Development of two novel nontoxic mutants of Escherichia coli heat-labile enterotoxin

Escherichia coli heat-labile enterotoxin (LT) is composed of catalytic A and non-catalytic homo-pentameric B subunits and causes diarrheal disease in human and animals. In order to produce a nontoxic LT for vaccine and adjuvant development, two novel derivatives of LT were constructed by a site-directed mutagenesis of A subunit; Ser 6 3 to Ty r 6 3 in LTS63Y and Glu 11 0 , G l u 11 2 were deleted in LT 11 0 / 112. The purified mutant LTs (mLTs) showed a similar molecular structural complex as AB 5 to that of wild LT. In contrast to wild-type LT, mLTs failed to induce either elongation activity, ADP-ribosyltransferase activity, cAMP synthesis in CHO cells or fluid accumulation in mouse small intestine in vivo . Mice immunized with m LTs either intragastrically or intranasally elicited high titers of LT-specific serum and mucosal antibodies comparable to those induced by wild-type LT. These results indicate that substitution of Ser 63 to Ty r 6 3 or deletion of Glu 11 0 and Glu 11 2 eliminate the toxicity of LT without a change of AB 5 conformation, and both mutants are immunogenic to LT itself. Therefore, both mLTs may be used to develop novel anti-diarrheal vaccines against enterotoxigenic E. coli .


Introduction
E n t e r o t o x i g e n i c Escherichia coli ( E T E C ) strain causes diarrheal disease in human and animals due to production of toxins such as heat-labile enterotoxin (LT) (Spangler, 1992). LT is a multimeric protein composed of two functionally distinct domains; the enzymatically active A subunit (LTA; Mr, ~30,000 daltons) with ADP-ribosylating a c t i v i t y, and the pentameric B subunits (LTB; Mr, ~11 , 6 0 0 daltons) that contain G M1 (monosialoganglioside) receptorbinding site (Bäckström et al., 1997). Upon thiol reduction, the A subunit dissociates into two polypeptide chains, A1 ( M r, 23,000 daltons) and A2 (Mr, 6,000 daltons) (Tsuji e t a l., 1985;Grant et al., 1994). The A1 subunit, in particular, intoxicates eucaryotic cells by catalyzing ADP-ribosylation of the protein Gs, a GTP-binding protein that regulates the levels of the second messenger cAMP (Guerrant et a l., 1974;Field et al., 1989). The resulting increase in cAMP levels causes secretion of water and electrolytes into the small intestine through interaction with two cAMPsensitive ion transport mechanisms including (i) NaCl co-transport across the brush border of villous epithelial cells and (ii) electrogenic Na + -dependent Cls e c r e t i o n by crypt cells (Guidry et al., 1997).
Both the cholera toxin (CT) from Vibrio cholerae and heat-labile enterotoxin (LT) from ETEC belong to the most potent mucosal adjuvants and immunogens known to date by oral and other mucosal routes, where most of antigens are unable to induce immune responses (Jackson et al., 1993;Takahashi et al., 1996). However, their toxi-cities have precluded usages in human (Douce et al., 1995). One approach to overcome the problem of toxicity is the generation of genetically detoxified derivatives of LT (Lobet et al., 1991;Dickinson and Clements, 1995) and CT (Fontana et al., 1995;Yamamoto et al., 1997b) by a site-directed mutagenesis of amino acids which are located on the βstrand that constitutes the 'floor' of NAD-binding cavity.
The most important factor for immunogenicity is shown to be the ability to bind to the receptor on eucaryotic cell (Nashar et al., 1996). In fact, a nonbinding mutant of the B subunit of LT was found to be nonimmunogenic (Guidry et al., 1997). Also, another group found that the ADPribosylating activity was unnecessary for immunogenicity because nontoxic derivatives of LT obtained by a sitedirected mutagenesis of the A subunit retained the immunological properties of the wild-type LT (Pizza e t al., 1994).
The attempt to define the role of ADP-ribosylating activity in LT adjuvanticity has generated conflicting results. For example, it was reported that a nontoxic derivative of LT (LT E 112K) when co-administered with keyhole limpet hemocyanin (KLH) by an oral route in mice, lacked the adjuvant properties, thus suggesting that the adjuvanticity of LT is linked to its ADP-ribosylating activity (Ly c k e et al., 1992). However, more recently, the adjuvant activity of the LTE112K was found to be identical to that of the LT holotoxin when delivered with influenza virus surface antigen by an intranasal route (Verweij et al. , 1998). On the other hand, other investigators showed that another LT derivative, LTK63, lacking enzymatic activity and toxicity was still able to elicit antibody responses against the co-administered antigen in mice immunized orally, intranasally, or intravaginally (Di Tommaso et al., 1996;Giuliani et al., 1998;Marchetti et al., 1998).
In an effort to develop anti-diarrheal vaccine or mucosal adjuvant, we constructed nontoxic mLTs, LTS63Y and LT∆11 0 / 112 by site-directed mutagenesis, each of which contains a single amino acid substitution and deletion of two amino acids, respectively, in the ADP-ribosyltransferase active center. We have demonstrated that in contrast to wild-type LT, both the LTS63Y and LT∆110/ 112 did not induce any toxic activities. Both the mutants elicited high and comparable levels of anti-LT antibodies when delivered either intragastrically or intranasally, inducing systemic and local responses in serum and fecal extracts. Thus, they might be useful for the development of a novel diarrheal vaccine in human and animals.

