Fused polycationic peptide mediates delivery of diphtheria toxin A ...

Proc. Natl. Acad. Sci. USA Vol. 93, pp. 8437-8442, August 1996 Cell Biology

Fused polycationic peptide mediates delivery of diphtheria toxin A chain to the cytosol in the presence of anthrax protective antigen STEVEN R. BLANKE*t, JILL C. MILNE*, ERICKA L. BENSON*, AND R. JOHN COLLIERt Department of Microbiology and Molecular Genetics, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115

Contributed by R. John Collier, March 28, 1996

delivery of two alternative A moieties; edema factor (EF, 89 kDa) and lethal factor (LF, 83. kDa). EF is a calmodulindependent adenylate cyclase, whereas LF acts by an unknown biochemical mechanisms to induce cytokine production in macrophages (and at high concentrations, lyses these cells) (3-6). According to the current model for anthrax toxin action (7), PA binds to a specific cell surface receptor (identity unknown) (8), and a cellular protease removes its amino-terminal 20-kDa domain (PA2o) (9), leaving the carboxyl-terminal 63 kDa (PA63) bound to the receptor. Removal of PA20 exposes a site on PA63 that can bind EF and LF competitively. The PA63-LF or PA63-EF complex is then internalized by receptor-mediated endocytosis and delivered to the endosome, where the low pH environment triggers translocation of EF and LF to the cytoplasm (10, 11). PA63 has multiple activities that are relevant to its translocation function. Besides binding EF, LF, and the receptor, PA63 also inserts and forms ion-conductive channels in artificial bilayers and cellular membranes (12-14). Like translocation, channel formation is dependent on acidic conditions (pH 5.0-5.5). Milne et al. (15) recently discovered that purified PA63 forms high molecular weight, dodecyl sulfate-resistant oligomers that appear in electron micrographs predominantly as heptameric rings. Similar or identical oligomers also form in cells incubated with PA, and their formation is blocked by inhibitors of internalization or endosome acidification that are known to block toxin action. Such correlations suggest that the oligomer is the form of PA63 that inserts and generates channels in membranes, and that it functions in the translocation of LF and EF. LF and EF are known to bind to PA63 by means of their respective amino-terminal domains (16-23). Genetic fusion of the LF 255-residue amino-terminal domain (LFN) to the catalytic domain of exotoxin A enabled the latter to bind PA63 and be delivered to the cytosol in the presence of PA, inhibiting protein synthesis and causing cell death (19). Recently, LFN was fused to either the amino or the carboxyl terminus of the diphtheria toxin catalytic domain (DTA), and the resulting fusion proteins were found to be equally effective in inhibiting protein synthesis in Chinese hamster ovary (CHO)-K1 cells in the presence of PA (23). Excess free LFN has been shown to block cytotoxicity of such fusion proteins, as well as the actions of EF and LF on cells (23). These findings imply that the binding of LFN to PA mediates entry of fused proteins to the cytosol. During our studies on the DTA fusion proteins we discovered that short, polycationic amino-terminal peptides could

The lethal factor (LF) and edema factor (EF) ABSTRACT of anthrax toxin bind by means of their amino-terminal domains to protective antigen (PA) on the surface of toxinsensitive cells and are translocated to the cytosol, where they act on intracellular targets. Genetically fusing the aminoterminal domain of LF (LFN; residues 1-255) to certain heterologous proteins has been shown to potentiate these proteins for PA-dependent delivery to the cytosol. We report here that short tracts of lysine, arginine, or histidine residues can also potentiate a protein for such PA-dependent delivery. Fusion of these polycationic tracts to the amino terminus of the enzymic A chain of diphtheria toxin (DTA; residues 1-193) enabled it to be translocated to the cytosol by PA and inhibit protein synthesis. The efficiency of translocation was dependent on tract length: (LFN > Lys8 > Lys6 > Lys3). Lys6 was '100-fold more active than Arg6 or His6, whereas Glu6 and (SerSerGly)2 were inactive. Arg6DTA was partially degraded in cell culture, which may explain its low activity relative to that of Lys6DTA. The polycationic tracts may bind to anionic sites at the cell surface (possibly on PA), allowing the fusion proteins to be coendocytosed with PA and delivered to the endosome, where translocation to the cytosol occurs. Excess free LFN blocked the action of LFNDTA, but not of Lys6DTA. This implies that binding to the LF/EF site is not an obligatory step in translocation and suggests that the polycationic tag binds to a different site. Besides elucidating the process of translocation in anthrax toxin, these findings may aid in developing systems to deliver heterologous proteins and peptides to the cytoplasm of mammalian cells.

