Toxin binding site of the diphtheria toxin receptor: loss and gain of diphtheria toxin binding of monkey and mouse heparin-binding, epidermal growth factor-like growth factor precursors by reciprocal site-directed mutagenesis

The transmembrane precursor of the monkey (Mk) heparin-binding, epidermal growth factor-like growth factor (proHB-EGF) functions as a diphtheria toxin (DT) receptor, whereas the mouse (Ms) precursor does not. Previously, using chimeric Ms/Mk precursors, we have shown that DT resistance of cells bearing Ms proHB-EGF may be accounted for by several amino acid substitutions between residues 122 and 148 within the EGF-like domain and that Glu-141 is an important amino acid residue for DT binding. In this study, reciprocal site-directed mutagenesis was performed on the major non-conserved residues in the region of 122–148, alone or in combination, between Mk and Ms precursors to identify more precisely which amino acid residues are important for DT binding. Two approaches were used. The first, more traditional approach was to destroy DT sensitivity and binding of Mk proHB-EGF by substitution(s) with the corresponding Ms residue(s). From the single mutations, the greatest loss of DT sensitivity was observed with Mk/Glu-141His (approximately 4000-fold) and the next greatest with Mk/Ile-133Lys (approximately fourfold). The double mutations Mk/Leu-127Phe/Glu-141His, Mk/Ile-133Lys/Glu-141His and Mk/His-135Leu/Glu-141His resulted in complete toxin resistance (> 100 000-fold). The second approach, both novel and complementary, was to gain DT binding and sensitivity of Ms proHB-EGF by substitution(s) with the corresponding Mk residue(s). Surprisingly, the single mutation Ms/His-141Glu resulted in the gain of moderate DT sensitivity (> 260-fold). The double mutation Ms/Lys-133Ile/His-141Glu and the triple mutation Ms/Lys-133Ile/Leu-135His/His-141Glu resulted in a progressive gain in toxin sensitivity (> 4700-fold and > 16 000-fold respectively) and affinity. This triple mutant cell line is essentially as sensitive (IC₅₀ = 3.1 ng ml^?1) as the highly toxin-sensitive monkey Vero cell line (IC₅₀ = 4 ng ml^?1), indicating that these three Mk residues enable the Ms proHB-EGF to act as a fully functional DT receptor. Taken together, these results indicate that Glu-141 plays the most critical role in DT binding and sensitivity and that two additional amino acid residues, Ile-133 and His-135, also play significant roles.     

Importance of the major extracellular domain of CD9 and the epidermal growth factor (EGF)-like domain of heparin-binding EGF-like growth factor for up-regulation of binding and activity

Heparin-binding epidermal growth factor (EGF)-like growth factor (HB-EGF) is a member of the EGF family of growth factors. The membrane-anchored form of HB-EGF (proHB-EGF) is mitogenically active to neighboring cells as well as being a precursor of the soluble form. In addition to its mitogenic activity, proHB-EGF has the property of binding to diphtheria toxin (DT), serving as the specific receptor for DT. Tetramembrane-spanning protein CD9, a member of the TM4 superfamily, is physically associated with proHB-EGF at the cell surface and up-regulates both mitogenic and DT binding activities of proHB-EGF. To understand this up-regulation mechanism, we studied essential regions of both CD9 and proHB-EGF for up-regulation. Immunoprecipitation experiments revealed that not only CD9 but also other TM4 proteins including CD63, CD81, and CD82 associate with proHB-EGF on the cell surface. However, these TM4 proteins did not up-regulate DT binding activity of proHB-EGF. Transfection of a series of chimeric constructs comprising CD9 and CD81 showed that the major extracellular domain of CD9 is essential for up-regulation. Assays of DT binding activity and juxtacrine mitogenic activity of the deletion mutants of proHB-EGF and chimeric molecules, derived from proHB-EGF and TGF-alpha, showed that the essential domain of proHB-EGF for up-regulation is the EGF-like domain. These results indicate that the interaction of the extracellular domains of both molecules is important for up-regulation.     

