Protein disulfide isomerase (PDI) family members are important enzymes for the correct folding and maturation of proteins that transit or reside in the endoplasmic reticulum (ER). The human PDI family comprises at least 19 members that differ in cell type expression, substrate specificity and post-translational modifications. PDI family A member 2 (PDIA2, previously known as PDIp) has a similar domain structure to prototypical PDI (also known as PDIA1), but the function and post-translational modifications of PDIA2 remain poorly understood. Unlike most PDI family members, PDIA2 contains three predicted N-linked glycosylation sites. By site-directed mutagenesis and enzymatic deglycosylation, we show here that all three Asn residues within the potential N-linked glycosylation sites of human PDIA2 (N127, N284 and N516) are glycosylated in human cells. Furthermore, mutation of N284 to glycosylation-null Gln increases formation of a highly stable disulfide-bonded PDIA2 dimer. Nevertheless, in HeLa cells, both wild-type and N127/284/516Q mutant PDIA2 proteins localize to the ER, but not the ER–Golgi intermediate compartment, suggesting that glycosylation is important for PDIA2 protein–protein interactions but not subcellular localization. Finally, we identified human major histocompatibility complex class 1 antigens (HLA-A,B,C) as potential binding partners of PDIA2, suggesting an involvement for PDIA2 in antigen presentation in addition to its previously described roles in autoimmunity and Parkinson's disease. These results further characterize this poorly defined member of the PDI family.
human major histocompatibility complex class 1 antigens
protein disulfide isomerase
- PNGase F,
peptide N-glycosidase F
Protein disulfide isomerase (PDI) family members are involved in the correct folding and maturation of many proteins that transit or reside in the endoplasmic reticulum (ER). At least 19 members of the human PDI protein family exist, all of which possess at least one thioredoxin-like domain and an ER targeting sequence . Many PDI family members are not fully characterized in terms of structure and function, although it is becoming increasingly recognized that each member has distinct substrate specificities [2, 3]. For example, whereas PDIA1 interacts with both glycosylated and non-glycosylated proteins, ERp57 (also known as PDIA3/Grp58) interacts specifically with glycosylated proteins . Also, while ERp57 is an essential protein, as shown by the non-viability of ERp57−/− mice , genetic ablation of either of the PDI family members AGR2 or ERdj5 produces viable mice, with specific defects in mucin production and salivary gland function, respectively, and increased ER stress in affected tissues [6, 7].
The PDI family member PDIA2 remains poorly characterized. PDIA2 was originally identified from a novel cDNA cloned from a pancreas-derived phagemid library, and was given the name pancreas-specific protein disulfide isomerase (PDIp) . As one of the most abundant microsomal proteins in dog pancreas , PDIA2 constitutes up to 0.1% of all cellular protein . Importantly, PDIA2 has also been detected in mouse stomach , human brain  and several other mouse tissues, including lung, kidney and spleen , indicating a broader pattern of expression than originally described. Recently, PDIA2 has been identified as an auto-antigen in a mouse model of type 1 diabetes  and in mice with autoimmune destruction of pancreatic acinar cells . Furthermore, PDIA2 is increased in a neuronal model of Parkinson's disease, and co-localizes with characteristic protein inclusions in the brain tissue of Parkinson's disease patients , suggesting an involvement of this PDI family member in multiple disease states.
PDIA2 is a 525 amino acid protein with a 21 amino acid ER signal sequence and a C-terminal ER retention signal, KEEL . Similar to PDIA1, PDIA2 has a domain structure comprising an N-terminal ER signal sequence, two CXXC di-cysteine active site motifs in the homologous ‘a’ and ‘a′’ domains, intervening ‘b’ and ‘b′’ domains and an x-linker region, and a C-terminal ‘c’ domain . Unlike most PDI family members, PDIA2 also contains three potential N-linked glycosylation sites with the consensus sequence Asn-Xaa-Ser/Thr at residues N127, N284 and N516 . Treatment of tissue extracts with the peptide N-glycosidase F (PNGase F), which cleaves N-linked glycans from proteins, causes a decrease in the apparent molecular weight of PDIA2 determined by immunoblotting [16-18]. The change in apparent molecular weight following deglycosylation of PDIA2 indicates that more than one site is glycosylated in vivo . However, which of the three potential N-linked glycosylation sites are glycosylated has not been experimentally determined.
