Functional expression, characterization, and purification of the catalytic domain of human 11-beta -hydroxysteroid dehydrogenase type 1.

11-beta-hydroxysteroid dehydrogenase type 1 catalyzes the conversion of cortisone to hormonally active cortisol and has been implicated in the pathogenesis of a number of disorders including insulin resistance and obesity. The enzyme is a glycosylated membrane-bound protein that has proved difficult to purify in an active state. Extracted enzyme typically loses the reductase properties seen in intact cells and shows principally dehydrogenase activity. The C-terminal catalytic domain is known to contain a disulfide bond and is located within the lumen of the endoplasmic reticulum, anchored to the membrane by a single N-terminal transmembrane domain. We report here the functional expression of the catalytic domain of the human enzyme, without the transmembrane domain and the extreme N terminus, in Escherichia coli. Moderate levels of soluble active protein were obtained using an N-terminal fusion with thioredoxin and a 6xHis tag. In contrast, the inclusion of a 6xHis tag at the C terminus adversely affected protein solubility and activity. However, the highest levels of active protein were obtained using a construct expressing the untagged catalytic domain. Nonreducing electrophoresis revealed the presence of both monomeric and dimeric disulfide bonded forms; however, mutation of a nonconserved cysteine residue resulted in a recombinant protein with no intermolecular disulfide bonds but full enzymatic activity. Using the optimal combination of plasmid construct and E. coli host strain, the recombinant protein was purified to apparent homogeneity by single step affinity chromatography. The purified protein possessed both dehydrogenase and reductase activities with a K(m) of 1.4 micrometer for cortisol and 9.5 micrometer for cortisone. This study indicates that glycosylation, the N-terminal region including the transmembrane helix, and intermolecular disulfide bonds are not essential for enzyme activity and that expression in bacteria can provide active recombinant protein for future structural and functional studies.

In mammalian tissues, two isozymes of 11-␤-hydroxysteroid dehydrogenase (11␤-HSD) 1 catalyze the interconversion of hor-monally active C 11 -hydroxylated corticosteroids (cortisol, corticosterone) and their inactive C 11 -keto metabolites (cortisone, 11-dehydrocorticosterone). The 11␤-HSD1 and 11␤-HSD2 isozymes share only 14% identity and are separate gene products with different physiological roles, regulation, and tissue distribution (1). 11␤-HSD2 is a high affinity NAD-dependent dehydrogenase that protects the mineralocorticoid receptor from glucocorticoid excess; mutations in the HSD11B2 gene explain an inherited form of hypertension, the syndrome of apparent mineralocorticoid excess in which cortisol acts as a potent mineralocorticoid (2). By contrast, 11␤-HSD1 is a relatively low affinity NADP-dependent enzyme that acts predominantly as a reductase in vivo, although extracted enzyme typically shows predominant dehydrogenase activity (3). By converting cortisone to cortisol, 11␤-HSD1 facilitates glucocorticoid hormone action in key target tissues such as liver and adipose tissue, and as such has been implicated in a number of disorders including insulin resistance and central obesity (4,5).
11␤-HSD1 is a member of the short chain alcohol dehydrogenase family, also known as the short chain dehydrogenase/ reductases (SDRs). SDRs typically exhibit residue identities only at the 15-30% level, indicative of early duplicatory origins and extensive divergence (6 -8). However, in contrast to other SDR members, 11␤-HSD1 is unusual in possessing a single transmembrane helix at the N terminus. This is intrinsic to the endoplasmic reticulum (ER) membrane, with a short 5-amino acid N-terminal region on the cytosolic side and the main catalytic domain of the protein facing the lumen of the ER (9,10). The importance of the transmembrane domain on 11␤-HSD1 activity has been studied but with inconclusive results. An N-terminally truncated variant of rat 11␤-HSD1 was expressed in COS cells and reported to be inactive (11,12). However, this construct encoded a protein that had lost more than just the transmembrane helix and therefore may have lost vital parts of the enzymatic domain. In addition, because the expression studies were performed in COS and Chinese hamster ovary cells, the truncated protein would have been targeted (because of the lack of signal sequence) to the cytosol and not the ER. The lumen of the ER promotes the formation of disulfide bonds, and studies have indicated that there are important intrachain disulfide bonds within the 11␤-HSD1 protein (9).
