Antibody-mediated Routing of Diphtheria Toxin in Murine Cells Results in a Highly Efficacious Immunotoxin*

The chemical coupling of diphtheria toxin to an anti- murine Thy1 antibody resulted in the most efficacious immunotoxin to date. At 1 pg/ml the immunotoxin in- hibited protein synthesis of a Thy+ AKR murine cell at a rate of 1.4 logs/h, within the order of magnitude of the efficacy of native toxins. This is unusual since murine cells are highly resistant to diphtheria toxin. The conjugate is highly specific; Thy- AKR cells display no intoxication at 1 pg/ml even after 18 h. The effects of ammonia, acid pulsing of external media, and low temperature reveal some similarities and some differences between intoxication of sensitive cells by toxin and of murine cells by the antibody-toxin conjugate. The differences that result in the high efficacy of the antibody-toxin conjugate appear to result from the antibody-mediated routing. These results imply that murine cells possess an acidic compartment which can mediate toxin cytosolic entry. Unlike the Thy antigen, the toxin receptor on murine cells is unable to route the toxin to this cellular site. lo6 cells in a volume of 100 pl into V-shaped 96-well plates (Nunc). At the appropriate time, the plates were centrifuged at 400 X g for 5 min in a Beckman TJ-6 centrifuge (TH-4 rotor). The media was then siphoned off, and the cells were resuspended upon addition of new medium with a micro-pipettor. The above techniques and other general procedures concerning immunotoxin synthesis and characterization have been described previously in greater detail (21).

Natural protein toxins, such as the bacterial diphtheria and pseudomonas toxin, and the plant toxins ricin and abrin possess a remarkable capacity to inhibit cellular protein synthesis. These toxins are composed of two subunits, A and B, coupled by a reducible disulfide bond. The high efficacy is due to the presence of multiple domains that function to move the toxin through sequential steps towards cellular intoxication. The A subunit possesses enzymatic activity; the bacterial toxins inhibit protein synthesis through ADP-ribosylation of elongation factor 2 (1); the plant toxins modify the 60 S ribosomal subunit (2) through an N-glycosidase activity (3). The initial interaction with a cell is the binding of the toxin via the B subunit to a plasma membrane receptor (4,5 ) .
Endocytosis follows, but the toxin remains within the lumina of the endosome or organelle, inaccessible to the cytosolic substrate. It is the second function ascribed to the B subunit, cytosolic entry, that effectively moves the enzymatic activity past the membrane barrier into the cytosolic compartment resulting in protein synthesis inhibition. For recent reviews on the topic of bacterial and plant toxins, see Refs. 6 and 7.
These toxins are opportunistic in the sense that they utilize cellular receptors and processes that evolved for normal cellular growth. One remarkable exception to toxin potency concerns rodents. Mice, rats, and their isolated cells share a * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Lte of Mental Health, Bethesda, Maryland 20892
remarkable tolerance to diphtheria toxin. Compared to cell lines derived from sensitive species, such as guinea pigs, monkeys, and man, murine cells require concentrations approximately IO5 higher for comparable intoxication (8). The insensitivity exists despite the fact that murine cells possess toxin-specific binding sites (9-11) and can internalize levels of radiolabeled diphtheria toxin equivalent to that seen with toxin-sensitive cells (12, 13). Morphological examination of gold-labeled toxin in murine LM cells has demonstrated movement of internalized diphtheria toxin into lysosomes (14). Additionally, murine elongation factor 2 is fully susceptible to enzymatic inactivation by the toxin's A chain (15). Experiments involving hybrid cells formed between murine and sensitive cells demonstrated that the factor(s) responsible for sensitivity was dominant (16); it was also shown that the gene product for human sensitivity was located on the chromosome 5 (17). It thus appears that murine cells lack an essential element required for cytosolic entry.
Recent work by O'Keefe and Draper (18) and Guillemot et al. (19) has demonstrated that by coupling diphtheria toxin to transferrin or concanavalin A, a lo4 and lo3 increase, respectively, in potency was seen. We report here the construction of an anti-murine T cell antibody-diphtheria toxin conjugate that displays the highest inactivation rate of any immunotoxin to date, reaching 1.4 logs/h on Thy+ AKR target cells, at a concentration that displays no toxicity on nontarget Thy-AKR cells. Because of the antibody specificity of this conjugate, we can differentiate the novel efficacious route in murine cells from the normally inefficient one.
