ERp44 Mediates a Thiol-independent Retention of Formylglycine-generating Enzyme in the Endoplasmic Reticulum*

Inside the endoplasmic reticulum (ER) formylglycine-generating enzyme (FGE) catalyzes in newly synthesized sulfatases the post-translational oxidation of a specific cysteine. Thereby formylglycine is generated, which is essential for sulfatase activity. Here we show that ERp44 interacts with FGE forming heterodimeric and, to a lesser extent, also heterotetrameric and octameric complexes, which are stabilized through disulfide bonding between cysteine 29 of ERp44 and cysteines 50 and 52 in the N-terminal region of FGE. ERp44 mediates FGE retrieval to the ER via its C-terminal RDEL signal. Increasing ERp44 levels by overexpression enhances and decreasing ERp44 levels by silencing reduces ER retention of FGE. Suppressing disulfide bonding by mutating the critical cysteines neither abrogates ERp44·FGE complex formation nor interferes with ERp44-mediated retention of FGE, indicating that noncovalent interactions between ERp44 and FGE are sufficient to mediate ER retention. The N-terminal region of FGE harboring Cys50 and Cys52 is dispensible for catalytic activity in vitro but required for FGE-mediated activation of sulfatases in vivo. This in vivo activity is affected neither by overexpression nor by silencing of ERp44, indicating that a further ER component interacting with the N-terminal extension of FGE is critical for sulfatase activation.

The biogenesis of sulfatases involves as an essential step the generation of a unique amino acid, C ␣ -formylglycine (FGly) 4 (1,2). FGly is the catalytic residue in the active site of nearly all eukaryotic and prokaryotic sulfatases (1)(2)(3)(4)(5), 17 of which are encoded in the human genome and 10 of which so far have been shown to fulfill highly specific functions (6 -9). The FGly participates as an aldehyde hydrate in the hydrolysis of sulfate esters according to a novel transsulfation/elimination mechanism (10 -12). In eukaryotes newly synthesized sulfatase polypeptides undergo FGly modification in the lumen of the endoplasmic reticulum (ER) by late co-or early post-translational oxidation of a critical cysteine residue that is part of a highly conserved consensus motif (C(T/S/C/A)PSR for human sulfatases) (13,14). The oxidation of this cysteine is catalyzed by the recently discovered FGly-generating enzyme (FGE) (15,16) through a novel mixed-functional oxygenase mechanism (17,18). 5 Failure to generate FGly, as found in patients with mutations in the FGE-encoding SUMF1 gene, leads to multiple sulfatase deficiency (MSD), a rare but fatal inherited disorder that is characterized by synthesis of catalytically inactive sulfatases (15-17, 19 -21).
FGE is an N-glycosylated single-domain protein with a novel fold (17,18). Although it has little secondary structure, the catalytic core of FGE is a compact monomeric molecule that is stabilized by two intramolecular disulfide bridges and two Ca 2ϩ ions. On its surface FGE harbors a binding groove for the sulfatase CXPSR substrate peptide with Cys 336 and Cys 341 of FGE being involved in FGly formation (17,18). From this compact globular structure (residues 73-374), an N-terminal extension of 40 residues, the structure of which is unknown, obviously sticks out and is trimmed off in case of FGE secretion (22) by a furin-like protease. 6 Recently we could show that residues 34 -88 (residues 1-33 are cleaved off by signal peptidase) fulfill two functions. On the one hand they confer ER retention to FGE, and on the other hand they are required for in vivo sulfatase activation through FGly generation. 7 Moreover, we could show that retaining FGE in the ER and ensuring the in vivo activity of FGE are two separate functional properties mediated by the N-terminal extension of FGE, as fusion with a C-terminal KDEL signal led to effective ER retention of FGE but not to biological activity. 7 The latter is surprising, because in vitro the N-terminal extension is not needed for FGly modification of synthetic sulfatase peptide substrates (15,22). It was even more surprising to find that the N-terminal extension of FGE in living cells at least partially can complement in trans an N-terminally truncated, and therefore inactive, FGE. 7 This complementation, however, was only observed when FGE1-88 was expressed as an N-terminal fusion with the paralog of FGE (pFGE), which itself is catalytically inactive because of the absence of the two catalytic cysteines in the substrate-binding groove (24,25). pFGE lacks an N-terminal extension but is otherwise structurally very similar to FGE and shares with it localization in the ER lumen (24,25). Obviously the structural similarity of pFGE to FGE enables the FGE1-88-pFGE fusion described above to present the N-terminal extension to its catalytic FGE partner and thereby to complement in vivo functionality, i.e. FGly generation in nascent sulfatase polypeptides inside the ER. 7 We could further show that a Cys-Gly-Cys motif in this N-terminal extension of FGE, which is fully conserved in all known eukaryotic FGE sequences, is critical for activation of sulfatases, whereas it is dispensible for ER retention of FGE. 7 This led us to postulate that the N-terminal extension of FGE mediates the interaction with ER components, which are required for the generation of FGly residues and the retention of FGE in the ER. In this study we describe the identification of an ER component interacting with the N-terminal extension of FGE, the biochemical basis of this interaction, and its direct functional relevance for FGE retention in the ER.

