Mutations in the FAD Binding Domain Cause Stress-induced Misoxidation of the Endoplasmic Reticulum Oxidoreductase Ero1β*

Disulfide bond catalysis is an essential component of protein biogenesis in the secretory pathway, from yeast through to man. In the endoplasmic reticulum (ER), protein-disulfide isomerase (PDI) catalyzes the oxidation and isomerization of disulfide bonds and is re-oxidized by an endoplasmic reticulum oxidoreductase (ERO). The elucidation of ERO function was greatly aided by the genetic analysis of two ero mutants, whose impairment results from point mutations in the FAD binding domain of the Ero protein. The ero1-1 and ero1-2 yeast strains have conditional and dithiothreitol-sensitive phenotypes, but the effects of the mutations on the behavior of Ero proteins has not been reported. Here, we show that these Gly to Ser and His to Tyr mutations do not prevent the dimerization of Ero1β or the non-covalent interaction of Ero1β with PDI. However, the Gly to Ser mutation abolishes disulfide-dependent PDI-Ero1β heterodimers. Both the Gly to Ser and His to Tyr mutations make Ero1β susceptible to misoxidation and aggregation, particularly during a temperature or redox stress. We conclude that the Ero FAD binding domain is critical for conformational stability, allowing Ero proteins to withstand stress conditions that cause client proteins to misfold.

(Cxx)CxxC motif, which is required for the structural stability of the protein (28) and is tethered by a long range intra-molecular disulfide bond (29). The (Cxx)CxxC motif is in contact with a non-covalently buried FAD molecule, to which electrons are passed en route to the final electron acceptors. The flavin fold found in Ero proteins is novel but is related to that of Erv2p, an alternative oxidoreductase of the yeast ER (30).
The ero1-1 mutation results in temperature-dependent lethality, whereas the ero1-2 mutation results in hypersensitivity to reducing agents (DTT) (16,17). Single amino acid substitutions in Ero1p cause the ero1-1 and ero1-2 phenotypes. These mutations are G229S (equivalent to Gly 252 in Ero1␤) and H231Y (equivalent to His 254 in Ero1␤). Both residues are within the Ero flavin fold, with His 231 directly contacting the ribose 5Ј-phosphate group of the FAD moiety (29). In Ero1p, residues Arg 187 , Thr 189 , Trp 200 , Ser 228 , His 231 , and Arg 260 form hydrogen bonds or salt bridges with the FAD cofactor and are conserved in Ero1␣ and -␤. The alteration of structure and charge within the flavin fold of the ero1 mutants is therefore likely to lead to loss of function by preventing electron transfer from occurring normally. However, the biochemistry of the FAD binding site mutants has not been fully examined.
Using Ero1␤, we show that FAD domain mutants misoxidise during redox or temperature stress. The unusual Ero FAD binding domain therefore has a dual function: to facilitate electron transfer and to confer protein stability. The flavin pocket safeguards protein oxidation pathways, preventing spurious electron transfer at times when client proteins will misfold, misoxidise or aggregate.
Transfections-Transfections with Lipofectamine 2000 (Invitrogen) were performed according to manufacturer's instructions. Subconfluent cells in 6-cm dishes were washed twice with PBS (Invitrogen) and transfected with 1 or 2 g of DNA for 6 h in the presence of Opti-MEM serum-free medium (Invitrogen). The medium was replaced after 6 h with complete medium and the cells analyzed 24 h post-transfection.
