Expression and Site-directed Mutagenesis of Human Protein Disulfide Isomerase in Escherichia coli TWO INDEPENDENTLY ACTING CATALYTIC SITES FOR THE ISOMERASE ACTIVITY*

THIS MULTIFUNCTIONAL POLYPEPTIDE HAS Protein disulfide isomerase (PDI, EC is a highly unusual multifunctional polypeptide, being identical to the B subunit of prolyl 4-hydroxylase, a cellular thyroid hormone binding protein and a com-ponent of the microsomal triglyceride transfer protein complex, and highly similar to a polypeptide acting in vitro as a glycosylation site binding protein. It has two -Cys-Gly-His-Cys- sequences which, it has been pro- posed, act as catalytic sites for the isomerase activity, but few data have been available to indicate whether one or both of them do indeed act as catalytic sites and whether the two presumed catalytic sites act independ- ently or cooperatively. We report here on the expression of human PDI in Escherichia coli with three dif- ferent signal sequences. All three polypeptide variants were secreted into the periplasmic space as fully active enzymes. Oligonucleotide-directed mutagenesis was used to convert either one or both of the -Cys-Gly-His-Cys- sequences to -Ser-Gly-His-Cys-. The PDI activity of both polypeptides containing a single modified se- quence was about 60% of that of the wild-type polypeptide, whereas the polypeptide with

vitro, depending on the reaction conditions, and is regarded as the in vivo catalyst for disulfide bond formation in the synthesis of secretory proteins (1)(2)(3). Cloning and nucleotide sequencing were first reported for the rat enzyme (4). Surprisingly, PDI was subsequently found to be identical to the B subunit of prolyl 4-hydroxylase, an cw2/3, tetramer (5, 6 ) , a cellular thyroid hormone binding protein (7,8), and a component of the microsomal triglyceride transfer protein complex (9), and highly similar to a polypeptide acting in vitro as a glycosylation site binding protein (10)(11)(12). Complete DNAderived amino acid sequences are now available for this multifunctional polypeptide from several sources (for reviews see The human PDI polypeptide consists of 491 amino acid residues and a signal sequence of 17 additional amino acids (5). The polypeptide has two thioredoxin-like regions, each containing a -Cys-Gly-His-Cys-sequence, which, it has been proposed (2,4), act as catalytic sites for the isomerase activity. Some observations support this proposal (2), but it is also possible that only one of the two -Cy$-Gly-His-Cys-sequences is involved in the catalysis (16). It is furthermore unknown whether the two presumed catalytic sites act independently or whether the polypeptide is folded in such a manner that the two -Cys-Gly-His-Cys-sequences come close to each other and act cooperatively (2,17).
We report here on expression of human PDI in Escherichia coli. Three signal sequences were tested in order to obtain secretion of an active enzyme into the periplasmic space. Sitedirected mutagenesis of the two -Cys-Gly-His-Cys-sequences was used to determine whether both of these sequences act as catalytic sites for the isomerase activity and whether they act independently or cooperatively.

MATERIALS AND METHODS
Unless otherwise indicated, the DNA manipulations were performed as described in Sambrook et al. (18). The oligonucleotides were synthesized in an Applied Biosystems DNA synthesizer (Department of Biochemistry, University of Oulu), and the mutations were created using an oligonucleotide-directed in vitro mutagenesis system (Amersham Corp.).
Construction of the Expression Plasmids pKK233-PDZ, pKK233-PDZ/ompAsig, and pTM2-PDZ-To construct pKK233-PDI (Fig.  lA), a 1993-bp EcoRI-HindIII fragment of a full-length cDNA clone for human PDI (5) was ligated to a similarly digested M13 mpl8. This construct was then used for site-directed mutagenesis to create an NcoI restriction site (NcoI* in Fig. lA) at the translation initiation site of PDI. Because of the internal NcoI site in the PDI cDNA, first the NcoI-HindIII fragment and then the NcoI*-NcoI fragment of this construct was ligated to the NcoI-HindIII site of the vector pKK233-2 (Pharmacia LKB Biotechnology Inc.).
In pKK233-PDI/ompAsig (Fig. lA), the original human PDI signal sequence was replaced by the slightly modified signal sequence of the E. coli ompA protein (19) (Fig. 1B). A 1905-bp BssHII-HindIII fragment from a PDI cDNA clone, a 58-bp NcoI**-BssHII fragment prepared from four oligonucleotides encoding the signal sequence, and an NcoI-HindIII-digested vector pKK233-2 were ligated in a 1:l:l molar ratio.
The vector pTM2-2 was a gift from Dr. Tiliang Deng (University of California at San Diego, La Jolla, CA) and is a slight modification of pTO-N (20). When constructing pTM2-PDI (Fig. lA), an NcoI restriction site (NcoI*** in Fig. lA) was created at the end of the signal sequence of PDI as described above for pKK233-PD1, thus changing the codon for Ala (the extreme C-terminal amino acid of the original signal sequence) into that for Met. First the NcoI-Hid11 fragment and then the NcoI***-NcoI fragment of the PDI insert of Expression of Human Protein Disulfide Isomerase in E. coli pKK233-PD1, pKK233-PDI/omp-