Domain analysis of the tetraspanins: studies of CD9/CD63 chimeric molecules on subcellular localization and upregulation activity for diphtheria toxin binding

CD9 and CD63 belong to a tetramembrane-spanning glycoprotein family called tetraspanin, and are involved in a wide variety of cellular processes, but the structure-function relationship of this family of proteins has yet to be clarified. CD9 associates with diphtheria toxin receptor (DTR), which is identical to the membrane-anchored form of heparin-binding EGF-like growth factor (proHB-EGF). CD9 upregulates the diphtheria toxin (DT) binding activity of DTR/proHB-EGF, while CD63 does not upregulate the DT binding activity in spite of the fact that this protein also associates with DTR/proHB-EGF on the cell surface. CD9 molecules localize on the cell surface, while those of CD63 localize predominantly at lysosomes and intracellular compartments. We made CD9/CD63 chimeric molecules and then studied their intracellular localization and upregulation activities. The C-terminal regions of CD63, which includes the lysosome sorting motif, showed a strong inhibitory effect on the expression of the chimeric proteins at the cell surface, while mutants lacking the lysosome sorting motif delivered more efficiently on the cell surface, indicating that the lysosome sorting motif contributes to the inhibitory effect of the C-terminal region. However, the N-terminal half of this family of proteins containing the 1st to 3rd transmembrane domains also seems to influence the cell surface expression. For the upregulation of DT binding activity the large extracellular loop (EC2) of CD9 was essential, while the remaining regions influenced the upregulation activity by changing the efficiency of cell surface expression. From these results we discussed the structure-function relationship of this family of proteins.     

The 27-kD diphtheria toxin receptor-associated protein (DRAP27) from vero cells is the monkey homologue of human CD9 antigen: expression of DRAP27 elevates the number of diphtheria toxin receptors on toxin- sensitive cells

Diphtheria toxin (DT) receptor associates with a 27-kD membrane protein (DRAP27) in monkey Vero cells. A cDNA encoding DRAP27 was isolated, and its nucleotide sequence was determined. The deduced amino acid sequence revealed that DRAP27 is the monkey homologue of human CD9 antigen. DRAP27 is recognized by CD9 antibodies. A human-mouse hybrid cell line (3279-10) possessing human chromosome 5, sensitive to DT, but not expressing CD9 antigen, was used for transfection experiments with DRAP27. When the cloned cDNA encoding DRAP27 was transiently expressed in 3279-10 cells, the total DT binding capacity was three to four times higher than that of untransfected controls. Transfectants stably expressing DRAP27 have an increased number of DT binding sites on the cell surface. Furthermore, the transfectants are 3-25 times more sensitive to DT than untransfected cells, and the sensitivity of these cells to DT is correlated with the number of DRAP27 molecules on the surface. However, when the cloned cDNA was introduced into mouse cell lines that do not express DT receptors, neither an increased DT binding nor enhancement of DT sensitivity was observed. Hence, we conclude that DRAP27 itself does not bind DT, but serves to increase DT binding and consequently enhances DT sensitivity of cells that have DT receptors. 12 proteins related to DRAP27/CD9 antigen were found through homology search analysis. These proteins appear to belong to a new family of transmembrane proteins.     

Extension of juxtamembrane domain of diphtheria toxin receptor arrests translocation of diphtheria toxin fragment A into cytosol

Diphtheria toxin (DT) binds to the EGF-like domain of the DT receptor (DTR), followed by internalization and translocation of the enzymatically active fragment A into the cytosol. The juxtamembrane domain (JM) of the DTR is the linker domain connecting the transmembrane and EGF-like domains. We constructed mutants of DTRs with altered JMs and studied their abilities for DT intoxication. Although DTR mutants with extended JMs showed normal DT binding activity, the cells expressing the mutants showed both reduced translocation of DT fragment A into the cytosol and reduced sensitivity to DT, when compared with cells expressing wild-type DTR. These results indicate that the JM contributes to DT intoxication by providing a space appropriate for the interaction of DT with the cell membrane. The present study also indicates that consideration of epitopes of an immunotoxins would be an important factor in the design of potent immunotoxins.     

Characterization of the diphtheria toxin receptor-binding domain

The carboxyl-terminal region of diphtheria toxin (DT) has been analysed in order to determine regions of receptor recognition. Biochemical cleavage of the toxin with hydroxylamine (HA) was used to generate the peptides HA9DT (residues 454–535), HA6DT (residues 482–535), and HA3DT (residues 454–461). Characterization of HA6DT demonstrated that the final 54 amino acids of DT are sufficient to constitute the receptor-binding domain of the toxin. Within HA9DT, the region encompassing HA3DT and containing the highly cationic polyphosphate-binding site did not contribute to the binding ability of HA6DT. Consistent with this observation, HA3DT itself did not compete for binding of radiolabelled DT to Vero cells. A 30-amino acid synthetic peptide composed of residues 506–535 did not block receptor binding of DT, indicating that residues toward the amino-terminus of HA6DT, or the entire HA6DT region, are required for receptor recognition.     