Western Blotting-Cells were lysed in lysis buffer supplemented with 20 mM N-ethylmaleimide or 20 mM iodoacetic acid to trap disulfide bonds. Nuclei were removed by centrifugation at 16,100 ϫ g for 10 min at 4°C. Post-nuclear cell lysates or immunoprecipitates were analyzed by SDS-PAGE. Immunoprecipitations were carried out using 8 l of ␣PDI or 0.5 l of ␣HA (HA-7) or ␣Myc (9B11) antibodies immobilized on 50 l of a 20% suspension of Protein A-Sepharose beads, followed by washing twice with lysis buffer. Proteins were transferred to polyvinylidene difluoride membranes (Millipore) at 150 mA for 2 h or 30 V overnight, and the membranes were blocked in 8% milk/PBS-Tween for 1 h or overnight. The primary antibodies were used at 1:5000 (␣Myc), 1:2000 (␣HA), 1:1000 (␣PDI) and 1:50 (␣Ero␤). After washing four times with PBST, membranes were incubated with secondary antibodies (DAKO) at 1:3000 for 1 h, washed extensively, and visualized by ECL (GE Healthcare/Amersham Biosciences) upon exposure to Biomax Light film (Eastman Kodak Co.). Protein markers were from Bio-Rad. Each blotting experiment was reproduced at least twice.
DTT and Temperature Treatments-Transfected HeLa cells were washed with PBS (Invitrogen) and incubated with complete medium containing 10 mM DTT (Sigma) or buffer only for 15 min at 37°C. For temperature treatments, transfected HeLa cells were incubated at 24, 37, and 42°C on water baths for 1 h in complete medium buffered with 10 mM HEPES, pH 7.4 (Invitrogen). Cells were washed and lysed as described above, and post-nuclear supernatants were subjected to analysis by 8% SDS-PAGE.
Trypsin Sensitivity Assay-Post-nuclear supernatants were incubated with 0, 0.25, 1.25, and 2.5 g/ml TPCK-treated trypsin (Sigma) for 30 min at 4°C. Proteolytic digestion was terminated by the addition of 200 g/ml soybean trypsin inhibitor, in addition to 10 g/ml each of chymostatin, leupeptin, antipain, and pepstatin A. Samples were taken up into sample buffer and analyzed by 10% SDS-PAGE. Five independent trypsin assays were performed.

RESULTS AND DISCUSSION
Limited Proteolysis of Ero1␤G252S and Ero1␤H254Y-We have previously characterized the human Ero1␤ protein and shown that Ero1␤ homodimerises and interacts with PDI by both non-covalent and covalent (disulfide-dependent) interactions (22). Ero1␤ is functionally equivalent to Ero1p (21), and both Ero1␤ and Ero1p bind to PDI (31) and homodimerise (32). Since Ero1p and Ero1␤ also share conserved FAD binding residues (Fig. 1A), we used our knowledge of Ero1␤ to study the effects of the ero1-1 and ero1-2 mutations on Ero gene products. We undertook site-directed mutagenesis to create tagged Ero1␤G252S and Ero1␤H254Y mutants. Myc-and HA-tagged human Eros are functional, reside in the ER, and can interact with PDI in the same way as their non-tagged counterparts (21,22,28,33).
We found that biosynthesis of Ero1␤ was largely unaffected by the G252S and H254Y mutations in the FAD domain (supplemental Fig. 1). However, it was possible that mutations in this region could cause conformational changes resulting in misfolding, particularly since Ero1p seems to expose buried residues during the oxidation cycle (32). To compare the overall conformation of wild-type Ero1␤ and the mutant proteins, we used a limited proteolysis approach. Partial trypsin digestion has been used to map conformational changes in a number of ER proteins, including polyomavirus (34) and cystic fibrosis transmembrane conductance regulator (35,36). Thus transfected HeLa cell post-nuclear supernatants were treated with a concentration range of TPCK-trypsin for 30 min on ice. The reaction was quenched with soybean trypsin inhibitor and the lysates were analyzed by reducing SDS-PAGE, Western blotting, and detection with either ␣Myc or ␣Ero1␤. Since the Myc tag is located at the C terminus of the protein and the ␣Ero1␤ epitope is located upstream of the (Cxx)CxxC motif (residues Tyr 329 to Leu 343 ), the use of these antibodies can provide some positional information about the fragments.
