5 creates a net interaction that stabilizes the relative position of the catalytic residues, even though these two residues are not in the segments of the protein that had their flexibility mostly affected by the presence of N-glycans. That is, the suggested stabilization of the general acid/base catalytic pair may stem from the confluence of multiple longrange interactions, in which the N-glycans play a major role.
The highest activities were observed in the presence of N5 along with either N1 or N7, indicating that N5 is essential for BxlB dynamics, but is structurally stabilized by N1 or N7. As in the case of the essential N-glycosylation site in BxlB wt , previous studies have also demonstrated the critical importance of a single N-glycosylation site on folding and enzymatic activity. In P. pastoris BglS, for instance, N224 is important for enzyme activity and production, and N428 is essential for N-acetylglucosamine-6-sulfotransferase-1 activity [13,40]. An increased number of positive correlations throughout the enzyme structure in BxlB N1;5 endorses the view that glycosylation may lead to a higher number of intramolecular contactsmainly van der Waals packing contacts [41]. As a result, the protein movements acquire more of a rigid-body quality when compared to BxlB non-glyc [42,43]. However, the N1 and N5 N-glycosylation pattern also introduced negative correlations in the BxlB structure, suggesting an important change in the pattern of the essential dynamics of the enzyme that may underlie substrate binding, product release, and the observed differences in catalytic efficiency between BxlB N1;5 and BxlB non-glyc . Additional simulations, with different glycosylation patterns and different conditions, would be necessary to further analyze the role of N-glycans in BxlB.

Changing BxlB wt N-glycosylation context impairs protein folding and secretion
Our results show that the mispositioning of N-glycosylation sites impairs the production of recombinant enzymes in A. nidulans. To understand the impairment of BxlB CC production, microscopy experiments to monitor enzyme secretion were performed showing that BxlB CC mutations affect folding and secretion. The importance of balance in translation phases has been previously reported [44]. Thus, BxlB CC misfolding likely triggered protein degradation, resulting in the secretion of only a small amount of a non-functional enzyme.
Proteins with difficulty in achieving the correct folding remain longer in the ER, which maintains strict quality control, either actively facilitating folding or degrading directing misfolded proteins [28,45]. Quality control in A. nidulans is performed by the calnexin cycle. Calnexin recognizes N-glycans attached to proteins and assists disulfide bond formation, consequently facilitating correct folding and secretion [46,47]. Here, confocal microscopy showed that BxlB CC is retained in the intracellular fraction, suggesting that the mispositioning of N-glycosylation sites flagged the protein for degradation.

Conclusions
The study of the effects of N-glycosylation patterns on fungal CAZymes contributes to understanding how N-glycan positioning affects protein structure, dynamics, stability, and function. Here, we demonstrate that N-glycosylation facilitates the correct folding of a GH3 β-xylosidase from A. nidulans by eliminating an intermediate state. Despite the relatively unaffected transcriptional levels, changing the N-glycosylation context hampered enzyme secretion. Secondary structures were preserved in BxlB glycomutants, but thermal stabilities were reduced. While BxlB wt secretion and kinetic parameters were affected by N-glycosylation site mutations, the non-glycosylated enzyme (BxlB non-glyc ) was secreted in a functional state. There is strong evidence that mispositioning BxlB N-glycosylation sites (BxlB CC ) results in unfolded/misfolded non-functional enzyme. At an individual level, the N-glycosylation site N5 is essential for BxlB improved enzyme catalytic efficiency, but requires additional complementation by N1 and/or N7 sites. Moreover, BxlB glycomutants demonstrate the complexity of the N-glycosylation effect, positively and/or negatively affecting the folding process, secretion, and kinetic parameters. N-glycosylation engineering is a promising tool for enhancing target enzyme secretion, activity, and thermal stability.

Site-directed mutagenesis of GH3 β-xylosidase
The bxlB wt gene was amplified from A. nidulans A773 genomic DNA by PCR using specific primers (Additional file 1: Table S6). The gene was cloned into the pEXPYR shuttle vector using the NotI and XbaI sites, as previously described [11].