The catalytic domain is glycosylated (13)(14)(15), which is in agreement with a lumenal orientation. Experiments to resolve the importance of glycosylation have also yielded varying results. Enzymatic deglycosylation of rabbit (9) and human (13) 11␤-HSD1 has indicated that glycosylation is not important for enzyme activity. However, partial inhibition of glycosylation of the rat enzyme by tunicamycin decreased dehydrogenase activity but not reductase activity (14), and mutation of the rat (15) and human (13) sequences at putative N-glycosylation sites resulted in reduced or abolished activity.
Expression of human (and squirrel monkey) clones of 11␤-HSD1 has been achieved in COS cells (11,12), HEK cells (16), and the yeast Pichia pastoris (13,17) using a variety of vectors. This has led to ambiguous kinetic results with over 10-fold variation in K m values and often significant differences in activity between whole cells and lysates. These systems have not yielded large amounts of pure recombinant protein, and no structural information has come from them. Overexpression of 11␤-HSD1 in bacterial cells has been reported (17), but the resulting protein was inactive. Failure to obtain activity was attributed to either insolubility of the protein, and subsequent refolding problems, or a lack of glycosylation. In this study we sought to maximize the production of soluble recombinant human 11␤-HSD1 within Escherichia coli by varying the expression construct, the host strain, and the incubation conditions. In particular, because 11␤-HSD1 is thought to contain disulfide bonds, we have assessed the value of E. coli strains that promote disulfide bond formation within the cytoplasm of the bacterium through mutations in the genes encoding thioredoxin reductase and/or glutathione reductase. We also tested the effect of thioredoxin fusions, histidine tags, glycosylation status, the presence of the transmembrane domain, and mutation of a nonconserved cysteine residue on the activity of human 11␤-HSD1. Through these measures, we arrived at an optimal construct and E. coli host combination for producing sufficient protein for purification.
Production of 11-␤-Hydroxysteroid Dehydrogenase 1 Expression Vectors-The human 11␤-HSD1 cDNA (3) was subcloned into pCDNA3.1 (16). This was further subcloned into pET21b(ϩ) and pET32b(ϩ) E. coli expression vectors (Novagen). Four distinct constructs bearing alterations to the N and C termini were synthesized (Fig. 1). Modifications to the inserts in the pET21b(ϩ) vector were made to truncate the hydrophobic N terminus of 11␤-HSD1, both with (pET21CDH) and without (pET21CD) a C-terminal 6-histidine tag (6xHis) sequence. Each of these modifications also introduced an NheI restriction site coincident with the ATG start codon. The pET32b(ϩ) vector was used to produce an N-terminal fusion protein between the vector-encoded thioredoxin gene (TrxA) and either the full-length 11␤-HSD1 gene (pET32FL) or an N-terminally truncated version containing only the catalytic domain (pET32CD). This strategy also incorporated an Nterminal 6xHis tag to assist purification and introduced an NcoI site coincident with the ATG start codon. The original full-length cDNA in pCDNA3.1 was used as template for polymerase chain reaction amplification and the forward primers (pET32CD 5Ј, pET32FL 5Ј, and pET21 5Ј; see Table I) in conjunction with the appropriate reverse primers (which allowed the introduction of an XhoI restriction site downstream). The resulting polymerase chain reaction products were subcloned into pGEM-T Easy vector (Promega). After digestion with the appropriate restriction enzymes, fragments were gel-purified and li-gated with the appropriate pET expression vector to give the final constructs (Fig. 1). The direction and nucleotide sequence of the inserted cDNAs were confirmed by sequencing. For expression studies the plasmids were subcloned into the E. coli strains BL21(DE3), AD494(DE3), and Origami(DE3) (Novagen).
Site-directed Mutagenesis-A mutation was introduced into the expression construct pET21CD by polymerase chain reaction using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Primers were designed to mutate the cysteine at position 272 of the human type 1 sequence to the corresponding residue of 11␤-HSD1 from the squirrel monkey, namely a serine (18). The oligonucleotides used were 5Ј-CAGAAATCCATCCAGGAAGATC-3Ј and 5Ј-GATCTTCCTG-GATGGATTTCTG-3Ј (the mutated nucleotides are in boldface). The resulting construct (designated pET21CD-C272S) was verified by sequencing.
