Core glycan in the yeast multicopper ferroxidase, Fet3p: A case study of N-linked glycosylation, protein maturation, and stability

Abstract

Glycosylation is essential to the maintenance of protein quality in the vesicular protein trafficking pathway in eukaryotic cells. Using the yeast multicopper oxidase, Fet3p, the hypothesis is tested that core glycosylation suppresses Fet3p nascent chain aggregation during synthesis into the endoplasmic reticulum (ER). Fet3p has 11 crystallographically mapped N-linked core glycan units. Assembly of four of these units is specifically required for localization of Fet3p to the plasma membrane (PM). Fet3 protein lacking any one of these glycan units is found in an intracellular high-molecular mass species resolvable by blue native gel electrophoresis. Individually, the remaining glycan moieties are not required for ER exit; however, serial deletion of these by N → A substitution correlates with these desglycan species failure to exit the ER. Desglycan Fet3 proteins that localize to the PM are wild type in function indicating that the missing carbohydrate is not required for native structure and biologic activity. This native function includes the interaction with the iron permease, Ftr1p, and wild type high-affinity iron uptake activity. The four essential sequons are found within relatively nonpolar regions located in surface recesses and are strongly conserved among fungal Fet3 proteins. The remaining N-linked sites are found in more surface exposed, less nonpolar environments, and their conservation is weak or absent. The data indicate that in Fet3p the N-linked glycan has little effect on the enzyme's molecular activity but is critical to its cellular activity by maximizing the protein's exit from the ER and assembly into a functional iron uptake complex.

Introduction

The precise role of co-translational core glycosylation of nascent polypeptides in directing them along a productive folding trajectory within the endoplasmic reticulum (ER) remains a fundamental unanswered question in cell and structural biology.1–6 That this protein modification is critical to the timing of protein folding is indicated by the fact that the oligosaccharyltransferase responsible for this reaction is a component of the translocon complex in the ER membrane through which the translated polypeptide is threaded into the ER lumen.6 Certainly, downstream of completion of the synthesis of the protein, the terminal glucose moieties on the core N-linked N-acetylglucosamine/mannose branches serve as the ligands for the ER lectins, glucosidases, and glucosyltransferase that monitor progress toward the native folded state, marking the protein for degradation if it remains disordered.4, 7 However, this role as tag does not obviously relate to mechanisms by which core glycan per se contributes to the folding pathway if it does at all. Equally unclear is the extent to which carbohydrate contributes to the thermal stability of the native glycoprotein.

How core, N-linked glycan might contribute to protein folding, trafficking, and stability has been evaluated in three ways. The first is the reverse genetic approach: a sequon's asparagine residue is replaced with another amino acid and the effect on the processing and release of the desglycan species from the ER is examined. For example, a number of studies on the secretion of erythropoietin have been reported in which one or more of the three Asn residues that are glycosylated in this protein have been replaced by Gln.8–10 Although the results of these studies are somewhat mixed, a consistent result was that the post-ER trafficking and secretion of the double N38Q/N83Q mutant was strongly inhibited and in two reports, this strong suppression of secretion was seen also for the N38Q and N83Q mutants.

Proteomics provides a second means to indicate how core glycan might support protein folding and/or stability. If one collates all N-linked sequons into two groups, those that are and those that are not used (as indicated by the presence or absence of glycan in the proteins' structures) the protein structural features that cohere to those sequons utilized can be quantified. The most detailed such analysis indicates that sequons utilized are flanked by a predominance of hydrophobic amino acid residues, whereas such residue types are underrepresented in the sequence flanking sequons that are not used.11 Also, occupied sites are often found adjacent to or within clefts or grooves in the accessible protein surface.11 These data suggested that core glycan serves to cover and perhaps stabilize hydrophobic patches of the protein surface as has been suggested by studies on the folding in vitro of ribonuclease A12 and binding of carbohydrate moieties to recombinant erythropoietin.13

Third is the forward genetic approach in which N-linked sequons are engineered into a protein that is otherwise not modified by the addition of carbohydrate. This approach has been applied in an in silico analysis of the effect of core glycan units on the folding landscape of the SH3 domain protein.14 The results of this computational mutagenesis were that the melting temperature increased by ∼1°C per core glycan unit, an increase due to an increase in the energy of the unfolded conformational space. This increase, in turn, was modeled by unfavorable changes in both enthalpy and entropy of this conformational array resulting in an increased rate of protein folding.

In this article, we have used the reverse genetics approach in a protein that has 11 crystallographically resolved glycan units. We have correlated the effect on protein maturation of mutation at any one of these sites to its sequence and topologic location to test two ways core glycan might be important to maturation of the native protein. First, core glycosylation at some critical site(s) could suppress aggregation of the nascent polypeptide chain during elongation and extrusion into the ER lumen. Second, addition of core glycan units could subsequently serve as a regulator of chain folding due to the association of the glycosylated chain with the ER lectins, calreticulin, and/or calnexin15; the latter (Cne1p) is the sole ER lectin in Saccharomyces cerevisiae.16, 17 This model, which is not mutually exclusive of the first, predicts that a systematic deletion of N-linked carbohydrate by N → A substitution would lead to a graded reduction in the normal forward vesicular trafficking of nascent chains from the ER. This is a nonspecific mechanism whose efficacy should correlate with the total number of glycosylation sites in the nascent chain rather than on any specific one.