Given the lack of understanding of PDIA2 expression and function, we aimed to investigate the N-linked glycosylation and subcellular distribution of PDIA2 in human cells. Here we demonstrate that human PDIA2 is glycosylated at all three potential N-linked glycosylation sites. Interestingly, glycosylation at N284 was found to decrease the formation of a stable disulfide-bonded dimer, suggesting an important role for glycosylation at this site for PDIA2 protein interactions. We also show that PDIA2 co-localizes with PDIA1 and calreticulin in human cells, indicating retention in the ER, and that N-linked glycosylation does not affect subcellular distribution. Finally, we identify human major histocompatibility complex class 1 antigens (HLA-A,B,C) as potential interactors with PDIA2, suggesting an involvement for PDIA2 in antigen presentation. Overall, these results indicate that N-linked glycosylation is important for modulation of PDIA2 dimerization, and suggest a broader role for PDIA2 than originally predicted.
Characterization of the original PDIA2 clone
DeSilva et al.  identified what was predicted to be a full-length clone of PDIp/PDIA2, encoding a 511 amino acid protein that unusually lacked a typical ER signal sequence. This clone, termed Pa-1, contained an aberrant ATG methionine-encoding start codon at the 15th amino acid of the full-length protein. The Pa-1 PDIA2 clone also includes an additional 27 bp upstream of the identified ATG start codon. BLAST analysis of this 27 bp upstream region revealed reverse orientation homology with the first 27 bp of the genomically determined sequence immediately following the genomic ATG start codon (Fig. S1). This finding implies that the original PDIA2 clone was not a full-length clone, but was most probably the result of an aberrant recombination event that resulted in a truncated sequence with introduction of an aberrant ATG start codon. All studies reported here were performed using the full-length genomically determined 525 amino acid PDIA2 sequence (SwissProt number Q13087), in contrast to some previous reports that used the aberrantly truncated form of the protein [8, 10].
PDIA2 is N-linked glycosylated at asparagines 127, 284 and 516
Human PDIA2 has three potential N-linked glycosylation sites, at N127, N284 and N516 (Fig. 1A). The NetNGlyc program  was used to predict the likelihood of glycosylation at each of the three potential N-linked glycosylation sites of human PDIA2. NetNGlyc provides prediction scores for glycosylation that range from 0.0 (not glycosylated) to 1.0 (glycosylated) . This analysis revealed that PDIA2 was likely to be glycosylated at N127 (score 0.7532) and N284 (score 0.6526), but was predicted as unlikely to be glycosylated at N516 (score 0.3874).
In order to experimentally determine which of the three potential N-linked glycosylation sites of PDIA2 are glycosylated in human cells, individual asparagine residues were mutated to glutamine singly or in combination, and the apparent molecular weights of the resultant proteins were analysed by immunoblotting. Mutation of each of the three asparagine residues individually resulted in faster migration of the resultant protein, and this size shift was additive in the PDIA2 proteins with two or all three asparagine residues mutated (Fig. 1B). Variable levels of unglycosylated PDIA2, found as the lower approximately 57 kDa band, were detected in separate experiments for all samples, most likely due to slight variation in the high level of over-expression of the proteins, variation in exposure times of immunoblots and differences in the degree of separation of the bands by electrophoresis. These results indicate that PDIA2 is N-linked glycosylated at all three putative glycosylation sites: N127, N284 and N516. HEK293T cell lysates expressing wild-type or glycosylation-null mutant PDIA2 proteins were also treated with the deglycosylation enzymes endoglycosidase H (EndoH) and PNGase F (Fig. 1C). Deglycosylation with EndoH resulted in a decrease in the apparent molecular weight of the individual N127Q, N284Q and N516Q mutants (Fig. 1D), as well as the double mutants N127/284Q, N127/516Q and N284/516Q mutants (Fig. 1E). Deglycosylation with either EndoH or PNGase F resulted in a similar decrease in the apparent molecular weight of wild-type PDIA2 but no change in N127/284/516Q mutant PDIA2 (Fig. 1F). Sensitivity to both PNGase F and EndoH indicates that the N-linked glycans of PDIA2 are not modified by enzymes that are resident in the Golgi apparatus, implying efficient ER retention of the protein.
