Structure-Function Analysis of the Diphtheria Toxin Receptor Toxin Binding Site by Site-directed Mutagenesis*

Diphtheria toxin (DT) binds to the epidermal growth factor (EGF)-like domain of human membrane-anchored heparin-binding EGF-like growth factor (proHB-EGF), the human DT receptor (DTR). DT does not bind to mouse proHB-EGF because of amino acid substitutions within the EGF-like domain. We made 10 independent mutants, replacing a single amino acid within the EGF-like domain of human DTR/proHB-EGF with the corresponding amino acid residue in mouse proHB-EGF. The mutant proteins were transiently expressed in mouse L cells either expressing or not expressing DRAP27/CD9, and DT binding was measured. DT binding activity of GST fusion proteins containing the mutated EGF-like domain was also determined by a cell-free binding assay. The largest effect was seen with E141H, and second largest effects were seen with F115Y and L127F in all of the assay systems. We conclude that Phe115, Leu127, and Glu141 are critical amino acid residues for DT binding. A computer model of the tertiary structure of the EGF-like domain of human DTR/proHB-EGF was made. The model predicts that three amino acid residues critical for DT binding activity, Phe115, Leu127, and Glu141, are all located on the same face of the EGF-like domain, suggesting that this face of DTR/proHB-EGF interacts with the receptor-binding domain of DT.

Diphtheria toxin (DT) 1 (M r ϭ 58,342) is secreted by toxigenic strains of Corynebacterium diphtheriae (for reviews, see Refs. 1 and 2). DT inhibits cellular protein synthesis in eukaryotes by inactivating elongation factor 2 through ADP-ribosylation. DT is composed of two fragments, A and B, which are covalently linked by a disulfide bond. The crystal structure of DT reveals that DT is actually composed of three distinct domains, the catalytic domain (C-domain), which is A-fragment itself, the transmembrane domain (T-domain), which is essential for the translocation of the C-domain to the cytoplasm, and the receptor-binding domain (R-domain), which binds to DT receptor (DTR) (3). Entry of the A-fragment into the cytoplasm is required for the cytotoxic action of DT (4). DT binds to a specific receptor on the cell surface through the R-domain (5,6) and is internalized by receptor-mediated endocytosis (7,8). A conformational change in DT takes place in an acidic compartment, resulting in the insertion of DT into the endosome membrane through certain regions in the T-domain (9 -14), and finally the enzymatically active C-domain is translocated to the cytosol (15).
The DTR is the principal protein involved in binding DT to cells, and the DT sensitivity of cells is determined primarily by the presence or absence of DTR (16). DTR was purified from monkey Vero cells (17,18), one of the cell lines most sensitive to DT (19). A cDNA encoding DTR was cloned from this cell line (20). DTR is identical to the precursor form of heparin-binding EGF-like growth factor (proHB-EGF) (20, 21) that was originally identified as the heparin-binding member of the epidermal growth factor family (22,23). Although proHB-EGF is cleaved by an unidentified protease on the cell surface to yield the soluble mature growth factor (HB-EGF) (24), a significant amount of proHB-EGF is left uncleaved on the cell surface of many cell types, where it acts as a membrane-anchored growth factor (25) and as the DTR (21). It has also been shown that heparin and heparan sulfate proteoglycan binds to DTR/ proHB-EGF and influences its DT binding activity (26).
Although proHB-EGF is ubiquitously expressed in species including human, monkey, rat, and mouse (30), cells from mice and rats are resistant to DT (19,31,32). Transfection of human or mouse proHB-EGF cDNAs into mouse L cells revealed that mouse cells are insensitive to DT because mouse proHB-EGF does not bind the toxin (33). Use of human/mouse proHB-EGF chimeras demonstrated that DT binds to the EGF-like domain of human proHB-EGF but not to the EGF-like domain of mouse proHB-EGF (33). There are 10 amino acid differences between the EGF-like domain of human DTR/proHB-EGF and that of mouse proHB-EGF. Thus, mouse proHB-EGF is a natural mutant form of DTR. These nonconserved amino acid residues are good candidates for site-directed mutagenesis to study the relationship of proHB-EGF primary structure to its function as the DTR.
We introduced several mutations within the coding region for the EGF-like domain of human DTR/proHB-EGF and identified amino acid residues required for DT binding activity. A computer model of the tertiary structure of the human EGF-like domain of HB-EGF, based on the NMR structure of TGF-␣ (34), was made to examine the spatial arrangement of residues involved in DT binding.
