Domain architecture and sequence characteristics of PDI8

The Arabidopsis PDI8 gene contains five exons and encodes a deduced polypeptide of 440 amino acids [20]. The first 22 amino acids of the deduced PDI8 sequence are predicted by SignalP-4.1 to serve as a cleavable signal peptide (mean S value = 0.936), with the resulting mature PDI8 protein having a calculated molecular weight of 47.4 kDa and a theoretical pI of 5.01. PDI8 is predicted by TMHMM v. 2.0 to contain a single TMD, spanning residues 378-400 of the PDI8 preprotein sequence. Secondary structure prediction of the PDI8 preprotein by SPIDER2 revealed an alternating pattern of α-helices and β-strands, including three intervals with the thioredoxin structural fold, βαβαβαββα (Fig. 1a). Protein domains belonging to the thioredoxin fold class are identified on the basis of their secondary structural elements, rather than actual sequence homology to the cytoplasmic redox protein, thioredoxin [4]. Despite their predicted structural resemblance to thioredoxin, the three thioredoxin-fold domains of PDI8 do not share significant sequence homology to each other, and only the first domain (domain a in Fig. 1a) shares homology to canonical thioredoxin proteins.

By convention, PDI redox-active thioredoxin-fold domains are referred to as a domains, while redox-inactive thioredoxin-fold domains are termed b domains [3]. The N-terminal-most thioredoxin-fold domain of PDI8 is an a-type domain sharing 42 % and 35 % sequence identity with the a and a’ domains of human PDI, respectively, and contains the CGHC redox active site motif found in the a and a’ domains of the classical PDIs from human and yeast. The other two thioredoxin-fold domains of PDI8 do not contain any potentially redox-active Cys residues, and were thus designated as b-type domains (b, b’). BLAST searches of the Arabidopsis TAIR10 protein database indicated that the b and b’ domains of PDI8 do not share significant homology to other proteins from Arabidopsis, including the b and b’ domains of other members of the PDI family. Furthermore, although PDI8 shares a similar domain arrangement to human thioredoxin-related membrane protein 3 (TMX3; Fig. 1b), no homology was found at the amino acid level between the b-type domains of PDI8 and TMX3 in pairwise sequence similarity comparisons using the BLAST algorithm.

PDI8 promoter expression analysis using the GUS reporter system

To examine the spatial expression pattern of PDI8 in planta, we generated transgenic Arabidopsis plants harboring the ~2.3-kb region immediately upstream of the PDI8 start codon (including the PDI8 promoter and 5’ untranslated region) transcriptionally fused to the reporter gene, β-glucuronidase (GUS). A total of 11 independent transgenic lines were analyzed to establish the consensus expression pattern of the PDI8 _pro :GUS fusion in seedlings and flowering plants. Histological staining of 7-day-old seedlings revealed strong expression of the GUS transgene in the emerging first true leaves, cotyledons, roots, and the base of the hypocotyl (Fig. 2a). In cotyledons, GUS staining was primarily detected in the vasculature and guard cells (Fig. 2b). In roots, GUS staining was observed exclusively in the vasculature, both at the mature zone (Fig. 2c) and the root tip (Fig. 2d). The staining pattern of 14-day-old PDI8 _pro :GUS plants (Fig. 2e) was similar to that of 7-day-old seedlings, although GUS staining in older (expanded) leaves was primarily confined to the vasculature (Fig. 2f), whereas strong GUS staining was observed throughout younger (emerging) leaves (Fig. 2g). However, we did not observe significant GUS staining at the shoot apical meristem (Fig. 2g).

In 6-week-old reproductive-stage plants, expression of the PDI8 _pro :GUS transgene was detected at the style, and in the vasculature of petals, sepals and stamen filaments (Fig. 2h). Strong GUS expression was also present in pedicels, although the pedicels of older flowers exhibited decreased GUS staining near the pedicel/flower junction (Fig. 2h). We also detected significant GUS expression in siliques (Fig. 2i), and the pedicel/stem junction (Fig. 2j).