Glycosylation at N284 modulates formation of an unusually stable disulfide-bonded PDIA2 dimer
Immunoblot analysis of HEK293T cell lysates expressing untagged or tagged wild-type or glycosylation-null mutant PDIA2 proteins also revealed an approximately 135 kDa band that was detected using antibodies against either PDIA2 or FLAG (Fig. 2A,B). The antibody against PDIA2 also recognized a probably non-specific protein at approximately 50 kDa, and for this reason the antibody against FLAG was also used in all experiments to confirm specificity for over-expressed PDIA2. Interestingly, the N284Q glycosylation-null protein showed an increased abundance of the 135 kDa band, as detected by immunoreactivity to both antibodies against PDIA2 or FLAG (Fig. 2A,B). The N284Q PDIA2 mutant also showed several additional PDIA2- and FLAG-immunoreactive bands of 100–120 kDa. Levels of the higher-molecular-weight PDIA2 proteins were also similarly increased in double and triple glycosylation-null proteins containing N284Q (data not shown). These results suggest that glycosylation at N284 inhibits formation of the 135 kDa species and also prevents formation of high-molecular-weight complexes.
The 135 kDa band detected here was present even under mildly reducing conditions comprising 5 min incubation at room temperature with 10% β-mercaptoethanol (β-ME). This was in contrast to a previous report in which a PDIA2 produced recombinantly or in an artificially truncated form formed a homodimer that was readily eliminated by incubation with 10 mm dithiothreitol (DTT) . In order to further characterize the PDIA2 dimer, we compared expression of N127/284/516Q PDIA2 under control non-reducing conditions and in samples reduced with 10% β-ME as previously, or with 200 mm DTT, or highly denatured with 4 m urea (Fig. 2C). In non-reduced (control and 4 m urea) samples there was a high-molecular-weight smear of protein that was immunoreactive for PDIA2 but was completely absent from reduced samples, suggesting the presence of easily dissociated mixed disulfide-bonded complexes between PDIA2 and other interacting partners or substrates. There was a concomitant increase in the approximately 62 kDa monomeric PDIA2 protein in reduced samples. Unexpectedly, the levels of the approximately 135 kDa PDIA2 dimer were largely unaltered in reduced or denatured samples, indicating formation of a highly stable complex. However, heating at 95°C for 15 min with either of the reductants β-ME, DTT or tris(2-carboxyethyl)phosphine (TCEP) either drastically decreased or eliminated the 135 kDa band in both wild-type (Fig. 2D) and the N127/284/516Q mutant (Fig. 2E), with a concomitant increase in the amount of monomeric PDIA2. These results indicate that, in mammalian cells, PDIA2 forms a stable disulfide-bonded dimer, and that this dimer is more stable than high-molecular-weight complexes formed between PDIA2 and other proteins.
In order to confirm the identity of the 135 kDa PDIA2-immunoreactive species, we performed immunoprecipitation using an antibody against FLAG followed by excision of the 135 kDa band identified by silver staining with tryptic digestion and mass spectrometry analysis. Peptides corresponding to human PDIA2 were identified with up to 14.1% sequence coverage from the approximately 135 kDa band from cell lysates expressing either wild-type or glycosylation-null PDIA2 proteins; however, peptides derived from no other proteins were consistently detected in this band. These results suggest that the 135 kDa band corresponds to a PDIA2 homodimer.
PDIA2 co-immunoprecipitates HLA-A,B,C
An additional unexpected band of approximately 50 kDa was identified from silver-stained gels of wild-type PDIA2-expressing cell lysates immunoprecipitated for PDIA2 using anti-FLAG antibody. We detected multiple peptides derived from HLA-A,B,C (UniProt ID Q95365, and others) in minor amounts from this approximately 50 kDa band, suggesting an interaction between PDIA2 and HLA-A,B,C. In order to confirm this interaction, we performed co-immunoprecipitation experiments using an antibody against HLA class 1 in HEK293 cell lysates over-expressing untagged wild-type PDIA2 or tagged N127/284/516Q mutant PDIA2, followed by immunoblotting for PDIA2. Both wild-type and mutant PDIA2 were detected following immunoprecipitation with antibody against HLA class 1, but were not detected in control reactions containing either no antibody or lysates of wild-type PDIA2 over-expressing cells and non-specific IgG, confirming the interaction of PDIA2 with HLA-A,B,C detected by mass spectrometry (Fig. 3).