EXPERIMENTAL PROCEDURES
Fishing of FGE-interacting Proteins and Identification of ERp44-HT1080 cells stably expressing His 6 -tagged FGE (22) were incubated for 5 min with PBS, pH 7.4, containing 150 mM NaCl and 20 mM NEM (buffer I) prior to harvesting by trypsinization and lysis by sonication (3 ϫ 20 s on ice) in buffer II, containing 20 mM Hepes, pH 7.6, 100 mM NaCl, 10 mM CaCl 2 , 5 mM MgCl 2 , 20 mM NEM, and protease inhibitor mixture (Sigma). The lysate was cleared by centrifugation at 100,000 ϫ g for 1 h. The supernatant was incubated with Affi-Gel-10 (Bio-Rad) for 1 h at 4°C to remove proteins that interact with Affi-Gel matrix. After spinning, the supernatant was incubated with either Affi-Gel-10 matrix derivatized with the ASA scrambled peptide (PVSLPTRSCAALLTGR) or Affi-Gel-10 matrix derivatized with the ASA-Ser 69 peptide (PVSL-STPSRAALLTGR) (22) for 2 h at 4°C. The beads were extensively washed with lysis buffer I and eluted with 400 M ASA-Ser 69 peptide. The eluted material was subjected to precipitation with 10% trichloroacetic acid and boiled in SDS-PAGE sample buffer with or without 100 mM dithiothreitol. The samples were resolved by SDS-PAGE (10% acrylamide), visualized by silver staining, and identified as indicated.
Expression Plasmids-The ERp44 cDNA was synthesized from total RNA, isolated from HeLa cells, by reverse transcription using the Omniscript RT kit (Qiagen) and an oligo(dT) primer. The first strand cDNA was amplified by PCR with primers ERp44-Nhe-fwd and ERp44-EcoRV-rev (see below).
ERp44⌬RDEL-His was generated by PCR amplification with pBI-MycERp44 as template and primers ERp44-Nhe-fwd (CTA-GCTAGCATGCATCCTGCCGTCTTCCTATCC) and ERp44-6His-EcoRVrev (CGGATATCTTAATGATGGTGGTGATGG-TGATGCGATCCATCCCTCAATAGAGTATACC). All of the PCRs were performed with Pfu-Ultra polymerase (Stratagene). The resulting constructs were analyzed by full-length sequencing of coding regions to preclude any PCR-derived errors.
Cell Culture and Transfection-The immortalized fibroblasts from an MSD patient (kindly gifted by Prof. Andrea Ballabio, Naples, Italy) and HT1080 cells were maintained at 37°C under 5% CO 2 in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum (PAN Biotech GmbH). Transient transfections were performed with Lipofectamine 2000 as recommended by the manufacturer (Invitrogen). For single or coexpression experiments, the HT1080 Tet-On or MSDi Tet-On cells 7 were transfected with 2 g of pBI vector carrying one or two cDNAs driven by a bidirectional tetracycline responsive promoter. For triple expression, the MSDi Tet-On cells were transfected with a mixture of two plasmids: 2 g of pBI vector containing steroid sulfatase alone and 2 g of pBI vector containing two different cDNAs. 6 h after transfection, the medium was replenished with medium containing 2 g/ml doxycycline (BD Biosciences), and after 30 h, the cells and medium were collected for further analysis.
An HT1080 Tet-On cell line stably expressing c-Myc-ERp44 under control of a doxycycline-responsive promoter was established by cotransfecting HT1080 Tet-On cells (neomycin-resis-tant) with pBI-MycERp44 and the puromycin resistance vector pSV.pac (10:1 ratio). Transfectants were selected in normal medium with increasing concentrations of puromycin (Invitrogen) from 0.1 to 0.8 g/ml and 800 g/ml neomycin. The stable clones were screened for expression of Myc-ERp44 by Western blotting, upon induction of the cells with 0.1 g/ml doxycycline for 24 h.