The expression pattern of PDI8 was also examined by mining publicly-available microarray data through the Bio-Analytic Resource ePlant Browser (https://bar.utoronto.ca/eplant/; [33]). Consistent with our GUS reporter expression analyses, PDI8 transcripts were detected across many plant tissues, including roots, leaves, flowers and siliques (Additional file 1). The highest mean expression values corresponded to expanding siliques, heart and globular-stage embryos, pedicels, 24 h imbibed seeds, and the 2^nd internode of the inflorescence stem, while the lowest mean expression value corresponded to mature pollen.

PDI8 localizes primarily to the ER

The subcellular localization pattern of PDI8 was examined using two different approaches: 1) transient expression of a PDI8 fusion to the green fluorescent protein variant, GFP(S65T), in Arabidopsis leaf protoplasts, and 2) detection of native PDI8 in wild-type Arabidopsis ultra-thin sections by transmission immunoelectron microscopy. For the first approach, since PDI8 potentially contains both a signal peptide at its N-terminus and an ER retrieval signal at its C-terminus, we generated two constructs expressing GFP(S65T) at different positions relative to the PDI8 open reading frame (Fig. 3a). In the spGFP-PDI8 fusion, GFP(S65T) is positioned internally between the signal peptide and mature peptide sequences of PDI8. In the PDI8-GFP-KKED, GFP(S65T) is positioned at the C-terminus of the PDI8, with the C-terminus of GFP(S65T) modified to include the putative ER retention sequence of PDI8, KKED. When transiently co-expressed in protoplasts with a marker for the ER, both the spGFP-PDI8 and PDI8-GFP-KKED fusions exhibited a subcellular distribution pattern that strongly overlapped with that of the network-like localization pattern of the ER-mCherry, whereas unfused GFP(S65T) displayed a distribution pattern that was noticeably more diffuse than the ER-mCherry marker (Fig. 3b).

To facilitate the higher-resolution subcellular localization of PDI8, a PDI8-specific polyclonal antiserum was raised in rabbits against a truncated version of PDI8 containing the b-b’ region (PDI8^bb’; Fig. 1b) of the protein. The reactivity and specificity of the anti-PDI8 antiserum was examined by immunoblot analysis against recombinant PDI8^bb’, and against total protein samples extracted from 7-day-old wild-type (WT) Arabidopsis seedlings and transgenic plants expressing the PDI8 cDNA under the strong constitutive CaMV 35S promoter in either the sense orientation (35S _pro :PDI8) or antisense orientation. The anti-PDI8 antiserum strongly detected the recombinant PDI8^bb’ protein (Additional file 2a), and a protein with a MW of ~54 kDa in both WT and 35S _pro :PDI8 lines (Additional file 2b). The 54-kDa protein was detected very strongly in 35S _pro :PDI8 overexpression lines relative to WT, indicating that this protein corresponds to PDI8 in plants. We did not observe any phenotype associated with either overexpression or antisense suppression of PDI8. However, analysis of transcript levels in the PDI8 antisense lines by quantitative reverse transcription PCR (RT-qPCR) showed that the endogenous PDI8 gene was only partially suppressed in these lines (40–50 %), indicating that the obtained antisense lines were not true knockouts (data not shown).

For a high resolution analysis of the localization pattern of native PDI8, we prepared specimens for immunogold labeling from the shoot and root apices of wild-type Arabidopsis seedlings using high-pressure freezing and freeze-substitution. After sectioning the specimens, they were labeled with the anti-PDI8 antiserum, followed by secondary labeling with a gold-conjugated anti-rabbit antiserum. In shoot apical cells, prominent labeling of the ER by the anti-PDI8 antiserum was observed (Fig. 4a). This antiserum also labeled the ER in root apical cells (Fig. 4b). We did not detect significant anti-PDI8 labeling of any other sub-cellular structures. No labeling was observed using the pre-immune serum on sections from wild-type seedlings nor using the anti-PDI8 antiserum on the antisense line (Additional file 3a, b, c). Thus, the ER labeling observed using the anti-PDI8 antiserum (Fig. 4a) was specifically detecting PDI8. Sections from 35S _pro :PDI8 overexpression lines labeled with the anti-PDI8 antiserum displayed strong labeling of the ER, indicating that the overexpression of PDI8 in plants does not lead to mislocalization of the protein (Additional file 3d).