The Fet3 protein was used as subject to test our two hypotheses. Fet3p is the multicopper oxidase that supplies the plasma membrane (PM) ferroxidase activity needed for high affinity iron uptake in S. cerevisiae18 and, as noted, was chosen because it has 11 structurally mapped core glycan units.19, 20 We have used the native, Type Ia membrane protein and a truncation of this protein that lacks the carboxyl-terminal transmembrane domain. This soluble form, sFet3p, is therefore secreted from the cell; the thermal stabilities of desglycan forms of this soluble Fet3 protein have been quantified in vitro.20 Examining the native, membrane-tethered form, we evaluated also the possible role that Ftr1p, the iron permease partner in iron uptake, might play in the maturation of Fet3p in vivo; a variety of data indicates that these two proteins associate first in the ER.21–23

As no more than 3% of structures in the protein data base include resolved glycan,11 with 11 such sites sFet3p provides a unique platform to carry out this analysis. The results presented are consistent with our two hypotheses. First, we have identified one and only one N-linked site that is essential for protein folding of both forms of Fet3p. Three other sites are required for trafficking of the soluble but not the Ftr1p-interacting, membrane-tethered form suggesting a role for the iron permease in Fet3p maturation in the ER. Second, a trafficking deficit is exhibited also by sFet3p that lacks more than 4 of the 10 “nonessential” N-linked glycan moieties reflecting, perhaps, the concept that overall surveillance of core glycan units by ER lectins does serve to alter the landscape of protein maturation in the ER. Our data for the multiple, occupied N-linked sequons in a single protein reflect the pattern of glycosylation observed in archived protein structures while adding the functional information such structures alone cannot provide.

Results

Four of 11 N-linked glycan units are required for wild type sFet3p trafficking

The encoded Fet3 protein has 636 residues including a 21-residue signal recognition sequence that is cleaved during translocation into the yeast ER.19, 20 The native protein is a Type 1a membrane protein with a 23-residue helical transmembrane domain [residues 561–583, Fig. 1(A), boxed]. The protein is predicted to have 13 N-linked and possibly one O-linked glycosylation sites; glycan is resolved at 11 of the 13 N-linked sites but not at the single O-linked one. These several sequence motifs are given in Figure 1(A). The predicted N-linked sites are noted with a solid circle, whereas the putative O-linked Ser is marked with an open circle; glycan-occupied asparagines are indicated by an asterisk. The four canonical sequence motifs that provide the Cys and His ligands to the four copper atoms in this multicopper oxidase are indicated also (underlining). These four Cu atoms are represented by the four colored spheres in Figure 1(B).

Figure 1. Predicted N- and O-linked glycosylation sites in relation to structural motifs in sFet3p. (A) The primary sequence is annotated as follows: the signal sequence cleavage site following A21 is indicated by an arrow; the four Cu-binding motifs are underlined; the carboxyl-terminal transmembrane domain is boxed; predicted N-linked sites are indicated by a solid circle; the single predicted O-linked site is indicated by an open circle; those sites at which glycan is resolved in the Fet3p structure are marked by an asterisk. The vertical arrow indicates the truncation that yields the soluble, secreted form of Fet3p. (B) The structure of sFet3p highlights the 3-cupredoxin fold assembly and the resolved glycan units; this carbohydrate is concentrated in domains 1 (4 sites) and 2 (5 sites). The solid spheres represent the four copper atoms; Type 1 (blue), Type 2 (green), and Type 3 Cu-pair (yellow).

To assess the contribution glycan makes to Fet3p structure and stability, we took advantage of a soluble form of the protein. This sFet3p is secreted from the yeast cell due to a truncation before the transmembrane domain (at G555); an immunoreactive FLAG tag was introduced at the new carboxyl-terminus.20 By mass spectral analysis, this protein is estimated to contain ∼30% (w/w) carbohydrate; even after EndoH treatment, it retains ∼11% by weight CHO.20 Only this relatively deglycosylated form provided diffraction quality crystals and the 2.8 Å data and structure shown in Figure 1(B).19 The ribbon diagram includes the N-acetylglucosamine moieties (at least) that were resolved at 11 of the predicted 13 N-linked sites.

Whether glycosylation at any one of the 14 (putative) N- and O-linked sites was required for trafficking and secretion of this soluble Fet3p was evaluated first by making a family of mutant sFet3p alleles containing single N → A (or at T307, T → A) substitutions. These proteins were expressed downstream from the native FET3 promoter under control of the constitutively active Aft1pUP transcription factor.24 The conditioned media containing the FLAG-tagged sFet3 protein was concentrated 20-fold and together with extracts of the cell pellets was analyzed by immunoblot. The results of this screen are shown in Figure 2(A,B).

Figure 2. Immunoblot analysis of sFet3p species with Ala substitutions at predicted N- and O-linked glycan sites. AFT1-1^UP strain 3260 expressed FLAG-tagged soluble wild type and mutant forms of sFet3p as indicated. Conditioned media (A) and soluble cell extracts (B) were fractionated on SDS-PAGE gels, blotted, and probed with HRP-conjugated polyclonal rabbit anti-FLAG antibody; the immune complexes were visualized by film using Pierce SuperSignal chemiluminescent substrate (Thermo Scientific). The conditioned media was concentrated 20-fold to normalize its volume in comparison with the cell extracts. Note that the smaller molecular mass of the intracellular sFet3p species is consistent with the limited glycosylation expected for proteins that have not left the ER [cf. secreted sFet3p species in (A) to intracellular species in (B)]. VC, empty vector control.