N-linked glycosylation does not affect ER localization of human PDIA2 proteins in human cells
Since other PDI family members, including PDIA1 and ERp57, are present in subcellular compartments in addition to the ER , we next aimed to determine the subcellular distribution of human PDIA2 and investigate whether glycosylation altered the subcellular distribution. HeLa cells were transiently transfected with plasmids encoding wild-type or glycosylation-null mutant PDIA2, and cells were processed for immunocytochemistry, using calreticulin as a marker of the ER and FLAG to detect the exogenously expressed tagged PDIA2. As controls, FLAG immunoreactivity was not detected in HeLa cells transfected with untagged wild-type PDIA2, and immunofluorescence for FLAG completely co-located with PDIA2 in FLAG-tagged wild-type PDIA2 expressing cells, indicating the specificity of the FLAG antibody (Fig. 4A,B). Wild-type and glycosylation-null PDIA2 proteins (single mutant N127Q, N284Q, N516Q and triple mutant N127/284/516Q) were completely co-located with calreticulin, a classical ER marker, and all showed similar reticular distribution reminiscent of the ER (Fig. 4C–G′). Similarly, PDIA2 co-located with PDIA1 but not with ERGIC53, a marker of the ER–Golgi intermediate compartment (Fig. 5). These results, combined with the finding that PDIA2 N-linked glycans are not modified by Golgi-resident enzymes (Fig. 1), indicate that PDIA2 is retained within the ER and does not transit to the Golgi apparatus. Furthermore, the ER location of PDIA2 is not affected by the presence or lack of N-linked glycosylation.
PDI family members are abundant proteins that are important for modulating disulfide bond formation in many proteins, with differing substrate specificity [2, 9]. Here we describe the N-linked glycosylation and expression of PDIA2, a poorly defined member of the PDI family. Human PDIA2 was shown to be glycosylated at the asparagine residues of all three potential N-linked glycosylation sites (N127, N284 and N516), and prevention of glycosylation at N284 was found to increase the formation of a highly stable PDIA2 dimer. By immunocytochemistry and confocal microscopy, we showed that PDIA2 co-localized with both PDIA1 and calreticulin in a typical reticular pattern indicative of the ER, and mutation of N-linked glycosylation sites did not affect the subcellular distribution. PDIA2 did not co-localize with a marker of the ER–Golgi intermediate compartment, indicating that the protein is retained within the ER. Finally, we identified HLA proteins as potential interacting partners of PDIA2.
The finding that the PDIA2 dimer increased in abundance when the N284 glycosylation site was mutated implies that glycosylation at this site may inhibit formation of this dimer endogenously. It has been suggested that formation of a PDIA2 dimer is increased under conditions of oxidative stress, and that the PDIA2 dimer has increased chaperone activity compared to the monomeric form . Therefore, regulation of glycosylation at N284 may be one mechanism by which the activity of PDIA2 is endogenously controlled. Recently, PDIA1 has also been shown to form a homodimer via b′–b′ domain interactions, indicating that dimerization of PDI family members may be a general mechanism of regulation of PDI function . Indeed, the only glycosylation site that affects dimerization of PDIA2, N284, is within the b′ domain of PDIA2, suggesting that PDIA2 dimerization is also dependent on the b′ domain. The b′ domain of PDIA1 is involved in hydrophobic ligand interactions , suggesting that N-linked glycosylation at N284 may prevent or modulate PDIA2 substrate binding, although, as PDIA1 lacks N-linked glycosylation sites, this is likely to be a unique property of PDIA2. PDIA2 was shown previously to interact with other proteins both containing and lacking cysteine residues, similar to the protein interactions of PDIA1 [9, 22]. This suggests that either the initial interaction of PDIA2 with disulfide isomerase substrates is independent of disulfide bonding, or that PDIA2 is able to act as a chaperone independently of protein disulfide bonds . N-linked glycosylation at N284 may be one mechanism by which PDIA2 substrate binding is modulated in vivo.
