The N-terminal Anchor Sequences of 11β-Hydroxysteroid Dehydrogenases Determine Their Orientation in the Endoplasmic Reticulum Membrane*

11β-Hydroxysteroid dehydrogenase enzymes (11β- HSD) regulate the ratio of active endogenous glucocorticoids to their inactive keto-metabolites, thereby controlling the access of glucocorticoids to their cognate receptors. In this study, the topology and intracellular localization of 11β-HSD1 and 11β-HSD2 have been analyzed by immunohistochemistry and protease protection assays ofin vitro transcription/translation products. 11β-HSD constructs, tagged with the FLAG epitope, were transiently expressed in HEK-293 cells. The enzymatic characteristics of tagged and native enzymes were indistinguishable. Fluorescence microscopy demonstrated the localization of both 11β-HSD1 and 11β-HSD2 exclusively to the endoplasmic reticulum (ER) membrane. To examine the orientation of tagged 11β-HSD enzymes within the ER membrane, we stained selectively permeabilized HEK-293 cells with anti-FLAG antibody. Immunohistochemistry revealed that the N terminus of 11β-HSD1 is cytoplasmic, and the catalytic domain containing the C terminus is protruding into the ER lumen. In contrast, the N terminus of 11β-HSD2 is lumenal, and the catalytic domain is facing the cytoplasm. Chimeric proteins where the N-terminal anchor sequences of 11β-HSD1 and 11β-HSD2 were exchanged adopted inverted orientation in the ER membrane. However, both chimeric proteins were not catalytically active. Furthermore, mutation of a tyrosine motif to alanine in the transmembrane segment of 11β-HSD1 significantly reducedV max. The subcellular localization of 11β-HSD1 was not affected by mutations of the tyrosine motif or of a di-lysine motif in the N terminus. However, residue Lys5, but not Lys6, turned out to be critical for the topology of 11β-HSD1. Mutation of Lys5 to Ser inverted the orientation of 11β-HSD1 in the ER membrane without loss of catalytic activity. Our results emphasize the importance of the N-terminal transmembrane segments of 11β-HSD enzymes for their proper function and demonstrate that they are sufficient to determine their orientation in the ER membrane.

Sequence alignment and molecular modeling of human 11␤-HSD1 1 and human 11␤-HSD2, using the known three-dimen-sional structures of human dihydropteridine reductase and Streptomyces hydrogenans 20␤-HSD as templates, indicated that the structures of these members of the short chain dehydrogenase reductase family of proteins are very similar, despite only 18% sequence identity between their entire sequences (1). Although both 11␤-HSD enzymes control the conversion of biologically active glucocorticoids (cortisol in humans and corticosterone in rats and mice) to their inactive 11-keto forms (cortisone and 11-dehydrocorticosterone), there are important functional differences such as cofactor specificity, substrate affinity, or direction of the reaction.
The isoform 11␤-HSD1 is expressed in a wide array of tissues, with highest levels in the liver, from where it was purified originally (2). It catalyzes both the oxidation and reduction of glucocorticoids but acts predominantly as an oxidoreductase, thereby increasing the concentration of active glucocorticoids (3)(4)(5)(6)(7)(8). Studies on the purified protein demonstrated glycosylation and existence of a disulfide bond, suggesting that the bulk of 11␤-HSD1 is oriented to the ER lumen (9). By converting 11-keto-into 11␤-hydroxyglucocorticoids, 11␤-HSD1 plays an important role in the glucocorticosteroid receptor-mediated anti-inflammatory response of glucocorticoids (10). Mice deficient in 11␤-HSD1 were found to resist hyperglycemia provoked by obesity or stress (11). Other investigators provided evidence that 11␤-HSD1 plays a role in detoxification processes (12) and in the reductive metabolism of xenobiotics (13).
To understand the differences in the physiological functions of 11␤-HSD1 and 11␤-HSD2, it is of great interest to know the exact topology and intracellular localization of these enzymes. Therefore, we evaluated the role of the N-terminal anchor sequences of 11␤-HSD1 and 11␤-HSD2 on their topology, in-tracellular localization, and catalytic activity. Our results demonstrate that the N-terminal sequences are critical for proper function of 11␤-HSD enzymes and that they determine their orientation in the ER membrane.

MATERIALS AND METHODS
Constructs for Expression-For cloning into and expression from the pcDNA3 plasmid (Invitrogen), the 5Ј end of both the human 11␤-HSD type 1 gene (35) and the type 2 gene (18) was modified first by the introduction of a BamHI site and second by changing the context 5Ј to the initiator ATG codon to a "Kozak" consensus sequence (36). At the 3Ј end, an XbaI site was introduced immediately 3Ј of the stop codon. Thus, the 11␤-HSD1 gene was modified to 5Ј GAGGATCCGCCATGG-CTTTT 3Ј to 5Ј AAGTAGGATCTAGATT 3Ј and the 11␤-HSD2 gene was changed to 5Ј GAGGATCCGCCATGGAGCGC 3Ј to 5Ј GCTCGGTGAG-CCATCTAGATT 3Ј (restriction endonuclease sites, initiator, and stop codons are underlined). 11␤-HSD1 was cloned by reverse transcriptase-PCR from total RNA isolated from primary culture cells of human aortic smooth muscle, a kind gift of Dr. K. Plü ss (37,38). 11␤-HSD2 was amplified by PCR from the original clone kindly provided by Dr. Z. S. Krozowski (see Ref. 18). The clone used for expression of the mineralocorticoid receptor (MR) was a generous gift of Dr. R. M. Evans (see Ref. 39).