Synthesis of 0 x 7 and Transferrin-Diphtheria Toxin Conjugates-To purified diphtheria toxin (at 20 mg/ml) a 5-fold molar excess of 2-iminothiolane stock was added (at 1 mg/ml) in pH 8.0, 0.16 M borate. Following 90 min at room temperature, the thiolated toxin was desalted on a Zorbax GF-250 column and concentrated back to the original volume in Centricon 30 microconcentrators. To 0x7 or transferrin (at 15 mg/ml in phosphate-buffered saline) was added a 5-fold molar excess of rn-maleimidobenzoyl N-hydroxysuccinimide ester (stock at 1 mg/ml in dimethylformamide). After 20 min, the The abbreviation used is: HEPES, 4-(2-hydroxyethyl)-l-piperazineethane sulfonic acid. 15993 modified proteins were desalted on a centrifuge column and immediately added to a 10 M excess of thiolated toxin. After 1 h, residual thiol groups were modified with N-ethylmaleimide. Purification of the conjugates was achieved on a 21.5 X 600-mm TSK-G 3000 column.
Characterization of Conjugates-Peaks from the TSK-G 3000 column were concentrated with Centricon 30 microconcentrators and then examined on preformed 2-16% acrylamide gradient gels, as described previously (21). The conjugates used in this report contained one toxin per antibody molecule. Protein concentrations were determined by the bicinchoninic acid method (22; Pierce Chemical CO.).
Protein Synthesis Assay-The effect of immunotoxin or toxin on protein synthesis on cultured cells was determined as described previously (21). Typically, lo5 cells in leucine-free RPMI 1640 medium containing 2 mM NaHC03, 25 m M HEPES, pH 7.4, and 0.1% bovine serum albumin were added to the desired level of immunotoxin and lactose (100 mM), so that the final volume was 0.1 ml. This was done in a 96-well round-bottom plate. After the designated times, a 1 in 10 dilution of stock ~-[U-"C]leucine (0.01 mCi/ml) into leucinefree RPMI was added in a 10-pi volume to the cells. After a 60-or 30-min incubation, the cells were harvested on glass fiber filters with a Titertek cell harvester (Flow Laboratories) and then counted. All protein synthesis assay time points were done in quadruplicate. Unless otherwise specified in the figure legend, data points are the mean value with standard deviations of 10% or less of the mean value. When ['4C]leucine incorporation was determined at 15 "C the incubation was extended from 1 to 4 h.
Media Changes for Acid Pulse Experiment-The mid-incubation media changes in 96-well plates were achieved by aliquoting the lo6 cells in a volume of 100 pl into V-shaped 96-well plates (Nunc). At the appropriate time, the plates were centrifuged at 400 X g for 5 min in a Beckman TJ-6 centrifuge (TH-4 rotor). The media was then siphoned off, and the cells were resuspended upon addition of new medium with a micro-pipettor. The above techniques and other general procedures concerning immunotoxin synthesis and characterization have been described previously in greater detail (21).

RESULTS
Diphtheria toxin was chemically coupled to either transferrin or an anti-Thy1.1 antibody (0x7; Ref. 20) as described under "Materials and Methods." This antibody binds to murine T cell lymphocytes or leukemic cells. Incubation of either conjugate on a murine AKR leukemic cell line (Thyl.l+ SL 2) resulted in a rapid loss of protein synthesis (Fig. 1). Under similar conditions, the native toxin at 1000 ng/ml displayed no inhibition up to the testing period of 4 h (data not shown). Like the intoxication of sensitive cells by diphtheria toxin, there was a lag period during which no loss in protein synthesis occurred, followed by an exponential decline in cellular capacity to incorporate labeled leucine. These effects were concentration-dependent (data not shown) and thus displayed the kinetics generally associated with the efficacious entry process of native toxins.