In each of five experiments, wild-type Ero1␤-Myc was digested into two major fragments (Fig. 1). The largest fragment of ϳ55 kDa appeared with 1.25 g/ml trypsin (fragment A, Fig. 1B, lane 3). This fragment must contain at least the (Cxx)CxxC motif and the C-terminal Myc tag (Fig. 1C, lane 3) and is therefore likely to have lost the N terminus of the protein.
The smaller product (fragment B, Fig. 1B, lane 4) was an ϳ45-kDa species that lacked the C-terminal Myc tag (absent from Fig. 1C, lane 4) but is likely to contain the (Cxx)CxxC motif based on size and the presence of the Tyr 329 -Leu 343 epitope. Given the sizes of the fragments generated and the clustering of lysine residues in an exposed loop in this region, the results suggest that Ero1␤ was selectively proteolysed between the two redox-active domains. Furthermore, a trypsin digestion prediction suggests that Ero1p is also likely to be cleaved in this region.
The digestion patterns of the H254Y mutant (Fig. 1, B and C, lanes [5][6][7][8] and the G252S mutant ( Fig. 1, B and C, lanes 9 -12) and their overall sensitivity to trypsin were comparable with wild-type Ero1␤, suggesting that the gross conformation of each protein was similar. However, there were some differences between the substrates, such as the appearance of a shadow band under the full-length protein in the mutant digestions probed with ␣Myc (Fig. 1C, lanes 7, 8, 11, and 12) and the lower abundance of fragment B in the G252S digestions (Fig. 1B, lane  12). The H254Y and, in particular, G252S mutation may have localized effects upon Ero structure and folding. However, given the overall similarity of the digestion product sizes, we conclude that gross conformational change and increased lysine exposure is unlikely to result from point mutations in the FAD binding site under normal steady state conditions. Differences in the Covalent and Non-covalent Interactions of Ero1␤G252S and Ero1␤H254Y with PDI-Having shown that the G252S and H254Y mutations did not cause extreme structural changes, we investigated the PDI binding properties of the mutants. Previously, we have shown that wild-type Ero1␣ and Ero1␤ both form intermolecular disulfide bonds with PDI (covalent interactions) and both co-immunoprecipitate with PDI (representing a combination of covalent and non-covalent interactions) (22,28). Upon non-reducing SDS-PAGE, Ero1␤ resolves as a monomer and as a collection of disulfide-bonded forms that include PDI-Ero1␤ and Ero1␤-Ero1␤ complexes (22).
Thus HeLa cells were transfected with either G252S ( Fig. 2A) or H254Y (Fig. 2B), and the post-nuclear supernatants were analyzed by reducing and non-reducing SDS-PAGE. Under non-reducing conditions, G252S ran as a monomeric population (OX) and as a collection of trapped disulfide-bonded species ( Fig. 2A, lane 1), which disappeared upon reduction of the samples with DTT ( Fig. 2A, lane 3). Note the background band that was also present in the mock transfectant ( Fig. 2A, lanes 2  and 4). G252S specifically co-immunoprecipitated with the PDI antiserum ( Fig. 2A, lanes 8 and 9), but disulfide-dependent Ero-PDI dimers were not evident when compared with the background (antibody) bands from the mock transfectant ( Fig. 2A,  lanes 6 and 7). Like G252S, the H254Y mutant also ran as a monomer, but the trapped, higher molecular weight species were more abundant and less diffuse than seen with G252S (Fig.  2B, lane 1, compare with Fig. 2A, lane 1). H254Y specifically co-immunoprecipitated with PDI (Fig. 2B, lanes 9 and 10), but unlike G252S, H254Y became trapped in a disulfide-dependent complex with PDI under non-reducing conditions (compare Fig. 2B, lanes 6 and 7 with Fig. 2A, lanes 6 and 7 ).
