The thiol-disulfide oxidoreductases of the PDI (protein disulfide isomerase) family assist in disulfide-bond formation in the ER (endoplasmic reticulum). In the present study, we have shown that the previously uncharacterized PDI family member TMX4 (thioredoxin-like transmembrane 4) is an N-glycosylated type I membrane protein that localizes to the ER. We also demonstrate that TMX4 contains a single ER-luminal thioredoxin-like domain, which, in contrast with similar domains in other PDIs, is mainly oxidized in living cells. The TMX4 transcript displays a wide tissue distribution, and is strongly expressed in melanoma cells. Unlike many type I membrane proteins, TMX4 lacks a typical C-terminal di-lysine retrieval signal. Instead, the cytoplasmic tail has a conserved di-arginine motif of the RXR type. We show that mutation of the RQR sequence in TMX4 to KQK interferes with ER localization of the protein. Moreover, whereas the cytoplasmic region of TMX4 confers ER localization to a reporter protein, the KQK mutant of the same protein redistributes to the cell surface. Overall, features not commonly found in other PDIs characterize TMX4 and suggest unique functional properties of the protein.
- endoplasmic reticulum
- protein disulfide isomerase (PDI)
- thioredoxin-like transmembrane protein 4 (TMX4)
The relatively oxidizing conditions inside the lumen of the ER (endoplasmic reticulum) provide an appropriate and supportive environment for oxidative protein folding . In the ER, the correct formation of disulfide bonds is most often a prerequisite for the production of native proteins. During folding, pairs of cysteine thiols are oxidized to form disulfide bonds. Since this process frequently leads to non-native combinations of cysteine residues, disulfide-bond reduction and isomerization (reshuffling) are critical reactions and often necessary steps en route to the correct structure .
Although disulfide bonds form in vitro when oxygen is present, it is an inefficient process in the absence of a catalyst. In the ER, oxidoreductases of the PDI (protein disulfide isomerase) family catalyse oxidation, isomerization and reduction by thiol–disulfide exchange. The PDI-like proteins contain at least one domain similar to thioredoxin, a cytosolic oxidoreductase of approx. 12 kDa. Typically, the active-site residues constitute a CXXC sequence motif (where X denotes any amino acid residue). The nature of the two variable residues is known to influence the redox (reduction–oxidation) potential, i.e. the propensity of the active-site cysteine residues to be reduced or oxidized, and thereby the redox activity of a given enzyme .
The human PDI family has nearly 20 members that differ in size, number and arrangement of thioredoxin-like domains, and, in part, in tissue distribution . Their biological function remains unclear for many. Apart from possible differences in their redox activity, these enzymes probably act on different substrates. For instance, ERp57 probably binds primarily to certain cysteine-containing glycoproteins . The function of the PDI family members is not restricted to oxidative folding. For example, ERp44 regulates Ca2+ release by the inositol 1,4,5-trisphosphate receptor 1, a Ca2+ channel of the ER membrane, through a direct interaction that probably depends on the redox state of the receptor . It is also clear that some of the PDIs perform redox-unrelated functions, since they lack both cysteine residues of the active site. An example is the murine ERp29 that is important for transport across the ER membrane to the cytosol of murine polyoma virus during infection . Here, in a step potentially required for membrane penetration, ERp29 promotes partial unfolding of the major structural viral protein VP1.
Although most of the PDI family members are soluble ER-luminal proteins, some span the ER membrane. This latter subset of the family is designated as the TMX (thioredoxin-like transmembrane) proteins, and includes TMX, TMX2 and TMX3. All contain a single thioredoxin-like domain, but are otherwise not closely related. TMX has been shown to suppress apoptosis induced by brefeldin A when overexpressed in HEK-293 (human embryonic kidney) cells . Recently, the interaction between TMX and the misfolded MHC class I heavy chain was demonstrated to protect the latter protein from proteasomal degradation, indicating a function of TMX in ER quality control . Although the knowledge about TMX2 is restricted to its cDNA sequence and tissue distribution , TMX3 has been well studied in vitro in terms of its structure–function relationships [11,12].