A Cre-inducible diphtheria toxin receptor mediates cell lineage ablation after toxin administration

A new system for lineage ablation is based on transgenic expression of a diphtheria toxin receptor (DTR) in mouse cells and application of diphtheria toxin (DT). To streamline this approach, we generated Cre-inducible DTR transgenic mice (iDTR) in which Cre-mediated excision of a STOP cassette renders cells sensitive to DT. We tested the iDTR strain by crossing to the T cell- and B cell-specific CD4-Cre and CD19-Cre strains, respectively, and observed efficient ablation of T and B cells after exposure to DT. In MOGi-Cre/iDTR double transgenic mice expressing Cre recombinase in oligodendrocytes, we observed myelin loss after intraperitoneal DT injections. Thus, DT crosses the blood-brain barrier and promotes cell ablation in the central nervous system. Notably, we show that the developing DT-specific antibody response is weak and not neutralizing, and thus does not impede the efficacy of DT. Our results validate the use of iDTR mice as a tool for cell ablation in vivo.     

Interaction of diphtheria toxin B subunit with sensitive and insensitive mammalian cells

The recombinant fluorescent derivative of diphtheria toxin (EGFP-SbB) obtained by the replacement of toxin A subunit by enhanced green fluorescent protein (EGFP) has been used for visualization of the interaction of diphtheria toxin (DT) with sensitive and insensitive cells. It was shown that EGFP-SbB could interact with cell surface of both toxin-sensitive monkey cells (Vero cell line) and toxin-resistant mouse cells (3T3 cell line). The affinity of this protein for receptors of Vero cells was three times higher as compared with 3T3 cells. It was demonstrated that fluorescent derivate was able to interact with receptors of both cell lines and to internalize into these cells. Internalization of EGFP-SbB into the cells was inhibited by endocytosis inhibitor phenyl arsine oxide. We suppose that diverse sensitivity to DT of monkey and mouse cells can be explained not only by differences in their receptor affinity for DT but also by the processes that occur after internalization of the toxin into the cells.     

CD9 amino acids critical for upregulation of diphtheria toxin binding.

CD9 associates with a diphtheria toxin receptor (DTR) that is identical to the membrane-anchored form of heparin-binding EGF-like growth factor. We determined the region of CD9 important for upregulation activity. Human and monkey CD9 upregulates DT binding activity of DTR, while mouse CD9 has no upregulation activity. Transfection of chimeric constructs comprising monkey and mouse CD9s showed that the human sequence between Ala156 and Asp183 is essential for the upregulation activity. Studies of mutants, replacing a single amino acid within the region between Ala156 and Asp183 of monkey CD9 with the corresponding amino acid residue in mouse CD9, revealed that substitution of Gly158 is critical for the reduction of the upregulation activity and secondly for the substitution of Val159 and Thr175. These three amino acid residues were deduced to be located on the head domain of the second extracellular loop, suggesting that interactions of CD9 with DTR or DT at the domain containing these three amino acids were important for the upregulation of DT binding.     

Quantal entry of diphtheria toxin to the cytosol

The rate-limiting step in diphtheria toxin (DT) intoxication of Vero cells has been determined utilizing cycloheximide as an inhibitor of the intoxication process. Cycloheximide is shown to inhibit the toxin catalyzed ADP-ribosylation of elongation factor 2 (EF-2). The inhibition is blocked by puromycin thus establishing the ribosome as the location of cycloheximide protection. Washing cells free of cycloheximide rapidly reverses the protective effect. The initial rates of protein synthesis inhibition observed after removal of cycloheximide from DT-intoxicated cells are 5 to 12-fold greater than rates observed in unprotected cells and are shown to reflect ADP-ribosylation of EF-2 by cytosolic DT. Ten to thirty minutes after cycloheximide removal, the rate of protein synthesis inhibition abruptly changes to values identical to those of unprotected cells. Both the initial rates and extent of the initial rapid inactivation are directly related to toxin concentration and time of incubation with DT in the presence of cycloheximide. We concluded that: the rate-limiting step in protein synthesis inhibition by DT is not the ADP-ribosylation of EF-2 by cytosolic toxin but rather the earlier entry step of DT into the cytosol. DT enters the cytosol as a bolus of sufficient size to rapidly inactivate all EF-2 in that cell. It is inferred from 1 and 2 that the first order inactivation rate exhibited by DT is the result of the probability of the release of a bolus of toxin to the cytosol of any cell in the population per unit time. Autoradiographic analysis of intoxicated cell populations support this two-population state model. The size of a single bolus or quantum of DT is calculated from data over the range of 10(-11) to 10(-9) M DT and is found to remain constant. We suggest that the cytosolic entry mechanism of DT results from a unique ability of the internalized toxin molecules to destabilize the vesicular membrane resulting in a random release of a bolus of toxin into the cytosol. Because the bolus size remains constant over a 50-fold change in receptor occupancy the possibility is raised that DT undergoes a post-receptor packaging process, package size remaining a constant and package number increasing with receptor occupancy.     