A sequence coding for the FLAG epitope (Sigma) was attached either to the N terminus (F1 and F2), creating the N-terminal sequence NH 2 -MDYKDDDD-M 1 , or to the C terminus (1F and 2F), creating the C-terminal sequence -MDYKDDDD-COOH. Attachment was through the polymerase chain reaction (PCR) using primers with 5Ј-add-on sequences containing a restriction endonuclease site and a sequence coding the FLAG epitope.
Chimeric constructs, where the N-terminal domain containing the transmembrane segment was exchanged, were obtained by PCR amplification. Sequences were exchanged at the beginning of the conserved Rossmann-fold motif (40,41) by introducing a PinAI restriction endonuclease site at the conserved threonine and glycine residues, not affecting the amino acid sequence (11␤-HSD1, Thr 40 Gly 41 , ACCGGC to ACCGGT; 11␤-HSD2: Thr 88 Gly 89 , ACAGGG to ACCGGT). The chimeric proteins were tagged with the FLAG epitope at the C terminus to allow facilitated detection. Constructs 1F and 2F were further subjected to site-directed mutagenesis using the Quick Change Mutagenesis kit from Stratagene. All constructs were verified by sequencing.
Immunostaining-HEK-293 cells, grown to a confluence of 70% in DMEM with 10% fetal calf serum, 4.5 g/liter glucose, 50 units/ml penicillin/streptomycin, 2 mM glutamine, were split 1:15 and transferred onto glass coverslips placed in six-well plates. In experiments where cells were cotransfected with 2F and MR, the molar ratio was 3:1, 1:1, or 1:3, respectively, with a total amount of 3 g of DNA per well. The medium was replaced 14 h later, and cells were transfected with 2.5 g of the corresponding DNA per well. Eight hours later, the medium was replaced to remove the Ca 2ϩ -phosphate precipitate. Immunostaining was performed 72 h post-transfection. The cells were fixed for 10 min at 25°C using 3.7% paraformaldehyde in NAPS buffer (100 mM sodium phosphate, pH 7.4, 120 mM sucrose), washed three times, and incubated for 30 min at 25°C with NAPS buffer containing 1% milk powder and 0.5% Triton X-100. Incubation with monoclonal antibody M2 raised against the FLAG epitope (Sigma) was for 1 h at 25°C. Glass coverslips were washed three times, followed by incubation with fluorescein-conjugated goat anti-mouse IgG (Molecular Probes). After three wash steps, the cells were mounted using the Slow Fade Antifade kit (Molecular Probes). For selective permeabilization of the plasma membrane, Triton X-100 was replaced by digitonin at a concentration of 25 M or by 0.05% saponin. Samples were analyzed on a LSM 400 confocal microscope (Carl Zeiss). Images were analyzed using NIH Image 1.60b7 software.
To assess the percentage of stained cells in Triton X-100 versus digitonin-permeabilized cells, fields of well spread cells were chosen by using the transmitted light, total cells were counted, and the number of fluorescent cells was determined under UV light. Data represent mean Ϯ S.D. of at least five experiments from independent transfections.
Assay for 11␤-HSD-HEK-293 cells were transfected according to the calcium phosphate precipitation method (42). The medium was replaced by charcoal-treated DMEM 48 h later, and cells were harvested 72 h post-transfection. The cells were washed once with phosphate-buffered saline and centrifuged. The cell pellet was subjected to a freeze/thaw cycle and resuspended in TG1 buffer (20 mM Tris-HCl, pH 7.4; 100 mM NaCl; 1 mM EGTA; 1 mM EDTA; 1 mM MgCl 2 ; 20% glycerol), and activity assays were started immediately.
Oxidative activity of 11␤-HSD enzymes was measured by determin-ing the rate of conversion of corticosterone to 11-dehydrocorticosterone in the presence of NADP ϩ to analyze 11␤-HSD1 derivatives, or NAD ϩ , for 11␤-HSD2 derivatives. The reaction was started by mixing cell extract corresponding to 2-10 g of total proteins with 10 l of reaction mixture. The assay was performed in a final volume of 20 l containing 400 M NADP ϩ , 30 nCi of [ 3 H]corticosterone, and corticosterone at different concentrations ranging from 25 nM to 2 M. Samples were incubated for 10 -30 min at 37°C. The reaction was stopped by adding methanol and an excess of unlabeled steroids. Corticosterone and 11␤dehydrocorticosterone were separated by thin layer chromatography and analyzed as described (10). In measurements using whole cells, TG1 buffer was replaced by charcoal-treated DMEM. The cells were washed and resuspended in prewarmed (37°C) charcoal-treated DMEM. No cofactor was added for the whole cell assay. When measuring reductive activity of 11␤-HSD enzymes, NADPH (only in assays using extracts), [ 3 H]11␤-dehydrocorticosterone, and dehydrocorticosterone were used. Enzyme kinetics were analyzed by the Eadie-Hofstee linear transformation of the Michaelis-Menten equation. K m and V max values, calculated using the Lineweaver-Burk plot, were in line with the values obtained from the Eadie-Hofstee transformation. Constants were calculated by unweighted linear regression analysis with mean values of at least three experiments from independent transfections. Only conversion rates between 10 and 75% have been considered for calculation. Significance was tested by Student's unpaired t test.