In the present study, we have characterized TMX4, a novel transmembrane member of the PDI family. We identified TMX4 as a paralogue of TMX, and established that it is an N-glycosylated type I transmembrane protein. We have also shown that in place of a classical ER-retrieval motif, an RQR sequence in the cytosolic region of TMX4 confers ER localization to the protein. The results suggest that regulated sorting based on sequence features in the C-terminal tail of the protein could modulate the function of TMX4.
Amino acid sequences of TMX and TMX4 from various organisms were obtained from the Ensembl Genome Browser (http://www.ensembl.org/index.html). Full-length sequences were aligned on the http://ch.embnet.org/ server, using ClustalW-XXL. The alignments were carried out using the BLOSUM (block substitution matrix) series of matrices with an open gap penalty of 10 and an extending gap penalty of 0.05. Phylogenetic trees were constructed using PhyML  and Phylip (phylogeny inference package) . Human ERp57 was selected as the outgroup.
The human melanoma cell lines A375M, A375P, LT5-1 and DX3 were provided by Dr Elena Sviderskaya (Cancer Research UK London Institute, London, U.K.). All cells were maintained at 37 °C in MEM (alpha-minimum essential medium; Invitrogen) supplemented with 10% FCS (fetal calf serum) under 5% CO2.
Tissue Northern blot
A human multiple tissue Northern blot (Ambion) was hybridized with radioactive probes generated from a NotI–XhoI fragment obtained by restriction digestion of pcDNA3.1/TMX4-myc. The gel-purified fragment was labelled with [α-32P]dCTP using the Strip-EZ DNA kit (Ambion), followed by purification using a Micro Bio-Spin 30 column (Bio-Rad Laboratories) to remove unincorporated nucleotides. Hybridization was carried out at 65 °C in ULTRAhyb buffer (Ambion) following the manufacturer's protocol. The probed blot was visualized by autoradiography. The blot was then stripped according to the Strip-EZ DNA kit protocol and hybridized with probes derived from the β-actin mouse DNA template that was supplied with the kit.
RT–PCR (reverse-transcription–PCR) analysis
Total RNA was isolated using the GenElute Total RNA kit (Sigma). The concentration of total RNA was adjusted and the mRNA was reverse-transcribed (Enhanced Avian Reverse Transcriptase kit; Sigma). PCR was performed with the TMX4-cyt forward and reverse primers that span two TMX4 exons, yielding a product of 412 bp. As a control, PCR was performed with the primers Actin for and Actin rev (see Supplementary Table S1 at http://www.BiochemJ.org/bj/425/bj4250195add.htm). The products were analysed on 1% agarose gels.
Gel electrophoresis and Western blotting
Samples were separated on Hoefer minigels (10 cm×10.5 cm; 10% gel) (GE Healthcare), with the exception of the CD4–TMX4 chimaeras, for which 7.5% polyacrylamide gels were used. Western blotting was performed as described previously  using the ECL Advance system for detection (GE Healthcare). The band intensities obtained by autoradiography or Western blotting were quantified using ImageJ .
Transfection and antibodies
Vero cells, as used in Figure 4(B), were transfected using an AMAXA electroporator following the manufacturer's recommendations for each cell line. For all other experiments, plasmids were transfected into HeLa cells with Lipofectamine™ 2000 (Invitrogen).