Immunofluorescence Microscopy-HT1080 Tet-On cells were grown on coverslips overnight and transfected with plasmids using Lipofectamine 2000. The cells were induced with 1 g/ml doxycycline 6 h after transfection. After 24 h of induction, the cells were washed twice with PBS, fixed with 4% paraformaldehyde for 20 min at room temperature, and treated with 50 mM NH 4 Cl for 10 min. The cells were permeabilized with 0.5% saponin in PBS and incubated for 1 h at room temperature with primary antibodies, namely rabbit anti-FGE antiserum, either rabbit polyclonal or mouse monoclonal antibodies against ERp44 (kind gift from Prof. Roberto Sitia, Milan, Italy) and mouse monoclonal antibodies against PDI (Stressgen Biotechnologies) or GM130 (BD Biosciences). After washing with PBS containing 0.1% saponin, the primary antibodies were decorated with appropriate secondary antibodies coupled with either Cy2 or Alexa 633 (Molecular Probes) for 1 h at room temperature. The coverslips were mounted on glass slides with fluorescent mounting medium (Dakocytomation), and confocal images were acquired with Leica TCS SP2 AOBS laser scanning microscope.
Coimmunoprecipitation-HT1080 Tet-On cells transfected with appropriate plasmids were induced for protein expression with 2 g/ml doxycycline 6 h after transfection. After induction for 24 h, the cells were washed with PBS and incubated with buffer I for 5 min at 37°C, washed once again with PBS, and harvested by trypsinization. NEM was included in the lysis buffer to prevent any reshuffling of disulfide bonds during lysis. The cell pellet was resuspended in buffer I, further containing a protease inhibitor mixture (Sigma), and the cells were lysed by sonication (3 ϫ 20 s on ice). The cell lysate was centrifuged at 100,000 ϫ g, and the supernatant was incubated either with rabbit preimmune serum or rabbit anti-FGE serum for 2 h at 4°C, followed by incubation with Pansorbin for 1 h at 4°C with end-over-end rotation. The Pansorbin pellet was washed five times with buffer I containing 0.1% Triton X-100 and washed finally once with buffer I without Triton. The pellet was boiled in 1ϫ sample buffer without ␤-mercaptoethanol. 10% of the starting material prior to immunoprecipitation (load) and 10% of the pellet were resolved by SDS-PAGE either under nonreducing or reducing conditions and transferred to a polyvinylidene difluoride membrane for Western blotting and probed with appropriate antibodies, as indicated in each figure.
ERp44 RNAi Experiments-The siRNA duplex names and their target sequences were exactly as described earlier (26), namely siControl-C (5Ј-AAGUAGUGUAUGCUAGAGUGG-3Ј), siERp44-C (5Ј-AAGUAGUGUUUGCCAGAGUUG-3Ј), siControl-3U (5Ј-AACAGCACCAUCGACCAACGU-3Ј), and siERp44-3U (5Ј-AACAGCAGCAUCAACCUACGU-3). The siRNA duplexes were purchased from IBA (Göttingen, Germany). Transfections of siRNA duplexes were performed using Lipofectamine 2000 as recommended by the supplier. For HT1080-Tet-on cells, the siRNA duplexes were transfected at 50 -60% confluency, and after 24 h, the cells were split and seeded in such a way that they were again 50 -60% confluent during the second siRNA duplex transfection, which was done ϳ48 h after the first transfection. After another 8 h, the indicated cDNAs were transfected as pBI constructs (see above). Six hours later the cells were induced for protein expression with 2 g/ml doxycycline. After 16 h of induction, the cells and medium were collected and analyzed.
For immortalized MSD (MSDi) cells, the siRNA duplexes were transfected at 30 -50% confluency using RNAiMax reagent (Invitrogen) as recommended by the supplier. After 30 h the indicated cDNAs were transfected as pBI constructs. 6 h after transfection, the cells were induced for protein expression with 2 g/ml doxycycline. After 36 h of induction (72 h after siRNA treatment), the cells were harvested and analyzed.
Activity Assay for Steroid Sulfatase and Western Blotting-Activity assays of steroid sulfatase were performed as described earlier (31). For Western blot analysis, polyclonal antisera against FGE, ERp44 or steroid sulfatase, and monoclonal antic-Myc (Santa-Cruz Biotechnology), anti-HA (Covance), or anti-Hsc70 (rat monoclonal, 1B5, Abcam) antibodies were used as primary antibodies. Horseradish peroxidase-conjugated goat anti-rabbit, anti-mouse, or anti-rat antibodies were used as secondary antibodies. Western blot signals were quantified using the AIDA 2.1 software package (Raytest). FGE amounts were calculated on the basis of known amounts of purified FGE present on each Western blot. Signals of steroid sulfatase are given as relative amounts, i.e. related to signal intensities detected in cells expressing steroid sulfatase only. Relative specific sulfatase activities were calculated, i.e. catalytic activity divided by the Western blot signal (arbitrary units) and referred to that of cells expressing the sulfatase only (relative specific sulfatase activity ϭ 1). Absolute values for this reference are given in the legends.