PDI8 is a type I integral membrane protein

To further the molecular characterization of PDI8, the orientation of the PDI8 protein in microsomal membranes was investigated. Since overexpression of PDI8 under the CaMV 35S promoter does not lead to mislocalization of PDI8 in stably transformed plants (Additional file 3d), or when transiently expressed in mesophyll protoplasts in the form of the spGFP-PDI8 or PDI8-GFP-KKED fusions (Fig 3b), microsomes were prepared from 35S _pro :PDI8 plants due to the strong PDI8 signal these lines exhibited on immunoblots. Separation of the 35S _pro :PDI8 protein sample into soluble and microsomal membrane protein fractions revealed that PDI8 was exclusively associated with the microsomal membrane fraction (Fig. 5a, upper panel, lanes 2 and 3 from the left). The microsomes were also tested for the presence of a microsomal marker protein, the soluble ER lumen protein, BiP, by using a polyclonal antibody recognizing BiP. BiP was primarily found in the microsomal fraction (Fig. 5a, middle panel), but a minor amount of BiP was also detected in the soluble protein fraction, which presumably was due to the escape of some proteins from the ER lumen during the mechanical fragmentation of the ER network to produce microsomes. Coomassie staining of an SDS-PAGE gel loaded with equivalent volumes of the total protein, soluble protein, and microsomal protein fractions demonstrated that the large subunit of Rubisco (which serves as a marker for soluble proteins) was present in both the total protein and soluble protein fractions in similar amounts, but was absent in the microsomal fraction (Fig. 5a, lower panel).

Since PDI8 is predicted to contain a single TMD near its C-terminus, we sought to address whether the N-terminal a-b-b’ region of PDI8 was lumenal (type I membrane protein) or cytoplasmic (type II). 35S _pro :PDI8 microsomal membranes were treated with proteinase K to ascertain if the PDI8 N-terminal region was located within the interior of microsomes, and would therefore be protected from degradation. As shown in Fig. 5b, treatment of 35S _pro :PDI8 microsomes with protease caused a downward shift in the apparent MW of PDI8 to ~48 kDa (compare lanes 1 and 2), while treatment with both protease and a detergent (Triton X-100) to disrupt the microsomal membranes resulted in the complete degradation of PDI8 (lane 4). Treatment with detergent alone had no effect on the apparent MW of PDI8 (lane 3). Since the C-terminal tail of PDI8 (residues 401–440) contributes a theoretical ~5 kDa to the total MW of PDI8, the observed minor decrease in the MW of PDI8 following proteinase K treatment is consistent with the C-terminal tail being located on the outside of microsomes. A model of the PDI8 polypeptide oriented in the ER membrane is shown in Fig. 5c, indicating that the catalytic domain (a’) and thioredoxin fold domains (b, b’) are oriented into the lumen of the ER.

Heterologous expression of PDI8 functionally complements the dsbA ⁻ mutation in E. coli by reconstituting alkaline phosphatase activity

To gain further insight into the molecular function of PDI8, we examined if PDI8 can functionally complement the E. coli oxidative protein folding mutant, dsbA ⁻ . The E. coli thioredoxin-fold protein, DsbA, plays a crucial role in the oxidative folding of proteins within the bacterial periplasm by catalyzing the formation (dithiol oxidation) of protein disulfide bonds. Loss-of-function mutations of dsbA disrupt the proper folding of several proteins, including alkaline phosphatase (PhoA), which in its native state is a homodimer containing two disulfide bonds in each of its subunits [28]. PhoA activity is substantially reduced in a dsbA ⁻ null mutant background, but can be restored by expressing human PDI in the periplasm of dsbA ⁻ cells [14].