Many protein toxins from both prokaryotic and eukaryotic sources act by delivering their enzymic moieties to the cytosol of target cells and covalently modifying intracellular substrates. Although the gross outline of the entry process is known for many such toxins, there is no toxin for which the mechanism of membrane traversal is understood in detail. Data elucidating such mechanisms are of interest in understanding pathogenesis at the molecular level, in using toxins to develop targeted therapeutic agents, and for delivering heterologous proteins or peptides to the cytoplasm. Intracellularly acting protein toxins are generally described as being composed of two functionally distinct moieties, A and B (1). The A moiety is responsible for enzymic activity, whereas B binds the toxin to its receptor and may also bear other determinants that facilitate delivery of A to the cytosol. Anthrax toxin is a member of the so-called binary class of bacterial toxins, in which the A and B moieties are synthesized and released from the bacteria as discrete proteins, which subsequently interact at the surface of the target cell (2). Among the known binary toxins anthrax toxin is unique in using a single B moiety, protective antigen (PA, 83 kDa), for

Abbreviations: LF, lethal factor; EF, edema factor; PA, protective antigen; PA63, carboxyl-terminal 63-kDa fragment of PA; LFN, aminoterminal 255 residues of LF; DT, diphtheria toxin; DTA, catalytic domain of diphtheria toxin; LFNDTA, fusion protein composed of LFN and DTA; CHO, Chinese hamster ovary; EF-2, elongation factor-2. *S.R.B., J.C.M., and E.L.B. contributed equally to this work. tPresent address: Department of Biochemical and Biophysical Sciences, University of Houston, Houston, TX 77204-5934. *To whom reprint requests should be addressed.

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.

8437

8438

Cell

Biology: Blanke et al.

substitute for the LFN moiety of LFNDTA fusions in promoting PA-dependent translocation. In this report, we document and characterize this phenomenon, and discuss its relevance to the process of translocation and the development of systems for delivering heterologous proteins and peptides to the cytosol.