Plasmid Constructions-All point mutations to alter amino acid residues were introduced into pTHG-1, which contains the entire human DTR/proHB-EGF coding region (33), by site-directed mutagenesis using an in vitro mutagenesis kit (Amersham, Ltd., Tokyo, Japan). Four chimeric plasmids containing multiple substitutions were constructed as follows. Three plasmids, pTMHG-1, pTMHGDS-1, and pTMHGDK-1, which encode mouse proHB-EGF, human/mouse HB-EGF chimera H(106 -136), and human/mouse HB-EGF chimera H(106 -186), respectively (33), were digested with PvuII and BalI, and the PvuII-BalI fragment of human DTR/proHB-EGF cDNA was inserted. This segment encodes the recognition site for anti-human proHB-EGF antibody (H-6) (21), which can be used as an immunological tag to measure the surface expression of proHB-EGF. The resulting plasmids were named pT-MHG-1tag, pTMHGDS-1tag, and pTMHGDK-1tag, respectively. Two plasmids, pTMHGDS-1tag and pTMHGDK-1tag, were digested either with DraII and SacI or with SacI and SmaI then ligated with synthetic DNAs encoding the corresponding regions of mouse proHB-EGF, except for Phe 127 . Compared with human proHB-EGF, the resulting plasmids have four types of multiple mutations within the EGF-like domain (Fig.  1). The BstXI-NotI fragment of each plasmid was introduced into the expression vector pRc/CMV (Invitrogen Corp., San Diego, CA), and construction of all mutant plasmids was confirmed by DNA sequencing as described previously (33).
Glutathione S-transferase (GST) fusion proteins containing the EGF-like domain of DTR/proHB-EGF were produced in Escherichia coli as follows. DNA fragments encoding the EGF-like domain, Asp 106 -Pro 149 , were amplified from plasmids for wild-type human DTR/proHB-EGF and single mutants using the polymerase chain reaction. The  Human DTR/proHB-EGF cDNA was transfected into LC cells. After incubation for 1 day, transfected cells were detached from dishes and mixed with untransfected LC cells at various ratios, and samples of 1 ϫ 10 5 cells were replated on new dishes. After further incubation for 1 day, the amount of proHB-EGF molecules on the cell surface was determined using anti-HB-EGF antibody and 125 I-labeled secondary antibody as described under "Experimental Procedures." The cell-associated radioactivity of the secondary antibody was plotted against the number of transfected cells (OE). The amount of DT bound to the cells was concomitantly determined (E). primers used were forward primer 5Ј-GTGGGATCCCCGGTGGATCA-GACCCATGTCGGAAATAC-3Ј and reverse primer 5Ј-ATGAATTCTAT-GGGAGGCTCAGCCCATGACACC-3Ј, except in the case of the S147T mutant for which the reverse primer containing the appropriate mutated sequence was used. The amplified fragments were digested with BamHI and EcoRI, and the resulting fragments were ligated into the corresponding restriction enzyme sites in the expression vector, pGEX-3X (Pharmacia). Consequently, the new sequence Gly-Gly-Ser was created between the C terminus of GST and the N terminus of DTR. Construction of all recombinant plasmids was confirmed by DNA sequencing.
Cell Culture and Transfection-Mouse L cells and LC cells, stable transfectants of mouse L cells expressing DRAP27 (27), were maintained in Dulbecco's modified Eagle's medium supplemented with 100 units/ml penicillin G, 100 g/ml streptomycin, and 10% fetal bovine serum (ICN Biomedical, Inc., Costa Mesa, CA) and were used as recipient cells for the transfection experiments. Transfection of plasmids into recipient cells was done as described previously (21) using calcium phosphate (35). Transfected cells were cultured for 48 h and then used for further studies as described previously (33).
DT Binding Assays for Transfectants-Binding of 125 I-labeled DT to cells was measured as described (21), except that 100 ng/ml 125 I-DT was used. This concentration (1.7 nM) is close to the concentration required for half-saturation of DT binding of human DTR (21,33). Nonspecific binding of 125 I-DT was assessed in the presence of a 1,000-fold excess of unlabeled DT. Specific binding was determined by subtracting the nonspecific binding from the total binding obtained using 125 I-DT alone.