Immunoblotting-To compare the expression level of the different constructs, cells were extracted with 0.5% Triton X-100 for 15 min on ice, centrifuged to remove debris, and an amount of 50 g of total proteins was separated by 12.5% SDS-PAGE. Proteins were transferred electrophoretically to nitrocellulose and incubated with the monoclonal antibody M2 raised against the FLAG epitope (Sigma). Antibody binding was visualized using the ECL Western detection system (Amersham Pharmacia Biotech). The data were analyzed by scanning densitometry using NIH Image 1.60b7 software.
In Vitro Transcription/Translation and Protease Protection Assay-The pcDNA3 plasmids carrying the relevant constructs were subjected to in vitro transcription by T7 RNA polymerase and translation in reticulocyte lysate in the presence of dog pancreas microsomes (TNT Quick system, Promega). In a typical assay, 500 ng of linearized plasmid DNA, 20 Ci of [ 35 S]methionine (Amersham Pharmacia Biotech), and 3 l of canine pancreatic microsomal membranes (Promega) were incubated for 60 min at 30°C in a final volume of 25 l. The reaction was chilled on ice, followed by addition of 10 mM Tris-CaCl 2 , pH 8.0. Translocation of polypeptides to the lumenal side of the microsomes was assayed by proteinase K treatment. An aliquot of 8 l from the in vitro transcription/translation product was incubated for 30 min on ice with 0.2 mg/ml proteinase K (Roche Molecular Biochemicals) in the presence or absence of 0.5% Triton X-100. Proteinase K was inactivated by the addition of 1 mM phenylmethylsulfonyl fluoride and subsequent boiling for 5 min. Proteins were subjected to 12.5% SDS-PAGE and analyzed by autoradiography and scanning. Data were analyzed using the NIH Image software (version 1.60b7).
Immunostaining using monoclonal antibody M2 specific for the FLAG epitope and FITC-labeled goat anti-mouse IgG revealed a pattern typical for localization to the ER membrane for both 11␤-HSD1 (1F) and 11␤-HSD2 (2F) (Fig. 1.). There was no indication of nuclear staining or staining at the plasma membrane. Cells were analyzed by confocal laser scanning microscopy 72 h post-transfection. There was no difference in the expression pattern when cells were stained 96 or 120 h post-transfection (data not shown). Densitometric analysis of over 100 transfected cells revealed that the fluorescent light from the area inside the nucleus was only 2.0 Ϯ 1.5% that of the light detected from the area containing the ER. A similar level of background staining was obtained in previous experiments using the same technique with the sarcoendoplasmic reticulum Ca 2ϩ -ATPase (SERCA1) and sarcolipin, two proteins known to be expressed exclusively in the ER membrane, with nuclear signals ranging from 1 to 3% (43). Furthermore, coexpression of 11␤-HSD2 and mineralocorticoid receptor in the presence of 5 nM aldosterone or 25 nM cortisol did not result in an increase of nuclear signal, with 1.8 Ϯ 1.4 and 2.0 Ϯ 1.6% of the signal detected from the area containing the ER, respectively (data not shown).
Orientation of 11␤-HSD1 and 11␤-HSD2 in the ER Membrane-To determine the topology of 11␤-HSD enzymes, we immunostained cells expressing constructs with a FLAG epitope attached either to the N terminus (F1 and F2) or to the C terminus (1F and 2F). Expression was detected using anti-FLAG antibody (M2) and FITC-labeled secondary antibody. Approximately 32% of total cells were stained after complete permeabilization of cells using 0.5% Triton X-100 (Fig. 2, A and C). The transfection efficiency was 32 Ϯ 2% (Table I). About the same number of fluorescent cells expressing constructs F1 (not shown) or 2F (Fig. 2D) were detected with anti-FLAG antibody after treatment with 25 M digitonin to permeabilize the plasma membrane selectively, leaving the ER membrane intact. Under these conditions anti-FLAG antibody was unable to interact with the FLAG epitope attached to the C terminus of 11␤-HSD1 (1F, see Fig. 2B) or to the N terminus of 11␤-HSD2 (F2, not shown), suggesting lumenal orientation of the FLAG epitope in 1F and F2 (Table I). Selective permeabilization of the plasma membrane was achieved at digitonin concentrations between 10 and 75 M (43). Similar results were obtained in experiments where the plasma membrane was selectively permeabilized with 0.05% saponin (data not shown). These results strongly suggest that the N-terminal sequence of 11␤-HSD1 faces the cytoplasm and the C-terminal part protrudes into the lumen of the ER. In contrast, the N terminus of 11␤-HSD2 is directed toward the lumenal compartment and the C terminus containing the catalytic moiety is facing the cytoplasm.