Gene cloning and A. nidulans transformation
Genes cloned into the pEXPYR plasmid were transformed into calcium competent E. coli cells by heat shock, and confirmation was carried out using colony PCR. Positive colonies were cultivated overnight and then plasmids were extracted and used for fungal transformation.
Fungal transformation was carried out as previously described, with modifications [49]. Conidia from A. nidulans A773 were inoculated on YG medium (20 g·L −1 glucose, 5 g·L −1 yeast extract, 1 × trace elements, 1 mg·L −1 pyridoxine, 2.5 mg·L −1 uracil/uridine) and were incubated at 30 °C and 130 rpm for 13 h. Protoplasts were prepared by hydrolysis of the fungal cell wall with 125 mg lysozyme from chicken egg white (L7651, Sigma) and 1.02 g Vinotaste Pro (Novozymes) for 2 h. Approximately, 10 µg of recombinant DNA was mixed with the protoplast solution and 25% PEG (w/v). Protoplast recovery was performed on MM plates supplemented with 1.2 M sorbitol and pyridoxine, incubated at 37 °C for 3 days.

LC-MS/MS validation of N-glycosylated sites
Purified N-glycomutants were incubated with 10 units of endoglycosidase-H (New England Biolabs) at 37 °C during 24 h for deglycosylation under denaturing conditions. Samples were then prepared for LC-MS/MS analysis as previously described [4]. To increase coverage and expand the number of peptide sequences, protein mutants were digested in gel with either 1 µg trypsin (Promega) at 37 °C overnight or 1 µg GluC (promega) at 37 °C overnight, and both simultaneously. The resulting peptides were concentrated by SpeedVac and 1 µL of each sample was injected for identification. Peptide loads were separated in an ACQUITY UPLC HSS T3 nano-ACQUITY Column (100 Å, 1.8 µm, 75 µm X 150 mm, Waters Corporation) in one dimension across three steps of acetonitrile (7%, 40%, 85%) over 36 min at a flow rate of 500 nL/min directly into a Synapt G2-Si mass spectrometer (Waters Corp., Milford, USA). MS/MS analyses were performed by nano-electrospray ionization in positive ion mode with a NanoLock Spray (Waters Corporation) ionization source and collected in DIA mode. The lock mass channel was sampled every 30 s. The spectrometer was calibrated with an MS/MS spectrum of [Glu1]fibrinopeptide B human (Glu-Fib) solution (100 fmol/µL) delivered through the reference sprayer of the NanoLock Spray source.
The N-glycosylation sites were validated by N-acetylglucosamine detection on asparagine residues by adding N-acetylglucosamine (N) as a variable modification (N + 203.0794) to the identification step performed by ProteinLynx Global Server (version 3.0.2; Waters Corp.). A manually created databank was used for identification with the peptide sequences of all expected protein variants. The FDR was set to 4% and was calculated using a reverse databank created on-the-fly. Relevant sequence coverage was validated within PLGS software (Additional file 6: Table S7).