To determine if PDI8 can likewise restore PhoA activity in dsbA ⁻ mutant cells, the coding sequence for the lumenal portion of PDI8 (PDI8^abb’; Fig. 1b) was cloned into the bacterial expression vector, pFLAG-CTS, between the vector sequences coding for the OmpA signal peptide (for bacterial periplasmic localization) and C-terminal FLAG epitope tag. The resulting plasmid, pFLAG-PDI8^abb’, was transformed into E. coli strain RI90, which harbors the dsbA null mutation, dsbaA1::kan1. As shown in Fig. 6, PhoA activity in the dsbA ⁻ strain (column 2) or dsbA ⁻ strain transformed with the pFLAG-CTS empty vector (column 3) was substantially reduced relative to the isogenic wild-type (dsbA ⁺ ) parental control strain RI89 (column 1), whereas dsbA ⁻ cells expressing PDI8^abb’ exhibited levels of PhoA activity similar to that of wild-type dsbA ⁺ cells (column 4). Thus, the lumenal portion of PDI8 can functionally substitute for the disulfide oxidase role of DsbA in E coli.

Bioinformatic analyses and identification of PDI8 homologs

To identify homologs of PDI8, BLAST (Basic Local Alignment Search Tool) searches were performed against both the National Center for Biotechnology Information (NCBI) non-redundant (nr) protein sequence database (http://blast.ncbi.nlm.nih.gov/Blast.cgi) and the Phytozome v10 (http://phytozome.jgi.doe.gov), using the bb’ region of PDI8 as the search query sequence due to its absence of homology to the other 13 PDIs from Arabidopsis. Whenever possible, incomplete or incorrectly annotated protein sequences were corrected based on available expressed sequence tag (EST) sequences. All sequences and their corresponding accession numbers are provided in Additional file 4, with alterations to the original source sequences highlighted in yellow.

Signal peptide cleavage site prediction for PDI8 was performed using the program SignalP (v. 4.1) (http://www.cbs.dtu.dk/services/SignalP/; [23]). The predicted locations of TMDs were obtained using the hidden Markov model-based membrane protein topology prediction program, TMHMM (v. 2.0) (http://www.cbs.dtu.dk/services/TMHMM/; [18]). Protein secondary structure predictions for α-helices and β-strands were obtained using the program SPIDER2 (http://sparks-lab.org/yueyang/server/SPIDER2/; [13]).

Generation of Arabidopsis transgenic plants

A fragment containing the PDI8 coding sequence was inserted between the cauliflower mosaic virus (CaMV) 35S promoter and nopaline synthase (NOS) terminator sequences of binary vector pCAMBIA1302 to create the PDI8 over-expression construct, pC1302[35s _pro :PDI8]. The PDI8 coding sequence was amplified from first-strand cDNA prepared from 7-day-old Arabidopsis ecotype Columbia-0 (Col-0) seedlings using a forward primer containing an engineered SphI restriction site (5’-TCG GCA TGC GTT CGT TAA AGT TAC TCC TTT G-3’), and a reverse primer containing a BstEII site (5’-AAG GGT CAC CAA ACT AGT CCT CTT TTT TGT CAC-3’). The incorporated restriction sites in these primer sequences (and all subsequent primers described in this report) are underlined. The PDI8 cDNA fragment was ligated between the NcoI and BstEII restriction sites of pCAMBIA1302. For antisense expression of PDI8, a PDI8 cDNA fragment was amplified from cDNA (as above) using a forward primer containing a BstEII restriction site (5’-AAA TGG TGA CCT CAT GAG ATC GTT AAA GTT ACT CCT TTG TTG-3’), and a reverse primer with an SpeI site (5’-AAG TGG GTC AAC ACT AGT CCT CTT TTT TGT CAC T-3’). The cDNA fragment was ligated in the antisense orientation between the SpeI and BstEII cloning sites of pCAMBIA1302 to generate the construct pC1302[35S _pro :antisensePDI8].