Plasmid construction and mutagenesis
A 1.5-kb BamHI DNA fragment including LT gene from porcine-origin, entero-toxigenic E. coli K88ac strain was cloned into pBluescript KSvector (pBlueKS -/ r LT). Singlestranded DNA was prepared from the culture supernatant of E. coli CJ 236 transformed with pBlueKS -/ r LT and then subjected to a site-directed mutagenesis using Mutan K (Takara Biomedicals, Kyoto, Japan). The sequences of oligonucleotides used for the substitution (S63Y) and deletion (∆110/112) were 5'-ATATGATGACGGATATGTTTC C A C T TA C C T TA G T T T G A G A A G T G C T C A C T T G-3' and 5'-A G G C G TATA C A G C C C T C A C C C ATAT C A G G T T T C T G C G T TA G G T GGAATACCAT-3', respectively. Serine at position 63 was substituted for tyrosine at LTS63Y and glutamic acids in position 110 and 112 were deleted in LT∆11 0 / 112. These residues are in proposed ADP-ribosyltransferase active center of LT and their substitutions or deletions have been shown to inactivate ADP-ribosyltransferase activity and enterotoxicity. We confirmed the changes of DNA sequences using Sequenase Version 2.0 sequencing kit (Amersham Life Science, USA).

Purification of recombinant mLTs
pBluescript KSvectors containing mutant LT genes were transformed into E. coli Top10F' (Invitrogen, USA). The mutant LTs were purified from cultures grown overnight. The cells were harvested by centrifugation, resuspended in TEAN buffer (0.2 M NaCl, 50 mM Tris, 1 mM EDTA and 3 mM NaN 3 [pH 7.5]), and lysed with microfluidizer (Microfluidics Corporation, USA). The lysates were clarified by centrifugation and then filtered using 0.45µm membrane (Micro Filtration Systems, Japan) prior to chromatography on immobilized D-galactose column (Pierce, USA) (Uesaka et al., 1994). The mLTs were eluted with 0.3 M galactose in TEAN buff e r. Holotoxin ( A B 5 ) fraction was separated from the free B-subunit pentamers by size exclusion chromatography using FPLC Superdex 200 column (Pharmacia, Sweden).

Cell elongation assay
The ability of mLTs to induce morphological changes in cultured Chinese hamster ovary-K1 (CHO-K1) (AT C C , USA) cells was tested as previously described (Guerrant et al ., 1974). CHO-K1 cells were incubated for 24 h as monolayer cultures in minimal essential medium alpha ( M E M -α) (GibcoBRL, USA) supplemented with 10% fetal bovine serum (FBS) in a humidified, 5% CO 2 atmosphere at 37˚C. The cells were washed once with H a n k s 's balanced salt solution (HBSS) and then removed from the flask by incubation of the cells with 0.1% trypsin for 5 min. After centrifugation, they were washed once, and then resuspended in the growth medium. To each well of a 48-well tissue culture plate, the same numbers of CHO-K1 cells (10 4 in a 200-µl volume per well) were added. The cells were allowed to adhere for 4 h prior to the addition of the toxin dilutions and then incubated for 24 h in a humidified, 5% CO 2 atmosphere at 37˚C. Cells were then washed with phosphate-buffered saline (PBS), fixed with methanol, and stained with 0.04% Trypan Blue Stain (GibcoBRL). After staining, the cells were washed, air dried, and analyzed for morphological changes by light microscopy.