Identification of ERp44 as an in Vivo Interaction
Partner of FGE-To identify interacting partners of FGE, we aimed at pulling out complexes of FGE with its partners from cells overexpressing FGE. We took advantage of an affinity column with an immobilized ASA peptide that binds with high affinity to the active site of FGE (15). To control for specificity, we used in parallel an affinity column carrying an immobilized peptide of identical amino acid composition but having the three key residues of the FGly modification motif of sulfatases in scrambled order (see Ref. 15).
HT1080 cells stably overexpressing FGE were treated with NEM prior to lysis to stabilize disulfide bonds that had formed in vivo. Using the original FGE purification protocol (15), we observed binding of FGE only to the sulfatase peptide column (Fig. 1, lanes 3 and 4). In the eluate, obtained with free ASA peptide, two additional bands in the range of 70 -75 kDa were detected (Fig. 1, lane 3), which contained no FGE, as evidenced by Western blotting (not shown). Tryptic digestion and MALDI-TOF mass spectrometry identified them as the two Hsp70 chaperone isoforms Hsp70-5 and Hsp70-9. These chaperones were also detectable in the eluate from the control affin-ity matrix, albeit at much lower levels ( Fig. 1, lanes 1 and 2), indicating that their binding is not FGE-dependent. Running the eluates under nonreducing conditions revealed in the eluate from the sulfatase peptide-derivatized affinity matrix a further band at ϳ85 kDa (Fig. 1, lane 4). Analysis of its tryptic peptides by MALDI-TOF mass spectrometry clearly identified two components, namely FGE and ERp44 that made up a stable complex. When NEM pretreatment of the cells prior to lysis was omitted (not shown) or dithiothreitol was present during SDS-PAGE, the 85-kDa band was not detectable (compare lanes 3 and 4 in Fig. 1), indicating that the complex is stabilized by disulfide bridges.
ERp44 has been described as a prominent retention factor for Ero1, unpolymerized immunoglobulin chains, and adiponectin and also as a redox-dependent regulator of the inositol 1,4,5triphosphate receptor type 1 (23,(27)(28)(29)(30). To further analyze the interaction of FGE and ERp44, we coexpressed both proteins in differentially tagged forms, i.e. FGE-HA and c-Myc-ERp44, in HT1080 cells. From these cells we could pull down the heterodimeric FGE⅐ERp44 complex by immunoprecipitation using either anti-FGE ( Fig. 2A, lanes 3 and 6) or anti-c-Myc antibodies (Fig. 2B, lane 9). Additionally, two bands of ϳ170 and 340 kDa were pulled down, which were shown by mass spectrometry to consist of FGE and ERp44 and are regarded as heterotetrameric (FGE⅐ERp44) 2 and heterooctameric (FGE⅐ERp44) 4 complexes, respectively. Western blot analysis of the immunoprecipitates revealed that ϳ50% of FGE were associated with ERp44, with the heterodimer being favored over the heterotetrameric and heterooctameric complexes. The remaining FGE was either monomeric or dimeric FGE, and also here the monomer was favored over the dimer ( Fig. 2A, right  panel). FGE dimers are stabilized through disulfide bonding of Cys 50 and/or Cys 52 , as has been described earlier (22). Thus, the majority of FGE is engaged in different covalent complexes either with itself or with ERp44.
ERp44 and FGE Form Noncovalent and Disulfide Bridge-stabilized Complexes-Under the given expression levels in the cells (FGE exceeding ERp44), approximately half of FGE is recovered as a complex with ERp44 ( Fig. 2A, lane 6), whereas ERp44 was quantitatively present in the complexes with FGE. In most experiments monomeric ERp44 barely was detectable (Fig. 2, lanes 1 and 3, Fig. 3, lanes 3 and 4). In vivo, the complexes are highly labile, because no complexes could be recovered from cell extracts when the NEM treatment of the cells prior to lysis was omitted (Fig. 3, lanes 1 and 2). The complexes are readily split into their monomeric constituents by ␤-mercaptoethanol (Fig. 3, lanes 7 and 8) or other thiols. This suggests that upon cell lysis the disulfide bridges stabilizing the complexes are readily reduced by free thiols. Only through blocking free thiol groups prior to lysis by alkylating reagents such as NEM, can the complexes be recovered (Fig. 3, lanes 3 and 4).
The noncatalytic N-terminal extension of FGE carries two cysteine residues in positions 50 and 52. When ERp44 was coexpressed with N-terminally truncated FGE lacking residues 34 -68, none of the disulfide bonded FGE⅐ERp44 complexes could be detected (Fig. 4, compare lanes 1 and 2 with lanes 4 and

5).