Structure based models and computational analysis
Structure Based Models (SBM) [50] were employed to evaluate the folding mechanisms and stability of non-glycosylated (BxlB non-glyc ) and wild-type (BxlB wt ) enzymes. In this model, the enzyme is represented by C α -beads and the N-glycosylations by beads in the center of mass of the saccharide rings and attached to amino acid (see details in SI). Although simple, C α -SBM is a suitable choice since they accurately describe the folding mechanisms for several proteins [51]. The native structure (or functional tridimensional form) was constructed using Swiss Model [52], based on A. nidulans β-xylosidase XlnD PDB 6Q7I (45.5% identity and 62.4% similarity).
The contact map of C α -SBM was defined by CSUalgorithm [51] and N-glycans were not considered. The simulations were carried out using the molecular dynamics software developed and tested by Whitford et al. [53]. Free energy (F) profiles and constant volume specific heat values were calculated using the weighted histogram analysis method (WHAM) [54] through simulations at various temperatures. The number of native contacts formed during the simulations (Q) was employed as a reaction coordinate to describe folding mechanisms. When Q is normalized by the total number of native contacts possible, values close to 1 indicate the ensemble of native conformations (nativeness).
The evaluation of how mutations interfere in the folding pathways was done by calculating φ-values, using the following equation [51,55]: where ΔF is the free energy variation calculated by perturbation theory, i.e., evaluating the free energy difference of contacts in the presence and absence of residue i. These differences are related to the transition state (ΔF TS ), native state (ΔF N ) and unfolded state (ΔF U ). φ-values close to 1 suggest poor mutation choices, since they can destabilize the transition state or folded state, affecting the folding pathway or folding rate. In some cases, the mutation of these residues can strongly interfere with the folding process, not allowing the protein to reach its functional state. Values close to 0 suggest that mutations of these residues may not interfere in the folding process, making them good candidates for mutations, especially for protein engineering using protein-surface residues. Furthermore, the φ-value calculations help to understand folding patterns, since they show contacts formed during the transition state for each residue.

Molecular dynamics simulations
Molecular dynamics simulations were performed on the BxlB in both its N1;5 and non-glycosylated forms using the NAMD package [56] with CHARMM36 [57] force field for proteins and carbohydrates, and TIP3P water model [58]. The BxlB structure was modeled using the I-TASSER server [59], with the PDB structure 6Q7I [20] (45.5% identity and 62.4% similarity to BxlB) as a template and further refined with the GalaxyWEB server [60]. The hydrogen atoms of the enzyme were added according to the protonation state predicted by the H++ server [61,62] at pH 5.5. The following residues were considered protonated (i.e., Asp/Glu residues electrically neutral and His residues positively charged): His67, His88, His144, Asp148, His175, His275, Asp288, His299, His300, Glu333, Glu489, Glu490, Glu510, Asp530, Glu582, Glu605, His616, and Asp649. The BxlB N1;5 glycosylated structure was modeled using the GLYCAM server [63], by covalently linking a Man 5 GlcNAc 2 moiety to the Asn63 and Asn458 residues. Both BxlB N1;5 and BxlB non-glyc structures were modeled with a xylobiose molecule in the active site, which was obtained from the PDB structure 5AE6 [to be published]. The simulation boxes contained, respectively to BxlB N1;5 and BxlB non-glyc , 31,178 and 29,621 water molecules, 119 and 115 sodium ions, and 88 and 84 chloride ions. Chloride and sodium ions were added to render the system electrically neutral at a salt concentration 0.15 mol L −1 . The simulations were performed using periodic boundary conditions and the particle mesh Ewald algorithm [64] with a 12 Å cutoff and a smoothing function starting at 10 Å. Bonds involving hydrogen atoms were constrained at their equilibrium values and the simulation time step was 2.0 fs. All simulations were performed at a constant temperature of 310 K and a constant pressure of 1 bar, using the Langevin thermostat and pressure control. After initial relaxation and thermalization, both glycosylated and non-glycosylated systems underwent a production run of 400 ns each. Three additional runs 50 ns-long were carried out with both systems to check for unbinding events.

Biochemical characterization of β-xylosidase and glycomutants Enzymatic activity and protein determination
Fresh conidia from A. nidulans A773 and recombinant strains (10 7 -10 8 per ml) were inoculated for 36 h at 37 °C in MM pH 6.5 supplemented with 2% (w/v) maltose. After growth, the mycelial mass was separated by filtration and the crude filtrate was centrifuged (10,000×g, 20 min 4 °C); mycelia were then washed and the intracellular content was extracted by maceration with liquid nitrogen, which was suspended in RIPA buffer and centrifuged (10,000×g, 20 min 4 °C). Protein concentration in the crude (non-purified) extracellular filtrate, as well as in the intracellular extract, was quantified by the Bradford method [65] and BCA method [66], respectively. About 20 µg of total protein from the supernatant was run in 10% SDS-PAGE [67]. Standard β-xylosidase activity was determined in 0.05 M ammonium acetate buffer pH 5.0 at 50 °C using 5 mM ρNP-X as a substrate. The reaction was stopped with 1 M sodium bicarbonate and the released ρ-nitrophenolate (ρNP) was determined at 400 nm. One unit (U) of β-xylosidase activity was defined as the amount of enzyme releasing 1 µmol of ρNP per minute under the assay conditions.

