Expression of the full-length and deletion mutant GAs in S. cerevisiae
To investigate the function of the linker region of RoGA, several constructs containing full-length, truncated or partially deleted GA fragments were generated (Figure 1). All constructs used in this study contained a leader signal sequence of the N-terminal 25 residues of RoGA. When the full-length GA was expressed in S. cerevisiae, secreted enzyme activity was detected in the culture medium following incubation at 30°C (Figure 2A, lane 2 and 2B, lower panel, lane 2). In contrast, removal of the interdomain 36-amino acid linker of RoGA (GAΔ132–167) led to generation of a mutant with no detectable secreted GA as determined by plate assay or Western blot analysis (Figure 2A, lane 3 and 2B, lower panel, lane 3), suggesting that the linker played an important role in the function of RoGA. To further examine whether the linker region was essential for the formation of active starch-binding or catalytic domains, four deletion mutants devoid of either the substrate binding or catalytic domain in the presence or absence of the linker region were engineered to generate plasmids pS1-GAΔ168–604 encoding the SBD with the C-terminal linker region; pS1-GAΔ132–604 encoding the SBD alone; pS1-GAΔ26–131 encoding the linker region preceding the catalytic domain; and pS1-GAΔ26–167 encoding the catalytic domain only. It was found that two SBD-containing clones, one, GAΔ168–604, containing the linker sequence and the other, GAΔ132–604, without the linker sequence were both successfully expressed and secreted (Figure 2B, upper panel, lanes 3 and 4). Since no effect was observed in the absence of the linker in both cases, the linker region appeared to be not essential for the formation and secretion of the SBD. As for the catalytic domain-containing variants, GAΔ26–131 was successfully expressed and secreted in yeast, but GAΔ26–167 lacking of the linker sequence was not expressed at all (Figure 2B, lower panel, lanes 4 and 7), indicating that the linker region was specifically required for formation and secretion of the catalytic domain. To further define which region in the linker was crucial for its role in facilitating formation and secretion of the catalytic domain, several constructs encoding partial deletions of the linker region fused to the catalytic domain were subsequently expressed and characterized. Interestingly, both GAΔ26–145 and GAΔ26–160, unlike GAΔ26–167, were successfully expressed in secretory forms (Figure 2B, lower panel, lanes 5 and 6), indicating that the linker region from residues 161–167 was essential for in vivo folding and secretion of the active catalytic domain. It is of interest that a recently published article  on Rhizopus oryzar found a second GA gene, amyB, lacks the N-terminal SBD. The amyB gene contains the conserved residues in the catalytic domain important for starch hydrolysis, and the gene also contains a stretch of linker sequence (33 bp) preceding the catalytic domain. This finding also indicates that the small region of linker plays an important role in function of RoGA.
Characterization of recombinant GAs
In order to characterize RoGA and to study the function of the linker region, the recombinant proteins including full-length GA, GAΔ168–604, GAΔ132–604, GAΔ26–131, GAΔ26–145, and GAΔ26–160 were individually expressed in the budding yeast S. cerevisiae and were grown in SD media at 30°C. Extracellular (secreted) proteins were concentrated from culture supernatants and purified by cation-exchange chromatography. As shown in Figure 3A, purified full-length GA, GAΔ168–604, GAΔ132–604, GAΔ26–131, GAΔ26–145, and GAΔ26–160 exhibited single protein bands with purity higher than 90%. SDS-PAGE and/or mass spectrometric determination revealed that the respective molecular mass of full-length GA, GAΔ168–604, GAΔ132–604, GAΔ26–131, GAΔ26–145, and GAΔ26–160 was approximately 78, 21, 12, 69, 60, and 56 kDa (Table 2). The differences between the calculated and observed molecular weights ranged from 3% to 28%, indicating that post-translational modifications occurred in some of the recombinant proteins, presumably due to differential degree of glycosylation. There are two main types of protein glycosylation: N-glycosylation and O-glycosylation. The former refers to the attachment of oligosaccharides to a protein through the amide of asparagine residues, whereas the latter involves attachment of sugars to the hydroxyamino acids serine and threonine via their hydroxyl groups . Potential N-glycosylation sites in RoGA analyzed using the on-line prediction server NetNGlyc version 1.0  revealed that RoGA contained three putative N-linked glycosylation consensus sites, one (Asn167-Ser-Thr) located in the linker region and two (Asn230-Thr-Thr and Asn236-Lys-Thr) in the catalytic domain (Figure 1). Although no specific consensus sequence for O-linked glycosylation has yet been reported [33,34], O-glycosylation may occur on any of the Ser or Thr residues within a short peptide region . Interestingly, the program NetOglyc version 3.1  predicted that potential O-glycosylation sites in RoGA were clustered only in the linker region. Of the residues in the 36-amino-acid linker region, 61% were Ser and Thr, among which some might serve as potential sites for O-linked glycosylation. MALDI fingerprint mass spectrum from the tryptic digestion fragments of RoGA (data not shown) also showed that no peptide fragment was mapped to the linker region, presumably due to the presence of heterogeneous glycosylation within this region.