The fluorescence signal observed from construct F1 in the presence of Triton X-100 was considerably weaker than that from other constructs analyzed (not shown). Also, F1 could barely be detected in immunoblotting (Fig. 3.). In contrast, the fluorescence signal intensity in the presence of digitonin was comparable to that from other constructs. In addition, the enzyme activity per mg of total protein of homogenates from cells expressing F1 was similar to that of 1F or wild type 11␤-HSD1 (Table II). A possible explanation could be that interaction of detergent molecules like Triton X-100 or SDS with the hydrophobic amino acids in the N terminus of 11␤-HSD1 may prevent proper epitope recognition by the anti-FLAG antibody.
Attachment of the FLAG epitope to either the N or the C terminus did not affect the activity of 11␤-HSD1 or 11␤-HSD2 (Table II, Fig. 3.). Furthermore, all four constructs showed a pattern typical for localization to the ER membrane and thus did not affect intracellular localization.
The N-terminal Anchor Sequences of 11␤-HSD Enzymes Determine Their Orientation in the ER Membrane-To investigate if the N terminus of 11␤-HSD1 or 11␤-HSD2 is sufficient to determine its orientation in the ER membrane, we constructed chimeric proteins where the first 39 amino acids of 11␤-HSD1 and the first 87 amino acids of 11␤-HSD2 were exchanged. For the site of exchange we have chosen the two conserved residues threonine and glycine, whereby the latter is the first residue of the Gly-(Xaa) 3 -Gly-Xaa-Gly motif highly conserved in all members of the short chain dehydrogenase reductase family (41).
Analysis by confocal microscopy revealed that anti-FLAG antibody no longer recognized the FLAG epitope attached to the C terminus of the chimeric protein 12F, consisting of the N-terminal 39 amino acids of 11␤-HSD1 and amino acids 88 -405 of 11␤-HSD2, in cells where the plasma membrane was selectively permeabilized by 25 M digitonin (Table I). The FLAG epitope of 12F was accessible in the presence of Triton X-100. These experiments suggest that the C terminus of 12F, unlike that of 2F, is protruding into the lumen of the ER. The FLAG epitope of construct 21F, consisting of the N-terminal 87 amino acids of 11␤-HSD2 followed by amino acids 40 -292 of 11␤-HSD1, was accessible in the presence of 25 M digitonin, suggesting cytoplasmic orientation. This is different from construct 1F, where the C terminus is protected by the ER membrane.
To confirm further these findings, we performed protease protection assays of constructs that were in vitro transcribed and translated in the presence of dog pancreas microsomes (Fig. 4). Constructs 2F and F2 showed a band at 44 kDa, as expected, that was completely degraded after incubation with proteinase K. For both constructs 1F and F1 a band at 29 kDa and three additional bands of higher molecular mass (31, 33, and 35 kDa), corresponding to glycosylated products, were detected. 11␤-HSD1 contains three putative N-glycosylation sites at amino acids Asn 123 , Asn 162 , and Asn 207 . The contribution of each site to the observed glycosylated products was not investigated; however, in the absence of microsomes or after treatment with endo-␤-N-acetylglucosaminidase H (7 milliunits, 30 min, 37°C) only the 29-kDa band could be detected (data not shown). The protein products of 1F and F1 were resistant to degradation when the microsomes were treated with proteinase K. After disruption of the microsomes with the detergent Triton X-100, no protease-resistant material remained. The chimeric protein 12F was protected from degradation by proteinase K, whereas 21F was completely degraded.
These findings confirm the results obtained from confocal microscopy experiments that the N-terminal anchor of 11␤-HSD1 is sufficient to determine the orientation of the catalytic domain toward the ER lumen. In contrast, the N-terminal sequence of 11␤-HSD2 directs its catalytic moiety to the cytoplasm.
Expression of chimeric proteins 12F and 21F in HEK-293 cells yielded the expected size of bands at 39 and at 33 kDa, respectively (Fig. 3.). However, both constructs were catalytically inactive in measurements of oxidation and reduction of corticosterone using cell homogenates or in whole cell assays. These data support the finding of previous investigators that the N-terminal sequence is important for proper function of FIG. 1. Localization of 11␤-HSD1 and 11␤-HSD2 to the endoplasmic reticulum membrane. HEK-293 cells were transfected with 11␤-HSD1 tagged with the FLAG epitope at the C terminus (1F) (A) or with C-terminally FLAG-tagged 11␤-HSD2 (2F) (B). Immunostaining was performed 72 h post-transfection. The cells were labeled with mouse anti-FLAG antibody against the C-terminally attached FLAG epitope in the presence of 0.5% Triton X-100. Expression was visualized using FITC-conjugated goat anti-mouse IgG, and FITC fluorescence was detected at 525 nm.
The Effect of Mutations in the N Terminus of 11␤-HSD1 on Its Topology and Intracellular Localization-We investigated whether the tyrosine motif in the transmembrane helix of 11␤-HSD1 which is also present in a microsomal 50-kDa esterase/N-deacetylase (47) acts as an ER lumenal targeting sequence. Tyrosine residues 18 -21 of 11␤-HSD1 were mutated either to phenylalanine (Y18 -21F), to alanine (Y18 -21A), or the second and fourth tyrosine of the motif were mutated in an alternating way to alanine (Y19A,Y21A). Fig. 5 shows that all mutants were expressed in HEK-293 cells to a level comparable to that of 11␤-HSD1 (1F) or 11␤-HSD2 (2F). Analysis of the catalytic activity to oxidize corticosterone revealed no significant difference between mutant Y18 -21F and 1F. However, the V max of both mutant Y18 -21A and mutant Y19A,Y21A was reduced to about 20% compared with 1F, whereas the K m was unaltered (Table III).