A rabbit polyclonal antiserum (R504) recognizing the cytosolic domain of TMX4 was raised against the peptide sequence DGVTREEVEPEEAEE, comprising residues 305–319 (Open Biosystems, Huntsville, AL, U.S.A.). The quality of the antiserum was verified on A375P cellular lysates treated with siRNAs (small interfering RNAs) directed against TMX4 (see Supplementary Experimental section at http://www.BiochemJ.org/bj/425/bj4250195add.htm). The anti-TMX3 antibody has been described previously . The mouse monoclonal antibodies HA.11 (clone 16B12) against the HA (haemagglutinin) epitope and clone 9E10 against the c-Myc epitope were from Covance Research Products. The polyclonal rabbit antiserum against CNX (calnexin) was provided by Ari Helenius (ETH Zurich, Zurich, Switzerland) and the polyclonal anti-E-cadherin antibody by M. Pasdar (Department of Cell Biology, University of Alberta, Edmonton, AB, Canada). Additional antibodies against the following were also purchased: CNX (Assay Designs), 14-3-3ζ (Santa Cruz), mitochondrial complex II (Mitosciences), TMX (Sigma), Alexa Fluor® 594 goat anti-mouse and Alexa Fluor® 488 goat anti-rabbit (Molecular Probes) and horseradish-peroxidase-coupled goat anti-mouse and rabbit IgG (Pierce).
Mammalian expression plasmids
Detailed information on the various constructs generated in the present study is provided in the Supplementary Experimental section.
The experimental conditions for fluorescence microscopy are given in the Supplementary Experimental section.
Endo H (endoglycosidase H) digestion
Endo H (New England Biolabs) digestions were performed on A375P cell extracts after lysis in loading buffer (180 mM Tris/HCl, pH 6.8, 0.5% SDS and 1% 2-mercaptoethanol; see Figure 4), or after lysis in lysis buffer (80 mM Tris/HCl, pH 6.8, and 1% SDS; see Figure 5). Samples were denatured at 96 °C for 15 min and then supplemented with one-tenth volume of 50 mM sodium citrate, pH 5.5, and treated with Endo H for 1 h at 37 °C.
ER and Golgi membranes were fractionated on a continuous OptiPrep® gradient (Axis-Shield, Dundee, Scotland) using 25, 20, 15, 10 and 5% OptiPrep®. A375P cells or HeLa cells transfected with HA-tagged TMX4 were lysed in homogenization buffer (0.25 M sucrose, 10 mM Hepes/NaOH, pH 7.4, and 1 mM EDTA) and passed 15 times through a ball-bearing homogenizer (Isobiotec, Heidelberg, Germany) with 18 μm clearance. Cell debris and nuclei were pelleted by centrifugation at 1000 g for 10 min. The post-nuclear supernatant was overlayed on to the continuous gradient and centrifuged at 32700 rev./min (SW55 Ti rotor, Beckman) for 3 h at 4 °C. Six equal fractions were collected from the top of the gradient and precipitated with acetone. Fractions were probed by Western blotting for endogenous TMX4 and HA–TMX4, CNX [marker for rER (rough ER)/MAM (mitochondria-associated membrane)], E-cadherin (plasma membrane), 14-3-3ζ (Golgi) and complex II (mitochondria).
AMS (4-acetamido-4-maleimidylstilbene-2,2-disulfonic acid) modification
A375P cells were soaked in buffer containing NEM (N-ethylmaleimide) before lysis to block free cysteine residues, as described previously . Subsequently, the cell extract was reduced with 10 mM TCEP [tris-(2-carboxyethyl)phosphine] for 15 min and treated with 15 mM AMS for 1 h at room temperature (20 °C). After separation by SDS/PAGE, TMX4, TMX and TMX3 were visualized by Western blotting.
Membrane association and proteinase K protection assay
To study the membrane association of TMX4, HA–TMX4-transfected HeLa cells at 60% confluency were harvested, fractionated and extracted in sodium carbonate, as described previously . To determine the topology of TMX4, HeLa cells were transfected with constructs to express HA–TMX4, TMX4–Myc or ERp57–HA. After transfection (12 h), the cells were pulsed overnight at 37 °C with 125 μCi of [35S]methionine/cysteine (PerkinElmer) in 0.5 ml of starvation medium [DMEM (Dulbecco's modified Eagle's medium) without methionine and cystine; Sigma] with 1% FCS per 35-mm-diameter dish. After the pulsing, the dishes were transferred on to ice where all further steps were performed. The cells were washed twice with PBS, scraped off with a rubber policeman in homogenization buffer [20 mM Hepes/NaOH, pH 7.5, 0.25 M sucrose and 1 mM DTT (dithiothreitol)] and passed 10 times through a 27-gauge needle. After removal of cell debris by centrifugation at 300 g for 3 min at 4 °C, the supernatant was centrifuged at 25000 g for 1 h at 4 °C. The pellet was resuspended in assay buffer (25 mM Tris/HCl, pH 8, 500 mM NaCl and 1 mM DTT) and treated with 60 μg/ml proteinase K (Roche Molecular Biochemicals) on ice for 1 h in the presence or absence of 1% Triton X-100. After inhibition of proteinase K with 1 mM PMSF, detergent-free samples were solubilized with 1% Triton X-100 before the epitope-tagged proteins were immunoprecipitated, separated by SDS/PAGE and visualized by autoradiography.