MATERIALS AND METHODS Cell Culture. The CHO-Ki cell line was obtained from the American Type Culture Collection (CCL 61). Cells were grown in Ham's F-12 medium supplemented with 10% calf serum, 500 units/ml penicillin G, and 500 units/ml streptomycin sulfate (Life Technologies, Gaithersburg, MD). Cell cultures were maintained as monolayers and grown in a humidified atmosphere of 5% C02/95% air. Peptide Synthesis. The Lys6 peptide, KKKKKKGSGCG, was synthesized at the BCMP Biopolymers Facility, Harvard Medical School. Construction, Expression, and Purification of the Fusion Proteins. Standard protocols were used for all genetic manipulations (24). An entirely synthetic gene encoding DTA (S.R.B. and R.J.C., unpublished data) plus a 16-residue aminoterminal polyhistidine fusion peptide was used to generate the constructs described in these investigations. The polycationicDTA fusion proteins were generated by PCR reactions with primers containing the desired sequence and designed for annealing to the amino terminus of the synthetic DTA gene. All constructs were cloned into the Escherichia coli expression vector, pETl5b (Novagen), replacing the NcoI-BamHI fragment. The ligation products were transformed into E. coli XL1-Blue (Stratagene). The plasmid DNA was amplified, purified, and screened for the presence of the appropriate sequence. Those gene constructions possessing the confirmed sequences were then transformed into the E. coli expression host, BL21(DE3) (25). For protein expression, cultures of BL21(DE3)/pETl5b-peptide-DTA were grown in Luria broth/ampicillin (100 ,g/ml) to an OD600 of 0.6-1.0; protein expression was induced by addition of isopropyl f-Dthiogalactopyranoside (1 mM) for approximately 3 h. Cell lysates were then prepared by sonication and cleared by centrifugation prior to column desalting and/or purification. For experiments involving crude protein preparations, lysates were first desalted by passage through a PD-10 gel filtration column (Pharmacia) equilibrated with 10 mM Tris HCl (pH 7.5), 2 mM DTT, 100 mM NaCl. Concentrations of the DTA fusion proteins were then estimated by Western blot comparison with known amounts of standard protein. Two steps of column chromatography were employed to purify His6DTA and Lys6DTA. The hexahistidine region enabled His6DTA to be purified by affinity chromatography on a Nil '-charged column. This chromatography step, using the Qiagen system (Chatsworth, CA), was performed according to the methods determined previously for the purification of His6DTA (26). All buffers and resins were as specified by the manufacturer. The column was washed and the protein was eluted with imidazole. The eluted protein was desalted and further purified by anion-exchange chromatography in 20 mM Tris HCl (pH 8)/2 mM DTT (Mono Q column on the fast protein liquid chromatography system; Pharmacia). Approximately 10 mg of purified protein was obtained from one liter of culture. The hexalysine region enabled Lys6DTA to be purified by cation exchange chromatography. Crude lysates were resolved by means of high salt elution on a Whatman P-11 resin (Pharmacia) in 20 mM MOPS (pH 7.5)/2 mM DTT. The eluted protein was desalted and further purified by anionexchange chromatography as above. Approximately 1 mg of purified protein was obtained from one liter of culture. NAD:EF-2 (Elongation Factor-2) ADP-Ribosyltransferase Assay. The NAD:EF-2 ADP-ribosyltransferase assay measures the initial rates of incorporation of the ADP ribose moiety of

Proc. Natl. Acad. Sci. USA 93 (1996)

[32P]NAD into the trichloroacetic acid (TCA)-precipitable EF-2 fraction of the reaction mixture. The assay was performed essentially as described (26), with initial rates determined by the collection of three linear time points in duplicate. Reaction mixtures contained 50 mM Tris HCl (pH 8.0), 1 mM EDTA, 10 mM DTT, 50 jtg/ml-' BSA, 50 ,tM NAD, 0.5 ,LM EF-2, and enzyme. The reactions were incubated at 25°C and aliquots were removed from duplicate samples at 2, 3, and 4 min and pipetted directly onto 3 MM filter paper (Whatman). The filter pads were placed immediately into ice-cold 5 % TCA, and washed 3-5 times for 15 min by gentle agitation on a platform rocker until no radioactivity could be detected in the discarded wash solutions. The filter pads were then washed twice for 5 min in ice-cold methanol, dried, and counted with 3 ml of Beckman Ready Safe Liquid Scintillation Cocktail in a 1209 Rackbeta scintillation counter (LKB). Initial rates were calculated based on the increase in counts (minus background) over 5 min, with less than 10% of the reactants having been used. Protein Synthesis Inhibition Assay. CHO-Ki cells were plated at a density of 4 x 104 cells per well, in Costar 96-well cluster plates, approximately 18 h prior to the start of an experiment. PA (2 x 10-8 M) and fusion proteins (concentrations indicated in the figure legends) were added to cells in Ham's F-12 medium. After 24 h at 370C, the medium was removed, the cells were washed with PBS (GIBCO/BRL), and L-leucine-deficient medium (GIBCO/BRL) supplemented with L-[3,4,5-3H]-leucine (1 ,tCi ml-'; 1 Ci = 37 GBq; DuPont/NEN) was added. After 1 h, the cells were washed with ice-cold PBS followed by ice-cold TCA (10%). Protein synthesis was measured as the incorporation of radioactivity into acid insoluble material and expressed as the percentage of incorporation by unintoxicated control cells. All assays were performed in duplicate. Variations of this assay are indicated in the appropriate figure legends. RESULTS DTA is an especially useful reporter molecule for studying protein translocation to the cytosol. When delivered to this compartment, DTA catalyzes the highly specific NADdependent ADP ribosylation of elongation factor-2 (EF-2) and produces a dramatic and easily measured inhibition of protein synthesis. However, because DTA is virtually incapable of penetrating to the cytosol in the absence of its complementary B fragment, the purified protein only produces an effect on protein synthesis at concentrations (- 0-7 M) many orders of magnitude higher than those of the holotoxin. While characterizing the translocation of LFNDTA fusion proteins into CHO-Ki cells, we discovered that DTA with a hexahistidine affinity tag fused to its amino terminus (His6DTA; Fig. 1) exhibited higher PA-dependent cytotoxicity than unmodified DTA (Fig. 2). In the absence of PA, however, His6DTA showed no greater cytotoxicity than unmodified DTA. The specific ADP ribosyltransferase activity of purified His6DTA was within a factor of two of that of unmodified DTA, eliminating enzymic activity as a basis of the difference in cytotoxicity (data not shown). This was also true of the other DTA fusion proteins characterized in later experiments (data not shown). These findings suggested that the His6 tract enhanced the PA-dependent cytotoxicity of DTA by facilitating its delivery to the cytosol. To confirm that the PA-dependent inhibition of protein synthesis by His6DTA was in fact due to the NAD:EF-2 ADPribosyltransferase activity of the DTA moiety, we characterized additional His6DTA fusion proteins, each containing a single active-site mutation known to attenuate the DTA enzymic activity. The mutations tested-H21N, H21R, H21A, Y65A, and W50A-reduced ADP ribosyltransferase activity in vitro by approximately 3-, 70-, 120-, 670-, and 200,000-fold, respectively. Reductions in PA-dependent cytotoxicity were