ADP-ribosyltransferase activity test
For the preparation of crude membranes, CHO-K1 cells were maintained in monolayer culture by serial passages in MEM-α medium supplemented with 10% FBS (Locht et al., 1987). The cells were detached from the flask, resuspended in PBS (pH 7.2) and then sedimented by centrifugation at 1,000 g for 10 min. The cells were resus-pended in ice-cold 25 mM Tris-HCl (pH 7.5) containing 5 mM MgCl 2 , allowed to equilibrate for 15 min on ice and then homogenized. The homogenate was centrifuged at 6 0 0 g for 10 min at 4˚C to remove nuclei and intact cells. The postnuclear supernatant fractions were centrifuged at 18,000 g for 7 min to yield a microsomal or membrane pellet. The pelleted material was suspended in 50 mM Tris-HCl (pH 8.0) and centrifuged at 18,000 g for 7 min. The final washed membrane pellet was resuspended in 50 mM Tr i s -H C l (pH 8.0) at a concentration of 1 mg of protein per ml and stored at -70˚C until used. ADP-ribosyltransferase activity was determined as the ability to catalyze the transfer of labeled ADP-ribose from [a d e n y l a t e-3 2 P]NAD to the 41 kDa G protein in CHO-K1 membranes (Locht et al., 1987). Reaction mixtures (100 µl) containing 32 µM [adenylate-32 P]NAD (2 µCi) (NEN, USA), 10 mM thymidine, 100 µM ATP, 20 mM DTT, 100 µM GTP, 50 µg of CHO-K1 membrane proteins, 50 mM Tris-HCl (pH 8.0) and 10 µg of wildtype LT or mLT were incubated at 37˚C for 30 min. The reactions were terminated by the addition of 1 ml of icecold 50 mM Tris-HCl (pH 8.0), and the membranes were sedimented by centrifugation at 15,000 g for 7 min at 4˚C. The membrane pellet was resuspended in ice-cold Tris-HCl (pH 8.0) and sedimented once more by centrifugation before solubilization in 50 µl of electrophoresis sample buffer containing 5% β-mercaptoethanol. The samples were heated to 95˚C for 5 min and then analyzed by SDS-PAGE and autoradiography.

Measurement of intracellular cAMP accumulation
CHO cells (ATCC) were maintained in MEM-α medium supplemented with 10% FBS in 24-well plate at a concentration of 5 1 0 4 cells per well, grown to near c o n f l u e n c y, and incubated in MEM-α containing 1% FBS and 1 mM IBMX for 30 min prior to addition of toxins (Grant et al., 1994). Either CT, CTB, trypsinactivated wild-type LT, LT S 6 3 Y, or LT∆11 0 / 112 was added to each well and incubated for 18 h. The cells were then washed three times with PBS. Intracellular cAMP was extracted by adding 200 µl of 50 mM HCl to each well and incubating the plates in -70˚C deep freezer for 20 min. cAMP was measured with a Biotrak cAMP enzyme-immunoassay (EIA) system (Amersham Life Science) as described by manufacturer's instructions.

Assessment of toxicity using mouse ileal loops
The enterotoxicity of mLTs was examined using a mouse ileal loop test (Yamamoto et al., 1997b). Groups of mice were anesthetized, and 100 µl of PBS containing diff e r e n t doses of toxins were injected into ileal loops (LT, 100 ng or 1 µg per mouse; mLT, 10 µg or 100 µg per mouse). The mice were killed 18 h after the injection, and the ratio of fluid to length was determined and defined as positive when the ratio was more than 40 µl/cm.

Mice and their immunization
Female Balb/c mice aged 6 weeks old were purchased from Charles River (Japan). For intragastric immunization, antigens were resuspended in PBS (pH 7.2) buff e r containing 0.35 M NaHCO 3 and delivered in a volume of 0.5 ml per mouse. Mice were immunized intragastrically with 25 µg of each toxin on days 0, 7, 14, and 21 (Takahashi et al ., 1996). For intranasal immunization, mice were delivered with a 20-µl aliquot (10 µl per nostril) containing 2 µg of each toxin on days 0, 7, and 14 (Yamamoto et al., 1997a).