DT Binding Assays for Soluble Recombinant DTRs-GST fusion proteins containing the wild type and mutated forms of the EGF-like domain of human DTR/proHB-EGF were purified by using glutathione-Sepharose 4B (Pharmacia) according to the manufacturer's instruction. The purities of all the fusion proteins were estimated to be about 90% by SDS-polyacrylamide gel electrophoresis. For binding assays, glutathione-Sepharose 4B gel (2.5-l bed volume) was suspended in 480 l of DT binding buffer (phosphate-buffered saline supplemented with 1 mg/ml bovine serum albumin, 2 mM phenylmethylsulfonyl fluoride, and 3 mM sodium azide) and incubated with 1.5 g of purified recombinant GST or GST fusion proteins in a rotator. The gel was washed three times with DT binding buffer and then was incubated with 50 ng/ml 125 I-DT in 1 ml of DT binding buffer in the presence or absence of unlabeled DT at 4°C for 6 h with gentle rotation. The gel was washed three times with phosphate-buffered saline supplemented with 1 mg/ml bovine serum albumin and 3 mM sodium azide, and the radioactivity that remained bound to the gel was counted by a ␥-counter. Nonspecific binding of 125 I-DT was assessed in the presence of a 1,000-fold excess of unlabeled DT. Specific binding was determined by subtracting the nonspecific binding from the total binding obtained using 125 I-DT alone.
Measurement of the Amount of proHB-EGF Expressed on the Cell Surface-L cells or LC cells transiently transfected with various mutant plasmids were incubated at 4°C for 2 h with 5 g/ml H-6 antibody in binding medium (Dulbecco's modified Eagle's medium containing nonessential amino acids, 20 mM HEPES-NaOH (pH 7.2) supplemented with 10% calf serum). Cells were washed with chilled washing buffer (phosphate-buffered saline supplemented with 0.5 mM CaCl 2 , 0.5 mM MgCl 2 , 5 mM NaN 3 , and 1% calf serum) three times and then incubated at 4°C for 2 h with 1 g/ml 125 I-goat anti-rabbit IgG antibody in binding medium. Finally, cells were washed with washing buffer three times and the cell-associated radioactivity was counted.
Computer Modeling of HB-EGF--Computer modeling was interactively performed using the Homology program (Biosym Technologies, San Diego, CA). Side chains of the minimized average structure of TGF-␣ (34) were replaced with those of corresponding residues in human HB-EGF. Conformations of side chains were selected from a rotamer library (36,37), avoiding severe steric repulsion with other parts of the molecule. Amino acid residues that were identical in the two amino acid sequences were not changed. The model was bathed in 1723 TIP-3P water (38). The whole system was fully energy-minimized by program AMBER 4.1 (39) with dielectric constant ϭ 1, cut-off of nonbonded interaction ϭ 12 Å, and a truncated octahedral periodic boundary condition applied to generate the final model of human HB-EGF.

RESULTS AND DISCUSSION
The 10 independent mutants of human DTR/proHB-EGF with substitutions of single amino acid residues from the sequence of mouse proHB-EGF and three chimeric proteins with multiple substitutions studied here are shown in Fig. 1. The mutant proteins, wild-type human DTR/proHB-EGF, and mouse proHB-EGF were transiently expressed in recipient cells, and the DT binding activity of the cells was measured. Both mouse L cells and LC cells were used as the recipient cells. L cells do not express DRAP27/CD9 (25); thus effects of the amino acid alternations on DT binding could be observed independently of effects on the interaction of DTR/proHB-EGF with DRAP27/CD9. The LC cell line is a stable transfectant of mouse L cells that express DRAP27/CD9. Using this cell line allows the effects of the amino acid alternations to be observed under conditions in which DTR can interact with DRAP27/CD9.
Although the transfection efficiency for the recipient cells used in this study was usually 30 -50%, the efficiency varied from experiment to experiment. Furthermore, the efficiency of DTR/proHB-EGF expression on the cell surface may differ among mutants. Therefore, to compare DT binding activities of the mutant proteins with that of wild type we normalized DT binding to the amount of proHB-EGF expressed on the cell surface. ProHB-EGF on the cell surface was measured for each set of transfected cells using an anti-human proHB-EGF antibody and 125 I-labeled secondary antibody as described under "Experimental Procedures." The anti-human proHB-EGF antibody used here was obtained by immunizing rabbits with a synthetic peptide corresponding to amino acids 54 -73 of human DTR/proHB-EGF. This antibody reacts with all of the mutant proteins used.