The formation of a PDIA2 homodimer, involving a disulfide bond between the C18 residues of two PDIA2 proteins, has been reported in a previous study that used recombinantly produced or artificially truncated forms of the protein . In this previous study, the PDIA2 dimer was readily eliminated by incubation with 10 mm DTT. However, it should be noted that this previous study investigated either recombinantly produced purified protein from bacteria or PDIA2 lacking the most N-terminal 14 amino acids over-expressed in COS7 cells . The ER signal sequence of PDIA2 has been predicted to span the first 21 amino acids of the protein , suggesting that PDIA2 lacking the first 14 amino acids is unlikely to be correctly targeted to the ER. As the ER is the subcellular site of formation of most disulfide bonds, this may account for the differences in the apparent stability of the PDIA2 dimer between the two studies; in the study by Fu and Zhu , the levels of PDIA2 disulfide bond formation is predicted to be lower due to the lack of the ER targeting sequence. It is also important to note that the cysteine residue identified by Fu and Zhu  as responsible for PDIA2 homodimerization, C18, is also within the 21 amino acid ER-targeting sequence of PDIA2, which is predicted to be cleaved upon entry to the ER . Therefore, the exact mechanism of the disulfide-bonded dimer formation remains to be determined. It will be important to fully define the domain boundaries of PDIA2 and determine the involvement of each of the six cysteine residues of PDIA2 in formation of the dimer to fully understand the molecular interactions of the protein. Further investigation of the role of dimerization in controlling PDIA2 isomerase activity is also warranted.
One step towards understanding the function of PDIA2, and identifying whether or not this protein has different specificities compared to other PDI family members, is identification of the endogenous protein substrates. In the present study, immunoprecipitation and mass spectrometry data demonstrated that HLA-A,B,C may be endogenous substrates of PDIA2. This finding is of particular interest given previous reports of the involvement of PDIA2 as an auto-antigen in a mouse model of type 1 diabetes  and in mice with a fatal lymphoproliferative disease caused by lack of cytotoxic T-lymphocyte antigen 4 . While it is possible that our immunoprecipitation experiments detected peptides derived from PDIA2 that were bound to HLA as presentation antigens, it is most likely that PDIA2 binds to HLA as a chaperone or disulfide isomerase. Previous studies have identified the PDI family members ERp57 and TMX as important for the assembly, maturation and retrotranslocation of HLA class 1 proteins [5, 23, 24], and our results suggest that PDIA2 is similarly involved in HLA processing. This indicates that PDIA2 may be involved as an auto-antigen in disease states as well as part of the normal process of HLA-A,B,C folding. However, given the limited tissue distribution of HLA class 1 proteins, the physiological significance of the interaction of PDIA2 with HLA in the mammalian body requires further investigation.
The findings presented here indicate that PDIA2 is an unusual PDI family member in which N-linked glycosylation is involved in modulating the formation of a highly stable disulfide-bonded dimer, and suggest that PDIA2 may play a broader role in protein folding than originally described.
A pCMV6 vector containing the open reading frame of full-length human PDIA2 (GenBank ID NM_006849.2) with C-terminal myc and DDK/FLAG tags was supplied by Origene (Rockville, MD, USA). This vector contains the PDIA2 sequence inserted between the SgfI and MluI restriction enzyme sites. To allow expression of full-length untagged human PDIA2, a stop codon was introduced in to the PDIA2–myc-DDK/FLAG-encoding construct immediately upstream of the myc-DDK/FLAG tags, using the primers 5′-CCAAGGAGGAACTGTAGCGTACGCGGCCGC-3′ (forward) and its reverse complement by QuikChange site-directed mutagenesis (Stratagene/Agilent, Santa Clara, CA, USA) according to the manufacturer's instructions. Recent evidence suggests that the six most C-terminal amino acids may be involved in KDEL receptor-mediated ER protein retention . It remains unknown whether or not PDIA2 requires interaction with the KDEL receptors for normal ER retention. Therefore, to allow expression of full-length myc-DDK/FLAG-tagged PDIA2 with a far C-terminal ER retention sequence (termed ‘wild-type’ throughout), the DNA sequence for the six most C-terminal PDIA2 amino acids (GSKEEL) was inserted following the tags and immediately upstream of the stop codon, using the pCMV6 PDIA2–myc-DDK/FLAG vector as a template by site-directed mutagenesis with the primer 5′-CGACGATAAGGTTGGGTCCAAGGAGGAACTGTAAACGGCCGG-3′ and its reverse complement. To mutate each of the three potential PDIA2 N-linked glycosylation sites from asparagine to glutamine, site-directed mutagenesis was performed using the wild-type PDIA2–myc-DDK/FLAG–GSKEEL construct as template, using the following primers and their respective reverse complements: N127Q, 5′-TCTTCCGCAATGGGCAGCGCACGCACCCG-3′; N284Q, 5′-CTGCTGCTGTTTGTCCAGCAGACGCTGGCTGCG-3′; N516Q, 5′-GGAGCCACCGGCCCAGTCCACTATGGGGT-3′. Double and triple glycosylation site mutants were produced by subsequent rounds of site-directed mutagenesis. All vectors were sequenced in both the forward and reverse directions to confirm correct insert and flanking sequences.