Cell



8439

OTA ..S..--gQ-..L-vp R| 08aHa H3--

Hi.60TA

C

0

II

H*DOTA

U)

A1960TA

;s-s -R- R- R -R-R- R-3s-,S-(-L-VP- R-

Lys6OTA

G-S-s-K-K-K-K- K--s-s-a-r-v-P-t- l

GhJ6OTA

f G-S--E-E -E-E-E- -S-3-S-L-V-P-R-

(SSG)2OTA

G-S-S-S-S-G-S-S-G-s-s-G-L-V-P-R-

In

U,s 20

aLys30TA

G-s-s- - -K-

s-s-K-K-

LysSOTA

--V-

p-R1

K- K-K-RKR-3-G-L-v-P- -L

10o IZl

FIG. 1. Schematic representation of the DTA fusion proteins characterized. The amino acid sequences shown were fused to the amino terminus of the catalytic domain of diphtheria toxin (DTA; residues 1-189).

seen that corresponded roughly to the changes in enzymic activity (Fig. 3). These results imply that the cytotoxicity of His6DTA is dependent on the ADP ribosyltransferase activity of the DTA moiety within the cytosol. The hypothesis that the activity of the His6 fusion peptide was related to the partial positive charge of the histidine residues (pKa 6.5) at neutral pH was tested by replacing these residues with others of similar or different charge. The histidines were replaced with lysines (pKa -10.5) or arginines (pKa >12), both of which should be fully positively charged under the same conditions; with glutamates, which should be negatively charged; or with the sequence Ser-Ser-Gly-Ser-Ser-Gly (SSG)2, a motif predicted to possess no charge or significant secondary structure (Fig. 1). Consistent with the hypothesis, only polycationic tracts of amino acids enhanced PA-mediated translocation of DTA. Lys6DTA exhibited 100-fold greater activity in inhibiting protein synthesis than His6DTA, whereas Arg6DTA exhibited similar activity to His6DTA (Fig. 4). In contrast, Glu6DTA and (SSG)2DTA showed no greater activity than unmodified DTA. We suspect that the lower activity of Arg6DTA relative to Lys6DTA was due to proteolysis of Arg6DTA in the medium or at the cell surface. Western blot analysis of fusion proteins remaining in the culture medium after a 24-h incubation with cells in the absence of PA revealed that Arg6DTA had been converted to a lower molecular weight species similar in size to unmodifed DTA, whereas the other fusion proteins were unaltered (data not shown). 0 a