Lysates were centrifuged at 25000 g for 30 min at 4 °C, and the supernatants were rotated for 2 h at 4 °C with 50 μl of Protein A–Sepharose beads (GE Healthcare), which had been pre-adsorbed with either 2 μl of HA.11 or 2 μl of 9E10 antibodies. After washing with lysis buffer (100 mM sodium phosphate, pH 8, and 1% Triton X-100) and 100 mM sodium phosphate (pH 8.0), proteins were eluted from the beads by boiling in sample buffer.
We identified the human gene encoding TMX4 in a GenBank® BLAST search using a consensus sequence for a thioredoxin-like domain at the NCBI website (http://www.ncbi.nlm.nih.gov). The open reading frame encodes a protein of 349 residues (39 kDa, pI 4.3), including a signal sequence for entry into the secretory pathway with a predicted cleavage site between residues 23 and 24 (Figure 1). Sequence alignment with thioredoxin-like domains of known three-dimensional structure from PDI, ERp57 and ERp72 initially suggested that the TMX4 thioredoxin-like domain would cover residues 39 to 136. However, comparison with the NMR structure of the closely related TMX protein  revealed that the TMX4 thioredoxin-like domain, similar to that of TMX, probably possesses an additional C-terminal α-helix and a short β-strand compared with a canonical thioredoxin fold (α5 and β6 in Figure 1B). The thioredoxin-like domain is followed by a stretch of 41 residues preceeding a predicted transmembrane helix spanning residues 188–210 (see also Figure 6A). The C-terminal region (residues 211–349) is highly negatively charged with 54 acidic residues and contains no classical ER-localization motif of the K(X)KXX type that targets many type I membrane proteins to the ER .
Phylogenetic analysis of the TMX4 and TMX sequences
Sequence alignments with other members of the human PDI family revealed that TMX4 is most closely related to TMX. Specifically, the thioredoxin-like domains show 53% sequence identity. In contrast, the C-terminal region after the predicted transmembrane helix is considerably less well conserved, with the most obvious difference being the length (78 residues in TMX compared with 139 residues in TMX4). Transcripts for both proteins are present in a variety of mammals, amphibia, fish and birds, but not in plants and yeasts, as shown by database analysis (see also below).
To gain more knowledge regarding the evolutionary relationship between the two proteins, we subjected a variety of TMX and TMX4 sequences to phylogenetic analysis by the maximum likelihood method. The evolutionary tree was rooted using ERp57, a close homologue of PDI (Figure 2). The proteins clustered in three main groups. The cluster shown at the bottom first diverged from a common ancestor and comprises organisms with only one TMX/TMX4 protein. One of these proteins is DPY-11 from Caenorhabditis elegans . We reason that a gene duplication then gave rise to the paralogues TMX and TMX4 which cluster in separate groups. Since the branch lengths represent the number of sequence changes that occurred prior to the next level of separation, the longer branches in the TMX4 cluster indicate that this protein diverged more from the common ancestor than TMX. The results of our analysis were supported by the TreeFam database , where the TMX/TMX4 phylogenetic tree has the same three groups as presented in Figure 2.