To demonstrate the specificity of the antibody-toxin conjugate, an 18-h incubation of the 0x7-diphtheria toxin conjugate at varied concentration was carried out on both target (Thyl.l+) AKR SL2 and non-target (Thy-; Ref. 23) AKR K-36 cells. The cells were also exposed to varied levels of the native toxin. The resultant dose-response curve is shown in Fig. 2. Whereas the two cell lines display similarly low sensitivities to the native toxin, there is a 3.6 log difference in the conjugate potencies. With the Thy-K-36 cell intoxication by the conjugate is through interactions of the toxin moiety with the cell. The potency of the conjugate is less than the native toxin since the toxin moiety has been modified and is now sterically hindered. The specificity for the Thy+ cell demonstrates that it is not the chemical alteration which rendered the toxin efficacious, but rather some feature supplied by the antibody.
Prior to intoxication of sensitive cells, diphtheria toxin requires intravesicular acidification, resulting in a conforma- At the designated times ["C]leucine was added, and after 1 h, the extent of protein synthesis was determined (see "Materials and Methods"). The 0x7-toxin conjugate, composed of 1 toxin molecule/IgG molecule, reached 1.4 log/h; the monodiphtheria toxin-transferrin conjugate yielded 1 log/h. tional change in the bound toxin molecule that leads to an increased association with, or insertion into, the vesicular membrane (24,25). The critical nature of this process is demonstrated by the toxin resistance displayed by mutant cells which lack endosomal acidification (26), as well as by the protection of intoxication by the addition of lipophylic amines, such as ammonia (27). As shown Fig. 3, the antibodytoxin conjugate is also rendered nontoxic by the presence of ammonia. The protection acquired by a sensitive cell through the addition of ammonia can be bypassed, however, through acidification of the medium (28)(29)(30); that is, if sensitive cells with prebound toxin are pulsed with a low pH medium, the presence of ammonia offers little protection. Medium acidification increases toxicity by decreasing the lag period for diphtheria toxin intoxication of sensitive cells. times indicated in Fig. 4, the cells were pulsed with ["C] leucine to determine their ability to incorporate label. When the AKR cells were preincubated with the 0x7-toxin conjugate plus ammonia, an acid pulse did not result in the intoxication of the ammonia-protected cells. Additionally, the lag phase was not shortened; in fact, under the conditions specified, 1 h was added to the lag. Acidic p H values were found not to alter the binding of the antibody to the AKR cell. Binding of the iodinated antibody was unaffected in the tested pH range of 4.5-8.0 (data not shown).
T o further characterize the acidification process, we ex- FIG. 4. Effect of acid pulse on ammonia protection. 0x7diphtheria toxin conjugate at 500 ng/ml was bound to AKR SL2 cells (lo5) at 4 "C in leucine-free RPMI medium, pH 7.4, in 96-well plates.
The cells were then centrifuged and resuspended in either pH 7.4 (squares) or pH 5.0 (triangles) medium at 4 "C. After 10 min, the cells were again centrifuged and resuspended in pH 7.4 medium (open symbols) or the same medium plus 10 mM NH4CI (filled symbols). The cells were then warmed to 37 "C at time 0. Protein synthesis was then determined as outlined in Fig. 1 using 30-min pulses. Data represents the mean of four determinations with a standard deviation of 15% or less of the mean value.
amined the timing of the passage of the toxin conjugates through the acid dependent (ammonia protection) step. This method, as originally described for native diphtheria toxin (31), was used to determine the time point at which the first antibody-and transferrin-toxin conjugate molecules pass through the ammonia protection process. Extrapolation of the kinetic data for both conjugates to control values (the protection level afforded cells by early presence of ammonia) of protein synthesis indicate that both conjugates started passing through this step at 15 min (Fig. 5). This value is longer (by a factor of 3-4) than that seen for native diphtheria toxin intoxication of sensitive cells (31). The inability to achieve the acidification step by a low pH pulse of the medium, as well as the delayed timing of emergence through the ammonia protection step, suggests that the routing which leads to the efficacious intoxication of murine cells by the diphtheria toxin conjugate varies from that of the efficacious intoxication of the sensitive cells by the native toxin.