By using confocal laser scanning microscopy, all three tyrosine mutants showed an expression pattern typical for localization to the ER membrane, and anti-FLAG antibody did not recognize the C-terminally attached epitope in the presence of 25 M digitonin as observed before for 11␤-HSD1 (1F). Further-  D). Immunostaining was carried out as indicated in Fig. 1. Cells were either completely permeabilized with 0.5% Triton X-100 (A and C) or the plasma membrane was selectively permeabilized using 25 M digitonin (B and D), allowing restricted access of the antibody to the cytosolic compartment.

TABLE I Topology of 11␤-HSD enzymes and chimeric constructs
Transfected HEK-293 cells were either completely permeabilized with 0.5% Triton X-100, or the plasma membrane was selectively permeabilized with 25 M digitonin, allowing restricted access of the antibody to the cytosolic compartment. Cells were incubated with an antibody against the FLAG epitope. Immunostaining was carried out as described under "Materials and Methods." Numbers represent the percentage of fluorescent cells relative to total cells from five independent experiments. In a typical experiment 300 -500 cells were counted. The transfection efficiency was 32 Ϯ 2%. Data represent mean Ϯ S.D. more, the tyrosine mutants were indistinguishable from 1F in protease protection assays, with the catalytic domain protruding into the microsomes (data not shown).
In order to investigate the role of the two positively charged amino acids Lys 5 and Lys 6 in the short cytoplasmic tail of 11␤-HSD1 on the topology and the intracellular localization, we replaced them by Ser or Arg. The expression level of all six mutants in HEK-293 cells was similar to that of 1F (Fig. 5). All mutants showed localization to the ER membrane and did not affect intracellular localization (not shown). The catalytic activity of homogenates from cells expressing the mutants was comparable to that of 1F, except that mutant K6R, for unknown reasons, had an increased K m (Table III). The activities of the lysine mutants were not significantly different from that of 1F when oxidation of corticosterone was measured on whole cells or when reduction of 11-dehydrocorticosterone was determined (Table IV).
Next, we analyzed if mutation of these lysines affects the orientation of 11␤-HSD1 in the ER membrane in HEK-293 cells. The cells were either completely permeabilized with 0.5% Triton X-100 or the plasma membrane was selectively permeabilized using 25 M digitonin. The orientation of the catalytic moiety of 11␤-HSD1 mutants was determined by detection of the FLAG epitope attached to the C terminus using confocal microscopy ( Table V). Replacement of residue Lys 6 to Ser or Arg had no effect on the orientation of 11␤-HSD1, and the epitope was not accessible to anti-FLAG antibody in the presence of digitonin. Also, mutation of Lys 5 to Arg or a double mutant where both lysines were changed to arginine (K5R,K6R) did not have any effect on the orientation. In contrast, anti-FLAG antibody recognized its epitope in mutants K5S and K5S,K6S, indicating cytoplasmic orientation of the  b For calculation of V max the amount of 11␤-HSD protein per mg of total proteins was determined by densitometric analysis of Western blots (Fig. 3), and the values of mutant proteins were compared to that of C-terminally tagged 11␤-HSD1 (1F) or 11␤-HSD2 (2F). V max of wild type proteins and of construct F1 is given in nM/h ⅐ mg of total protein and has not been determined by densitometric analysis of immunoblots using anti-FLAG antibody because of lacking or weak signals. c Measurements were below detection limit.

FIG. 4. The orientation of 11␤-HSD1 and 11␤-HSD2 in the ER membrane is determined by the N-terminal anchor sequence.
FLAG epitope-tagged 11␤-HSD1 (1F and F1), 11␤-HSD2 (2F and F2), and chimeric constructs (12F and 21F) were expressed in vitro in rabbit reticulocyte lysate in the presence of dog pancreas microsomes. Subsequently, microsomes were incubated in the absence (Ϫ) or presence (ϩ) of 0.2 mg/ml proteinase K (PK) and 0.5% Triton X-100 (T). Proteins were separated on 12.5% SDS-PAGE, and dried gels were subjected to autoradiography. The positions of protein size markers (kDa) are indicated on the left (upper panel). Models showing the orientation of 11␤-HSD enzymes in the ER membrane are given in the bottom panel. 11␤-HSD1 consists of six amino acids facing the cytoplasm (Cyt) followed by one transmembrane spanning helix and the catalytic moiety which protrudes into the ER lumen (Lum). The three glycosylation sites at Asn 123 , Asn 162 , and Asn 207 are indicated. 11␤-HSD2 consists of three amino acids on the luminal side, followed by a segment with three membrane spanning helices and the catalytic domain which faces the cytoplasm. The N-terminal anchor sequences of 11␤-HSD1 and 11␤-HSD2 are exchanged in chimeric proteins 12F and 21F and determine their orientation in the ER membrane. The FLAG epitope is indicated (F). A, constructs F2, 2F, and 12F; B, constructs 21F, 1F, and F1.