G M1 -ELISA
LT-specific antibodies were measured with a G M 1 c a p t u r e enzyme-linked immunosorbent assay (G M 1 -E L I S A ) (Spiegel, 1990;Douce et al., 1997). Plates were coated with 150 ng of G M 1 (Sigma, USA) suspended in PBS per well of 96-well EIA/RIA plate (Costar, USA) and then incubated at 37˚C for 1 h. Plates were washed three times with PBS containing 0.05% Tween 20 (PBST) and blocked for 1 h at 37˚C with 2.5% skim milk (Difco, USA) in PBST. After washing with PBST three times, 100 ng of wild-type LT was added into wells and plates were incubated for 1 h at 37˚C and washed three times with PBST. Sera or fecal samples obtained (Jackson et a l ., 1993;Yamamoto et al., 1997a) from each mouse were tested by using two-fold serial dilutions and incubated for 2 h at 37˚C. After washing with PBST six times, the plates were incubated for 1 h at 37˚C with an appropriate anti-mouse immunoglobulin G (IgG) or IgA antibody (KPL, USA) conjugated with horseradish peroxidase (HRP) and washed as described above. Bound antibodies were visualized by adding 3,3',5,5'tetramethylbenzidine (TMB) substrate. Absorbancies were read at 450 nm and ELISA titers were arbitrarily determined as the dilution of serum which gave an optical density value above the level measured in preimmune samples.

Expression and purification of mLTs
S e r 6 3 was substituted to Ty r 6 3 in LTS63Y and Glu 11 0 and Glu 112 were deleted in LT∆110/112. These residues located in or near the NAD-binding site of LT have been shown to be essential for the ADP-ribosyltransferase activity of LT . We expressed the mLTs, LTS63Y and LT∆11 0 / 112, in the plasmid vector (pBlueKS -/ r LTS63Y or pBlueKS -/ r LT∆11 0 / 11 2 ) , containing the coding (1.2 kb) and regulatory (160 bp of 5'-and 197 bp of 3'-noncoding genes) regions. The recombinant proteins were purified by immobilized Dgalactose column chromatography. The homogeneity of LTS63Y and LT∆110/112 was confirmed by SDS-PAGE, as shown in Figure 1. When the purified mLTs were analyzed without boiling, two protein bands were appeared; one band with the size of 70-100 kDa corresponding to the holotoxin and LTB pentamers, and the other band with the size of about 30 kDa corresponding to the LTA subunit. When the purified mLTs were boiled for 5 min, the holotoxins were dissociated into two bands of about 30 and 11 kDa, corresponding to the A and B subunits of LT, respectively. Since the mobilities of mLTs were identical to those of the wild-type LT (Figure 1), the molecular weights of the mLT subunits were presumed to be identical to those of wild-type LT. These results suggested that the innate structure of the A subunit associated with pentameric B subunits of LT was not affected by substitution of tyrosine for Ser 63 or deletion of Glu 11 0 and Glu 11 2 residues on NAD-binding pocket. Moreover, it was demonstrated that the binding ability of the B subunit of mLTs to G M1 ganglioside was similar to that of the normal B subunit using a G M 1 -ELISA, and mLTs were reacted with anti-LT antibody in Western blot analysis (data not shown). These results imply that m LTs retain the AB 5 conformation similar to wild-type LT.

Assays for biologic, enzymatic, and toxic activities of mLTs
Enzymatic and biologic characterizations of mLTs were carried out to compare their properties with wild-type LT, including cell elongation assay, ADP-ribosyltransferase activity test, cAMP assay and mouse ileal loop test. The morphological changes on the CHO-K1 cells (Grant et al., 1994) were used to detect the toxic activity of mLTs. As little as 100 ng/ml of LT induced longitudinal growth of approximately 90% of the CHO-K1 cells, a response previously shown to be dependent upon adenylate  10%< at 10 µg Negative at 100 µg a 10 4 cells of CHO-K1 cells were cultured with 100 ng of wild-type LT or 10 µg of each mLT for 24 h and a positive toxin effect on the CHO-K1 cells was defined as elongation of > 20% of the cells according to published criteria (Yamamoto et al., 1997b). b The enterotoxicity of mLTs was examined using an ileal loop test, where mice were anesthetized, and 100 µl of PBS containing 100 ng of wt LT or 100 µg of each mLT were injected into an ileal loop. Loops were examined 18 h later and the ratio of fluid to length was defined as positive when the ratio was > 40 µl/cm (Yamamoto et al., 1997b).  (Guerrant et al. , 1974). However, the cells treated with mLT at the level of 10 µg/ml showed no morphological changes of the CHO-K1 cells (Figure 2 and Table 1).