Tissue distribution of TMX4
The EST (expressed sequence tag)-based expression profile of the NCBI UniGene cluster Hs. 169358 suggests a broad tissue distribution of TMX4. We first confirmed this result by examining the transcript levels in a variety of human tissues by Northern blotting (Figure 3A). To this end, 32P-labelled probes for TMX4 were used to hybridize a commercial tissue Northern blot. By autoradiography, we detected a signal at the predicted size of 6.1 kb for the TMX4 transcript in all tissues tested, although with different intensities. We further corroborated the broad tissue distribution using RT–PCR analysis to identify TMX4 transcripts in a series of cultured cell lines (Figure 3B).
Next, we probed different human cell lines for the presence of TMX4 by Western blotting using our own polyclonal antibody (R504) raised against the cytosolic domain of TMX4 (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/425/bj4250195add.htm and Figure 4A for specificity of R504). We detected relatively low levels in common cell lines such as HeLa and HEK-293 (results not shown). Instead, a number of melanoma cells contained higher amounts of the protein (Figure 3C), which is in good agreement with the RT–PCR analysis showing the transcript to be abundant in A375 and Meljuso cells, and the finding that TMX4 expression is almost 10-fold up-regulated upon the transformation of melanocytes to melanoma cells .
TMX4 is an N-glycosylated ER protein
The TMX4 sequence contains one consensus site for N-glycosylation on Asn46 (Figure 1B). To investigate the glycosylation status of endogenous TMX4, we performed Western blot analysis of lysates of A375P melanoma cells treated with Endo H, a glycosidase that removes high-mannose-type sugars. The results showed that TMX4 contained a glycan sensitive to Endo H treatment (Figure 4A), as did the N-terminally HA-tagged protein HA–TMX4 when overexpressed in HeLa cells (see Supplementary Figure S2 at http://www.BiochemJ.org/bj/425/bj4250195add.htm). To experimentally verify the predicted ER localization of TMX4, we expressed HA–TMX4 in Vero cells. By immunofluorescence microscopy, HA–TMX4 showed a typical reticular ER staining pattern and was found to co-localize with the ER marker CNX (Figure 4B). Moreover, we confirmed that endogenous TMX4 co-fractionated with CNX on an OptiPrep® gradient. Here, TMX4 showed a distribution that differed slightly from that of CNX (Figure 4C), most likely because CNX not only localizes to the rough ER but also to MAMs .
The active site of TMX4 is mainly oxidized in cells
Knowledge about the cellular redox state of a given PDI is a prerequisite for further detailed analysis of its redox properties. To determine the oxidation state of endogenous TMX4, A375P cells were subjected to an alkylation procedure where free cysteine residues are modified with NEM and cysteine residues present in disulfides with AMS . As a result of the size difference between the two reagents, the AMS-modified form has decreased gel mobility. On a Western blot for TMX4, we detected two bands corresponding to the oxidized and reduced forms of the active site (Figure 5, top panel). For comparison, we also determined the redox states of TMX and TMX3 (Figure 5, middle and bottom panels). The results showed that, whereas TMX4 is primarily oxidized, as also seen for HA–TMX4 expressed in HeLa cells (see Supplementary Figure S3 at http://www.BiochemJ.org/bj/425/bj4250195add.htm), TMX is primarily reduced in accordance with recent findings in other cell lines . As we have seen previously for TMX3 in a variety of cultured cells, this protein was also primarily reduced in the A375P cells [11,16,23].
Membrane topology of TMX4
We next established that TMX4 is a membrane protein as predicted by the TMHMM algorithm  (Figure 6A). Isolated crude membranes from HeLa cells expressing HA–TMX4 were extracted with sodium carbonate at high pH, and soluble and membrane-associated proteins were separated by high-speed centrifugation and probed by Western blotting. HA–TMX4 was detected exclusively in the membrane fraction, similar to the type I transmembrane protein CNX (Figure 6B). The soluble marker protein ERp57 partly fractionated in the insoluble fraction, as observed previously for another soluble ER protein, calreticulin, possibly due to the association with membrane proteins [11,25].