Cell culture and transfection
Human HeLa and HEK293T cells were maintained in Dulbecco's modified Eagle's medium (Gibco/Life Technologies, Grand Island, NY, USA) with 10% fetal bovine serum, 100 μg·mL−1 penicillin and 100 μg·mL−1 streptomycin. Cells were plated 24 h prior to transfection with the indicated plasmids using Lipofectamine 2000 (Invitrogen/Life Technologies) according to the manufacturer's instructions. Cells lysates for immunoblotting were collected in TN buffer (50 mm Tris/HCl pH 7.5 and 150 mm NaCl) with 0.1% SDS, 0.1% Nonidet P-40 and 1% protease inhibitor cocktail (Sigma-Aldrich, St Louis, MO, USA) by incubation on ice for 10 min. Cell lysates for immunoprecipitation were collected in TN buffer with 0.1% Nonidet P-40 and 1% protease inhibitor. The supernatant (SDS-soluble fraction) was cleared by centrifugation at 16 100 g for 10 min. Protein concentrations were determined using the BCA assay (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions.
HeLa cells grown on 13 mm glass cover slips were washed for 10 min with NaCl/Pi 48 h post-transfection, and fixed with 4% paraformaldehyde in NaCl/Pi for 10 min. Cells were permeabilized in 0.1% Triton X-100 in NaCl/Pi for 10 min, blocked for 30 min with 1% BSA in NaCl/Pi, and incubated with primary antibodies [FLAG, 1 : 500, mouse F3165, Sigma-Aldrich; PDIA1, 1 : 100, mouse 2792, Abcam (Cambridge, MA, USA), PDIA2, 1 : 100, goat LS-C2830, LifeSpan (Seattle, WA, USA), calreticulin, 1 : 200, rabbit 120-29070, Abcam; ERGIC53, 1 : 100, rabbit E1031, Sigma-Aldrich] for 16 h at 4 °C. Cells were washed for 10 min three times in NaCl/Pi, and then incubated with secondary antibodies (AlexaFluor goat anti-rabbit IgG-488/594, goat anti-mouse IgG-488/594 or donkey anti-goat IgG-488, all from Molecular Probes/Life Technologies, Grand Island, NY, USA) at 1 : 5000 dilution in NaCl/Pi for 1 h at room temperature. Cells were treated with Hoechst 33342 (1 : 10 000, Molecular Probes/Life Technologies) prior to mounting in fluorescent mounting medium (Dako Glostrup, Denmark). Images were acquired for constant gain and offset using a Fluoview 1000 inverted confocal laser-scanning microscope (Olympus, Tokyo, Japan).
For deglycosylation, soluble proteins extracted from HEK293T cells (5 μg) were denatured in 0.5% SDS and 40 mm DTT, heated at 100 °C for 10 min, and incubated with EndoH (1000 U, where 1 U equals the amount of enzyme required to remove over 95% of carbohydrate from 10 μg of denatured RNase B in 1 h at 37 °C in a total reaction volume of 10 μL) or PNGase F (1000 U) (New England Biolabs, Ipswich, MA, USA) at 37 °C for 1 h according to the manufacturer's instructions. Control samples were denatured and processed without addition of enzyme. All samples were subsequently analysed by immunoblotting.