Expression of Recombinant 11␤-HSD1-Overnight cultures of E. coli expressing the pET constructs in LB medium containing 50 g/ml carbenicillin (for host strain BL21(DE3)), supplemented with 15 g/ml kanamycin for AD494(DE3) and Origami(DE3), and 12.5 g/ml tetracycline for Origami(DE3) were seeded (0.1%) into fresh LB medium containing appropriate antibiotics and then incubated at 37°C for 6 h (corresponding to an approximate absorbance at 600 nm of 1.0). Expression of recombinant 11␤-HSD1 was then induced by adding 1 mM isopropyl-␤-D-thiogalactopyranoside (IPTG). Control incubations without IPTG induction were also performed. Incubation was continued for 16 h, at 15°C unless otherwise stated, with shaking at 230 rpm.
Preparation of Cleared Lysates-Bacterial cultures from induced and uninduced E. coli cultures were pelleted by centrifugation. The cells were disrupted by resuspension in BugBuster reagent (Novagen) containing protease inhibitors (Mini-Complete EDTA-free, Roche Molecular Biochemicals) and benzonase DNase (Novagen) using 50 l of lysis reagent/ml of original culture. For fuller lysis of the cells, lysozyme was included to a final concentration of 200 g/ml. After incubation at room temperature with shaking for 25-30 min, the cell debris was pelleted by centrifugation at 11,000 ϫ g for 10 min. The supernatant was removed for activity assays and protein determination (Bio-Rad protein reagent).
Activity Assays-Cleared lysates and other enzyme preparations (typically 1-10 l) were incubated in 0.5 ml of phosphate buffer (0.1 M, pH 7.6) containing 50,000 cpm [ 3 H]cortisol, 100 nM unlabeled cortisol, and 200 M NADP for 30 min at 37°C to assess dehydrogenase activity or [ 3 H]cortisone (generated as reported previously (19)), 100 nM unlabeled cortisone, 200 M NADPH, and a regeneration system (10 mM MgCl 2 , 5 mM glucose-6-phosphate, and 10 units glucose-6-phosphate dehydrogenase) (20) to assess reductase activity. Steroids were partitioned into 10 volumes of dichloromethane and separated by TLC using ethanol/chloroform (8:92) as the mobile phase. The TLC plates were analyzed on a Bioscan radioimaging detector, and the fractional conversion of cortisol to cortisone or cortisone to cortisol was used to estimate enzyme activities.
Activity was also assessed in intact E. coli cells. Bacterial cultures (1 ml) were centrifuged, and the resulting pellet was resuspended in 0.5 ml of phosphate buffer (0.1 M, pH 7.6) containing either 100 nM cortisol plus [ 3 H]cortisol tracer, to assess the levels of dehydrogenase activity, or 0.5 ml of phosphate buffer (0.1 M, pH 7.6) containing 100 nM cortisone and [ 3 H]cortisone tracer, to measure reductase activity.
Kinetic Analysis-Enzyme activities were assayed in the standard reaction mixture containing cleared lysates or purified protein, the appropriate cofactor (NADP or NADPH plus the regeneration system), and varying substrate concentrations (0.25-60 M). In each case, linearity of enzyme activity versus time was ensured. K m value estimations were averaged from Lineweaver-Burke plots derived from three experiments as previously reported (21).
Western Blot Analysis-SDS-PAGE was performed using the Laemmli method (22) with 12.5% acrylamide minigels using a Bio-Rad

Expression of 11-␤-Hydroxysteroid Dehydrogenase in E. coli
Mini-Protean II apparatus. 10 g of protein from bacterial cleared lysates, human liver homogenates, or mouse liver homogenates (produced as reported previously (23,24)) were loaded either in sample buffer containing ␤-mercaptoethanol, to completely reduce any disulfide bonds, or in sample buffer without ␤-mercaptoethanol, to retain the disulfide bonds. Gels were stained with Coomassie Brilliant Blue (R-250) to investigate the purity and amount of protein in the extracts. Western blotting was performed as reported previously (23). Briefly, after electrophoresis, the proteins were transferred to Immobilon-P membranes (Millipore Corp., Bedford, MA). Nonspecific protein binding was blocked by incubating the membranes in 20% nonfat milk and 0.1% Tween 20 in phosphate-buffered saline at 25°C for 1 h. The membranes were then incubated with a validated polyclonal antibody to human 11␤-HSD1 (The Binding Site, Birmingham, UK) at a dilution of 1:1000 for 16 h at 4°C. After three 10-min washes in phosphate-buffered saline/0.1% Tween 20, the membranes were incubated with a secondary antibody (goat anti-sheep IgG peroxidase conjugate (The Binding Site)) at a dilution of 1:75,000 for 1.5 h at room temperature. Bound peroxidase-conjugated Ig was visualized using an ECL detection kit (Amersham Pharmacia Biotech) by exposing the membranes to x-ray film (Kodak). Relative band intensities were analyzed by laser scanning densitometry.