In addition to the transmembrane domain (residues 188–210), two other hydrophobic stretches were detected in TMX4 (Figure 6A). The first, at the very N-terminus, corresponded to the signal sequence. The second comprised residues 147–187, i.e. the region that links the thioredoxin-like domain to the transmembrane region. As we could not rule out that this linker region was also buried in the membrane, thereby creating a topology with both termini of TMX4 in the ER and a short cytosolic loop, we determined the membrane topology of TMX4 using a proteinase K protection assay (Figure 6C).
Crude membrane extracts from [35S]methionine/cysteine-labelled HeLa cells expressing HA–TMX4 were left untreated or treated with proteinase K, followed by immunoprecipitation of the expressed protein. The untreated sample showed a signal at the expected size for HA–TMX4 of approx. 45 kDa. Upon proteinase K treatment, the band shifted to 25 kDa, a size closely matching that of the 21 kDa expected for HA–TMX4 devoid of the C-terminal cytosolic region. When the ER membrane was solubilized by addition of Triton X-100, proteinase K gained access to the lumen and the signal disappeared. Thus the HA-tagged N-terminus was localized in the ER as expected from the presence of the N-glycan on Asn46. To examine the orientation of the C-terminus, we performed the proteinase K assay on HeLa cells expressing C-terminally Myc-tagged TMX4. Here, the TMX4–Myc signal disappeared almost completely, even in the absence of detergent, showing that the Myc-tagged C-terminus was not protected inside the ER. The soluble luminal ERp57 protein was used as a control for membrane integrity. Taken together, the results demonstrated that TMX4 is an N-glycosylated type I transmembrane protein of the ER.
A non-classical sequence motif contributes to ER targeting of TMX4
Although TMX4 resides in the ER, it lacks a classical K(X)KXX-type ER-localization signal. In multiple sequence alignments of TMX4, we noticed an LRQR sequence conserved among species, with only very few exceptions, that was located close to the C-terminus (Figure 1B). In a number of type II and multi-spanning membrane proteins a Φ/Ψ/R-R-X-R sequence, in which Φ/Ψ and X represent an aromatic/bulky hydrophobic and any residue respectively, acts as an ER-localization signal. So far, only one type I membrane protein, the ER lectin VIPL (VIP36-like), has been shown to have an arginine-based ER-localization motif .
To test the possible involvement of the RQR sequence in the ER localization of TMX4, we produced two HA-tagged mutants (Figure 7A): (i) a truncation variant lacking the C-terminal 121 residues (Δtail) and (ii) a mutant with the two arginine residues of the RQR sequence substituted by lysine residues (KQK). We also wanted to assess a possible influence of the luminal region of TMX4 in localizing the protein to the ER. As a construct comprising only the luminal domain of TMX4 was unstable [half-life of approx. 30 min compared with >10 h for the wild-type protein (data not shown)], a fusion protein containing the ER-luminal part of TMX4 combined with the transmembrane and cytosolic regions of CD4 was made (CD4-tail; Figure 7A). Here, CD4 was chosen as a fusion partner, since it localizes to the plasma membrane and hence does not contain ER retention/retrieval information itself .
After transfecting HeLa cells, we analysed the cellular localization of the wild-type protein and the three mutants by confocal immunofluorescence microscopy. We found that, unlike the wild-type protein, HA–TMX4 lacking the cytosolic tail partially escaped the ER and reached the plasma membrane. The KQK and CD4-tail mutants showed the same staining pattern as the Δtail mutant, also being detected on the cell surface. In conclusion, the RQR sequence was necessary for complete retention of TMX4-fusion proteins in the ER, whereas neither the luminal region of TMX4 nor the tail lacking RQR had this ability.
Most probably due to the continuous overexpression of protein, the steady-state localization of all TMX4 variants also included the ER. To remove the newly synthesized TMX4 variants of the ER, we treated HeLa cells with cycloheximide, a drug that inhibits protein synthesis (results not shown). Unfortunately, this strategy proved unfeasible, because export occurred so slowly that the prolonged cycloheximide treatment induced apoptosis . However, protein aggregation was probably not the cause for the partial ER localization, since the immunofluorescence images showed no visible aggregates. Moreover, we analysed the oligomeric state of HA–TMX4 by sucrose gradient velocity centrifugation and found no evidence for Triton X-100-insoluble high-molecular-mass complexes (see Supplementary Figure S4 at http://www.BiochemJ.org/bj/425/bj4250195add.htm). We further ruled out that the plasma membrane localization of the TMX4 mutants was a result of overloading a retention/retrieval mechanism, because the expression levels of the various constructs used were similar (results not shown). Taken together, we found that the RQR sequence in the C-terminal region contributes to the ER localization of HA–TMX4.