Q 0-n

.c

,7

1

'14

10 .12

1 0.10

10-

104

Protein (M) FIG. 2. LFNDTA and His6DTA inhibit protein synthesis in CHO-Kl cells. Purified DTA (A), LFNDTA (El), or His6DTA (0) was added to CHO-Kl cells in the presence (solid symbols) or absence (open symbols) of PA (2 x 10-8 M). After 24 h at 37°C, protein synthesis was measured as described. Results are expressed as the percentage of protein synthesis in the absence of added proteins.

10 lo

lo I

10 lo

0

14 lo

lo;

10

4

Protein (M) FIG. 3. Active site mutations in DTA reduce the cytotoxicity of His6DTA fusion proteins. Purified His6DTA (-), His6DTAH21N (a), His6DTAH21R (-), His6DTAH21A (Ol), HiS6DTAY65A (A), or HiS6DTAw5oA (A) was added to CHO-Kl cells in the presence of PA (2 x 10-8 M). Experimental details are as described in the legend to

Fig. 2.

Next we prepared additional DTA fusion proteins containing 3, 8, 10, or 12 lysine residues in the amino-terminal peptide to investigate the dependence of translocation activity on length of the polycationic tract (Fig. 1). Expression levels were satisfactory with the Lys3DTA and Lys8DTA constructs, but LysioDTA and Lys12DTA were produced in amounts insufficient for activity measurements. Crude E. coli lysates containing the fusion proteins were assayed for PA-dependent cytotoxicity in CHO-Ki cells. As shown in Fig. 5, inhibition of protein synthesis was dependent on tract length. The activity of Lys8DTA was several-fold higher than that of Lys6DTA and only about 10-fold lower than that of LFNDTA. The cytotoxicity of Lys3DTA, the least active member of the series, was similar to that of His6DTA. Lys6DTA purified by cation- and anion-exchange chromatography showed the same activity as Lys6DTA in crude extracts (data not shown), demonstrating that E. coli constituents in the extracts did not affect the activity measurements.§ These results reveal an increase in activity with tract length through Lys8. The ability to test longer lysine tracts will be dependent on identifying systems for stable expression of the corresponding fusion proteins. The simplest way in which polycationic leader peptides might substitute for LFN in mediating the PA-dependent entry of DTA would be by binding to the same site on PA63 as LF and EF. If this mechanism were operative, then free LFN should block cytotoxicity by competing for binding to PA63, as it does with EF, LF, and LFN-containing proteins. As shown in Fig. 6A, however, free LFN did not protect cells from the cytotoxic effects of the Lys6DTA fusion protein, even at 1 ,tM, a concentration 100-fold higher than that needed to show protection from LFNDTA. In complementary experiments, a short synthetic peptide containing an Lys6 tract (KKKKKKGSGCG) was found to provide measurable protection against Lys6DTA, but only at peptide concentrations approaching 1 mM (Fig. 6B). No protection against LFNDTA was seen, even at the highest concentrations tested. These findings support the conclusion that entry of Lys6DTA depends on its binding to saturable sites distinct from that used by EF and LF. The high concentration of KKKKKKGSGCG required to inhibit the activity of Lys6DTA suggests a low affinity of the synthetic peptide (and Lys6DTA) for the Lys6DTA-binding site, or the presence of a large number of binding sites.

§The fused Lys6 tract permitted DTA to bind to cation-exchange resins

at neutral pH, thereby providing an effective "affinity handle" for

purification.