The data indicated that glycosylation of otherwise native sFet3p at Asn residues 27, 74, 88, 198, 244, 265, 292, 300, or 381, or at the putative O-linked T307, was not required for the relatively normal secretion of sFet3p (not less than 50% of wild type, WT). Among this group of predicted glycosylation sites, electron density for N-acetylglucosamine (at least) was seen at Asn residues 27, 88, 198, 244, 292, 300, and 381 (Fig. 1) showing these sites are core glycosylated in wild type sFet3p; our data here show that the absence of carbohydrate at any one of these sites does not impair the efficient maturation of sFet3p in the ER or in subsequent compartments. In contrast, sFet3p N77A, N113A, N194A, and N359A species apparent in the medium concentrate was strongly reduced (<10% of WT); furthermore, the total amount of these species present in the cells was <50% of wild type with N194A the most strongly reduced. The sFet3p structure gave evidence for carbohydrate at all four of these Asn residues (Fig. 1).19

Serial deletion of “nonessential” core glycan blocks sFet3p trafficking

The data in Figure 2 indicate that glycosylation at several N-linked sites is not required for normal trafficking; however, the results did not rule out the possibility that in combination some of them do make a contribution to the maturation, trafficking, and/or stability of wild type sFet3p. We tested this possibility by progressively deleting “nonessential” sites, testing each multiple mutant for secretion. No striking phenotype was observed for triple mutants, but a decrease in secretion was characteristic of sFet3p lacking four members or more of this group of Asn residues [Fig. 3(A)]. Also, mutants that lacked five or more of these putative CHO chains were retained within the cell in a core-glycosylated (ER) form only [Fig. 3(B)]. This state of glycosylation was indicated by the steady decline in apparent molecular mass due to the accumulated absence of core glycan units in this series of Fet3 proteins.

Figure 3. Immunoblot analysis of trafficking of sFet3p species with multiple alanine substitutions at “nonessential” asparagine residues. AFT1-1^UP strain 3260 expressed FLAG-tagged soluble wild type and mutant forms of sFet3p. Conditioned media and soluble cell extracts were fractionated on SDS-PAGE gels, blotted, and probed with HRP-conjugated polyclonal rabbit anti-FLAG antibody. Fully glycosylated sFet3p is indicated as natFet3p; sFet3p containing core glycan only due to ER retention is designated as preFet3p. The notation NA1–NA7 indicates the sequential addition of N → A substitutions at asparagine residues 74 (NA1), 88, 198, 244, 292, 300, up to 381 (NA7).

Trafficking-defective sFet3 mutants accumulate in high-molecular mass species

The possible cellular status of sFet3p species that exhibited limited or no trafficking and secretion was evaluated for some of these mutants by blue native gel electrophoresis (BN-PAGE) in which proteins commonly retain their cell oligomeric state.25 We considered that trafficking-defective proteins possibly would be retained in the ER as aggregates and/or complexed with the components of the ER quality control system (ERQC). We first examined cell extracts containing the N113A and N194A single mutants. Three controls were included: (1) cell extract and (2) secreted samples of wild type sFet3p, and (3) EndoH-treated, secreted sFet3p, the species whose structure is illustrated in Figure 1(B). The results are given in the immunoblot shown in Figure 4(A).

Figure 4. Blue native analysis of sFet3 species. AFT1-1^UP strain 3260 expressed FLAG-tagged soluble wild type and mutant forms of sFet3p as indicated. (A) Wild type and secretion-deficient mutants (left) were compared with EndoH-treated secreted sFet3p (right). Cell extracts (left) and conditioned media (right) were fractionated on BN-PAGE gels, blotted, and probed with HRP-conjugated polyclonal rabbit anti-FLAG antibody. EndoH-treated secreted WT is included to provide a marker for the electrophoretic mobility of sFet3p carrying a full complement of core GlcNAc. Note the correlation between increased high molecular mass forms and decreased monomeric ones. (B) Cell extracts containing wild type and sFet3p mutants NA2 to NA5 (see Fig. 3) were examined by BN-PAGE and immunoblotting. A fraction of each mutant protein failed to enter the gel; however, these samples lacked the ∼ 720 kDa molecular mass species observed with the secretion-deficient mutants.

Two forms of sFet3 are detected in the cell samples. The faster moving, lower molecular mass species is likely monomeric [(Man)_5/6-(GlcNAc)₂]_n-sFet3 species (where n = number of core glycan units, up to 11 in WT), whereas the slower migrating form is sFet3 protein in a higher molecular mass complex. Native, secreted sFet3p migrates with a M_r ≅ 130 kDa due to the elaboration of the core glycan. EndoH-treatment removes much but not all of the branched-chain glycan in sFet3p [cf. Fig. 1(B)]; therefore, this species appears with a molecular mass less than the monomeric, core-glycosylated protein in the cell. The higher mass species (∼720 kDa) accumulated as the monomeric forms destined for secretion diminished, whereas little if any of this aggregated material was detected in wild type or other secretion-competent sFet3p forms. Clearly, these experiments provide no information about the nature of the high molecular weight complexes that appear to correlate with retained desglycan species but they do support the conclusion that Fet3 species lacking one of the apparently essential glycan units either aggregates or fails to escape the components of ERQC.