Promoter expression studies were performed using stably transformed transgenic lines harboring the construct PDI8 _pro :GUS, which contains a 2.3-kb PDI8 promoter fragment transcriptionally fused to the β-glucuronidase (GUS) reporter gene, gusA. The PDI8 promoter fragment was amplified from Arabidopsis (Col-0) genomic DNA, using a forward primer with a SacI site (5’- TTT GAG CTC GTA GAA GTT TGC TTG AAT ATT CA-3’) and a reverse primer with an NcoI site (5’-AAC CCA TGG CGA TCT GAT TTT CAG ACC AAA C-3’). The gusA gene was amplified from pCAMBIA1304 using a forward primer with an NcoI site (5’-TGA CCA TGG TAG ATC TGA CTA GTT TAC GTC-3’) and a reverse primer with a BstEII site (5′-CTC CGG TCA CCT ATT GTT TGC CTC CCT GCT GCG-3′). The PDI8 _pro :GUS fusion was assembled in pCAMBIA1302 by cloning the gusA PCR fragment between the NcoI and BstEII sites of the vector to create the intermediate construct pC1302[35S _pro :GUS]. The PDI8 promoter fragment was then cloned between the SacI and NcoI sites of pC1302[35s _pro :GUS] to produce the final construct, pC1302[PDI8 _pro :GUS].

The pC1302[35S _pro :PDI8] and pC1302[PDI8 _pro :GUS] constructs were introduced into wild-type Arabidopsis (Col-0) plants by Agrobacterium-mediated transformation, using the floral dip method [7]. Initial transformants were obtained by selecting for hygromycin resistance in the T₁ generation, and the presence of an intact transgene determined by PCR. Homozygous transgenic lines were subsequently identified by screening for the occurrence of 100 % hygromycin-resistance in the T₃ generation.

GUS expression analysis

PDI8 _pro :GUS seedlings were grown vertically on 1/2× LS agar plates [0.8 % (w/v) Gellan Gum (Sigma-Aldrich, St. Louis, MO), 1/2x Linsmaier & Skoog media (Caisson Laboratories, Smithfield, UT) and 1.5 % (w/v) sucrose] for 7 or 14 days at 22 °C under a 16 h-light/8 h-dark cycle. Shoot inflorescences were obtained from 6-week-old PDI8 _pro :GUS plants grown on Farfard Super-Fine Germinating Mix (Sun Gro Horticulture, Agawam, MA) under a 16 h-light/8 h-dark cycle at 25 °C. GUS staining was performed as described [17]. Briefly, the tissue samples were fixed in 90 % ice-cold acetone for 20 min at 25 °C, then washed with staining buffer (50 mM sodium phosphate buffer (pH 7.0), 0.2 % Triton X-100, 2 mM potassium ferrocyanide, and 2 mM potassium ferricyanide) three times on ice, then submerged in staining buffer containing 1 mM 5-bromo-4-chloro-3-indoxyl-β-D-glucuronide cyclohexylammonium salt (X-gluc). The tissues were vacuum infiltrated briefly, then incubated O/N at 37 °C. After staining, the samples were incubated in 70 % ethanol to extract soluble pigments, repeating with fresh 70 % ethanol as necessary. Images of GUS staining in roots and stomata were acquired on an Olympus BX-51 upright microscope, with the samples mounted on glass slides in 50 % glycerol. All other images were taken on an Olympus SZX-12 stereomicroscope, with samples submerged in 70 % ethanol in a petri dish.