To determine how much of the molecular weight discrepancy for each recombinant protein was derived from the addition of glycosyl groups, each protein was enzymatically deglycosylated by PNGase F or Jack bean α-mannosidase followed by SDS-PAGE analysis. PNGase F is an amidase that cleaves between the innermost GlcNAc and the asparagine residue of complex N-linked oligosaccharides from glycoproteins . The fully glycosylated RoGA possessed a molecular mass of 78 kDa. After removal of N-linked carbohydrates with PNGase F, its molecular mass was reduced to 65 kDa (Figure 3B), indicating that RoGA produced by S. cerevisiae was highly glycosylated and the N-linked carbohydrates contributed approximately 13 kDa to the molecular weight. Enzymatic deglycosylation of asparagine-linked glycans in the catalytic domain variants GAΔ26–131, GAΔ26–145, and GAΔ26–160 also showed that these proteins were efficiently glycosylated to various degrees (Figure 3B).
The linker sequences code for a 36 amino acid extension at the C-terminal end of GAΔ168–604, which correspond to a molecular mass of approximately 3.5 kDa. The difference in molecular mass between GAΔ168–604 and GAΔ132–604 was 8.7 kDa, implying that the linker region contributed to the hyperglycosylation in GAΔ168–604. However, treatment of GAΔ132–604 and GAΔ168–604 with the N-glycan-specific PNGase F did not reduce its apparent molecular weight; hence there seemed to be little or no N-glycosylation on either protein (data not shown). At present, no enzyme comparable to PNGase F is available for removing intact O-linked sugars . In general, O-glycans are short linear oligosaccharides consisting of one to five mannose residues in yeast [24,34,39]. Therefore, α-mannosidase, an exoglycosidase capable of removing terminal mannose, was used to treat the recombinant proteins. If O-linked carbohydrates were present, the overall mass of the protein would decrease after treatment with α-mannosidase. The SBD variant GAΔ132–604 did not respond to the treatment with α-mannosidase, such that no change was detected in the electrophoretic mobility (Figure 3C), suggesting that no major O-glycosylation occurred in this case. On the other hand, a clear shift in the mobility of GAΔ168–604 was observed after treatment with α-mannosidase, indicating that presence of the linker region led to a considerable degree of modification by high-mannose-type O-glycans (Figure 3C). Taken together, our data indicate that the linker region of RoGA is modified by both N- and O-linked glycosylation, but the exact positions and functions are not clearly understood and require further investigation.
Effects of mutations in the linker region of RoGA
The minimal active linker region Ala161 to Asn167 was previously demonstrated to be extremely important for GA function. In addition, N-terminal sequencing analysis of the GAΔ26–160 yielded the following sequence: Ala161-Thr162-Phe163-Pro164-Xaa165-Gly166-Xaa167, where Xaa represented unidentified residues. Generally speaking, biochemically modified amino acid residues were not identifiable by regular Edman degradation; hence residues Thr165 and Asn167 were possibly modified after translation. To abolish the glycosylation and evaluate the consequence, Thr165 and Asn167 in the linker region were individually mutated to generate T165A and N167D, respectively, and their effects on protein secretion and enzymatic activity were examined at 30°C. T165A mutation showed no effect on the phenotype as compared with wild-type full-length GA (Figure 4A and 4B, upper panel, lanes 2 and 5), while a significant decrease in GA expression and function was observed in the N167D mutant (Figure 4A and 4B, upper panel, lane 6). Site-specific mutation at Asn167 led to much less active GA secretion, presumably due to either misfolding of RoGA, or faster degradation of the nascent protein. In support of the misfolding hypothesis, several different methods were used to improve the production of the mutant GA, N167D. One evidence showed that N167D exhibited enhanced expression at 20°C (Figure 4B, lower panel, lane 6), and about 0.5 mg protein could be obtained from each liter of yeast culture at 20°C. The activity of the N167D mutant could also be detected directly on starch-agar plate with prolonged incubation at 4°C (Figure 4C, No. 6). Similar phenomena were also observed for the mutants GAΔ132–167 and GAΔ26–167 after prolonged incubation at low temperature (Figure 4C, No. 3 and No. 4).
Moreover, elimination of the N-glycosylation site at position Asn167 (N167D) resulted in a faster relative electrophoretic mobility and lower apparent molecular weight (Figure 4D), providing additional line of evidence that the only putative N-glycosylation site in the linker region, Asn167-Ser-Thr, was indeed modified in S. cerevisiae. The biological function of carbohydrates in GA is not, however, completely understood, although it directs protein folding, facilitates secretion, and enhances stability of Aspergillus GAs . In this study, we have demonstrated that N-glycosylation at Asn167 in the linker is required for functional expression of RoGA, whereas mutation abolishing N-glycosylation in the linker, N167D, decreased the efficiency of protein expression. Furthermore, similar result was obtained when Asn167 was converted to Gln (data not shown). N-linked carbohydrates play important roles for a variety of structural and functional activities of glycoproteins . Here, we have characterized the N-glycans attached to the linker region of RoGA to be important for folding process, especially for its catalytic domain.