In contrast, Fet3 proteins lacking one or several of the seven apparently nonessential glycan units do not accumulate in high-molecular mass species. The immunoblot in Figure 4(B) provides this result. The protein species used were based on Fet3p(N74A) (designated as NA1 in the figure); subsequent N → A substitutions were serially made up NA7. Comparison of the gels shown in Figures 3(B) and 4(B) demonstrates that these desglycan species migrate with the same relative apparent molecular mass whether in a denaturing or nondenaturing condition. Unlike the trafficking defective N113A and N194A, none of these desglycan species were found in the high-molecular mass, ∼720 kDa species.

Trafficking-defective desglycan Fet3 mutants retain molecular activity

These studies with sFet3p do not address the possibility that glycan moieties at any of these N-linked sites are essential to some function in vivo, for example, in the assembly and function of the Fet3p ferroxidase, Ftr1p iron permease complex in the yeast PM. To address this question directly, N → A substitutions in the full-length, membrane-targeted form of Fet3p were made; this Fet3p had green fluorescent protein (GFP) appended to its cytoplasmic, carboxyl-terminal tail. This fluorescent Fet3p fusion, which is wild type in its iron-uptake properties, has been invaluable in the examination of the mechanism of the trafficking of the Fet3p, Ftr1p complex to the yeast PM.18 Using this fusion, the trafficking of glycan-mutants in the yeast cell was followed, and the iron uptake activity associated with these mutants was quantified.

The results were revealing in that the apparent trafficking defect observed with alanine substitution at some asparagines in sFet3p was not observed in the full-length, membrane-tethered protein. This result is illustrated in Figure 5(B,C) for the N77A and N113A mutants whose steady-state localization in the PM was not different from WT. In contrast, the full-length Fet3p(N194A) failed to localize to the PM indicating that it exhibited a trafficking defect comparable with that exhibited by the corresponding sFet3p form [Fig. 5(D)]. As a negative control, we examined the trafficking of wild type Fet3p expressed in the absence of its trafficking partner, Ftr1p [Fig. 5(E)]. The ⁵⁹Fe uptake activity correlated with the localization patterns of these Fet3:GFP species: uptake was wild type with the N77A and N113A mutants (and the N198A one as well), but essentially absent in the case of Fet3p(N194A) (Table I, first column).

Figure 5. Trafficking of wild type and N-linked glycan mutants of Fet3:GFP. Fet3p and Ftr1p species were produced in strain AJS-05 (fet3Δftr1ΔAFT1-1^UP) from low-copy vectors under control of FET3p. Cells in log-phase growth were examined by epifluorescence using a Zeiss Axioimager Z1 fluorescence microscope. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Table I. ⁵⁹Fe-Uptake Rates and Fe(II) Turnover Kinetic Constants for Wild Type and N → A Mutants of Native and Secreted Forms of Fet3p

Fet3p species	^59Fe uptake (native Fet3p) pmol ^59Fe/20 min/10^7 cells	Fe(II) turnover (sFet3p)
		kcat (s^−1)	Km^Fe(II) (μM)
^59Fe-uptake assays were performed at pH 6.0 in the presence of 20 mM ascorbate in strain AJS-05 (fet3Δftr1ΔAFT1-1^UP) with [Fe] = 0.2 μM.25 Expression Fet3p and Ftr1p species was under control of the pFET3 in low copy vectors. Steady-state kinetic constants for Fe(II) turnover were obtained by O2-uptake at pH 6.0 under air.26 All data were evaluated using Prism 5 software which provided the standard errors shown. a ns, not secreted. b Control value (fet3^−ftr1^− parental strain, AJS-05), 7 ± 5 pmol ^59Fe/20 min/10^7 cells.
WT	115 ± 8	56.0 ± 2.9	7.7 ± 1.6
N77A	92 ± 5	52.9 ± 1.4	4.7 ± 1.1
N113A	88 ± 7	51.1 ± 1.2	8.8 ± 1.5
N198A	97 ± 9	54.7 ± 1.3	9.7 ± 1.4
N194A	7 ± 3b	nsa	nsa

The results with the N77A and N113A mutants indicated that Ftr1p might assist in the maturation of the native, membrane-anchored form of Fet3p; several studies indicate that the two proteins interact in the ER, an interaction that is required for the subsequent trafficking of both proteins to the PM.18 Our results also suggested that these desglycan Fet3p forms may have retained some measure of normal folding and stability even in the sFet3p forms evaluated above. The possibility that sFet3p was lost as a result of the 20-fold concentration of the conditioned media was tested by chromatographic analysis of the media as in the purification of sFet3p.20 A similar strategy was successful in recovering relatively unstable N → A mutants of lecithin-cholesterol acyltransferase.27

In fact, sFet3p mutants N77A and N113A could be purified from the conditioned media albeit in ∼50% yield in comparison with wild type; sFet3p(N198A) recovery was ∼80% of wild type. In contrast, no sFet3(N194A) protein could be recovered using these purification protocols. The recovered mutants exhibited UV-visible absorbance spectra that were quantitatively identical to wild type with the characteristic transition at 608 nm (ε = 5200 M⁻¹/cm) due to the “blue” or Type 1 Cu(II), and transitions at 330 and 750 nm (ε = 5000 and ∼750 M⁻¹/cm) due to the binuclear, Type 3 Cu(II) cluster (spectra not shown).20 The steady-state kinetic parameters for Fe(II) oxidation by these proteins was wild type as well; these data are included in Table I and correlate with the ability of the membrane form of these mutants to localize with Ftr1p at the PM and to support iron uptake.