It should be noted that in the cells used for this experiment (stably expressing ERp44 and transiently expressing wild-type FGE or ⌬34 -68FGE), large amounts of monomeric ERp44 were detectable (Fig. 4, lanes 1 and 4). In these cells traces of monomeric ERp44 also were detected in the immunoprecipitates obtained with anti-FGE antibodies (Fig. 4, lanes 2 and 5). This indicates that ERp44 also forms noncovalent complexes with both full-length and truncated FGE, which after SDS-PAGE are disassembled into their monomeric constituents. They are detectable, however, only under conditions of ERp44 excess over FGE.
Cys 50 and Cys 52 of FGE Mediate Disulfide Bonding with Cysteine 29 of ERp44 and with One Another-To identify the cysteines engaged through mixed disulfide linkage in FGE and ERp44, the following alanine mutants were generated: FGE-C50A, FGE-C52A, FGE-C50A/C52A, ERp44-C29A, and ERp44-C63A. Expression of FGE-C50A or FGE-C52A led to efficient recruitment of coexpressed wild-type ERp44 into heterodimeric covalent complexes, i.e. with similar efficiency as observed for wild-type FGE (Fig. 5). As observed earlier (22) FGE homodimer formation requires Cys 50 and Cys 52 . Notably, the heterotetrameric and heterooctameric complexes were not found with each of the cysteine mutants. With the double mutant FGE-C50A/C52A no covalent complex with ERp44 was formed at all. Thus, the two cysteines Cys 50 and Cys 52 are equally well capable of forming mixed disulfide bonds with ERp44, but both cysteines are required for formation of the higher heterooligomeric complexes. It should be noted that in the FGE-C50A/C52A-coexpressing cells ERp44 is forming higher oligomeric complexes with itself (Fig. 5, top of lane 10). This obviously is the consequence of lacking disulfide stabilization of complexes with FGE, at least under the applied in vitro   using either anti-c-Myc or anti-FGE antibodies, as described above (Fig. 4). SDS-PAGE was run under nonreducing (ϪSH, upper panels) or reducing (ϩSH) conditions (lower panels). conditions (cell lysis and immunoprecipitation). The observation, on the other hand, of significant amounts of ERp44 that were immunoprecipitated by the anti-FGE antibody even from FGE-C50A/C52A-coexpressing cells again demonstrates that noncovalent heteromeric complexes between FGE and ERp44 are formed, which, however, are disrupted during SDS-PAGE (Fig. 5, lane 11).
Cys 29 of ERp44 has been shown to form disulfide-bonded complexes with other proteins. Also Cys 63 of ERp44 was proposed to have a surface-exposed localization and found to be accessible for NEM (30). In cells coexpressing FGE and ERp44-C63A, heteromeric complex formation between FGE and ERp44 was as efficient as in cells coexpressing wild-type ERp44 (Fig. 6, lanes 5, 6, 11, and 12), whereas in cells coexpressing FGE and ERp44-C29A, none of the heteromeric complexes were detectable (Fig. 6, lanes 3, 4, 9, and 10). Thus, disulfide bonding occurs between Cys 29 of ERp44 and either Cys 50 or Cys 52 of FGE.
ERp44 Is a Retention Factor for FGE-A fraction of FGE is constitutively secreted. This fraction increases upon overexpression of FGE (22). During secretion the majority of FGE (42 kDa) is processed by a furin-like activity to a 37-kDa form (⌬34 -72FGE). We established HT1080 Tet-On cells that allowed to coexpress FGE and ERp44 from a bidirectional doxycycline-responsive promoter. After induction for 24 h, the cells and medium were analyzed by Western blotting. 15-20% of total FGE are found intracellularly, whereas the rest is secreted largely as ⌬34 -72FGE (Fig. 7, sample 1). Coexpression of ERp44 led to an up to 4-fold increase of intracellular FGE (Fig. 7, sample 2). ERp44 itself was fully retained intracellularly. ERp44 lacking the C-terminal RDEL retention signal was largely secreted and unable to increase FGE retention (Fig. 7, sample 3), indicating that the increase of ER retention by ERp44 relies on retrieval of the FGE⅐ERp44 complexes through KDEL receptors.
To determine whether disulfide bonding between ERp44 and FGE is required for mediating ER retention, we examined the retention when the cysteine mutants C50A and C52A of FGE were coexpressed with wild-type ERp44 and also wild-type FGE with the C29A mutant of ERp44. The results clearly show that ERp44-mediated retention of FGE is independent of disulfide bonding between ERp44 and FGE (Fig. 7,  samples 4 -7).