The starch-binding domain
The SBD, the non-catalytic module of GA that binds raw starch , is found at the N terminus in RoGA and shares only 13.5% identity with that of AnGA . Characterization of the key functional groups of this domain has been accomplished using sequence-based structure alignment and NMR spectroscopy [22,23]. The effects of the linker region on SBD adsorption on to insoluble starch were thus investigated. The binding assay was conducted at pH 4.5 using insoluble corn-starch as the affinity matrix. Figure 5 showed the binding isotherms for the interaction between corn-starch and the SBD variants (GAΔ132–604 and GAΔ168–604). The binding isotherms were used to calculate binding parameters as described in the Materials and Methods section. The Kd values for GAΔ132–604 and GAΔ168–604 were determined to be 3.98 μM and 5.99 μM, respectively; and the Bmax values of GAΔ132–604 and GAΔ168–604 were measured as 35.12 μmol/g and 17.24 μmol/g, respectively. It was apparent that the 106-residue SBD, GAΔ132–604, possessed stronger ligand affinity and higher capacity than those of the 142-residue SBD, GAΔ168–604, indicating that the presence of the linker region in the SBD construct did not increase but instead slightly interfered with the raw starch-binding affinity. The difference in starch binding affinity between GAΔ132–604 and GAΔ168–604 might be due to higher sugar content in the linker region or steric hindrance caused by the linker tail. Binding of the starch-degrading enzyme to its substrate is a critical step in starch hydrolysis because it involves the phase transfer of a soluble enzyme to the insoluble substrate . In our case the linker sequence influences the binding of the SBD to starch, possibly by affecting the ligand transfer process.
The catalytic domain
GA activity was assayed by measuring the reducing sugar released from the reaction on starch. The standard assay was performed as previously described with minor modification . The specific activities of the wild-type RoGA and the truncated mutants are listed in Table 3. Full-length GA and GAΔ26–131 exhibited similar activities, 4.58 × 103 U/μmol and 4.88 × 103 U/μmol, respectively, indicating that the absence of the N-terminal SBD had little or no effect on the ability to digest soluble starch. Moreover, comparison of the activities of the three catalytic domain variants GAΔ26–131, GAΔ26–145 and GAΔ26–160 revealed that longer size of the linker region was correlated with enhanced catalytic activity.
Figure 6 showed the effects of temperature and pH on enzyme activity. Regarding the thermal stability, the full-length RoGA remained stable at 40°C for 30 min at pH 4.5 with almost 100% of its activity remained, whereas at the same pH only approximately 35% and 5% residual activity was respectively detected at 50°C and 60°C (Figure 6, solid curve, open square). The three catalytic domain derivatives possessed inactivation profiles similar to that of the full-length GA (Figure 6, solid curves), indicating that the linker region did not contribute as much as expected to the thermal stability. In addition, the pH stability of the full-length RoGA at 25°C was found to be quite high over the pH range between 4.0 and 6.0, and more than 70% and 50% of the activity remained after 2 h incubation (Tm, 25°C) at pH 3.0 and pH 8.0, respectively (Figure 6, dotted curves). The truncated enzymes showed similar trends in stability over the pH range tested, indicating that the linker region was not crucial for the pH stability of GA either. Our data imply that the catalytic domain can function independently in terms of digesting soluble starch and the addition of linker sequences does not affect the thermal and pH stabilities of the enzyme.
The three-dimensional structure of RoGA is currently unavailable; circular dichroism spectroscopy has been widely used to study the secondary structure, with the goal of understanding the role of the linker in protein structure. In this study, the far-UV spectrum of full-length RoGA displayed two broad minima at 208 and 222 nm, characteristic of the presence of a mainly α-helical structure (Figure 7A, open square). The circular dichroism spectra of catalytic domain variants with different linker lengths were also indicative of a high content of helical conformation. In addition, the spectra of GAΔ26–131, GAΔ26–145, and GAΔ26–160 were very similar to each other, mainly differing in absorbance intensity (Figure 7A, closed symbols). GAΔ26–131 possessed a higher fraction of ordered secondary structure than did the other mutants, suggesting that the linker sequence might increase the stability of the secondary structural motifs. It was thus concluded that the linker sequences or their high degree of glycosylation facilitated stabilization of conformation of the catalytic domain of RoGA.
As for the SBD, Figure 7A also showed the calculated circular dichroism spectra of GAΔ132–604 and GAΔ168–604 (open triangle and circle, respectively). The negative ellipticity peak centered at 215 nm indicated that the secondary structure of the protein was predominantly β-sheet, but the latter showed a higher β-sheet conformation content. Similar to the case for the catalytic domain variants, the presence of the linker peptide in the recombinant SBD induced a substantial conformational change to a more ordered state. In addition, the thermal denaturation of GAΔ132–604 and GAΔ168–604 was monitored by circular dichroism at 215 nm. Figure 7B revealed that the denatured state of GAΔ168–604 possessed a greater negative ellipticity at 215 nm than did GAΔ132–604, suggesting that the presence of linker sequence might induce the SBD to form a more compact structure at high temperature. Taken together, although the 36-amino-acid linker alone of RoGA is predicted to be disordered, it strongly influences the structural ordering of the independent functional domains of RoGA.