FIG. 5. Expression of mutants of 11␤-HSD1 in HEK-293 cells.
Mutants of 11␤-HSD1 with a FLAG epitope attached to the C terminus were expressed in HEK-293 cells, and immunoblotting was carried out as indicated in Fig. 3. catalytic domain. We further investigated these findings by in vitro transcription of mutant cDNA, translation in the presence of microsomes, and subsequent subjection to protease protection assays (Fig. 6.). Mutants K5S and K5S,K6S were completely degraded, and the bands from the glycosylated proteins were missing, confirming the cytoplasmic orientation of the catalytic domain observed in immunohistochemistry. In contrast to the findings from microscopy experiments, both mutants K5R and K5R,K6R were mostly degraded, suggesting inversion of the orientation of the catalytic domain toward the cytoplasm. However, both mutants were partially protected, indicated by the presence of glycosylated proteins and by protected proteins upon treatment with proteinase K. A rough estimation of the amount of protected protein from mutants K5R and K5R,K6R was obtained by densitometric analysis from three independent experiments and was in the range of 10 -30%. Mutants K6S and K6R were, like 11␤-HSD1, protected by the microsomal membrane from degradation by proteinase K, suggesting lumenal orientation. DISCUSSION Human embryonic kidney cells (HEK-293) have proven to be a suitable system for the transient expression of epitope-tagged constructs of 11␤-HSD enzymes, since there is no endogenous 11␤-HSD1 or 11␤-HSD2 activity (Ͻ1% of transfected cells, see Ref. 33). Conditions to permeabilize the plasma membrane of HEK-293 cells selectively by the use of digitonin or saponin were established previously, and it was shown that the antibody raised against the FLAG epitope did not recognize any other proteins in untransfected cells (43).
Immunostaining of HEK-293 cells transfected with 11␤-HSD1 or 11␤-HSD2 tagged with the FLAG epitope at the C terminus revealed very similar expression patterns for both proteins, with high expression in the ER membrane system and the nuclear envelope but no expression in the nucleus or at the plasma membrane (Fig. 1). This expression pattern is highly similar to that observed in an earlier study for the sarcoendoplasmic reticulum Ca 2ϩ -ATPase (SERCA1) and for sarcolipin (43), two proteins known to localize exclusively to the ER membrane. Restricted expression of 11␤-HSD1 to the ER membrane is in line with the finding that 11␤-HSD1 activity was detected

TABLE V Effect of mutations on the topology of 11␤-HSD1
Transfected HEK-293 cells were permeabilized with 0.5% Triton-X 100 or 25 M digitonin and immunostaining was carried out as indicated in table I. All mutants were epitope-tagged at the C-terminus and detected by an antibody against the FLAG epitope.

Mutant of 11␤-HSD1
Permeabilization conditions Triton X-100 Digitonin 6. Mutation of residue Lys 5 to Ser causes 11␤-HSD1 to insert with an inverted N ext -C cyt topology into microsomes. Mutants of 11␤-HSD1 tagged with the FLAG epitope at the C terminus were expressed in vitro in rabbit reticulocyte lysate in the presence of dog pancreas microsomes. Microsomes were incubated in the absence (Ϫ) or presence (ϩ) of 0.2 mg/ml proteinase K (PK) and 0.5% Triton X-100 (T). Proteins were separated by 12.5% SDS-PAGE. Protein size markers (kDa) are indicated on the left of the autoradiograph (upper panel). Mutants K6S and K6R adopt wild type topology, with the catalytic domain facing the ER lumen (Lum) and showing the typical glycosylation products. Cyt, cytochrome. In contrast, mutants K5S and K5S,K6S show inverted insertion into the ER membrane (bottom panel). A, mutants K6S, K5S, and K5S,K6S. The mutation of Lys to Ser is indicated by a square. B, mutants K6R, K5R, and K5R,K6R. The mutation of Lys to Arg is indicated by a circle. only in microsomal but not nuclear fractions obtained from human decidua (34). However, reports on the subcellular localization of 11␤-HSD2 are controversial.