ER localization conferred by the TMX4 cytosolic tail depends on the RQR sequence
To further characterize the importance of the RQR sequence of TMX4, we created three new fusion proteins (Figure 8A). All were HA-tagged chimaeras between the cytosolic tail of TMX4 and the luminal region of CD4. Two fusion proteins contained the wild-type sequence for the cytosolic tail of TMX4, but differed with respect to the origin of the transmembrane domain (TMX4 or CD4), whereas the last chimaera contained the CD4 luminal and transmembrane domains fused to the KQK mutant of the TMX4 cytosolic tail.
The chimaeras and wild-type HA–CD4 were transfected into HeLa cells that were analysed by confocal immunofluorescence microscopy (Figure 8B). As expected, we found HA–CD4 strongly stained the plasma membrane. In contrast, the two CD4 chimaeras fused to a wild-type TMX4 tail sequence did not reach the cell surface. The nature of the transmembrane domain, whether from CD4 or TMX4, did not influence the localization pattern. However, the chimaeric protein of HA–CD4 fused to the cytosolic tail of TMX4 containing the KQK mutation was clearly detected on the plasma membrane and showed a distribution very similar to wild-type HA–CD4. Overall, the results demonstrated that the cytosolic tail of TMX4 is sufficient for ER targeting and that the RQR sequence is required for this ability.
In the present study, we provide the first characterization of the transmembrane PDI family member TMX4. Using phylogenetic analysis we have shown that TMX4 and TMX are paralogues with closely related N-terminal ER-luminal regions, including a single thioredoxin-like domain, but quite different C-terminal cytosolic tails. Our analysis revealed that an evolutionarily older version of the two proteins is the DPY-11 protein from C. elegans. In C. elegans, this protein is expressed exclusively in the hypodermis and is required for body and sensory organ morphogenesis . Curiously, DPY-11 does not have the RQR motif, but instead a KKTK sequence, at the C-terminus, suggesting a different ER-localization mechanism than TMX4 and TMX. Also, human TMX cannot rescue a dpy-11 mutant in C. elegans .
Human TMX4 and most of its orthologues contain a CPSC active-site sequence, and all TMX orthologues have a CPAC motif. These are unusual active-site sequences among the PDIs, where the CGHC sequence is the most common with 16 occurrences among 25 domains with a CXXC sequence, whereas only two other domains contain a CPXC sequence . The enzymes with the common CGHC motif generally seem to have moderate reduction potentials that are neither especially oxidizing nor reducing with values of approx. −160 mV, as measured for TMX3, ERp57 and PDI [11,29–31]. The cellular redox state of these proteins and TMX is partially oxidized ([9,11,16,23]; Figure 5), but not nearly to the extent as seen in TMX4. Since the active-site sequence motif of TMX is very similar to that of TMX4, it is, however, unlikely that the uncommon CPSC motif in TMX4 can account for the observed redox state. Rather, currently unknown features of the TMX4 sequence that influence the redox properties of the protein, or its cellular redox regulation, set it apart from the other PDI family members investigated to date. Although the unusually oxidized redox state of TMX4 could have interesting functional implications, we do presently not know the exact cellular function of the protein. However, it could potentially be involved in ER quality control similar to TMX .
Neither TMX4 nor TMX contain a C-terminal K(X)KXX-type ER-retrieval signal. Using chimaeric molecules with CD4 that normally localizes to the plasma membrane, we mapped the TMX4 ER-localization signal to the cytosolic tail. In the present study, we found that even the conservative substitution of the two arginine residues in the RQR motif by lysine residues resulted in partial cell surface expression of TMX4 (Figure 7). Indeed, the RQR sequence constitutes a bona fide ER-localization signal, as further supported by the requirement of the two arginine residues for ER localization of TMX4 chimaeras with CD4 (Figure 8).