Thermal unfolding analyses of these desglycan forms were performed to quantify their stability in comparison with wild type. The differential scanning calorimetry (DSC) data are shown in Figure 6 as excess heat capacity (ΔC_p) as a function of temperature and the analyses of these thermal denaturation profiles are compiled in Table II.

Figure 6. Thermal unfolding profiles for WT and N → A mutants of sFet3p. The excess heat capacity was monitored by DSC during the thermal denaturation of Fet3p species as indicated. The quantification of the data is given in Table II.

Table II. Thermal Denaturation Analysis for WT and Mutant sFet3 Proteins

Fet3p Species	Transition temperatures (°C)	Calorimetric enthalpy (total) (kJ/mol)
WT	54.3
	67.6	1480
N77A	Shoulder ∼ 51
	57.0	1350
	73.6
N113A	52.3
	54.5	1100
	65.7
N198A	Shoulder ∼ 50
	56.1	1230
	66.2

As reported, wild type sFet3p unfolding involves two thermal transitions. The one centered at 54°C is due to the unfolding of cupredoxin domain 2 (D2), which triggers the cooperative unfolding of domain 1 [cf. Fig. 1(B)]; the second transition is due to the unfolding of D3.28 All three mutants differ from WT in exhibiting a new thermal transition at ∼51°C. The lower temperature shoulder seen in the three N → A mutants is ascribed to a destabilized D2 with the T_m values for the melting of domains 1 and 3 relatively unchanged from wild type. Correlated to this destabilization of D2 and perhaps with a more thermally sensitive interaction between D2 and D1 is a significantly smaller overall calorimetric enthalpy of unfolding indicating an overall loss in stability for these three desglycan mutant proteins.

Fungal ferroxidase sequence and structural homology identifies essential core glycan

The data presented here for Fet3p indicates that specific core glycan is essential for the efficient trafficking and relative stability of this fungal protein; retention of this essential glycan by Fet3 ortho- and paralogs but not of nonessential glycosylation sites would support this inference. Taking first the 100 closest ScFet3 homologs, 17 were found to lack the ferroxidase-specific motifs, for example, the carboxylate residues that tune the redox potential of bound Fe²⁺ for efficient electron transfer to the Type 1 Cu(II) in these metallo-oxidase enzymes.29, 30 Of the remaining 83 proteins, only two are predicted to lack N-linked carbohydrate at the Fet3p N113 homologous residue, three are predicted to lack CHO at N194 and 10 are predicted to lack CHO at N198. In contrast, 47 lack the nonessential N74 and a comparable lack of other Fet3p NXS/T sequons that individually do not appear required for normal protein maturation and/or localization. In short, “essential” glycan is conserved in fungal Fet proteins, nonessential glycan is not.

This bioinformatics approach yielded a “negative” result, which actually provided insight as to the role that core glycan might play in the maturation of these cupredoxin domain-containing proteins. An unnamed protein from Podospora anserina (a filamentous ascomycete), accession number XP_001909918, was the only protein among the 83 homologs predicted to lack 194 and 198 core glycan (Supporting Information Fig. S1). The lack of core glycan in PaFet3 in this domain 2 region correlated with a short-DPDADE sequence in domain 1 unique to this Fet homolog (Supporting Information Fig. S2).

The structure of sFet3p (1ZPU) was used to model this PaFet protein; with respect to the sFet3p structure, this model had an RMSD = 1.8 Å. The ribbon diagram of this model is shown in Figure 7(A) with this unique acidic sequence motif noted; in Figure 7(B), sFetp is shown for comparison with the crystallographically resolved carbohydrate that lies within a cleft between domains 1 and 2 included in CPK representation. The images shown in panels C and D illustrate in CPK the difference in the structural basin between domains 1 and 2 in these two proteins that correlates with the sequence differences noted above. The model suggests the hypothesis that the structural role of the core glycan found at N194 and N198 in sFet3p, and its close homologues is fulfilled in this putative PaFet3 by the carboxylate side chains of D232, E233, and D235 interacting with the acidic loop in D1 [Fig. 7(C)]. This putative loop in the P. anserina protein contrasts with the homologous loop from D1 in ScFet3p that is dominated by nonpolar residues [Fig. 7(D)]. Reasonably, the core glycan positioned in this otherwise nonpolar cleft acts as a buffer preventing non-native, hydrophobic burial of these nonpolar surfaces.

Figure 7. Model of the D1/D2 interface in PaFet and in ScFet3p. The structure of the PaFet3 homolog (A) was built based on ScFet3p (1ZPU) (B) to an RMSD = 1.8 Å. The interface region between D1 and D2 contains the side chains from the sequences given in Supporting Information Figure S1 including the unique acidic loop, ⁶⁹DDPDDADE⁷⁵, and the N194/N198 homologous region, ²³¹NDEXD²³⁵. This region in PaFet3 is shown in CPK form (C) as is the homologous cleft between domains 1 and 2 in ScFet3p (D). This region in ScFet3p is dominated by nonpolar residues and the core (GlcNAc)₂ at N194.