Ero1␤ G252S and Ero1␤ H254Y Both Dimerize-Wild-type Ero1␤ can form homodimers, and mutating Cys 396 of the Ero1␤ active site impedes both function and dimerization (22). Ero-Ero interactions may therefore be important for Ero1␤ regulation and/or activity (32). To ask whether ero1-1 and ero1-2 phenotypes could be partly explained by a failure of mutant Eros to self-complex, we investigated whether the G252S and H254Y mutants interacted with wild-type Ero1␤. For this, we co-transfected HeLa cells with both HA and Myc-tagged versions of Ero1␤. Since these tagged proteins have different molecular weights, they can be discriminated by their migration on SDS-PAGE gels.
Having demonstrated that the G252S and H254Y mutants interacted with wild-type Ero1␤, we investigated whether mutant-mutant interactions could occur. This was of interest because an Ero1␤C396A mutant heterodimerizes with wildtype Eros but fails to form mutant-mutant dimers (22). Thus cells were transfected with individual HA-tagged mutants or co-transfected with both the Myc-and HA-tagged versions of the same mutant. The lysates were analyzed by SDS-PAGE to verify transfection and were subjected to immunoprecipitation with ␣Myc prior to immunoblotting with the anti-Ero1␤ serum.
Aberrant Oxidation of Ero1␤G252S and Ero1␤H254Y during Reducing and Temperature Stress-The yeast counterparts of G252S and H254Y are hypersensitive to temperature stress and reductants (16,17). We therefore examined the fate of the G252S and H254Y proteins when mammalian cells were subjected to similar conditions. First, we established the effects of DTT treatment on living HeLa cells transfected with wild-type Ero1␤. We found that at steady state, 10 mM DTT in the medium eliminated the majority of disulfide-dependent (Ero-PDI and Ero-Ero) complexes when the lysates were subsequently analyzed by non-reducing SDS-PAGE (Fig. 4A, lanes 1  and 2). A similar result was obtained with cells co-transfected with wild-type Ero1␤-HA and wild-type Ero1␤-Myc (Fig. 4A,  lanes 3 and 4). When the samples were analyzed by reducing SDS-PAGE, a single reduced band was recovered for each mutant, as expected (Fig. 4A, lanes 7-10). The small shift in mobility of Ero1␤ after exposing the cells to DTT was a consequence of increasing availability of -SH groups to the alkylating agent (Fig. 4A, compare lanes 7 and 8 and lanes 9 and 10).

Ero1␤ Misoxidation
Next, we examined the fate of H254Y and G252S after incubating cells with DTT. Cell lysates from treated and untreated transfectants were examined by non-reducing SDS-PAGE and were immunoblotted with ␣Myc. Wild-type Ero1␤ became mostly reduced and lost virtually all detectable disulfide-dependent interactions, as seen in Fig. 4A (Fig. 4C, lanes 1 and 2). However, the H254Y and G252S mutants behaved differently. H254Y maintained its complexes in the face of a DTT challenge (Fig. 4C, lanes 3 and 4, *). For G252S, DTT treatment caused a reproducible increase in the proportion of higher molecular weight complexes and aggregates (Fig. 4C, lanes 5 and 6, **). As an internal control, monomeric Ero1␤ became reduced upon DTT treatment (R, Fig. 4C, lanes 1, 3, and 5). All Ero1␤ molecules could be reduced in vitro when DTT was added to the sample buffer (Fig. 4D, lanes 1-6), demonstrating that the species seen in the non-reducing gels were disulfide-dependent complexes. Note the small shift in Ero1␤ mobility caused by increased exposure of Ero1␤ thiols to the alkylating agent (Fig.  4D, compare lanes 1 and 2, lanes 3 and 4, and lanes 5 and 6). Misoxidation was reversible, since the most severe mutant, G252S, could recover after DTT washout in normal medium prior to lysis (Fig. 4E, lane 3). Other stresses such as hydrogen peroxide challenge or deprivation of glutathione by treatment with butathione sulfoxamine did not result in oxidative misfolding of either mutant or wild-type Ero1␤ at steady state (not shown).