FIG. 1. Construction of plasmids
Asig, and pTM2-PDI. A shows the maps of the plasmids constructed. The PDI cDNA is a black arc in each vector. Ptrc, trc promoter; T7410, T7 gene 10 promoter; Amp', ampicillin resistance gene, respectively. B shows the signal sequences of the plasmids. Arrows indicate the differing amino acids between the signal sequence of pKK233-PDI and the original PDI and pKK233-PDI/ ompAsig and the authentic E. coli ompA protein, respectively. The first amino acid of the mature PDI polypeptide is indicated by +I.

M V R R A L L C L A V A A L V R A D A ATGGTGCGCCGCGCTCTGCTGTGCCTGGCCGTGGCCGCCCTGGTGCGCGCCGACGCC
SIGNAL SEQUENCE OF pKK233-PDI/ompAsig I 1 -1 + 1 + 2

M K K T A I A I A V A L A G F A T V V R A D A ATGAAAAAGACAGCTATCGCGATTGCAGTGGCACTGGCTGGTTTCGCTACCGTAGTGCGCGCCGACGCC
SIGNAL SEQUENCE OF pTM2-PDI -1 + 1 + 2 +3 this mutated construct was ligated to the NcoI-Hind111 site of the vector pTM2-2, as when constructing pKK233-PDI.