Discussion

The nascent polypeptide chain negotiates a successful folding trajectory only through the sure hand of a cohort of molecular guides. Indeed, without these chaperones, globular proteins would have little if any selective advantage; the very constituent properties that most strongly distinguish the free energy of the folded in comparison with the unfolded state—the nonpolar side chain van der Waals contacts that stabilize the former and the degree to which the very same residues in water destabilize the latter—would serve as well to promote the aggregation of these nascent chains, diverting them irreversibly from the productive folding pathway. In this context, one reasonably would expect to find core-glycan decoration of nascent polypeptides within or adjacent to relatively nonpolar stretches of amino acid sequence, those that in retrospect can be identified as aggregation-prone. In fact, statistical analysis of the sequence domains in which N-linked sites are found in proteins indicates that Met is the most “polar” residue disproportionately found adjacent to glycan-linked Asn residues (±6 residues).11

Four of the 11 core glycan units resolved in the crystal structure were required for wild type secretion of sFet3p; these were at N77, N113, N194, and N359. sFet3p(N381A) showed a somewhat less severe secretion defect (Fig. 2). Although we have provided no direct evidence that the N → A mutants at any one of these sites were subject to ER-associated degradation, the BN-PAGE analysis does indicate that these sFet3 species were in a high-molecular mass state within the cell. Given our experimental design, we cannot be specific about the composition of this protein complex, whether it is oligomeric sFet3p alone or a complex of sFet3p and other ER proteins. What our data do show, however, is that as this sFet3p species increases, there is a corresponding decrease in both monomeric sFet3p within the cell and secreted sFet3p in the conditioned media.

What distinguishes these latter four N-linked sites from the other seven resolved crystallographically? We examined two features of the sFet3p structure suggested by the proteomics analysis conducted by Petrescu et al., sequence context and conformational locale.11 These two features are represented in Table III by the transfer free energies of the sequon including residues −3 to +3 relative to each N-linked asparagine, and by a qualitative description of the secondary structure and surface topology contiguous to a given asparagine. The latter description uses the reference points made by Dwek and coworkers.11 This summary allowed us also to compare the distribution of these features in sFet3p to their distribution in the >600 sequons selected from the structure data base by these investigators.

Table III. Sequence Context of N-Linked Core Glycosylation Sites in Fet3p

N → A substitution	Traffic/stability	Sequence context (ΣHydro)	Topology 2°/3° structural context
The 11 Asn residues in Fet3p at which N-acetylglucosamine moieties were resolved in the crystal structure are listed and underlined within their sequence context. The trafficking of the N → A mutant proteins is indicated with reference to wild type using a descriptive notation (+++). The hydrophobic index of each sequon is given as the sum of the side chain water to octanol transfer free energies. Values for ΣHydro are given as the sum (in kcal/mol) of the transfer free energies of the corresponding amino acid side chains from water to octanol31 corrected for solute–solvent volume differences.32 The glycan-linked asparagine is not included in the calculation. The topologic context of the N-linked Asn side chain is taken from PDB 1ZPU.27
27	+++	HTPNWTT (−15.1)	β-Strand/solvent
77	−	NNTNTSM (−8.2)	Coil/recessed
88	+++	LFQNGTA (−13.9)	Coil/solvent
113	−	MLYNFTV (−22.6)	β-Strand/recessed
194	−	LIVNNTM (−18.9)	β-Strand → coil/groove
198	+	NTMNLTW (−18.2)	Coil → β-strand/solvent
244	+++	TEKNVTD (−8.4)	β-Strand/solvent
292	+++	LQLNATS (−14.4)	β-Strand/solvent
300	++	MVYNKTA (−14.7)	Coil → coil/surface groove
359	−	FFNNITY (−20.7)	β-Strand→coil/shallow recess
381	+	DQANNSE (−4.3)	Coil → helix/solvent

As the data in Table III show, Asn residues at positions 113, 194, 198, and 359 are found within relatively hydrophobic sequence patches. By way of comparison, N88, N244, N292, N300, and N381 are found in relatively hydrophilic sequence domains. Overall, 8 of 11 sequons containing glycan have water to octanol transfer free energies equal to or more negative than −14 kcal/mol. Thus, in regards to this characteristic, sequon occupancy in sFet3p mirrors usage in all structurally characterized glycoproteins. Topology at these sites in sFet3p exhibits a similar distribution, also. First, glycan is absent within helical motifs. On the other hand, at 5 of 11 occupied sites, glycan is located at a change in secondary structure. Overall, in the structure data base fewer than half of sequons located within helices are found to be occupied by glycan, whereas glycan is common at sequons positioned adjacent to a β-strand → coil transition, for example.11 Last, of the 11 core glycan GlcNAc units resolved in the sFet3p structure, five are in recesses or grooves in the protein surface, and six are in solvent. This ∼50:50 distribution is comparable with what is seen in glycoproteins as a group11 and, thus, overall, sFet3p is a good representative of this cohort of proteins.