Naray Fejes-Toth et al. (26) reported exclusive localization to the ER membrane and to the nuclear envelope in microscopy studies using a 11␤-HSD2 green fluorescent protein chimera that was expressed in either Chinese hamster ovary cells or in Madin-Darby canine kidney cells or in a study using a mouse polyclonal 11␤-HSD2 antibody to stain rabbit kidney sections (28). In contrast, Stewart and co-workers (32,33) reported that approximately 40% of the total immunostaining observed in renal collecting ducts and colonic epithelial cells is nuclear in origin and has an intranuclear localization. In fractionation assays they found 11␤-HSD2 activity in the nuclear fraction of colon, an MR target tissue, but not placenta (34). They suggest that the MR may be responsible for the intranuclear localization of 11␤-HSD2. However, we did not observe any difference in 11␤-HSD2 localization to the ER membrane in our experiments where MR and 11␤-HSD2 were coexpressed in HEK-293 cells in the absence or presence of aldosterone or cortisol. The presence of 11␤-HSD2 in the nuclear fraction of cell fractionation assays can be attributed to the expression of 11␤-HSD2 in the outer nuclear membrane, especially in the early stage of protein synthesis. We have, however, no explanation for their observation of soluble 11␤-HSD2 inside the nucleus. Our results demonstrate clearly that the N-terminal sequences of both 11␤-HSD1 and 11␤-HSD2 traverse the ER membrane. The bulk of 11␤-HSD1 is oriented to the ER lumen, whereas its N terminus is cytoplasmic. 11␤-HSD2 adopts an opposite topology with the N terminus on the lumenal side and the catalytic domain of the protein on the cytoplasmic side. This is supporting a model of 11␤-HSD2 with three transmembrane helices (45,48). If 11␤-HSD2 were to be soluble inside the nucleus, cleavage of the N-terminal anchor would have to occur. In that case, we would expect staining of the ER membrane and nuclear envelope with our construct F2, but nuclear staining should be detectable with the C-terminally tagged construct 2F. Also, attempts to make a truncated construct without the N-terminal transmembrane part to achieve better solubility of 11␤-HSD2 for the purpose of purification failed so far, since these constructs lost activity (45,46). 2 An influence of the FLAG epitope on the intracellular localization is unlikely since almost identical staining patterns were observed, no matter if the FLAG epitope was attached to the N or to the C terminus.
The signals that direct 11␤-HSD1 and 11␤-HSD2 to the ER membrane are still unknown, and none of the two proteins contains typical ER retrieval signals. ER retention signals often consist of a cluster of basic residues, e.g. in C-terminal di-lysine motifs or in N-terminal di-arginine motifs (for a review see Ref. 49). 11␤-HSD2 contains three regions that may be of interest, residues 278 -280 (KRK), 333-337 (RPRRR), and 359 -361 (RRR). Mutation of Arg 337 to Cys, which was found in apparent mineralocorticoid excess patients (21) and whose functional defect is not understood yet, did not affect subcellular localization. 2 A study by Ozols (47) proposed that a tyrosine motif present in the transmembrane helix of 11␤-HSD1 and of a 50-kDa esterase/N-deacetylase could act as a ER-targeting signal. When we replaced tyrosine residues 18 -21 by either phenylalanine or alanine, targeting to the ER was not affected, indicating that the tyrosine motif is not essential for targeting.
11␤-HSD1 contains a di-lysine motif, consisting of residues Lys 5 and Lys 6 , that is in a distance from the N terminus typically found in di-arginine motifs of type II membrane proteins (49). Therefore, we investigated whether these two basic residues could act as an ER-targeting signal. Mutation of Lys 5 or Lys 6 to Ser did not affect subcellular localization, showing that the di-lysine motif of 11␤-HSD1 does not act as an ERtargeting signal. This is supported by the fact that the FLAG epitope in F1 did not change the ER localization pattern even though it changes the distance of the di-lysine motif to the N terminus (which is critical for the function of di-arginine motifs). The ER retention signals known to date reside on the cytoplasmic face of proteins. However, since neither the dilysine motif nor the tyrosine motif seem to be essential for localization to the ER membrane, we propose that the ERtargeting signal of 11␤-HSD1 is on the lumenal side.
Our experiments using selectively permeabilized cells and immunostaining together with protease protection assays demonstrated that 11␤-HSD1 and 11␤-HSD2 insert with an opposite topology into the ER membrane. Using chimeric proteins, where the membrane anchors of both proteins were exchanged just upstream of the beginning of the conserved cofactor binding site, demonstrated that the transmembrane regions are sufficient to determine their topology. The chimeric protein 12F became protected by the ER membrane-like 1F, whereas the domain of 21F adopted cytoplasmic orientation, like 2F. This is supported by the lack of glycosylation products of 21F. Although 11␤-HSD2 contains a putative Asn-Xaa-Ser motif at position 394 -396 that could serve as a glycosylation site, no glycosylation products were detected for 12F, suggesting that this site is not accessible in the native conformation.
According to the generally accepted "positive inside" rule (50 -54), the topology of 11␤-HSD2 is probably determined by the three positively charged arginine residues in the short cytoplasmic loop between the proposed helices 1 and 2 and by the three arginine residues following the third transmembrane helix. 11␤-HSD1 has only a single N-terminal transmembrane segment, and it represents a suitable model to study the determinants of membrane protein topology. The charge distribution at the membrane suggests that the two positively charged lysine residues on the cytoplasmic side and the two negatively charged glutamate residues are critical for the orientation of 11␤-HSD1 in the ER membrane. To evaluate the importance of residues Lys 5 and Lys 6 , we studied mutants where lysine was changed either to arginine, leaving the positive charge intact, or to serine. Mutation of Lys 6 did not affect the topology of 11␤-HSD1. In contrast, mutant K5S and K5S,K6S adopted an inverted orientation in the ER membrane. This is shown by the accessibility of the FLAG epitope at the C terminus of the catalytic domain to its antibody in cells where the plasma membrane was selectively permeabilized with digitonin (Table  V). In addition, the bulk of the protein K5S or K5S,K6S was no longer protected by the microsomal membrane in protease protection assays. The cytoplasmic orientation of mutants K5S and K5S,K6S is supported by the lack of glycosylation products upon in vitro transcription and translation in the presence of microsomes. Mutant K5R and double mutant K5R,K6R expressed in HEK-293 cells were directed to the ER lumen, and immunostaining patterns were similar to that of 1F. However, protease protection assays revealed only partial protection of mutants K5R and K5R,K6R by the microsomal membrane. The protected bands represented approximately 10 -30% of the total protein synthesized, indicating that mutant proteins were inserted in both orientations but predominantly adopted cytoplasmic orientation. As a possible explanation for that discrepancy, a mechanism that controls insertion into the membrane may exist in HEK-293 cells which is absent or not fully functional in the rabbit reticulocyte lysate.