Finally, we asked whether varying the temperature could also invoke behavioral changes in the mutant proteins. Thus transfected HeLa cells were incubated at 24, 37, and 42°C for 1 h prior to lysis. The wild-type Ero1␤ protein was stable across the temperature range and maintained its intermolecular Ero-PDI and Ero-Ero disulfide bonds (Fig. 4F, lanes 1-3, *). In contrast, both H254Y and G252S lost their disulfide-dependent complexes at 24°C (Fig. 4F, lanes 4 and 7, *) and misoxidised at 42°C, with mutant complexes also being retained in the stacking gel (Fig. 4F, lanes 6 and 9, **). Temperature variation did not markedly influence the oxidation state of the remaining monomers, although in general, H254Y tended to have a higher proportion of monomers in the reduced form, regardless of treatment (Fig. 4F, lanes 4 -6, and Fig. 4C, lane 4). All Ero1␤ molecules could be reduced to a single band when DTT was added to the sample buffer, showing that all the complexes were disulfide-dependent (Fig. 4G, lanes 1-9).
We conclude from these experiments that both mutations in the FAD binding domain disrupt disulfide-dependent Ero1␤ interactions and result in misoxidation of the Ero1␤ protein when cells are subjected to either a reducing or temperature stress.
Implications of Ero1␤ Misoxidation-The human and yeast Eros have conserved Gly and His residues at the FAD binding site. Although the properties of the yeast ero1-1 and ero1-2 mutant proteins have not been examined in this study, one would predict that the equivalent mutations in Ero1p also result in stress-induced misoxidation and that the Ero1␤ FAD mutants will not restore viability to the ero1-1 and ero1-2 strains. We have shown that there is an inducible misoxidation defect in the equivalent Ero1␤ gene product upon temperature or reducing stress (Fig. 4). In the absence of FAD as a cofactor and electron acceptor, we suspect that the normal relationship between Ero's intramolecular disulfides and its inter-molecular disulfides with PDI cannot be maintained, resulting in -SH exposure and incorrect selection of S-S bridges by the mutant Ero proteins. The fact that misoxidation could be partly reversed (Fig. 4E) may explain why oxidizing agents such as diamide can restore viability to the ero1-1 yeast strain (16,17).
Recent work has suggested that Ero-Ero dimers contribute to Ero function (22,32). Given that an Ero1␤C396A mutant does not homodimerise efficiently, it was possible that mutations in the FAD binding domain also prevent dimerization. However, our data show that Ero-Ero associations do not depend entirely on the FAD binding domain (Fig. 3) and thus a failure to selfcomplex is not likely to explain the ero1-1 and ero1-2 phenotypes. Nevertheless, it remains possible that Ero-Ero intermolecular disulfide bond formation could be altered in the G252S and H254Y mutants. Structural and enzymatic studies will be required to investigate this question further.
In S. cerevisiae, the conditional ero1-1 strain is unable to oxidize PDI, and PDI remains in the reduced state in these yeast (31). Here, we show that while the equivalent G252S Ero1␤ mutant can bind to PDI non-covalently, intermolecular disulfide bond formation between Ero1 and PDI is largely prevented (Fig. 2). This result demonstrates that Ero1 and PDI can interact in a disulfide and flavin-fold independent manner. We show that residue Gly 252 is required for the establishment of PDI-Ero1 disulfide bonds (Fig. 2), even though the Ero-PDI docking site and the buried FAD binding site are not likely to be in direct contact. This fits with the idea that Ero1-PDI specificity might arise from both the active site cysteines and direct proteinprotein interactions (23) and that structural changes may occur during electron transfer to the Cxx(CxxC) site in Ero1p (32).
The properties of its novel FAD fold make Ero well placed to regulate oxidative protein folding in the ER. The dual function of FAD as an electron transfer agent and as a stabilizing cofactor provides the Ero protein with the ability to catalyze oxidation reactions and, at the same time, maintain appropriate proteinprotein interactions and structural stability in a rapidly changing, stressful environment.