However, what we offer here is a test of the significance to protein maturation and stability that sequence and structure analyses alone cannot provide. Indeed, our data show clearly that glycan located in a recess or groove and, therefore, at an asparagine out of the solvent in the folded protein is more important to these two dynamic aspects of protein structure than glycan located on the surface or in solvent. This perhaps is an unexceptional result in that if glycan does not change its state upon polypeptide folding, its presence or absence should not alter chain distribution within the folding landscape. On the other hand, our data show that topologic location is a stronger predictor than sequon hydrophobicity in that irrespective of this latter characteristic, removal of glycan located at a recessed site led to a strong defect in trafficking and stability, that is, the N → A mutants at 77, 113, 194, and 359. The only glycan partially occluded from solvent that was not required for wild type trafficking was located at N300. Thus, our data confirm the inference that utilized sequons cluster in patches that are more nonpolar than is typical of a protein's surface and, importantly, carbohydrate at those sequons make a more significant contribution to protein maturation and stability when topologically recessed.

The behavior of Fet3p(N77A) could reflect two ways glycan might be essential to protein maturation despite the relatively polar nature of sequence context at this position. First is its topologic location in a recess within the protein's surface. Second is the strongly favored H-bond interaction between asparagine side chains on polypeptide strands in a coil or parallel β-sheet conformational relationship33; a N, N pairing in this secondary structure motif is 2.7-fold more prevalent than any other same-residue pairing. During synthesis into the ER lumen, adjacent Fet3p chains would be in this conformation allowing the ⁷³MNNTATSM₈₀ region (in the N77A mutant) to participate in inter-strand H-bonding. In this context, the next most favored pairings in β-sheet conformations are S-S and T-T, also well represented in the N77 sequon region.33

As noted, Fet3p is assembled from three cupredoxin domains [Fig. 1(B)]. Nine of the 11 N-linked core glycan moieties are distributed in domains 1 and 2 with four and five sites, respectively [Fig. 1(A)]; except for the site at N359, all of the essential sites are located in the sequence region that extends from the middle of domain 1 to the middle of domain 2 (N77–N198). In as much as copper addition to Fet3p and other eukaryotic MCO proteins occurs late in the secretory pathway,34, 35 it is the apo-protein that folds in the ER. Therefore, a reasonable assumption is that the individual cupredoxin domains function as independent folding units during synthesis; this inference is consistent with the fact that apo-Fet3p unfolds in three steps, which correspond to the sequential unfolding of the three cupredoxin domains.28 Thus, to the extent that core glycan serves as a guide to direct protein folding, it would be needed primarily during the folding of the first rather than later cupredoxin motifs because the folding of domain 3, for example, could use the two-domain species as a folding template. Inspection of the Fet3p structure also reveals that alone among the resolved glycan, the NAG-NAG-MAN-(MAN, MAN) chain at N194 lies between strands of β-sheet associated with domains 1 and 2, perhaps serving as a buffer towards aggregation. This was shown in Figure 7(D), which highlighted the strongly nonpolar nature of the inter-domain cleft in which the core glycan at N194 resides.

Here, we show that the complete absence of core glycan at some N-linked sites does alter the thermal stability of sFet3p much as removal of a prosthetic Cu atom does.28 Two changes were apparent. First, a new transition was observed in the DSC traces with a relatively low T_m indicating a change in the unfolding mechanism and second, the overall enthalpy of unfolding was reduced by up to ∼400 kJ/mol. One must approach the DSC data with caution, however, because thermal unfolding of Fet3p is irreversible28; consequently, a difference between native and des-glycan species may reflect kinetic as well as thermodynamic stability factors, which are yet to be separately delineated. Nonetheless, the quantified calorimetric differences between wild type and mutant Fet3p forms correlated with the observation that thermally less stable des-glycan sFet3p species were relatively unstable to manipulation in the conditioned medium following their secretion from the cell and were found in proportionately larger abundance in high molecular weight forms in the cell.

At least with respect to those few desglycosyl Fet3p species we examined for both in vivo and in vitro function, full core glycosylation per se was not required for maturation of an apparently wild type Fet3p multicopper oxidase enzyme, that is, a ferroxidase that supported Fe-uptake at the cell membrane and catalytic turnover of Fe²⁺ and O₂. Although not a determinative result in regards to our postulate that core glycosylation serves as an intrinsic chaperone for aggregation-prone sequences, the wild type functional properties of these Fet3p species indicate at the least that selection for glycosylation has not been to hone molecular Fet3p activity. Rather, our data support the alternative view that glycosylation enhances cellular Fet3p activity by increasing the fraction of Fet3 polypeptides that pass the ERQC check-point. As a result, a larger fraction is trafficked to the post-ER network for activation by copper addition and final delivery to the PM in support of high-affinity iron uptake in this yeast.

Materials and Methods

Strains, media, and culture conditions

The strain used in the majority of the studies described was AJS05, which was derived from DEY1457 (MATα can1 his3 leu2 trp1 ura3 ade6).18 The AJS05 genotype is MATα can1 his3 leu2 trp1 ura3 ade6 fet3::HIS ftr1::TRP1 aft1::AFT1-1^upKAN. The AFT1-1^up allele codes for a constitutively active form of the Aft1p transcription factor that drives expression of the FET3 (and FTR1) locus.24 In this background, expression of episomally expressed wild type and mutant alleles of FET3 and FTR1 was maximized affording cultures that contained the protein partners in the PM as described. Early log phase cells (OD_660nm = 0.8 − 2.0) grown in selective media (6.67 g/L yeast nitrogen base w/o amino acids, 2% glucose plus the appropriate drop-out mixture of amino acids) were used for all experiments.