M K K T A I A I A V A L A G F A T V A Q A M D A
Construction of the Expression Plasmids pTM2-PDI0, pTM2-PDIb, and pTM2-PDIOb-The codon for the first Cys in both presumed catalytic sites of PDI (5) was changed to the codon for Ser by sitedirected mutagenesis using M13 mp19 containing either the PDI cDNA clone 0-210 (coding for the N-terminal catalytic site) or 0-202 (coding for the C-terminal catalytic site) as a template (5). When constructing pTM2-PDI. (see Fig. 3 in "Results"), the NcoI***-NcoI fragment of the PDI cDNA clone in M13 (see above) was cloned to NcoI-digested pTM2-2. The 847-bp AuaI-AuaI fragment encoding the N-terminal altered catalytic site from the mutated M13 p-210 vector was then used to replace the corresponding fragment in this construct. The NcoI***-NcoI fragment with the N-terminal mutated catalytic site was finally used to replace the NcoI***-NcoI fragment of pTM2-PDI. The 547-bp XhoI-XhoI fragment from the mutated recombinant pTM2-PDI and pTM2-PDI., thus yielding pTM2-PDIb and pTM2-M13 0-202 vector was used to replace the corresponding fragment in PD1.h (Fig. 3), respectively.
Expression and Localization of Human PDI in E. coli-E. coli cells harboring one of the expression plasmids or the corresponding parental plasmid were grown to an optical density of 0.6 at 600 nm, and expression was induced by IPTG for 24 h (or 5 h in the case of pTM2-PDI). Triton X-100 soluble proteins were released by suspending the cells in a buffer containing 50 mM Tris-HC1, pH 8.0, 50 mM NaC1, 1% Triton X-100, and 5 mM EDTA, with 22 mM N-ethylmaleimide, 2.7 mM p-aminobenzoic acid, 2.2 mM phenylmethylsulfonyl fluoride, 1 pg/ml aprotinin, 1 pg/ml leupeptin, 1 pg/ml pepstatin A, 50 pg/ml tosyllysine chloromethyl ketone, and 100 pg/ml tosylphenylalanine chloromethyl ketone as proteinase inhibitors. The cells were sonicated 4 X for 15 s with a 20-kHz cell disruptor, incubated for 2 h at room temperature, and centrifuged for 2 min a t 5000 X g to obtain the Triton X-100 soluble proteins. Periplasmic proteins were obtained by osmotic shock as described previously (21).
PDZActiuity Assay-PDI was assayed either by a method involving measurement of the rate of regeneration of incorrectly disulfidelinked RNase to the native form (22) or by the g1utathione:insulin transhydrogenase assay (23). When comparing the PDI activities of mutant polypeptides to the activity of recombinant wild-type PDI, the proportions of PDI polypeptide expressed were first quantified by radioimmunoblotting. The protein samples were run in SDS-PAGE and transferred to nitrocellulose, incubated with monoclonal antibody 5B5 (1.3 mg of IgG/ml) to human PDI (24), and transferred to a second antibody solution containing 1 pCi/ml 35S-labeled anti-mouse with the PDI polypeptides was quantified from the bands of the filter IgG (200 pCi/mmol, Amersham Corp.). The radioactivity associated in a liquid scintillation counter. The same amounts of wild-type and mutant PDI polypeptides were then assayed for PDI activity, and the activity obtained for the wild type PDI was taken as 100%.
The activity of the recombinant PDI was compared with that of pure prolyl4-hydroxylase (the / 3 subunit of which is identical to PDI), and the quantities of PDI in the assay samples were measured by radioimmunoblotting, as described above.
Other Assays-E. coli cells harboring pKK233-PDI or pKK233-PDI/ompAsig were labeled with ["S]methionine (25), and the produced PDI was immunoprecipitated from Triton X-100 soluble and periplasmic proteins. The radioactivity of the immunoprecipitate was quantified by liquid scintillation counting and compared with the radioactivity obtained from the total cell extracts in order to determine the proportion of PDI. Samples of immunoprecipitates were also subjected to 10% SDS-PAGE and treated for fluorography. The majority of the immunoprecipitable counts represented the PDI polypeptide, as fluorography indicated that only minor contaminant protein bands were present (not shown).
The expression level obtained with pTM2-PDI was estimated by densitometry from Coomassie staining of the SDS-PAGE of fractionated Triton X-100 soluble and periplasmic proteins using a Kortes K495000 densitometer.
N-terminal sequences were determined in an Applied Biosystems were fractionated by 10% SDS-PAGE, electroblotted onto a polyvi-model 477A with on-line 120A liquid-pulsed Sequencer. The proteins nylidene difluoride membrane, and visualized by staining with heparine/toluidine, and the band corresponding to PDI was cut out and subjected to sequencing.

RESULTS AND DISCUSSION
Expression of Human PDI in E. coli-Three expression plasmids varying in their signal sequences were constructed