Expression cloning of a diphtheria toxin receptor: identity with a heparin-binding EGF-like growth factor precursor.

A monkey cDNA (pDTS) encoding a diphtheria toxin (DT) sensitivity determinant was isolated by expression cloning in mouse L-M cells. Mouse cells are naturally resistant to DT, because they lack functional cell surface receptors for the toxin. Unlike wild-type L-M cells, pDTS-transfected mouse cells are extremely toxin sensitive and specifically bind radioiodinated DT. Intoxication of the transfected cells requires receptor-mediated endocytosis of the bound toxin. The cDNA is predicted to encode an integral membrane protein that is identical to the precursor of a heparin-binding EGF-like growth factor. The DT sensitivity protein is thus a growth factor precursor that DT exploits as a receptor.     

Membrane-anchored heparin-binding EGF-like growth factor (HB-EGF) and diphtheria toxin receptor-associated protein (DRAP27)/CD9 form a complex with integrin alpha 3 beta 1 at cell-cell contact sites

Heparin-binding epidermal growth factor-like growth factor (HB-EGF) is a member of the EGF family of growth factors, which interact with EGF receptor to exert mitogenic activity. The membrane-anchored form of HB- EGF, proHB-EGF, is biologically active, providing mitogenic stimulation to neighboring cells in a juxtacrine mode. ProHB-EGF forms a complex with diphtheria toxin receptor-associated protein (DRAP27)/CD9, a tetra membrane-spanning protein that upregulates the juxtacrine mitogenic activity of proHB-EGF. We explored whether other proteins associate with DRAP27/CD9 and proHB-EGF. Immunoprecipitation with anti-DRAP27/CD9 resulted in preferential coprecipitation of integrin alpha 3 beta 1 from Vero cell, A431 cell and MG63 cell lysates. Anti-integrin alpha 3 or anti-integrin beta 1 coprecipitated DRAP27/CD9 from the same cell lysates. Chemical cross-linking confirmed the physical association of DRAP27/CD9 and integrin alpha 3 beta 1. Using Vero-H cells, which overexpress HB-EGF, we also demonstrated the association of proHB-EGF with DRAP27/CD9 and integrin alpha 3 beta 1. Moreover, colocalization of proHB-EGF, DRAP27/CD9, and integrin alpha 3 beta 1 at cell-cell contact sites was observed by double-immunofluorescence staining. At cell-cell contact sites, DRAP27/CD9 was highly coincident with alpha- catenin and vinculin, suggesting that DRAP27/CD9, proHB-EGF, and integrin alpha 3 beta 1 are colocalized with adherence junction- locating proteins. These results indicate that direct interaction of growth factors and cell adhesion molecules may control cell proliferation during the cell-cell adhesion process.     

The effect of receptor rapid-internalization signals on diphtheria toxin endocytosis and cell sensitivity

Diphtheria toxin enters toxin-sensitive mammalian cells by receptor-mediated endocytosis employing the heparin-binding EGF-like growth factor precursor as its receptor. We reported previously (Almond and Eidels, 1994) that cytoplasmic domain mutants of the toxin receptor and cells expressing wild-type receptor internalize toxin slowly, the rate being approximately that of normal turnover of the plasma membrane. To determine whether it was possible to increase toxin sensitivity by increasing the rate of toxin internalization, we constructed diphtheria toxin cytoplasmic domain mutant cell lines containing rapid-internalization signals from either the low density lipoprotein receptor or from the lysosomal acid phosphatase precursor. Although cells transfected with mutant receptor genes internalized toxin at a faster rate than those expressing the wild-type receptor, they showed a decrease in toxin sensitivity. This decreased sensitivity may be accounted for by an observed decrease in the number of toxin-binding sites and by an increased rate of toxin internalization and degradation. These results suggest that the rate of toxin internalization may not be the rate-limiting step in the cytotoxic process.     