Transient expression of spGFP-PDI8 in protoplasts

The creation of the ER marker construct pBL(35S _pro :ER-mCherry), and the unfused green fluorescent protein (GFP) control construct pBL(35S _pro :GFP(S65T)), was described previously [6]. The construct pBL(35S _pro :spGFP-PDI8) was generated by cloning the following arrangement of DNA sequences between the KpnI and BstEII sites of pBL(35S _pro :GFP(S65T)): a CaMV 35S promoter fragment (KpnI/XhoI), a PDI8 signal peptide coding sequence-GFP(S65T) fragment (XhoI/XmaI), and a PDI8 mature protein cDNA fragment (XmaI/BstEII). The CaMV 35S promoter fragment was amplified from pCAMBIA1302 using a forward primer with a KpnI site (5’-TTC AGG GTA CCT TCA TGG AGT CAA AGA TTC A-3’), and a reverse primer with an XhoI site (5’-ATC TAC TCG AGT CAA GAG TCC CCC GTG-3’). The GFP(S65T) fragment, modified to include the signal peptide sequence at the N-terminus of GFP, was amplified from plasmid HBT95::sGFP(S65T)-NOS using a forward primer with an XhoI site (5’-TTT CTC GAG ATG CGT TCG TTA AAG TTA CTC CTT TGT TGG ATC TCG TTT CTT ACG TTA TCA ATC TCA ATC TCT GCA TCG TCA ATG GTG AGC AAG GGC GAG GAG CTG-3’), and a reverse primer with an XmaI site (5’-AAA CCC GGG CTT GTA CAG CTC GTC CAT GC-3’). The PDI8 mature protein cDNA fragment was amplified from a full-length PDI8 cDNA clone using a forward primer with an XmaI site (5’-ATA CCC GGG TCG TCA GAT GAT CAA TTC ACC CTC-3’) and a reverse primer with a BstEII site (5’-AAG GGT CAC CAA ACT AGT CCT CTT TTT TGT CAC TAG-3’). The construct pBL(35S _pro :PDI8-GFP-KKED) was generated by replacing the spGFP-PDI8 coding sequence of pBL(35S _pro :spGFP-PDI8), between restriction sites XhoI and BstEII, with a full-length PDI8 cDNA fragment (XhoI/XmaI) and a GFP(S65T)-KKED fragment (XmaI/BstEII). The PDI8 cDNA fragment was amplified from a PDI8 cDNA clone using a forward primer with an XhoI site (5’-CAG CTC GAG ATG CGT TCG TTA AAG TTA C-3’) and a reverse primer with an XmaI site (5’-ACA CCC GGG GTC CTC TTT TTT GTC ACT AGG CT-3’). The GFP(S65T) fragment, modified to include the KKED putative retention signal of PDI8, was amplified from plasmid HBT95::sGFP(S65T)-NOS using a forward primer with an XmaI site (5’-TAG TCC CGG GAT GGT GAG CAA GGG CGA GGA-3’), and a reverse primer with a BstEII site (5’-AGG ATG GTC ACC TAA TCC TCT TTT TTG CCG TGA GTG ATC-3’).

The procedure for isolating and transfecting protoplasts was adapted from Yoo et al. [36] and Wu et al. [35]. The abaxial epidermis of rosette leaves from four-week-old Arabidopsis plants was removed using the tape-sandwich method [35]. Mesophyll cells were released by incubating the peeled leaves in 10 mL of enzyme solution (1.5 % cellulase R10, 0.4 % macerozyme R10, 0.4 M mannitol, 20 mM KCl, 20 mM MES, pH 5.7) for 3 h, then mixed gently with 10 mL of W5 solution (154 nM NaCl, 125 mM CaCl2, 5 mM KCl, 2 mM MES, pH 5.7). The protoplasts were gently centrifuged at 100 g for 2 min, resuspended in fresh W5 solution to a density of 2 × 10⁵/mL, and incubated on ice for at least 30 min. The W5 solution was then removed, and the protoplasts resuspended in MMg solution (0.4 M mannitol, 15 mM MgCl₂, 4 mM MES, pH 5.7) to a density of 2 × 10⁵/mL. The protoplasts were transfected by gently mixing 200 μL of protoplasts in MMg solution with 20 μL of plasmid DNA solution (containing ~20 μg of each construct in H₂O), and 220 μL of PEG solution (40 % PEG, 0.2 M mannitol, 100 mM CaCl₂). After incubating at 25 °C for 5-10 min, transfection was stopped by adding 0.8 mL W5 solution. The protoplasts were centrifuged at 100 g for 2 min, and then resuspended in 1 mL WI solution (0.5 M mannitol, 20 mM KCl, 4 mM MES, pH 5.7). The transfected protoplasts were incubated in the dark at 22 °C for 18 h to allow for transgene expression. Fluorescence was visualized using an Olympus FV-1000 laser scanning confocal microscope at the Biological Electron Microscope Facility (University of Hawaii at Manoa, Honolulu, HI). The excitation/emission filters utilized for fluorescence detection were 488/505–525 nm for GFP(S65T) and 543/585–615 nm for mCherry.