Purification of Recombinant 11␤-HSD1-A further construct was designed for the purification of recombinant 11␤-HSD. The plasmid pET21CD was digested with NcoI and XhoI to release the 11␤-HSD insert and ligated to the similarly digested pET28b(ϩ) vector (Novagen). This gave the construct pET28HCD, which contained the catalytic domain of 11␤-HSD with an N-terminal 6xHis tag to aid purification. For expression, the construct was subcloned into E. coli strain BL21(DE3) and induced as described above. Cleared lysate (8 ml) was prepared from 200 ml of induced culture and mixed with 0.3 ml of nickel-nitrilotriacetic acid His-bind resin (Novagen) followed by gentle shaking at room temperature for 30 min. The mixture was loaded into an empty column, and unbound protein was removed by washing with 8 ml of 50 mM NaPO 4 buffer, pH 8, containing 300 mM NaCl. Bound protein was eluted in a stepwise fashion with the same buffer containing increasing concentrations of imidazole (2 ϫ 0.2 ml each of 40, 60, and 100 mM imidazole). The eluted fractions were tested for enzyme activity, and the purity was assessed by SDS-PAGE.

Expression of Different Variants of 11-␤-Hydroxysteroid
Dehydrogenase 1 in E. coli-Four constructs were designed ( Fig.  1) to test the effect of (a) removing the N-terminal transmembrane region of 11␤-HSD1, (b) including a 6xHis tag at the C terminus, and (c) fusing the N terminus of the catalytic domain to thioredoxin, a procedure reported to increase the solubility of recombinant proteins (25), particularly those requiring disulfide bonds. Initial comparisons used a thioredoxin reductasedeficient strain of E. coli, AD494(DE3), that has been reported to enhance the formation of disulphide bonds within the bacterial cytoplasm (26), particularly for thioredoxin fusion proteins. In addition, because initial tests indicated that very little soluble protein was produced by incubation entirely at 37°C, cultures were switched to 15°C after the addition of IPTG to increase the possibility of producing soluble protein (26). The overall level of expression from each of the four constructs under these conditions was estimated using SDS-PAGE gels and subsequent densitometric analysis. All constructs produced protein of the expected size, with levels varying from 15% of the total cell protein for the thioredoxin full-length 11␤-HSD1 fusion (pET32FL) to 35% for the nontagged catalytic domain (pET21CD).