Recently we could show that the N-terminal extension (residues 34 -68) of FGE is required for the retention of FGE in the ER. 7 After induction for 24 h, ϳ10% of the N-terminally truncated FGE (⌬34 -68FGE) were found intracellularly. Coexpression with ERp44 increased the retention ϳ2-fold (Fig. 7, samples 8 and 9), indicating that the N-terminal extension of FGE is one but not the only part of FGE that contributes to ERp44-mediated retention.
Overexpression of ERp44 clearly increased the intracellular retention of FGE. To determine whether FIGURE 6. Cys 29 of ERp44 is responsible for disulfide-bonding with FGE. NEM lysates of HT1080 cells coexpressing FGE-HA and c-Myc-tagged forms of ERp44, ERp44-C29A or ERp44-C63A were subjected to immunoprecipitation with rabbit anti-FGE antibodies (␣-FGE) and analyzed by SDS-PAGE under nonreducing conditions (ϪSH) and Western blotting (WB) using either antic-Myc or anti-HA antibodies, as described above (Fig. 4). To determine total amounts of wild-type and mutant forms of ERp44 in the starting material and the immunoprecipitates, Western blot analysis was also performed after SDS-PAGE under reducing conditions (ϩSH, see lower panels). The blot signals, shown for lanes 1-6, also apply to lanes 7-12, respectively, because they represent aliquots from the same samples. The open arrowhead indicates the ERp44 homodimer detectable for the cysteine mutants of ERp44 both in the cell lysate and the immunoprecipitate with anti-FGE antibodies (lanes 3-6). FIGURE 7. ERp44 retains FGE intracellularly. HT1080 Tet-On cells were transiently transfected with pBI plasmids containing FGE and ERp44 cDNAs, or mutants thereof, in the combinations indicated above the lanes (with FGE constructs coding also for a C-terminal HA tag and ERp44 constructs for an N-terminal c-Myc tag). 6 h after transfection, coexpression of FGE and ERp44 was induced with 2 g/ml doxycycline. After induction for 24 h, the cells and medium (at a ratio of 10:1 in samples 1-7 or 5:1 in samples 8 and 9) were analyzed for FGE and ERp44 by Western blotting with polyclonal anti-FGE and monoclonal anti-c-Myc antibodies. The amount of FGE in the cells and in the medium was determined by calibration of the Western blot signals with known amounts of purified FGE protein. The amount of FGE retained intracellularly is given below the lanes as percentage of total FGE in cells and medium. The asterisk (in sample 3) indicates endogenous ERp44, which migrates slightly slower than the recombinant ERp44⌬RDEL.
ERp44 is necessary for ER retention of FGE, we attempted to reduce ERp44 levels by silencing. By expressing different RNAi constructs (si-ERp44-3U and si-ERp44-ORF) for ERp44, the levels of ERp44 in HT1080 cells were reduced to less than 5 and 20% of control, respectively (Fig. 8, shown for si-ERp44-3U). Reducing the ERp44 level to 4% of control decreased the fraction of FGE recovered intracellularly 12 h after induction from 36 to 8% (Fig. 8). Likewise silencing of ERp44 in cells coexpressing FGE and ERp44 from the same inducible promoter diminished the retention of FGE from 57 to 19%, which agrees with the less complete ERp44 silencing in these cells (data not shown). Please note the shorter induction time (12 h) as compared with Fig. 7 (24 h), which was chosen to start at relatively higher intracellular FGE levels (36 instead of 18%, see lanes 1 in Figs. 8 and 7, respectively), i.e. with less FGE accumulating in the secretions. Taken together, these results clearly indicate that ERp44 is responsible for retention of FGE in the ER.
FGE Recruits ERp44 from Distal Compartments to the ER-To verify at the subcellular level the association of ERp44 with FGE, we examined the colocalization of FGE and ERp44 with one another and with PDI and GM130, markers for the ER and Golgi, respectively. In HT1080 cells endogenous ERp44 localizes mainly to perinuclear structures (Fig. 9, two cells on the left) that are positive for GM130 (not shown, see Ref. 29). Upon transient expression of FGE, ERp44 largely redistributes to the ER and colocalizes with FGE (Fig. 9, compare ERp44 localization in FGE-expressing cells marked by asterisks with that in nontransfected cells). FGE colocalizes with PDI as shown ear-lier (22). These observations clearly indicate that localization of ERp44 is influenced by the expression level of FGE, providing additional evidence for the interaction between ERp44 and FGE.