Anti-PDI8 antibody production

Affinity-purified polyclonal rabbit antibodies recognizing PDI8 were generated commercially through YenZym Antibodies, LLC (San Francisco, CA), using a truncated form of recombinant PDI8 as the antigen for both rabbit immunization and affinity purification of the antiserum. For production of the truncated PDI8 protein, a cDNA fragment encoding the central b-b’ region of PDI8 (PDI8^bb’, corresponding to residues 138-377 of the PDI8 preprotein sequence) was amplified by RT-PCR using a forward primer with an NdeI site (5’-GCC TAC GCA TAT GGT TGC TCC AGA TGT GCG G-3’) and reverse primer with a BamHI site (5’-CGT GGA TCC CTA TGA GTT GAT AAA TCC CAT GAA-3’). The PDI8 ^bb’ cDNA fragment was ligated between the NdeI and BamHI sites of the bacterial expression vector pET-15b (EMD Millipore, Billerica, MA), placing the PDI8 ^bb’ sequence in-frame with the 6xHis-tag of pET-15b. Expression of PDI8^bb’ was induced in Escherichia coli strain BL21(DE3) for 5 h at 28 °C by the addition of 0.2 mM IPTG. After induction, the E. coli cells were harvested by centrifugation and lysed using BugBuster Protein Extraction Reagent (EMD Millipore). The His-tagged PDI8^bb’ protein was purified from the bacterial lysate by nickel affinity chromatography.

Transmission electron microscopy and immunolabeling

For immunogold labeling analysis, developing roots and apical buds were preserved by high-pressure freezing/freeze-substitution techniques as described in [6]. For immunolabeling, 80 nm thick sections from Lowicryl HM20 resin embedded specimens were placed on formvar-coated gold or nickel slot grids and blocked for 30 min with 2 % (w/v) non-fat dried milk solution in 0.01 M phosphate-buffered saline pH 7.2 containing 0.1 % Tween-20 (PBST). The sections were washed and then incubated with a 10-fold dilution of the primary antibody, anti-PD8, for 2 h at RT. Sections were washed and transferred to a 25-fold dilution of secondary antibody goat anti-rabbit IgG-conjugated to 10 or 15 nm gold particles (Ted Pella, Inc) for 2 h at RT. Sections were washed and then stained with uranyl acetate solution for 2 min and lead citrate for 4 min. All observations were performed using a Hitachi H-7000 transmission electron microscope operated at 80 KV (Hitachi USA, OH).

Preparation of microsomal membranes and protease protection analysis

For extraction of microsomal membranes, 35S _pro :PDI8 Arabidopsis seedlings were grown vertically on 1/2× LS agar plates under a 16 h-light/8 h-dark cycle at 22 °C. 7-day-old seedlings were homogenized with a chilled mortar and pestle in ice-cold extraction buffer [40 mM HEPES pH 7.5, 0.4 % polyvinyl polypyrrolidone (PVP), 1 mM MgCl₂, 10 mM KCL, and 0.4 M sucrose], at a ratio of 1.5 μL extraction buffer per 1 mg of plant tissue. To remove insoluble debris, the homogenate was centrifuged twice at 1000 g and 4 °C for 3 min, collecting the supernatant after each spin. The total protein homogenate was separated into microsomal and soluble protein fractions by centrifuging as 150 μL aliquots at 21,000 g and 4 °C for 1.5 h [1]. The microsomal pellets were washed once with 150 μL of fresh extraction buffer, recovering the microsomes by spinning at 21,000 g and 4 °C for 45 min and removing the supernatant. Finally, the microsomal pellets were resuspended in a volume of fresh extraction buffer equivalent to the original sample volume (i.e. 150 μL).