Construction of Fet3p mutants

Mutant FET3 alleles were constructed directly in pDY148 (for secreted, soluble Fet3p)20 and in pDY133 expressing a Fet3::GFP fusion18 (for the native membrane-associated Fet3p) by site-directed mutagenesis using the QuikChange kit from Stratagene. Mutated FET3 sequences were confirmed by automated fluorescence sequencing on an ABI PRISM 377 instrument. Vectors expressing the mutant proteins were transformed into yeast strain M2* for soluble protein expression20 and in strain AJS05 for in vivo analyses of protein localization at the PM and quantification of ⁵⁵Fe uptake.18, 22

Expression and purification of soluble Fet3 proteins

S. cerevisiae strain M2* carrying plasmid pDY148 was used as the expression system for the purification of soluble Fet3p.20 This strain is MATα trp1-63 leu2-3,112 gcn4-101 his3-609 ura3-52 AFT1-1^up. Plasmid pDY148 is a high-copy vector that carries a recombinant FET3 gene truncated at nucleotide +1666 (at amino acid residue 555). This truncation removed the apparent membrane-spanning domain found in the carboxy-terminal region that is included in residues 559–586. Expression of Fet3p from pDY148 results in a Fet3p that is secreted directly into the growth medium rather than being retained in the PM. Fet3 protein production and purification have been described in some detail.20

Electrophoretic and western blot analyses

sFet3p in conditioned media was examined by by SDS-PAGE/western blot analysis either without or with 20-fold concentration before addition of standard loading buffer (with reducing agent). Cell extracts for western analysis were prepared from washed cells recovered from this media by direct addition of loading buffer. Samples were heated to 95°C and cooled before separation on 8% gels. Samples for BN analysis were prepared in a lysis buffer containing 500 mM 6-aminocaproic acid, 20 mM BisTris (pH 7.0), 2 mM EDTA, 10% glycerol, 1% Triton-X100, Coomassie blue G-250, and a protease cocktail that included PMSF and a fungal protease inhibitor mix (Sigma Chemical Corp., St. Louis, MO). Samples were run on 8% acrylamide gels with Tris/BisTris/Coomassie blue cathode and BisTris anode buffers, pH 7.0. Following electrophoretic separation, gels were soaked in 0.1% SDS followed by transfer to nitrocellulose for immunoblot analysis programmed by the carboxyl-terminal FLAG epitope carried by all Fet3 protein species examined in this report. Immunocomplex detection was provided by HRP oxidation of Pierce SuperSignal chemiluminescent substrate (Thermo Scientific).

Protein and kinetic characterization

Room-temperature UV-visible absorption spectra were recorded using a Varian Cary 50 spectrophotometer. Protein concentration was determined using the standard dye-binding Bradford assay using BSA as the protein standard.26 Steady-state kinetic analyses were based on oxygen consumption using an Oxygraph (Hansatech, www.hansatech-instruments.co.uk).30 Rates of O₂ uptake were evaluated using the OXYG32 software provided by Hansatech. All initial velocity, v versus [S] data were subsequently analyzed by direct fitting to the Michaelis–Menten equation using Prism 5 software (GraphPad Software, La Jolla, CA). Ferrous ammonium sulfate (Sigma-Aldrich, St. Louis, MO) was used for kinetic analysis of the metallo-oxidase activity of Fet3 proteins. Substrate stock solutions were freshly prepared in nitrogen-purged 100 mM MES at pH 6.0. All transfers from these stock solutions were done using gas-tight syringes. The buffer used for O₂-uptake measurements was air-saturated 100 mM MES at pH 6.0.

DSC analyses

Thermal unfolding experiments were performed on a VP-DSC microcalorimeter (MicroCal, Northampton, MA) at 0.5, 1, and 1.5 K/min scan rates.28 Protein concentration was 4 μM in 50 mM phosphate buffer, pH 7. Sample and reference solutions were degassed for 5 min and carefully loaded into the cells to avoid bubble formation. A pressure of 2 atm was kept in the cells throughout the heating cycles to prevent degassing. Background was subtracted from each protein trace. The reversibility of the transitions was assessed by performing second heating cycles after cooling. As reported,28 the thermal reactions were irreversible. Excess heat capacity curves were plotted and analyzed with Origin software (OriginLab, Northampton, MA).

Trafficking and ⁵⁹Fe-uptake activity of des-glycan Fet3p species

These experiments were carried out in strain AJS05 expressing a carboxyl-terminal Fet3::GFP fusion that has native Fet3p activity in vivo.18 des-Glycan mutants were constructed in this fusion complementing those made in pDY148. Yeast cells were grown into early log phase, pelleted, washed once with phosphate-buffered saline, and resuspended in 100 μL of the same buffer for examination of Fet3::GFP localization by epifluorescence using a Zeiss Axioimager Z1 fluorescence microscope. ⁵⁹Fe uptake in these cells was quantified at [⁵⁹Fe] = 0.2 μM at pH = 6.0 in the presence of 20 mM ascorbate (reductase-independent uptake). This concentration of iron is comparable with the kinetically determined K_m value for reductase-independent Fe-uptake in this strain background, 0.36 μM.36

Acknowledgements

The authors thank Julie Spix for her contributions to the kinetic analysis of sFet3 proteins.