FGE-mediated Activation of Sulfatases Is Independent of ERp44-The N-terminal extension of FGE is required for sulfatase activation. Mutation of Cys 52 to alanine abrogates the FGE-mediated activation of newly synthesized sulfatases, and mutation of Cys 50 to alanine reduced the activation to approximately half. 7 To examine whether ERp44 is involved in the activation of sulfatases, we overexpressed steroidsulfatase (STS) in immortalized MSD Tet-On cells (MSDi cells), which lack endogenous FGE. Cotransfecting these cells with bidirectional promotor-driven cDNAs coding for FGE and ERp44 and an STS encoding plasmid allowed for coexpression of all three proteins under the control of doxycycline (Fig. 10). Expression of STS alone led to the synthesis of STS polypeptides that were barely active. Coexpression of FGE increased the activity of STS 50 -100-fold. This effect is solely due to the activation of the STS polypeptides and not to an increase of STS synthesis. Coexpression of ERp44 with FGE did not further increase the activation of STS (Fig. 10). If at all, it decreased the activity of STS slightly (by 5-20% in three independent experiments). Coexpression of FGE with ERp44-C29A or ERp44-C63A, which potentially are dominant negative mutants of ERp44, did not affect STS activity (Fig. 10). A comparison of the expression levels of endogenous and recombinant ERp44 forms revealed that the level of endogenous ERp44 exceeded that of the recombinant forms at least 4-fold. The experiment was repeated with HT1080 Tet-On cells. In these cells the recombinant forms of ERp44 reach levels that are 2-fold higher than that of endogenous ERp44. However, in these cells, which express endogenous FGE, the overexpression of FGE increases the activity of STS only 2-3-fold. 7 Also in this system coexpression of ERp44 with FGE did not affect STS activity (not shown). These data indicate that ERp44 is not a limiting factor for activation of STS by FGE when the latter two are overexpressed.
Next we examined whether reducing the expression of ERp44 by silencing would affect the activation of STS by FGE. Using the RNAi construct si-ERp44-3U, we succeeded in reducing the level of endogenous ERp44 in MSDi Tet-on cells by more than 90%. This, however, did not affect the relative FIGURE 8. Silencing of ERp44 increases secretion of FGE. HT1080 Tet-On cells were treated with the indicated siRNA duplexes and transiently transfected with pBI plasmids containing FGE cDNA (see "Experimental Procedures"). After induction with 2 g/ml doxycycline for 12 h, the cells and medium (at a ratio of 5:2) were analyzed for FGE and ERp44 by Western blotting with polyclonal anti-FGE and monoclonal anti-c-Myc antibodies. The amount of FGE in the cells and in the medium was determined as described above (Fig. 7). Different from other experiments an induction period of only 12 h was chosen. This explains the clearly higher retention in sample 1 (36%) as compared with values obtained after 24 h of induction (Fig. 7, sample 1; 18% retention), because less FGE could accumulate in the medium. stimulation of STS activity by FGE (not shown). In HT1080 cells, the two RNAi constructs (si-ERp44-3U and si-ERp44-ORF) reduced the ERp44 level to less than 5% (Fig. 8) and ϳ20%, respectively. The 2-3-fold stimulation of STS activity by coexpression of FGE was not affected by silencing ERp44 expression (not shown). These data indicate that either ERp44 is not involved in FGE-mediated activation of STS or that the residual amount of ERp44 left after silencing is sufficient to support FGE-mediated activation of STS.

DISCUSSION
In a recent study 7 we found the N-terminal extension of FGE (residues 34 -88) to be involved in two separable and essential properties of FGE. First, the retention of FGE in the ER is significantly enhanced by its N-terminal noncatalytic extension. This effect is independent of the conserved cysteine residues Cys 50 and Cys 52 within the N-terminal extension. Second, the activation of sulfatases by FGE depends on the presence of the N-terminal extension, although the latter is dispensible for the generation of FGly residues under in vitro conditions. For the in vivo activation of sulfatases the cysteine residue Cys 52 is critical, whereas mutating Cys 50 reduces FGE-mediated activation of sulfatases only to approximately half.
Here we show that the N-terminal extension of FGE significantly enhances the complex formation with the ER protein ERp44. These complexes are stabilized by disulfide bridges between Cys 29 of ERp44 and Cys 50 or Cys 52 of FGE. The complex formation is shown to mediate the retention of FGE in the ER, whereas there is no evidence that activation of sulfatases by FGE depends on FGE⅐ERp44 complexes.