Structural characterization by circular dichroism (CD) spectroscopy
CD analysis was carried out using a JASCO 815 spectropolarimeter (JASCO Inc., Tokyo, Japan) equipped with a Peltier temperature control unit and a 0.1 cm path length cuvette, as previously described [69]. Data from 260 to 190 nm were collected at 100 nm/min scanning speed, 1 nm spectral bandwidth, and 0.5 s response time. Melting temperatures were evaluated by spectral measurement from 20 to 100 °C.

Effect of temperature and pH on enzyme activity
To determine the optimal temperatures of all recombinant proteins, enzyme activity was assayed as described above from 35 to 70 °C in 0.05 M ammonium acetate buffer, pH 5.0. Optimal pH was calculated by ranging the pH of a 0.1 M glycine-citrate-phosphate buffer from 3.0 to 12.5 at 50 °C, using ρNP-X as the substrate.

Kinetic parameters
Maximum velocity (V max ), Michaelis-Menten constant (K m ), the catalytic constant (k cat ), and catalytic efficiency (k cat /K m ) were determined for each of the mutants using different ρNP-X concentrations (1-20 mM). These analyses were performed at optimal conditions, 0.05 M ammonium acetate buffer, pH 5.0, and 60 °C.

RNA extraction, transcript analysis by qPCR (quantitative real-time PCR), and primer design
Aspergillus nidulans total RNA was extracted by first grinding mycelia frozen in liquid nitrogen with a mortar and pestle, then collecting with a Direct-Zol RNA Miniprep kit (Zymo Research), according to the manufacturer's instructions. Total RNA (DNA-free) was assayed for reverse transcription using the Maxima First Strand cDNA Synthesis Kit for RT-qPCR, with dsDNase (Thermo Fisher Scientific). cDNA samples were diluted, and each qPCR reaction containing cDNA (100 ng), SYBR Green (Life Technologies), forward and reverse primers (Additional file 1: Table S6), and nuclease-free water was carried out using the ViiA ™ 7 real-time PCR system (Life Technologies). All PCR reactions were performed in biological triplicate. Gene expression levels were determined using the ΔΔCt method with β-tubulin (tubC) as the reference gene.

Confocal microscopy
Microscopic analyses were carried out by confocal microscopy based on Fischer-Parton et al. with modifications [70]. Mycelium was obtained after cultivation of A. nidulans strains for 24 h and 48 h, at 37 °C and 200 rpm in MM. The images were obtained by an LSM 510 Axiovert 200 M (Carl Zeiss) confocal inverted microscope, fitted with an argon laser. Laser parameters were, respectively to cyan and red fluorescence, 365 and 590 nm excitation and 440 and 617 nm fluorescence emission, using the A. nidulans A773 strain to calibrate autofluorescence. All samples were spread over a microscope slide, in natura, then analyzed immediately after cultivation.

Western blot
WB analysis was carried out as described by Nutzmann et al. [71]. Total protein (60 μg) was separated by SDS-PAGE and then transferred to PVDF membrane using a wet blotting system (Bio-Rad). The membrane was blocked with 5% BSA, incubated overnight with primary antibody (BxlB) and then gently shaken at room temperature following the addition of immunoglobulin G anti-rabbit secondary antibody labeled with peroxidase. Protein detection was carried out using the Clarity Western ECL Substrate chemiluminescence detection kit (Bio-Rad), as described by the manufacturer.