Improvement of a recombinant anti-monkey anti-CD3 diphtheria toxin based immunotoxin by yeast display affinity maturation of the scFv

Recently, a bivalent recombinant anti-human CD3 diphtheria toxin (DT) based immunotoxin derived from the scFv of UCHT1 antibody has been made that shows enhanced bioactivity and is free from the side effects of Fc receptor interaction. In this case, the diminution of CD3 binding due to the placement of the scFv domain at the C-terminus of the truncated DT in single scFv immunotoxins was compensated by adding an additional scFv domain. However, this strategy was less successful for constructing an anti-rhesus recombinant immunotoxin derived from the scFv of FN18 antibody due to poor binding of the anti-rhesus bivalent immunotoxin. We report here that, by increasing the FN18 scFv affinity through random mutagenesis and selection with a dye-labeled monkey CD3epsilongamma recombinant heterodimer, we greatly improved the bioactivity of FN18 derived immunotoxin. The best mutant, C207, contained nine mutations, two of which were located in CDRs that changed the charge from negative to positive. Binding affinity of the C207 scFv to the monkey T cell line HSC-F increased 9.8-fold. The potency of the C207 bivalent immunotoxin assayed by inhibition of protein synthesis increased by 238-fold.     

The role of the diphtheria toxin receptor in cytosol translocation

The role of the receptor in the transport of diphtheria toxin (DT) to the cytosol was examined. A point-mutant form of DT, CRM 107 (CRM represents cross-reacting material), that has an 8,000-fold lower affinity for the DT receptor than native toxin was conjugated to transferrin and monoclonal antibodies specific for the cell-surface receptors T3 and Thy1. Conjugating the binding site-inactivated CRM 107 to new binding moieties reconstituted full toxicity, indistinguishable from native DT linked to the same ligand, indicating that the entry activity of the DT B chain can be fully separated from the receptor binding function. Like DT, the toxin conjugates exhibited a dose-dependent lag period before first-order inactivation of protein synthesis. Inactivation of the binding site of the toxin portion of the conjugate was found to have no effect on the kinetics of protein synthesis inactivation. The receptor used by the toxin determined the length of the lag period relative to the killing rate. Comparing the potency of CRM 107 conjugates with native DT, standardized for receptor occupancy, shows that new receptors can be as or more efficient than the DT receptor in transporting DT to the cytosol. The transferrin-CRM 107 conjugate, unlike native DT, was highly toxic to murine cells. All the data presented are consistent with a model that the DT receptor, other than initiating rapid internalization of the toxin to low pH compartments, is unnecessary for transport of the toxin to the cytosol and that membrane translocation activity is expressed by the DT B subunit independent of the receptor-binding site.     

Organization of diphtheria toxin in membranes. A hydrophobic photolabeling study

Diphtheria toxin (DT) is a disulfide linked AB-toxin consisting of a catalytic domain (C), a membrane-inserting domain (T), and a receptor-binding domain (R). It gains entry into cells by receptor-mediated endocytosis. The low pH ( approximately 5.5) inside the endosomes induces a conformational change in the toxin leading to insertion of the toxin in the membrane and subsequent translocation of the C domain into the cell, where it inactivates protein synthesis ultimately leading to cell death. We have used a highly reactive hydrophobic photoactivable reagent, DAF, to identify the segments of DT that interact with the membrane at pH 5.2. This reagent readily partitions into membranes and, on photolysis, indiscriminately inserts into lipids and membrane-inserted domains of proteins. Subsequent chemical and/or enzymatic fragmentation followed by peptide sequencing allows for identification of the modified residues. Using this approach it was observed that T domain helices, TH1, TH8, and TH9 insert into the membrane. Furthermore, the disulfide link was found on the trans side leaving part of the C domain on the trans side. This domain then comes out to the cis side via a highly hydrophobic patch corresponding to residues 134-141, originally corresponding to a beta-strand in the solution structure of DT. It appears that the three helices of the T domain could participate in the formation of a channel from a DT-oligomer, thus providing the transport route to the C domain after the disulfide reductase separates the two chains.     