The extraction of proteins by sonication indicated that some of the recombinant protein was in a soluble form (data not shown). Despite this, no enzyme activity could be detected in the sonicated lysates. However, incubation of intact bacterial cells (transformed with the pET32CD construct) with labeled cortisol and cortisone indicated that the bacteria had acquired both 11-␤-hydroxysteroid dehydrogenase and 11-oxo reductase activities (Fig. 2). This was specific to IPTG-induced cultures and hence indicative of the presence of active recombinant enzyme within the cells prior to sonication. Moreover, a gentler lysis of the cells by a mixture of a commercial detergent (Bug-Buster) and lysozyme produced an extract with clear 11␤-HSD activity for some of the constructs (Fig. 3). Examination of these extracts and the remaining pellets by SDS-PAGE indicated that all the constructs were expressing protein that could be detected easily in the pellets (Fig. 3B) but was not easily visible on Coomassie staining of the supernatants (Fig. 3A). However, Western immunoblots of the supernatant samples using a specific antiserum to 11␤-HSD1 clearly showed that soluble recombinant protein was being expressed for at least three of the constructs (Fig. 3C), with the order of expression being pET21CD (catalytic domain alone) Ͼ pET32CD (thioredoxin catalytic domain fusion) Ͼ pET21CDH (catalytic domain ϩ His tag). No expression was observed with the pET32FL (thioredoxin full-length fusion) construct. Production of active recombinant 11␤-HSD1 was assessed by measuring the ability of cleared lysates to convert cortisol to cortisone (11-␤-hydroxysteroid dehydrogenase reaction). Comparison of these enzyme activities (Fig. 3D) indicated that the construct encoding the catalytic domain alone (pET21CD) produced the highest levels of activity, concomitant with the highest levels of protein observed on the Western blots. Interestingly, for this construct the regulation of protein expression was poor, with similar protein and enzyme activity being observed in the presence and absence of IPTG. Fusion of the catalytic domain with thioredoxin (pET32CD), which included an intervening N-terminal 6xHis tag, also resulted in active soluble protein, albeit at a lower level than that seen with the catalytic domain alone. However, the addition of a C-terminal 6xHis tag (pET21CDH) resulted in almost no activity despite soluble protein being evident on Western blot analysis. Inclusion of the transmembrane domain in the thioredoxin fusion (pET32FL) resulted in no enzyme activity, which is in agreement with a lack of detectable protein on the blots. Additional constructs were pro- duced that encoded the full-length 11␤-HSD1 sequence in pET21b(ϩ) (i.e. not as a fusion), but these constructs repeatedly failed to express any recombinant protein in either a soluble or insoluble form (data not shown).
Expression of 11-␤-Hydroxysteroid Dehydrogenase 1 in Different Strains of E. coli-The two constructs that demonstrated enzyme activity (pET21CD and pET32CD) were then used to compare the effect of E. coli host strain on the production of active protein, with the particular purpose of comparing strains reputed to enhance disulfide bond formation by mutations in either thioredoxin reductase or glutathione reductase genes (26). Thus the thioredoxin reductase-deficient strain, AD494(DE3), was compared with the thioredoxin reductaseand glutathione reductase-deficient strain Origami(DE3) and a strain deficient in neither enzyme, BL21(DE3). Surprisingly, for the pET21CD construct, both activity assays (Fig. 4C) and Western analyses (Fig. 4B) clearly indicated that BL21(DE3) cells gave better protein expression and activity than AD494(DE3) or Origami(DE3), with the latter giving negligible levels of both protein and activity. As before (Fig. 3), the thioredoxin fusion construct pET32CD yielded lower amounts of protein and enzyme activity compared with the nonfusion con- Blue-stained SDS-PAGE gel of the pellet fractions of 11␤-HSD1 expression constructs. Recombinant 11␤-HSD1 from pET21 constructs is clearly visible at ϳ29 kDa (pET21-CD) and 30 kDa (pET21-CDH), whereas those for the pET32 constructs are at 45 kDa (pET32-CD) and 47 kDa (pET32-FL) because of the presence of the TrxA gene. C, Western blot of the supernatant fractions of 11␤-HSD1 expression constructs showing soluble protein was produced from pET21CD, pET21CDH, and pET32CD. D, activity data (cortisol to cortisone) for the 11␤-HSD1 expression constructs showing highest levels of activity for pET21CD. No activity was obtained from pET32FL, which is coincident with no protein expression. struct pET21CD, although in this case the host strain had much less effect on the level of expression (Fig. 4), with all three strains giving similar results. Examination of Coomassiestained SDS-PAGE gels indicated that the levels of soluble recombinant proteins in the lysates were very low (Fig. 4, A and  B), with only faint bands being evident in the highest expressing combinations of construct and host strain. Using E. coli strains BL21(DE3) and AD494(DE3), a comparison was made between protein extraction produced with detergent (Bug-Buster) alone and detergent with the addition of lysozyme, the latter combination having been shown microscopically to cause complete lysis of the cells and to release approximately twice the amount of protein from the bacterial cells when compared with detergent alone (data not shown). This analysis showed that almost all the active 11␤-HSD1 could be released by detergent alone, without lysozyme, resulting in a preparation in which high enzyme activity was accompanied by a clear band on Coomassie-stained gels (Fig. 5).