Our results suggest that residue Lys 5 is a critical determinant of the topology of 11␤-HSD1. Arginine can only partially complement the effect of lysine, indicating that both the charge and the side chain of Lys 5 are essential for proper insertion into the membrane. It is interesting to note that mutation of Lys 6 had no effect on the topology. Whereas Lys 5 of 11␤-HSD1 is conserved in all different species known to date, Lys 6 is replaced in squirrel monkey by a threonine residue accompanied by a histidine residue. That basic residues have a clear effect on membrane orientation of a number of proteins with a single N-terminal transmembrane segment has been shown by other investigators (55)(56)(57)(58)(59)(60)(61)(62)(63). All mutants of the di-lysine motif showed catalytic activity comparable to that of the wild type enzyme, except that mutant K6R had a slightly increased K m for the oxidation of corticosterone (Table III).
Since K5S and K5S,K6S are oriented to the cytoplasm and are not glycosylated, we also measured dehydrogenase and oxidoreductase activity using the substrates corticosterone and 11-dehydrocorticosterone in whole cell assays (Table IV). Although previous reports indicated that 11␤-HSD1 functions predominantly as an oxidoreductase in vivo (6 -8), the dehydrogenase showed, for unknown reasons, lower K m values than the oxidoreductase in the HEK-293 expression system. The activity of mutant K5S was similar to that of the wild type enzyme (Table IV). These results suggest that glycosylation does not significantly alter the activity of 11␤-HSD1, which is supported by the report of Ozols (9) that endo-␤-N-acetylglucosaminidase H treatment of purified 11␤-HSD1 did not significantly affect activity. On the other hand, other investigators reported decreased 11␤-HSD1 activity following incubation with tunicamycin, an inhibitor of glycosylation, or of proteins containing mutated glycosylation sites (4,64).
In our experiments using intact HEK-293 cells, inversion of the orientation of the 11␤-HSD1 mutant K5S did not significantly alter its K m value (Table IV), suggesting that similar concentrations of glucocorticoids were reached in the cytoplasm and in the ER lumen in these cells. This may be different in a specific tissue, in vivo, where mechanisms may exist for the control of intracellular glucocorticoid concentrations. The fact that the catalytic activity in mutants K5S and K5S,K6S was not affected by inversion of the orientation in the ER membrane also indicates that the membrane itself does not have an important impact on the activity of 11␤-HSD1. Recently, 11␤-HSD1 was expressed in yeast (65) and 11␤-HSD2 in bacteria (46). In both cases the composition of the membrane differs from that of the corresponding native tissue, but the kinetic parameters were not significantly different.
Furthermore, our results emphasize that the transmembrane region is critical for 11␤-HSD function. The N termini of 11␤-HSD1 and 11␤-HSD2, when exchanged immediately upstream of the conserved cofactor binding site, could not functionally complement each other, and both chimera were inactive. N-terminal deletion mutants of 11␤-HSD2 and a translation product of a short transcript of 11␤-HSD1, starting at Met 27 were not catalytically active (44). Also, mutation of the tyrosine residues in the transmembrane segment of 11␤-HSD1 to alanine resulted in a loss of V max . These results emphasize the essential role of the N-terminal membrane spanning region for proper function.
The impact of the differential topology of 11␤-HSD1 and 11␤-HSD2 and their restricted localization to the ER membrane on their biological functions is not fully understood. The mineralocorticoid aldosterone and the glucocorticoids cortisol and corticosterone have similar affinities to bind to the MR (8, 66, 67). Activation of the receptor occurs at low nanomolar concentrations. 11␤-HSD2 mediates access of aldosterone to the MR by inactivating glucocorticoids in the low nanomolar range (K m about 5 nM). One might anticipate that for efficient protection of the MR from glucocorticoid binding, 11␤-HSD2 should be localized in close proximity to the MR and should adopt a cytoplasmic orientation. Mutations that would alter intracellular localization or topology could, therefore, lead to a loss of the protective function of 11␤-HSD2 on the MR even if the catalytic activity of 11␤-HSD2 would not be significantly affected. On the other hand, 11␤-HSD1 should not be able to modulate access of glucocorticoids to the MR efficiently, since its catalytic domain is facing the ER lumen. In addition, the K m of 11␤-HSD1 for glucocorticoids is in the high nanomolar to micromolar range. Such high concentrations may be reached in tissues where glucocorticoids are metabolized, like in the liver where 11␤-HSD1 is highly expressed. The catalytic domain of 11␤-HSD1 is protruding into the ER lumen where glucocorticoids may be metabolized and concentrated for subsequent secretion.