Another method for examining the routing of proteins through a cell is to lower the temperature of the cells. Various intracellular transport/processing elements are effectively hindered below certain temperature ranges (see "Discussion"). The intoxication of sensitive (non-murine) cells by diphtheria toxin is relatively insensitive to temperature within the range of 15-37 "C (31, 32). The intoxication of murine leukemic cells by the native toxin (see Fig. 6, top) is also relatively insensitive to lowered temperatures. All processes are slowed down at 15 "C, including protein synthesis; thus, the ["C] leucine pulse is extended from 1 to 4 h; in all other respects, the cell plus toxin incubation is identical for all temperatures. When the two toxin conjugates were examined under identical conditions, the intoxication was diminished at lowered temperatures (Fig. 6, middle and bottom). The efficacy seen a t 24 "C and above was lost at 15 "C. In a separate, but shorter incubation (6 h), protection was evident at 19 "C (data not shown). The intoxication seen at 15 "C may be through the inefficient temperature-insensitive route seen with the native toxin. While the inefficient intoxication processes of the native toxin is largely unaffected by temperatures below 20 "C, the cellular process that renders the cell sensitive to the toxin moiety does display temperature sensitivity. Thus, the remarkable increase in toxicity coincides with the acquisition of a temperature-sensitive process, presumably routing, sup- A similar experiment was performed with the K-36 cell line. This cell is resistant to the native toxin and the anti-Thy antibody-toxin conjugate (Fig. 2); however, it is sensitive to the transferrin-diphtheria toxin conjugate. When exposed to varying levels of this conjugate under the identical conditions described in Fig. 6, 50% inhibition of protein synthesis at 24 "C is reached at concentrations below 10 ng/ml. Reduction of the temperature below 20 "C resulted in greater than a 4 log reduction in potency. Thus, the routing or processing steps that are essential for efficacious diphtheria toxin activity are ligand specificity for the K-36 cells.
The sequence of the ammonia-and temperature-sensitive steps was determined by the inhibitor exchange procedures previously used in the study of toxin entry (29,33). The results are outlined in Table I. The 0x7-diphtheria toxin conjugate was bound to the murine cells at 4 "C for 1 h. The cells were then washed and the temperature was brought up to 15 "C. After 7 h, the temperature was elevated to 37 "C, and the cells' capacity to synthesis protein was examined. If ammonia was excluded from all steps in the incubation, protein synthesis was decreased by greater than 80%; however, the addition of ammonia to the medium prior to the elevation to 37 "C resulted in complete protection. This result implies that the temperature-sensitive step precedes the ammonia-sensitive step.

DISCUSSION
The diphtheria toxin-antibody conjugate constructed here has the highest efficacy of any reported immunotoxin. At 4.8

TABLE I
Sequence to temperature and ammonia steps Aliquots of cells (106/ml) were incubated at 4 "C in the presence or absence of 500 ng/ml 0x7-diphtheria toxin for 1 h, followed by centrifugation and resuspension in fresh medium at 4 "C, lacking conjugate, in the presence (sample 1) or absence (samples 2 and 3) of 10 mM NH4Cl. The cells were then warmed to 15 "C for 7 h. NH4Cl was then added to sample 3, and the temperature was elevated to 37 "C for 1 h for all the samples, followed by a 1-h [14C]leucine pulse. Determinations were done in quadruplicate and values represent the mean f standard deviation. This value was compared to toxin-free controls (% control).  (36), as well as inherent differences in cell types. A more informative comparison can be made by examining the inhibition rates of other 0x7-toxin conjugates on the same cell line. Comparisons are accomplished at concentrations of lo-' M, the 90% saturation point for 0x7 binding to AKR cells. 0x7-ricin A chain and 0x7-diphtheria toxin A chain (37) conjugates reached a maximal 0.14 and 0.02 log/h rate of protein synthesis inhibition, respectively. The holoricin conjugates in the presence of 100 mM lactose achieved a 0.14 log/ h rate for the monoricin conjugate and 0.40 log/h for the biricin conjugate (21). At 5 X M (0.3 receptor occupancy) ricin inhibited AKR protein synthesis at a rate of 5.6 logs/h, whereas a %fold higher concentration of diphtheria toxin yielded no detectable toxicity after 6 h (37); a 20-fold increase in toxin concentration resulted in 0.05 log/h (lom6 M; estimated KO for diphtheria toxin: lo-' M, Ref. 10). Thus, the conjugation of these two natural toxins to the same antibody resulted in a paradoxical shift in efficacy. Coupling ricin to 0x7 decreases efficacy by approximately 1 log; coupling diphtheria toxin to 0x7 results in a greater than 1 log increase. It has been the objective of this report to characterize the cellular mechanisms responsible for this shift.