Although it would be difficult to determine the absolute amounts of proHB-EGF by such an indirect method, for the present purpose only, the relative amounts of proHB-EGF on the cell surface must be determined. To verify that this method allows accurate measurement of the relative amount of proHB-EGF on the cell surface, we performed a model experiment. LC cells transfected with human DTR/proHB-EGF cDNA were mixed with untransfected LC cells at various ratios, and the cells were treated with anti-human proHB-EGF antibody and 125 I-labeled secondary antibody. As shown in Fig. 2, the amount of 125 I-labeled secondary antibody bound is proportional to the number of transfected cells added. Similar results were obtained using L cells as recipients (data not shown). Table I shows the amount of proHB-EGF expressed on the cell surface and the amount of 125 I-DT bound to the L cells transfected with each mutant plasmid. The DT binding activities, obtained by dividing the specific binding of DT by the amount of proHB-EGF expressed on the cell surface, are also shown. Among the 10 single mutants, the largest effect was seen with E141H, for which no DT binding activity was detected using this binding assay. The second largest effects were seen with F115Y and L127F, for which DT binding activity decreased by about 86 and 81%, respectively. Smaller effects were seen with K122R, I133K, H135L, and S147T, for which DT binding activity decreased by about 31, 64, 51, and 16%, respectively. The other three single mutants, V124L, K125Q, and A129T, did not show significant effects on DT binding activity.
DT binding activities of the mutant DTR/proHB-EGF proteins expressed in LC cells are shown in Table II. The E141H substitution also showed the largest effect on the DT binding in these cells, with DT binding activity decreased by about 91%. F115Y and L127F also showed reduced binding activity, but the other seven single mutants, K122R, V124L, K125Q, A129T, I133K, H135L, and S147T, did not have significant effects. Effects of combinations of substitutions were also studied. The multiple mutants F115Y/K122R/V124L/K125Q/E141H/S147T and A129K/I133K/H135L/E141H/S147T showed DT binding activity about 3 and 2% that of wild type, respectively. These values are lower than that of the single mutant E141H but higher than that of mouse proHB-EGF. The triple mutant, A129T/I133K/H135L, showed DT binding activity about 62% that of wild type despite the fact that single amino acid alternations in these three residues (Ala 129 , Ile 133 , and His 135 ) did not show negative effects. This indicates that amino acid substitutions whose influence on DT binding was not apparent by testing of individual substitutions may collectively have significant effects.
To assess the effect of each single amino acid alternations introduced in the EGF-like domain by a different system, GST fusion proteins containing wild-type and mutant forms of the EGF-like domain of human DTR/proHB-EGF were made, and DT binding activities were determined in a cell-free system (Table III). As in the transfection assays, DT binding activity was greatly reduced with E141H, and F115Y and L127F showed reductions of 51 and 68%. Smaller effects were seen with K122R, A129T, I133K, and H135L. The other single mutants, V124L, K125Q, and S147T, did not show significant effect on DT binding activity in this assay.
Both human and mouse HB-EGF bind to EGF-receptor and have mitogenic activity. Substitution of amino acid residues in the EGF-like domain of human DTR/proHB-EGF by the corresponding amino acid residues in the mouse protein is therefore not expected to result in an undesirable change in the overall structure of the EGF-like domain. However, this approach has a limitation for evaluating the effects of substitutions at different sites. In the case of F115Y, K122R, V124L, L127F, and S147T, the amino acid substitutions are relatively homologous amino acid residues, whereas the substitutions in K125Q, I133K, H135L, and E141H involve nonhomologous or oppositely charged amino acid residues. Generally, substitution with a nonhomologous or oppositely charged amino acid would influence the activity more strongly than substitution with a homologous amino acid. Thus the failure to observe an effect of the substitutions used in this study does not necessarily indicate that the site does not play a role in DT binding, and further analysis is necessary to clarify the contributions of such sites.
Nevertheless, the reduced binding activities of F115Y, L127F, and E141H are prominent among the mutants studied here. In all three DT binding assays, transfection into L cells or LC cells or binding of recombinant GST fusion proteins, the E141H substitution showed the largest effect, as indicated in earlier studies (33,40). Reduced DT binding activity of F115Y and L127F was also observed in all three binding assays, even though both mutants involve substitution of a nonpolar amino acid residue by another residue of the same type. We conclude that these three substitutions are most critical for the loss of DT binding activity in mouse proHB-EGF.
DRAP27/CD9 associates with DTR/proHB-EGF and enhances its DT binding activity. When DT binding activities of the mutant DTR proteins expressed in L cells were compared with their binding activities in LC cells, enhancement of DT binding was observed for all of the single mutants in the presence of DRAP27/CD9 (compare Table I with Table II). EGF-like domains of DTR/proHB-EGF GST fusion proteins containing wild-type or mutated forms of the EGF-like domain of human DTR/proHB-EGF were adsorbed to glutathione-Sepharose 4B and incubated with 50 ng/ml 125 I-DT at 4°C for 6 h. The gel was washed, and the radioactivity associated with the gel was counted by a ␥-counter. Nonspecific binding of 125 I-DT was assessed in the presence of unlabeled DT (50 g/ml). Specific binding was determined by subtracting the nonspecific binding from the total binding obtained using 125 I-DT alone. Data are shown as averages of five experiments. Standard deviations are shown after each value. Thus, the single amino acid alternations introduced in this study do not block the enhancement DT binding by human DTR/proHB-EGF.