Mutational Analysis of the Enzyme-Because human 11␤-HSD1 contains four cysteine residues, three of which are conserved across all mammalian 11␤-HSD1 proteins, we tested for the presence of interchain disulfide bonds by probing Western blots of SDS-PAGE gels of lysates run under both reducing and nonreducing conditions (Fig. 6A). The results clearly showed the presence of both dimer and monomer bands in the nonre-ducing lanes, suggesting that some of the protein existed in an interchain disulfide-bonded dimeric form. Examination of human liver extracts indicated that the natural enzyme also consisted of a similar combination of monomeric and dimeric forms (Fig. 6B), although analysis of mouse liver extracts (Fig.  6C) showed the presence of only monomers. Because the human 11␤-HSD1 contains an additional cysteine (Cys-272) when compared with the other mammalian sequences reported to date, we investigated the effect of mutating this residue to the corresponding residue of 11␤-HSD1 from squirrel monkey, namely a serine (18). Interestingly, the expression of this mutant in E. coli produced very similar dehydrogenase and reductase activities to those observed with the wild type and no significant alteration in the respective K m values. However, Western blots of nonreducing gels indicated that the ability of the protein to form disulfide-bonded dimers had been abolished (Fig. 6D).
Purification of Recombinant 11␤-HSD1-Several attempts at purification of recombinant pET21CD from BL21(DE3) cells were unsuccessful. Several combinations of gel filtration, ion exchange, and ADP-agarose methods failed to yield sufficiently pure protein. Because the addition of sequences at the N terminus of the catalytic domain of 11␤-HSD did not seem to affect activity, a further expression construct based on our most active pET21CD plasmid was generated that incorporated an N-terminal 6xHis tag to allow purification by metal affinity chromatography. Use of this construct (pET28HCD) in BL21(DE3) cells resulted in lysates from which the enzyme could be purified to apparent homogeneity, as indicated by SDS-PAGE, in a single chromatographic step (Fig. 7). Activity measurements indicated that the recombinant human 11␤-HSD1 had been purified 159-fold with an overall yield of 28% With the cloning of a second "kidney-type" 11␤-HSD isozyme, the liver-type isozyme is now termed 11␤-HSD1.
The importance of these isozymes in the metabolism and clearance of glucocorticoids is well established; in addition, these enzymes are intricately involved in the pathogenesis of human diseases. For example, 11␤-HSD2 is implicated in hypertension and fetal growth retardation (1). Specifically, for 11␤-HSD1, emerging data have highlighted the role of this enzyme in modulating insulin sensitivity and visceral adiposity. Thus mice lacking the HSD11B1 gene are resistant to hyperglycemia of stress/feeding because of a failure to activate glucocorticoid within the liver and stimulate gluconeogenesis (4). Improvements in insulin sensitivity in normal volunteers given the 11␤-HSD1 inhibitor carbenoxolone support such a concept (27). Similarly in visceral adipose tissue, 11␤-HSD1 acts locally to generate active glucocorticoid concentrations, thereby stimulating adipogenesis (5,28). Defect in the activity of 11␤-HSD1 is also thought to underpin an inherited form of polycystic ovary syndrome, the syndrome of apparent cortisone reductase deficiency (29). Finally, further studies are investigating the role of the enzyme in central nervous system tissues and its relationship to neurodegenerative diseases (30). In each case though, the exciting concept has emerged that modulation of 11␤-HSD1 expression may represent a novel mechanism to modulate glucocorticoid action at the tissue level without changing circulating concentrations, thereby precipitating states of glucocorticoid excess or deficiency. Thus there is a clear clinical need to undertake a detailed characterization of the human 11␤-HSD1 isozyme with a view to its purification.