Because of the antigen specificity of the conjugate, we have been able to demonstrate that the antibody brings about an altered routing of the toxin moiety to a cellular compartment presumably unreached by the conjugate in Thy-cells or the native toxin in murine cells in general. The procedures we used to study routing that follows receptor-mediated endocytosis were acid pulsing of external medium, prevention of vesicular acidification by treatment with amines, and inhibition of specific vesicle fusion reactions by low temperature. Many receptor-bound ligands, such as insulin (38), asialoglycoproteins (39), epidermal growth factors (40), and low density lipoproteins (41) are endocytosed via their receptor and then degraded within lysosomes. Transferrin is unusual in this respect; it is spared from degradation (42,43). Upon vesicular acidification, the transferrin-bound ferric ions are released, and the receptor-bound apotransferrin recycles back to the cell surface (44,45). The routing of transferrin and its receptor through the cell has been extensively studied and may involve more than one non-lysosomal pathway (46). One route involves a juxtanuclear Golgi region. Evidence for this has been both morphological (47-49) and biochemical, as receptor-bound asialo-transferrin and the asialo-receptor are transported, in a limited way, to the sialyltransferase compartment of the Golgi (50,51), as well as the site of Golgi mannosidase I (52). Thus, a route for bound transferrin to the lysosome is lacking; whereas, movement through the secretory pathway has been established. It is of interest that the addition of anti-transferrin receptor antibody does bring about the delivery of the receptors to the lysosome (53,54). This knowledge suggests that the transferrin-diphtheria toxin conjugate's intracellular routing would not involve the lysosome, but could involve other cellular acidifying compartments, such as the Golgi and Golgi-associated vesicles.
The effect of temperature on cellular routings has also been extensively studied. Receptor-bound ligands are not internal-ized at temperatures below 10 "C, and their delivery to lysosomes is inhibited by temperatures below 20 "C (55,56). Movement of newly synthesized proteins is also affected by temperature. Transfer of protein from the rough endoplasmic reticulum to the cis face of the Golgi occurs only at temperatures above 10 "C, whereas movement through the Golgi cisternae and entry into condensing vacuoles is inhibited by temperatures less than 20 "C (57-59). It is of interest that the efficacious route of the diphtheria toxin conjugates is temperature-sensitive, whereas the efficacious route of the native toxin in sensitive cells lacks temperature sensitivity (31, 321, as does the poorly efficacious route of the native toxin in murine cells (Fig. 6). Thus, the conjugation of the toxin to an alternative binding site results in an altered routing of the toxin through the murine cell, possibly through the lysosomal or Golgi pathways. However, toxin bound to murine cells is delivered to the lysosomal compartment, but efficient intoxication of the cell is lacking. The known processes affected so demarcatedly by moving the temperature below 20 "C are routing functions, presumably involving select types of membrane fusions. This does not preclude a temperature-dependent nonrouting process that is essential for diphtheria toxin intoxication in murine cells.
Whereas the transferrin and antibody-toxin conjugates may be endocytosed by different processes, the two conjugates begin to pass through the ammonia-sensitive compartment at the same time, approximately 15 min. This value differs from that of the native toxin with sensitive cells, which reach that point at 4-5 min. Acidification processes that would be affected by the addition of ammonia are known to occur in both the endosomal/lysosomal pathway as well as the secretory pathway of the trans cisternae of the Golgi (60,61). Morphological examination of the routing of gold-labeled diphtheria toxin in murine cells has demonstrated rapid internalization and appearance in lysosomes within 2.5-5 min. The toxin is largely excluded from coated invaginations (unlike sensitive cells) as well as from the Golgi region (14,15). The finding that the Thy antigen, which is also excluded from coated pits (62), effectively routes the toxin to a productive (lethal) intracellular site indicates that the endocytotic mechanism is not a determinant in toxin resistance. Transferrin is internalized through the coated pits (63,64); thus the two conjugates are internalized by different routes but start completion of their acid-requiring step at the same time. This is consistent with the prior finding that in murine cells labeled a*-macroglobulin (endocytosis through coated invaginations) and diphtheria toxin are found in the same prelysosomal vesicles (11) and with more recent work directly examining the (common) intracellular pathway of ligands internalized through coated and noncoated invaginations (65).