Although the tertiary structure of the EGF-like domain of HB-EGF has not been determined, the solution structures of the EGF-like domains of EGF and TGF-␣ were determined by NMR studies (34,41,42). All members of the EGF family of growth factors, including HB-EGF, have six conserved cysteine residues and several other conserved amino acids in their EGFlike domains. Furthermore, HB-EGF, EGF, and TGF-␣ all bind to the EGF receptor. Therefore, we expect the main chain folds of the EGF-like domain of HB-EGF to be similar to those of EGF and TGF-␣. Because HB-EGF and TGF-␣ have the same number of the amino acid residues inside the EGF-like domain and EGF has one additional amino acid residue between the second and third cysteine residues, we chose TGF-␣ as the template structure for a computer model of the tertiary structure of the EGF-like domain of HB-EGF. Human HB-EGF and mouse HB-EGF bind to same EGF receptor, and mouse HB-EGF is as mitogenic for cells as human HB-EGF. Thus, the main chain structures of the EGF-like domain are not likely to be greatly affected by substitution of amino acid residues in the human DTR/proHB-EGF sequence with corresponding amino acid residues in mouse proHB-EGF. Fig. 3 shows two views of the model in different orientations. In the model, the three amino acid residues critical for DT binding activity, Phe 115 , Leu 127 , and Glu 141 , are all located on the same face. We propose that the face containing Phe 115 , Leu 127 , and Glu 141 is important for DT recognition by the EGF-like domain of DTR/proHB-EGF and that these residues play roles in DT binding. The other seven amino acid residues in the EGF-like domain that differ between human and mouse, except for Ala 129 , are located on other faces. Although Ala 129 is on the same face as Phe 115 , Leu 127 , and Glu 141 , A129T did not show significantly decreased DT binding activity. Ala 129 may not be involved in DT recognition or Ala 129 may be involved in DT recognition, but the homologous substitution Thr for Ala may not have a significant effect.
Reduced DT binding activity was observed with K122R, I133K, and H135L in the absence of DRAP27/CD9 (Table I and  III) but not in the presence of DRAP27/CD9 (Table II). The presence of DRAP27/CD9 may compensate for the negative effects of these mutants by interacting with DTR/proHB-EGF. The tertiary structure model predicts that amino acid residues Lys 122 , Ile 133 , and His 135 are located on the face opposite the one containing Phe 115 , Leu 127 , and Glu 141 . Therefore, it is intriguing to speculate that DRAP27/CD9 interacts with the face on which Lys 122 , Ile 133 , and His 135 are located and helps stabilize the structure of DTR/proHB-EGF, reducing the effects of the substitutions.
In human amphiregulin, amino acid residues Phe 115 , Leu 127 , and Glu 141 are conserved, but DT does not bind to human amphiregulin (33). Whereas amino acid residues Phe 115 , Leu 127 , and Glu 141 are important for the interaction of HB-EGF with DT, other residues within the EGF-like domain also affect DT recognition. Amino acid residues conserved in both human and mouse HB-EGF probably participate in DT recognition, especially amino acid residues located in the face including Phe 115 , Leu 127 , and Glu 141 .
Alanine replacements in DT indicated that residues Lys 516 and Phe 530 , located within a loop of the R-domain of DT, are involved in DTR recognition (5). All three amino acid residues that we identified as critical for DT binding activity are located in a loop in the EGF-like domain of HB-EGF. It is likely that DT-DTR interactions are mediated through these loops. It would be interesting to examine whether Lys 516 and Phe 530 of DT interact directly with Phe 115 , Leu 127 , or Glu 141 of DTR/ proHB-EGF. Further site-directed mutagenesis studies of DT and DTR/proHB-EGF can provide more precise understanding of DT-DTR interactions.
Another approach to the study of DT-DTR interaction would be structural analysis of the DT-DTR complex. An analysis of the crystal structure of DT⅐HB-EGF complex is under way by S. Choe and associates in the Salk Institute. Their results also suggest that Phe 115 , Leu 127 , and Glu 141 of the DTR/proHB-EGF are located in the face interacting with DT. 2