Expression of Human Protein
Disulfide Isomerase in E. coli   . 1). Plasmid pKK233-PDI contains the original signal sequence of human PDI with one conservative amino acid change, while pKK233-PDI/ompAsig and pTM2-PDI contain slightly modified signal sequences of the E. coli ompA protein (Fig. 1B). The E. coli JM105 cells carrying either pKK233-PDI or pKK233-PDI/ompAsig, or BL21(DE3) cells carrying pTM2-PDI were grown in the presence of IPTG. Triton X-100 soluble proteins of the cell sonicates were studied by SDS-PAGE followed either by Coomassie staining or Western blotting with a monoclonal antibody to human PDI. A distinct 55-kDa polypeptide corresponding to human PDI was seen in the Western blots with all three plasmids, whereas no band was detected in experiments with the parental plasmids pKK233-2 or pTM2-2 (Fig. 2). The expression level was 0.5-1% of the Triton X-100 soluble protein with both pKK233-PDI and pKK233-PDI/ompAsig and 10-15% with pTM2-PDI (Table I).
Several experiments using the JM105 cells nevertheless gave a 43-kDa band instead of the expected 55-kDa band, and both bands were present in some experiments (Fig. 2). Little or no 43-kDa form was seen in experiments with E. coli strains SF100, SF110, and BL21(DE3)(lacIq) which are deficient in the ompT protease (26).
Secretion of Human PDI into the Periplasmic Space as a Mature, Fully Active Polypeptide-E.
coli cells carrying pKK233-PD1, pKK233-PDI/ompAsig, or pTM2-PDI were grown and treated by the osmotic shock method to release their periplasmic proteins. These were then studied by SDS-PAGE followed by Western blotting. A 55-kDa band was seen with all three plasmids (Fig. 2). Quantification experiments using [35S]methionine labeling or Coomassie staining followed by densitometry indicated that PDI represented 2-4% of the periplasmic protein when using pKK233-PDI or pKK233- PDI/ompAsig and 35-40% when using pTM2-PDI (Table I).
After the osmotic shock only a small fraction of the PDI was still cell-associated, and >95% was found in the supernatant. The high expression level obtained with pTM2-PDI and almost 100% secretion into the periplasmic space makes this system suitable for large-scale production of PDI, e.g. for crystallographic work. When using the standard purification protocol for PDI, 1 ml of bacterial cell cultures harboring PTM2-PDI will yield about 30 pg of pure PDI. Determination of the N-terminal amino acid sequences of the expressed polypeptides containing either the authentic PDI signal peptide (coded by pKK233-PDI) or the ompA signal peptide (coded by pTM2-PDI) demonstrated that both variants of the signal sequences were correctly cleaved (not shown).
PDI activity was determined in the periplasmic and total Triton X-100 soluble protein. After correction of the values for the proportion of PDI, the specific activity was about twice that of the p subunit present in the pure prolyl4-hydroxylase tetramer (not shown). Since the p subunit present in the prolyl4-hydroxylase tetramer has 50% of the PDI activity of the free p subunit (6), the specific activity of PDI expressed in E. coli was the same as that of PDI isolated from human tissues.
Construction of the Mutant Expression Plasmids pTM2-PDL, pTM2-PDIb, and pTM2-PDb-To study whether the two -Cys-Gly-His-Cys-sequences represent catalytic sites for PDI activity, three modified plasmids were constructed from pTM2-PDI. In pTM2-PDI. the N-terminal presumed catalytic site was modified by replacing the codon for its first Cys by a codon for Ser (Fig. 3). In pTM2-PDIb the corresponding change was made to the sequence coding for the first Cys in the presumed C-terminal catalytic site, while in pTM2-PDIab  the respective changes were made to the codons for both presumed catalytic sites (Fig. 3).

PDIActiuity ofthe MutatedPolypeptides-BLZl(DE3) cells
carrying either pTM2-PDI or one of the mutated plasmids were grown in the presence of IPTG for 5 h. The PDI activities of the four polypeptide variants were then measured in the periplasmic or Triton X-100 soluble protein fraction by two different methods and were recorded either under standard assay conditions or as maximal velocities (Table 11). The PDI activity of both polypeptides containing a single modified sequence (coded either by pTM2-PDI, or pTM2-PDIb) was found to be about 50% of that of the wild-type polypeptide, while the polypeptide containing two modified sequences (coded by pTM2-PD1.b) had no PDI activity (Table 11). The PDI polypeptide has two thioredoxin-like regions, both of which contain the sequence -Cys-Gly-His-Cys-within a highly conserved stretch of amino acids (4,5,27). Thioredoxin from various sources contains the slightly different sequence -Cys-Gly-Pro-Cys-(13,28,29) which acts as its catalytic site. The essential reactive thiol group in thioredoxin has been identified as the first of the two cysteines (2, 28). Therefore it was the first cysteine of either the N-terminal or C-terminal -Cys-Gly-His-Cys-sequence or of both these sequences that was converted to serine here. As modification of one of these two sites led to loss of half of the PDI activity of the polypeptide and modification of both these sites led to a complete loss of PDI activity, it can be concluded that the two -Cys-Gly-His-Cys-sequences represent catalytic sites for the isomerase activity and that the polypeptide has two PDI catalytic sites which appear to operate independently of one another.
PDI is now known to be a highly unusual multifunctional polypeptide (see the Introduction). The present results map the isomerase activity to the two -Cys-Gly-His-Cys-sequences of the thioredoxin-like regions. No data are currently available on the critical regions needed for its other functions, and it is also unknown whether one or both of the two PDI catalytic sites are essential for some of these other functions. The availability of the mutated polypeptides reported here will make it possible to study the latter aspect. As all attempts to construct an active prolyl 4-hydroxylase tetramer by association from its subunits in uitro have been unsuccessful (15), experiments to investigate the effects of these mutations on prolyl 4-hydroxylase activity will require expression of the a subunit and the wild-type or mutated / 3 subunits in the same cell.