Our experimental results characterizing the effect of ammonia on the intoxication of a murine cell line by a diphtheria toxin conjugate are similar to and corroborate those of O'Keefe and Draper (18); however, we find differences in the effect of acidification of pre-bound conjugates as it pertains to bypassing ammonia protection. They found that media acidification could bypass the protection by ammonia. These differences may be due to the different murine cell types or different receptors. In the absence of ammonia, their transferrin-diphtheria toxin conjugate (1 pg/ml) resulted in a 1 log decrease in protein synthesis in less than 2 h, and in their 24h dose response curve, a 1 log decrease was achieved at levels less than 0.01 pg/ml. Media acidification of ammonia-protected cells resulted in a log decrease after 24 h (at a conjugate concentration of 1 pg/ml). This drop in efficacy might be explained by a complicating feature of transferrin; acidifica-tion of the medium induces release of bound ferric ions, which in turn decreases the affinity of the transferrin (now apotransferrin) conjugate for the cellular receptor (44,45), which would result in decreased toxicity. Pre-bound labeled 0x7 antibody displays no cellular dissociation upon acidification to below pH 5.0.
Heagy and Neville (66) found that acidifying the media did not bypass the ammonia protection of the native toxin bound to murine cells. In addition to the 0x7-diphtheria toxin conjugate, pseudomonas toxin and modeccin have an aminesensitive step when examined on murine cells, and like the conjugate, neither of the toxins are potentiated by media acidification (67, 68), as diphtheria toxin is on non-murinesensitive cell lines (30).
The sequence of the temperature-and amine-sensitive steps is defined by the results in Table I, suggesting that the temperature-sensitive process must occur prior to the aminesensitive process. This sequence, a temperature-sensitive process followed by vesicular acidification, is known to occur in two intracellular pathways, delivery of endosomal vesicles to the lysosome and movement into (or through) the trans Golgi cisternae. Early endosomal acidification, or for that matter, acidification of plasma membrane-bound conjugate on the murine cell, did not fulfill the acidification requirement as it does in sensitive, non-murine cells (30).
Media acidification of diphtheria toxin bound to sensitive cells results in the bypass of ammonia protection. Why media acidification is inadequate for intoxication of murine cells by the 0x7-diphtheria toxin conjugate is unknown; the toxin moiety may require processing prior to acidification or may require some cellular element for the actual translocation act. This element may by missing on the plasma membrane or early endosomal; presumably, it or an appropriate alternative is present at some distal intracellular site. The temperaturesensitive step may be this process or routing to that cellular compartment where acidification is productive for toxin entry. The lack of temperature sensitivity, as well as the ability of diphtheria toxin to gain immediate access to the cytosol of sensitive cells through medium acidification, implies that the essential processing in a sensitive cell occurs in a pre-lysosomal, pre-Golgi compartment. It is now evident that murine cells lack this early endosomal processing element, but nonetheless are able to translocate (through the membrane) the toxin's enzymatic activity, if it is transported through a temperature-sensitive process to a more distal intracellular site, such as the trans cisternae of the Golgi or beyond. The failure of this substitute site to render murine cells sensitive to the toxin is presumably due to the inability of the normal murine diphtheria toxin receptor to route to that productive site.
As the murine cell binding site for diphtheria toxin is relatively ineffective, the efficacious toxin-antibody conjugate has the specificity of the conjugated antibody. Thus, nature has supplied us with a model immunotoxin in vivo system. This laboratory is presently examining in greater detail the action of these conjugates in viuo, and in a syngeneic tumor system, this agent has achieved greater than 3 log kills.