It is interesting to consider why TMX4 (and potentially TMX, which also has a conserved arginine-based sequence motif; Figure 1B) uses an arginine-based signal and not the classical C-terminal K(X)KXX motif. It is now clear that both types of signals function by COPI (coatamer protein I)-mediated retrieval to the ER, although binding of the involved sequence motifs occurs through different COPI subunits [32,33]. Both motifs can prevent forward transport of unassembled subunits until they are sterically hidden by oligomerization in the ER [34,35]. However, unlike the K(X)KXX sequences that are mainly found in ER-resident proteins, arginine-based signals are often present on multimeric proteins destined for the plasma membrane. In addition to masking by multimeric assembly, arginine-based motifs can be overcome by cytosolic interaction partners, such as 14-3-3 proteins, which promote cell surface expression of ion channels and receptors [36–38]. In the case of TMX4 (and TMX) suppression of the arginine-based motif, potentially in response to specific cellular conditions, might therefore allow the relocalization of the protein.
Features in addition to the RQR motifs in TMX4 and TMX suggest that their trafficking could be regulated by multiple factors. The tails of both proteins contain a long stretch of negatively charged residues comprising two experimentally verified phosphorylation sites on Ser251 and Ser259 for TMX4 (Figure 1B), and on Ser247 and Ser253 for TMX . The combination of these sequence features has been shown to constitute a potential binding site for PACS-2 (phosphofurin acidic cluster sorting protein 2), a cytosolic sorting protein that uses COPI to target its substrate proteins to the ER dependent on their phosphorylation state [40,41]. Although speculative at present, it is conceivable that the subcellular localization and thereby the functions of TMX4 and TMX are modulated by binding (potentially phosphorylation-dependent) of regulatory sorting factors that target binding sites in the cytosolic tail.
Doris Roth, Emily Lynes, Thomas Simmen and Lars Ellgaard designed the research. Doris Roth, Emily Lynes, Jan Riemer, Henning Hansen and Nils Althaus performed the research. Doris Roth, Emily Lynes, Jan Riemer, Thomas Simmen and Lars Ellgaard analysed the data. Doris Roth and Lars Ellgaard wrote the paper.
The work in the Simmen laboratory is supported by the National Cancer Institute of Canada, Terry Fox Foundation [grant number 17291]; and the Alberta Heritage Foundation for Medical Research [grant number 200500396]. The work in the Ellgaard laboratory is supported by the Eidgenössische Technische Hochschule (ETH) Zurich; The Roche Research Foundation; and the Novo Nordisk Fonden.
We thank the Institute of Biochemistry, ETH Zurich, Switzerland for continued support, all members of the Ellgaard laboratory for critical reading of the manuscript, Zhila Nikrozi for technical support and I. Novak for help with immunofluorescence microscopy. Constructs, cells and antibodies were kindly supplied by A. Helenius, R. Sitia, E. Sviderskaya, H.P. Hauri, R. Hartmann-Petersen, S.T. Christensen, M. Pasdar and K. Karjalainen.
Abbreviations: AMS, 4-acetamido-4-maleimidylstilbene-2,2-disulfonic acid; CNX, calnexin; COPI, coatamer protein I; DTT, dithiothreitol; Endo H, endoglycosidase H; ER, endoplasmic reticulum; FCS, fetal calf serum; HA, haemagglutinin; HEK-293, human embryonic kidney; MAM, mitochondria-associated membrane; NEM, N-ethylmaleimide; PDI, protein disulfide isomerase; redox, reduction–oxidation; rER, rough ER; RT–PCR, reverse-transcription–PCR; TCEP, tris-(2-carboxyethyl)phosphine; TMX, thioredoxin-like transmembrane
- © The Authors Journal compilation © 2010 Biochemical Society