11␤-HSD1 belongs to the SDR superfamily, as defined on the basis of an N-terminal nucleotide binding motif, a central active site, and consensus sequence data. Sophisticated analytical approaches suggest that there are over 1000 members of this superfamily (6 -8), with only one residue (Tyr) being strictly conserved. A lysine 4 residues downstream and a serine 14 residues upstream are also largely conserved; all these residues are present in 11␤-HSD1. A model for catalysis of SDRs has been proposed on the basis of these residues (31). Binding of the coenzyme, NAD(H) or NADP(H), is in the Nterminal part of the molecule involving a common protein folding arrangement of ␣and ␤-strands ("Rossmann" fold) associated with a common Gly(Xaa) 3 GlyXaaGly motif (also found in 11␤-HSD1). The critically important tyrosine seems to maintain a fixed position relative to the scaffolding of the Rossmann fold and the cofactor position, whereas the substrate-binding pocket alters in such a way that the dehydrogenation/reduction reaction site is brought into bonding distance of the tyrosine hydroxyl group. The tyrosine therefore acts as a basic catalyst, the lysine binds to NAD/P(H) and lowers the pK a value of the tyrosine, and the serine plays a subsidiary role of stabilizing substrate binding (31, 32). Several groups have evaluated the importance of the Nterminal domain of 11␤-HSD1. Recently it has been shown that the orientation of the enzyme within the ER is determined by sequences close to the N terminus (10). Chimeric proteins where the N-terminal regions, including the membrane anchors, of the 11␤-HSD1 and 11␤-HSD2 enzymes were exchanged adopted inverted orientations in the ER membrane (10). Neither protein was catalytically active. However, mutation of a single lysine residue close to the N terminus of type 1 resulted in an inverted orientation without loss of activity. These results suggest that the N-terminal anchor is required for both activity and correct orientation, although it should be noted that the sequences exchanged in the chimeras included much more than just the transmembrane helix. Mercer et al. (12) reported that expression of an N-terminally truncated 11␤-HSD1 did not produce a soluble protein. However, these studies and others have employed mammalian expression systems where such constructs would not be appropriately targeted to the ER, and hence correct folding and disulfide bond formation may not have been facilitated.
In this study, using a series of bacterial expression constructs, we have shown that the activity of human 11␤-HSD1 does not depend on the N-terminal domain. Constructs where the N-terminal region had been removed (pET32CD and pET21CD) exhibited higher levels of expression and activity than constructs containing the entire 11␤-HSD1 sequence. Moreover, inclusion of the transmembrane domain, either with or without the thioredoxin fusion partner, failed to produce soluble active protein. This is in agreement with a study carried out by Blum et al. (13), in which the complete human 11␤-HSD1 sequence was expressed in E. coli and resulted in a protein that was virtually insoluble, difficult to purify, and completely inactive.
We also investigated expression systems in which thioredoxin, the product of the E. coli TrxA gene (25), was a fusion partner. In many cases heterologous proteins produced as thioredoxin fusion proteins are correctly folded and display full biological activity (25,(33)(34)(35). This has been thought to be caused by the small, highly soluble nature of thioredoxin, which also has robust folding characteristics (36). However, in our study proteins produced as a fusion with thioredoxin at the N terminus (pET32CD and pET32FL) showed no overall increase in the levels of soluble protein when compared with nonfusion constructs (pET21CD and pET21CDH), indicating that such fusions are not always profitable. Similarly, fusion of a 6xHis tag at the C terminus, as a means to simplify purification, was also detrimental to the solubility, and particularly the activity, of the enzyme. Residues close to the C terminus of SDRs may frequently be important in substrate binding (37), and modifications in this region may thus affect protein structure to the detriment of enzyme activity.
This study also clearly resolves the issue of whether glycosylation is required for the activity of the human enzyme. Studies on the rat 11␤-HSD1 enzyme indicated that partial inhibition of glycosylation with tunicamycin inhibited dehydrogenase activity by 50% but had no effect on reductase activity (14). Mutagenesis of the first of two potential N-glycosylation sites reduced dehydrogenase and reductase activities by 75 and 50%, respectively, whereas mutagenesis of the second site completely abolished activity (15). Conversely, studies carried out on the rabbit enzyme, which like the human homologue contains three potential glycosylation sites, suggest that glycosylation is not important for enzyme activity. No alteration in activity could be observed after complete deglycosylation of rabbit 11␤-HSD1 (9). Conflicting studies on the human enzyme have also been reported. Recently, human 11␤-HSD1 has been expressed in E. coli, where the biosynthesis of N-linked glycoproteins does not occur. This resulted in a recombinant protein that was completely devoid of enzyme activity (17). The same group also investigated the effects of deglycosylation on human 11␤-HSD1 purified from liver and recombinant protein produced by the yeast P. pastoris (13). Site-directed mutagenesis of the three potential glycosylation sites yielded an inactive protein from yeast cells as assessed using metyrapone and metyrapol as the substrates. However, the enzyme purified from human liver, upon complete deglycosylation, remained fully active. The results here agree with the latter experiment and clearly show that nonglycosylated enzymatically active 11␤-HSD1 can be generated within E. coli, with the recombinant enzyme possessing both reductase and dehydrogenase activities with similar kinetic properties to those reported previously from mammalian expression systems. Glycosylation is therefore not required for activity or protein folding, although it could still be important for protein stability within the endoplasmic reticulum.