cyclase-induced increases in cAMP
In general, the A1 fragment of LT is capable of binding NAD and catalyzing the ADP-ribosylation of Gs, a GTPbinding regulatory protein associated with adenylate cyclase (Spangler, 1992). The consequence is a sharp increase in cAMP production resulting in excessive accumulation of salts and water in the intestinal lumen (Field et al., 1989). A subunits of LT is known to catalyze ADP-ribosylation of the membrane-bound substrate Gproteins. As shown in Figure 3, when the membrane proteins (50 µg) from CHO-K1 cells were incubated with wild-type LT in the presence of [a d e n y l a t e-3 2 P]NAD, it   specifically ADP-ribosylated the Mr-41,000 proteins, which correspond to the α-subunits of the GTP binding Gs protein (lane 2 in Figure 3). In contrast, no ADPribosylation of this protein was detected in reaction mixtures incubated with the same amounts (10 µg) of LTS63Y (lane 3) or LT∆110/112 (lane 4). This result was identical to that of the negative control treated without toxins (lane 1). Therefore, the substitution of Ser 6 3 t o Tyr 63 or deletion of Glu 110 and Glu 112 in A1 subunit did cause changes in structural integrity of NAD b i n d i n g crevis that may be important for enzymatic activity of LT. To investigate cAMP accumulation induced by mLT, the levels of cAMP were determined in CHO cells treated with CT, CTB, LT, LTS63Y, or LT∆110/112. As shown in Figure 4, the addition of 50 ng/ml concentration of wildtype CT or LT caused about tenfold higher levels o f cAMP production than those of untreated cultures. On the other hand, cAMP formation in cultures treated with CTB, LTS63Y, or LT∆110/112 was undetectable even at a concentration as high as 5 µg/ml. These data showed that the presence of wild-type LTA subunit (accurately LTA1 subunit) is necessary for an increase in the intracellular cAMP concentration and the mutant derivatives, LTS63Y and LT∆11 0 / 112, devoid of enzymatic activity, are unable to form cAMP.
The toxicity of LTS63Y or LT∆11 0 / 112 was also assessed in a mouse ileal loop assay. One hundred nanogram of wild-type LT induced significant fluid accumulation in small intestine, however, no fluid accumulation was observed in the loop treated with thousand-fold higher levels (100 µg) of mLTs (Table 1). These data strongly indicate that the mLTs possess negligible enterotoxicity in vivo.
The derivatization of the wild-type toxin via substitution of a single amino acid or deletion of double amino acids known to be associated with the NAD-binding site on the A subunit resulted in complete loss of enzymatic, biologic, and toxic activities of LT. In particular, two novel mutants LTS63Y and LT∆110/112 that are shown to be devoid of ADP-ribosyltransferase activity, were unable to induce longitudinal growth of CHO-K1 cells, increase intracellular cAMP, and elicit fluid accumulation in mouse-ligated ileal loops. These results have revealed that LTS63Y or LT∆11 0 / 112 is qualitatively, physiologically distinct from wild-type LT.

Immunogenicity of LTS63Y and LT 110/112
The mucosal immunogenicities of LTS63Y and LT∆110/ 112 were tested via two immunization routes. Groups of mice were immunized intragastrically four times at weekly intervals with 25 µg of LTS63Y or LT∆11 0 / 112 and intranasally three times at the same intervals with 2 µg of the antigens. The control groups received PBS alone. The serum and fecal antibody titers to LT were determined using samples prepared on day 7 following the last immunization; the results are shown in Figure 5. The mice immunized with LTS63Y or LT∆11 0 / 11 2 contained high and comparable levels of anti-LT antibodies in sera and fecal extracts compared with those immunized with wild-type LT. The LTS63Y was slightly more immunogenic than LT∆11 0 / 112 on both intragastric and intranasal administration. On the other hand, titers of anti-LT in the serum or fecal extracts of mice intranasally immunized with wild-type LT or mLTs were slightly higher than those observed in mice intragastrically administered. Intranasal i m m u n i z a t i o n o ffers several advantages compared with other immunization routes. Lower doses of proteins are required to induce antibody responses and would decrease the cost for vaccine (Yamamoto et al., 1997a). As shown in Figure 5, when administered intranasally, only 6% of the quantity of mLT used in intragastric immunization was required to elicit slightly higher levels of secretory IgA responses and this dose also eff e c t i v e l y induced systemic IgG and IgA antibody responses. Thus, intranasal immunization using mLT could be an effective method for vaccination in human and animals.
In summary, this study has shown that both novel mLTs, namely LTS63Y and LT∆110/112, lacked toxicity but elicited mucosal immunogenicity via two mucosal routes and could be useful for the development of antidiarrheal vaccines. Particularly, both mLTs appeared to be more immunogenic on intranasal administration. We are currently assessing the mucosal adjuvanticity of these mLTs to Helicobacter pylori antigens. Further, the protection test using mLTs as adjuvants against H. pylori has been investigated in mice following intranasal immunization