ERp44 Is a Thiol-independent Retention Factor for FGE-ERp44 has been thoroughly characterized as a thiol-dependent ER retention factor for Ero1, the major oxidase for PDI, and for other oxidoreductases of the ER (23,(27)(28)(29)(30). ERp44-mediated retention relies on retrieval from post-ER compartments through KDEL receptors. The formation of mixed disulfides with Ero1 via Cys 29 of ERp44 was found to be essential for Ero1 retention (thiol-mediated retrieval, ref. 30). ERp44 also retains secretory proteins like immunoglobulin chains and adiponectin through disulfide bond formation during biogenesis to assist their assembly and to retrieve nonpolymerized subunits back to the ER (thiol-mediated assembly and quality control). Both thiol-mediated processes localize ERp44 to the exit sites of the ER, to the ER-Golgi intermediate compartment and to the cis-Golgi (29). We observed that ERp44 localizes to both PDI-and GM130-positive compartments and that overexpression of FGE shifts the distribution of ERp44 to the ER (Fig. 9). This relocation suggested that ERp44 and FGE physically interact. In fact, FGE⅐ERp44 complexes could be detected by coimmunoprecipitation and coaffinity purification (Figs. 1-6).
The complex formation between FGE and ERp44 serves to retrieve FGE back to the ER. Increasing ERp44 levels by overexpression (Fig. 7) improves and reducing ERp44 levels by silencing (Fig. 8) decreases the intracellular retention of FGE. The retention depends on the C-terminal RDEL retrieval signal of ERp44. Deletion of the latter fully abolished the retention of FGE. Interestingly, in case of ERp44⌬RDEL coexpression, FGE is mainly secreted without N-terminal truncation (Fig. 7, compare samples 1 and 3). Thus, ERp44⌬RDEL, which also is secreted, protects FGE from furin processing. This clearly indicates that FGE passes the furin-type protease in the secretory route as an FGE⅐ERp44⌬RDEL complex and corroborates the tight interaction of the two components.
It should be noted that ERp44 exerts its FGE retention function through a thiol-independent mechanism. FGE retrieval did not depend on the cysteine residues required to establish the disulfide bonds between ERp44 and FGE. The observation that FGE retention through coexpressed ERp44-C29A is fully functional in vivo (Fig. 7) clearly suggests that the noncovalent interaction between the two partners is strong enough to mediate retrieval to the ER. The partial retention of FGE lacking residues 34 -68 by ERp44 (Fig. 7, sample 9) suggests that noncovalent FGE-ERp44 complexes are formed even if the N-terminal 35 residues of the N-terminal extension of FGE (residues 34 -68) are lacking. It should be noted that a thiol-independent but ERp44-mediated retention of cargo proteins is not unique for FGE and has been observed for monomeric immunoglobulin K and J and mutant chains (27,30).
Sulfatase Activation by FGE Does Not Depend on Complexes with ERp44-ERp44 has been shown to mediate the redox regulation of the inositol 1,4,5-trisphosphate receptor type 1, a prominent ion channel in the brain for the release of Ca 2ϩ from the ER (26). The dependence of sulfatase activation on the N-terminal extension of FGE 7 and requirement of the cysteine residues Cys 50 and Cys 52 within this extension for the formation of covalent complexes with ERp44 prompted us to examine the involvement of such complexes in the FGE-mediated activation of sulfatases. Two experimental lines strongly suggest that complexes with ERp44 are not directly involved in FGEmediated activation of sulfatases. First, overexpression of ERp44 augmented the FGE-mediated stimulation of STS neither in MSDi cells nor in HT1080 cells. Second, reduction of the ERp44 level to less than 10% of control through silencing did not affect the FGE-mediated activation of STS. Although the result of the silencing cannot formally exclude a catalytic role of ERp44 in the activation of sulfatases, it clearly excludes a role for ERp44 as a coactivator required in stoichiometric amounts for FGE-mediated activation of sulfatases. ERp44, however, can affect FGE-mediated activation of sulfatases indirectly through increasing the retention of FGE in the ER.
Concluding Remarks-The N-terminal extension of FGE enhances its retention in the ER. Complex formation with the ER protein ERp44 is shown to mediate this retention. The complexes are stabilized by disulfide bridges involving Cys 50 and Cys 52 in the N-terminal extension of FGE and Cys 29 in ERp44. The functional significance of the covalent linkages between FGE and ERp44, which in vivo are highly labile, remains unclear. They are not required for retention. The N-terminal extension is furthermore required for the biological function of FGE, the activation of sulfatases by post-translationally generating an FGly residue in the catalytic site of newly synthesized sulfatase polypeptides. This function apparently is not mediated by the complexes of FGE with ERp44. Silencing of ERp44 did not diminish FGE-mediated activation of sulfatases. The latter critically depends on Cys 52 but not on Cys 50 within the N-terminal extension of FGE. The requirement for complex formation of FGE with ERp44 did not distinguish between Cys 50 and Cys 52 . Formation of noncovalent complexes was independent of either cysteine residue, whereas formation of covalent complexes was equally dependent on both cysteine residues. This indicates that for sulfatase activation the interaction of the N-terminal extension with a second, non-ERp44 component is required.