All the constructs used in this study gave only moderate levels of soluble protein but a high proportion of protein in an insoluble form. The lack of protein solubility in E. coli is a complex event with many contributing factors. Although fusion with heterologous proteins may sometimes help to redress many solubility problems, another factor that may be important is the inability of the recombinant protein to form key disulfide bonds in the reducing environment of the bacterial cytoplasm (38). Rabbit 11␤-HSD1 is known to contain an intrachain disulfide bond (9), and therefore we investigated the expression levels and activity of our recombinant proteins in a variety of host E. coli strains, some of which have been developed to promote disulfide bond formation. Within E. coli at least two systems are responsible for reducing disulfide bonds that form in the cytoplasm; the thioredoxin system that consists of thioredoxin reductase and thioredoxin and the glutaredoxin system that includes glutathione reductase, glutathione, and glutaredoxins. We evaluated this using three separate E. coli strains. It was anticipated that disulfide bonds, and therefore solubility and activity of the soluble protein, would be enhanced by the use of AD494(DE3) and particularly the Origami(DE3) strain. In effect, the reverse was observed with the highest levels of soluble protein and enzyme activity being observed using BL21(DE3) as the host strain. This result could imply that the intramolecular disulfide bond observed in the rabbit 11␤-HSD1 protein (9) is not present in the human enzyme, although this awaits experimental confirmation.
We also tested for the presence of interchain disulfide bonds by probing Western blots of SDS-PAGE gels of bacterial lysates run under both reducing and nonreducing conditions. Both dimer and monomer bands were identified in the nonreducing lanes, suggesting that some of the recombinant protein exists in an interchain disulfide-bonded form. Examination of human liver extracts indicated that the native 11␤-HSD1 enzyme existed in a similar combination of monomeric and dimeric forms, proving that the heterogeneity was not a consequence of expression in the bacterial system. However, this heterogeneity not only complicates the purification of 11␤-HSD1, as has been noted previously for native enzyme (9), but could also hinder crystallographic analysis because crystal growth requires pure protein in a homogeneous form. Tests on extracts of mouse liver, however, detected no intermolecular disulfide bridges. Because previous reports also suggested that rabbit 11␤-HSD1 contains no intermolecular disulfide bonds (9), we investigated the effect of mutating the additional cysteine (Cys-272), found only in the human sequence, to the corresponding residue from the most closely related 11␤-HSD1 sequence (squirrel monkey). Expression of the resulting mutant (pET21CD-C272S) in the optimized bacterial expression system resulted in a protein with kinetic properties indistinguishable from the wild-type recombinant protein. However, nonreducing gels showed that the ability to form the interchain disulfide bonds had been abolished. Previous structural studies on other SDRs suggest that they exist naturally as nondisulfide-bonded dimers or tetramers (reviewed in Ref. 37). The results here suggest that Cys-272 of human 11␤-HSD1 may be involved in disulfide bonds between adjacent polypeptide chains of the enzyme, possibly stabilizing the dimeric form. However, this property does not seem to be vital to the activity of the enzyme. Using a modified expression construct that included an Nterminal 6xHis tag, we developed a simple purification protocol that allowed 157-fold purification of recombinant human 11␤-HSD1 in one step from crude cell lysates. The purified protein demonstrated activity in both dehydrogenase and reductase directions with K m values of 1.4 M for cortisol and 9.5 M for cortisone.
In conclusion this study has shown that, despite reports to the contrary, bacterial expression systems have the potential to produce active soluble 11␤-HSD1 protein. The results also demonstrate conclusively that the N-terminal region, containing the transmembrane domain, glycosylation, and interchain disulfide bonds are not essential for the activity of this enzyme. For the first time, active soluble 11␤-HSD1 has been produced in vitro and purified to apparent homogeneity. This will now allow detailed functional analysis of the enzyme using E. coliproduced protein and facilitate future structure/crystallographic studies.