For immunoblot detection of PDI8 and BiP, protein samples were separated by SDS-PAGE (10 % polyacrylamide gels) and transferred onto nitrocellulose membranes. An equivalent amount (by volume) of the 35S _pro :PDI8 total and fractioned protein samples were loaded, equaling ~20 μg protein in the unseparated homogenate, ~14 μg protein in the soluble fraction, and ~7 μg protein in the microsomal fraction. Immunoblot analysis of PDI8 was performed using the anti-PDI8 antiserum at 1:100 dilution, and an anti-rabbit horseradish peroxidase (HRP)-conjugated secondary antibody at 1:2000 dilution supplied in the Amersham ECL Western Blotting Detection Kit (GE Healthcare Bio-Sciences, Pittsburgh, PA). Detection of BiP was performed using the goat anti-BiP primary antibody aC-19 (Santa Cruz Biotechnology, Inc., Dallas, Tx) at 1:1000 dilution, and a donkey anti-goat HRP-conjugated secondary antibody (Santa Cruz Biotechnology, Inc.) at 1:3000 dilution.

To determine the membrane topology of PDI8, 35S _pro :PDI8 resuspended microsomes were incubated at 37 °C for 30 min in extraction buffer alone (negative control), or with 50 μg/mL proteinase K and/or 0.1 % Triton X-100. Each reaction contained ~0.36 μg/μL microsomal protein in a total volume of 60 μL. Proteinase K digestion was stopped by adding 5 mM PMSF to all samples. SDS-PAGE and immunoblot detection of PDI8 was performed as described above, with each lane loaded with 20 μL of sample (~7.2 μg of microsomal protein).

Alkaline phosphatase activity assay

A construct for the heterologous expression of PDI8 in E. coli was generated by cloning the coding sequence for the lumenal portion of PDI8, containing the catalytic a domain and redox-inactive b and b’ domains (PDI8^abb’, Fig. 1) between the XmaI and SalI restriction sites of the bacterial expression vector pFLAG-CTS (Sigma-Aldrich, St. Louis, MO). The PDI8 ^abb’ gene fragment was amplified from a full-length PDI8 cDNA using a forward primer with an XmaI site (5’-TGT CCC GGG AGA TGA TCA ATT CAC CCT CGA C-3’) and a reverse primer with a SalI site (5’-AAT GTC GAC CAT TGA GTT GAT AAA TCC CAT G-3’).

The E. coli strains RI89 (dsbA ⁺ ) and RI90 (dsbA::kan1; RI89 genetic background) were obtained from the E. coli Genetic Stock Center (Yale University, New Haven, CT). The pFLAG-PDI8^abb’ construct and pFLAG-CTS empty vector were transformed into strain RI90. To measure alkaline phosphatase (PhoA) activity, the cells were grown at 37 °C in M9 minimal media to an OD₆₀₀ of 0.4–0.6, harvested by centrifugation, washed once with 50 mM Tris-HCl (pH 8.0), and lysed with 0.2 % Triton X-100. PhoA activity was determined using the QuantiChrom Alkaline Phosphatase Assay Kit (BioAssay Systems, Hayward, CA). Briefly, 150 μL of working solution (5 mM magnesium acetate, and 10 mM p-nitrophenyl phosphate in supplied assay buffer, pH 10.5) was added to 50 μL of lysed cells. After quickly mixing, the initial OD₄₀₅ (t = 0) was measured for each sample, and then re-measured after 4 min (t = 4). PhoA activity (IU/L) was calculated from the OD₄₀₅ values as described in the kit. The activities reported are averages (±standard deviation) derived from three independent trials.