Intracellular localization and degradation of diphtheria toxin

The internalization of surface-bound diphtheria toxin (DT) in BS-C-1 cells correlated with its appearance in intracellular endosomal vesicles; essentially no toxin appeared within secondary lysosomal vesicles. In contrast, internalized epidermal growth factor (EGF) was localized within both endosomal and lysosomal vesicles. Upon preincubation of cells with leupeptin, a lysosomal protease inhibitor, a threefold increase in the accumulation of EGF into lysosomes was observed. Under identical conditions, essentially all of the diphtheria toxin remained within endosomes (less than 2% of the intracellular diphtheria toxin accumulated in the lysosomal fraction), indicating that the inability to detect diphtheria toxin in lysosomes was not due to its rapid turnover within this vesicle. Following internalization of EGF or DT, up to 40% of the ligand appeared in the medium as TCA-soluble radioactivity. EGF degradation was partially leupeptin-sensitive and markedly NH4Cl-sensitive, indicating lysosomal degradation. In contrast, DT A-fragment degradation was resistant to these inhibitors, while B-fragment showed only partial sensitivity. These data suggest that the bulk of endocytosed diphtheria toxin is localized within endosomes and degraded by a pathway essentially independent of lysosomes.     

Enhancement of immunotoxin efficacy by acid-cleavable cross-linking agents utilizing diphtheria toxin and toxin mutants

We have utilized a new class of acid-cleavable protein cross-linking reagents in the construction of antibody-diphtheria toxin conjugates (Srinivaschar, K., and Neville, D. M., Jr. (1989) Biochemistry 28, 2501-2509). The potency of anti-CD5 conjugates assayed by inhibition of protein synthesis on CD5 bearing cells (Jurkat) is correlated with cross-linker hydrolytic rates. The maximum increase in potency of the cleavable conjugates over non-cleavable conventional conjugates is 50-fold and is specific for the CD5 uptake route as judged by competition with excess anti-CD5. The potency of conjugates made from diphtheria toxin and the anti-high molecular weight melanoma-associated antigen (HMW-MAA) is enhanced 3-10-fold by a cleavable cross-linker. However the potency of transferrin or anti-CD3 diphtheria toxin conjugates is only minimally enhanced (2-3-fold). Mutant diphtheria toxins, CRM103 and CRM9, previously shown to express less than 1/100 of the wild type in binding affinity were substituted into these conjugates as probes for possible intracellular toxin receptor interactions. Both mutants were equally as toxic to Jurkat target cells exhibiting 1/700 the wild-type potency. CRM9 non-cleavable conjugates were equally as potent as wild-type conjugates for transferrin and anti-CD3-mediated uptake but not for anti-CD5-mediated uptake where toxicity was reduced 60-fold over the wild-type analog. The cleavable cross-linker enhanced the toxicity of anti-CD5-CRM103 and anti-CD5-CRM9 conjugates, but potency was only 1/10 that of the analogous wild-type cleavable conjugate. These data are consistent with a model in which potentiation of toxicity of the anti-CD5 and anti-high molecular weight melanoma-associated antigen conjugates by the cleavable cross-linker occurs from an enhanced intracellular toxin-toxin receptor interaction that ultimately results in increased toxin translocation to the cytosol compartment. In contrast, these data indicate that the anti-CD3 and transferrin uptake systems do not require this interaction in agreement with previous work (Johnson, V.G., Wilson, D., Greenfield, L., and Youle, R. J. (1988) J. Biol. Chem. 263, 1295-1300).     

Characterization of high-order diphtheria toxin oligomers

In 3D domain swapping, a domain of a protein breaks its noncovalent bonds with the protein core and its place is taken by the identical domain of another molecule, creating a strongly bound dimer or higher order oligomer. For some proteins, including diphtheria toxin, 3D domain swapping may affect protein function. To explore the molecular basis of 3D domain swapping in a well-characterized protein system, domain-swapped oligomers of diphtheria toxin were produced by freezing and thawing under a variety conditions, including in various salts and buffers, and at various temperatures. Reaction yields were followed by high-performance size-exclusion chromatography. The traditional low pH pulse produced by freeze-thawing in mixed sodium phosphate buffer induces the oligomerization of DT, but the addition of alkali chloride salts was found to increase the yield in the order of Li(+) > Na(+) > K(+). Unexpectedly, oligomers also formed when DT was frozen and thawed in the presence of 1 M LiCl alone. Slower freezing and thawing of the mixture led to the production of more and larger oligomers. DT oligomers were also produced by exposure to acidic buffers, and were found by electron microscopy to adopt both linear and cyclized forms in a wide distribution of sizes. Upon the basis of these results, the model for the production of DT oligomers by freezing and thawing was expanded to include a salt-mediated pathway. We present a mechanism for the formation of high-order DT oligomers by acidification that takes into account domain swapping and hydrophobic interactions.