Immunoprecipitation and mass spectrometry
HEK293T cell extracts (20–200 μg) were incubated with antibodies against FLAG (1 : 1000) or HLA class 1 (1 : 500, mouse W6/32; Sigma-Aldrich) and BSA-blocked protein A–Sepharose CL-4B (Amersham/GE Healthcare, Little Chalfont, UK) in TN buffer with 0.1% Nonidet P-40 and 1% protease inhibitor at 4 °C for 16 h. Control samples comprised transfected cell lysates without antibody or with irrelevant mouse IgG, or untransfected cell lysates with the relevant antibody. Beads were washed for 15 min three times prior to elution in SDS sample buffer and SDS/PAGE. Gels were stained using a Pierce silver stain kit according to the manufacturer's instructions (Thermo Fisher Scientific). For mass spectrometry, protein bands were excised from the gels, reduced, alkylated and trypsinized. Tryptic peptides were extracted from the gel and separated by reversed-phase HPLC (Ultimate 3000, Dionex/Thermo Fisher Scientific) using a Vydac Everest capillary column (C18, 300 Å pore size, 5 μm particle size, 150 μm internal diameter, 150 mm length; Grace Discovery Sciences, Deerfield, IL, USA). Samples were rinsed by a 10 min isocratic wash with buffer A (5% v/v acetonitrile and 0.05% v/v aqueous trifluoroacetic acid) at 1 μL·min−1. Peptides were eluted using a gradient of 1% increment per min of buffer B (80% v/v acetonitrile in 0.05% aqueous trifluoroacetic acid) for 12 min, followed by 1.5% increment per min of buffer B for 25 min. The column eluate was fractionated onto an AnchorChip 600/384 plate (Bruker Daltonics) with a sheath flow of α-cyano-4-hydroxycinnamic acid dissolved in 30% v/v acetonitrile matrix using a Proteineer fc fraction collector (Bruker Daltonics) with 10 s collection time for each spotted fraction. Each spot was analysed using an UltraflexIII MALDI-TOF/TOF-MS (Bruker Daltonics Bremen, Germany) under the control of Bruker's proprietary software, WARP-LC. The acquired MS/MS spectra of peptides were searched against the human protein database from SwissProt (ExPASy, http://www.expasy.org/proteomics) using the Mascot search engine (Matrix Science, London, UK). A 50 p.p.m. mass window was used as mass tolerance in MS spectra, and 0.8 Da tolerance was used in the MS/MS spectra. Carbamidomethyl was used for the global modification of cysteine.
SDS/PAGE and immunoblotting
Protein samples (1–20 μg) were subjected to 9.25% or 12% SDS/PAGE and transferred to nitrocellulose membranes. Membranes were blocked with 3% BSA (Invitrogen, Carlsbad, CA, USA) in 10 mm Tris, 150 mm NaCl, 0.05% tween-20, pH 8.0, and then incubated with primary antibodies (FLAG, 1 : 4000; PDIA2, 1 : 2000; β-actin, 1 : 4000, mouse A5441, Sigma-Aldrich; Glyceraldehyde 3-phosphate dehydrogenase, 1 : 4000, mouse MAB-374, Millipore Billerica, MA, USA). Membranes were washed, then incubated for 1 h at room temperature with secondary antibodies [horseradish-peroxidase-conjugated donkey anti-sheep/goat (1 : 4000, AB324P) or goat anti-mouse (1 : 4000, AB326P) (both from Millipore)]. Signals were detected using enhanced chemiluminescent reagent (Roche Penzberg, Germany or Amersham/GE Healthcare) with Kodak Biomax MR film (Sigma-Aldrich) or a ChemiDoc XRS system (Bio-Rad, Hercules, CA, USA). Blots were stripped of antibodies using ReBlot Plus Strong solution (Millipore) and re-probed as above.
The NetNGlyc 1.0 program from the Center for Biological Sequence Analysis of the Technical University of Denmark was used to predict the likelihood of N-linked glycosylation (http://www.cbs.dtu.dk/services/NetNGlyc) .
This work was supported by the National Health and Medical Research Council of Australia (project grant 454749), the Amyotrophic Lateral Sclerosis Association (USA), the MND Research Institute of Australia, the Bethlehem Griffiths Research Council, a Henry H. Roth Charitable Foundation Grant for MND Research, and Australian Rotary Health. A.K.W. holds a National Health and Medical Research Council C.J. Martin Biomedical Early Career Fellowship (number 1036835).