8440

Cell



30 90 60

300-

0 10.12

10.11

1010

10 9

1 07

1

l0

10t3

FIG. 4. The amino acid composition of the amino-terminal fusion peptide affects cytotoxicity. E. coli lysate containing Lys6DTA (E), Arg6DTA (A), His6DTA (0), DTA (0), Glu6DTA (A), or (SSG)2DTA (E) was added to CHO-Kl cells in the presence of PA (2 x 10-8 M). Experimental details are as described in the legend to Fig. 2.

DISCUSSION

120 C

90

0In 60 C'a 20

a-

0

Io10

10o2

10

10

1'10

10

10

10

LFN (M)

12000

CL

60

Our results demonstrate that a short amino-terminal sequence of positively charged residues can functionally replace the 255-residue amino-terminal domain of LF in potentiating proteins for PA-dependent delivery to the cytosol. Fusion peptides containing tracts of six lysine, arginine, or histidine residues were shown to potentiate DTA for PA-dependent delivery to the cytosol, whereas a tract of six glutamic acid residues or a six-residue sequence composed of serine and glycine did not. Lysine proved to be the most active of the cationic residues, apparently because it is both fully charged and stable under the conditions of the cell culture assay. These charge effects likely take place at the neutral pH of the cell surface, where lysine is expected to be more positively charged than histidine. Differential effects based on charge are less likely at the low pH of the endosome, where both tracts would be fully positively charged. Tracts of eight lysine residues were more effective than shorter lysine tracts in potentiating DTA for cytosolic delivery, and were only about an order of magnitude less active than LFNHow can a short polycationic peptide functionally substitute for LFN? Currently the only well-established role of LFN is to bind its parent protein to PA63 at the cell surface, enabling the protein to be endocytosed and delivered to an acidic intracellular compartment (19-21, 23). LF has been shown to bind with high affinity to PA63 (Kd = 0.01 nM) (27). If the translocation event releases LF and other LFN-containing

0 C-

1o -12

.*

0

Protein (M)

10'd

i0

Protein (M) FIG. 5. The number of lysine residues in the amino-terminal fusion peptide affects cytotoxicity. E. coli lysate containing Lys3DTA (-), Lys6DTA (E), Lys8DTA (-), or LFNDTA (a) was added to CHO-Kl cells in the presence of PA (2 x 10-8 M). Experimental details are as described in the legend to Fig. 2.

30

0¾ 10 .12

10

10

4

10

-4

10

.4

Lys6 PepUde (M) FIG. 6. (A) LFN blocks cytotoxicity of LFNDTA, but not Lys6DTA. Purified LFN was added to CHO-Kl cells in the presence of PA (2 x 10-8 M) and either Lys6DTA (5 x 10-1 M, 0) or LFNDTA (1 x 10-11 M, *). (B) An Lys6 peptide reduces cytotoxicity of Lys6DTA, but not LFNDTA. An Lys6 peptide (KKKKKKGSGCG) was added to CHO-Kl cells in the presence of PA (2 x 10-8 M) and either purified Lys6DTA (5 x 10-10 M; a) or LFNDTA (1 x 10-11 M; a). Experimental details are as described in the legend to Fig. 2.

proteins into the cytosol, the affinity of these proteins for PA63

must presumably become attenuated after delivery to the

acidic compartment. This implies that an attenuation mechanism (perhaps pH triggered) must also be inherent to the system. LFN may serve an additional function besides binding to PA63 and releasing at the appropriate moment, but evidence for such a function is lacking. One hypothetical function would be to serve as a leader sequence for vectoral passage of LF across the membrane, but this seems unlikely because LFN mediates PA-dependent delivery to the cytosol regardless of whether it is fused to the carboxyl terminus or amino terminus of DTA (20, 23). If the minimal function of LFN is to bind PA63 under neutral conditions and undergo release under acidic conditions, a polycationic tag may fulfill such requirements through electrostatic interactions. Thus, if the tag were to bind electrostatically to a cluster of anionic residues on PA63 at the neutral pH of the external medium, it would be coendocytosed and delivered together with PA63 to an acidic intracellular compartment. There, the carboxylate groups forming the anionic region of PA63 should become partially protonated, attenuating the affinity of the polycationic tail for PA63. Dissociation would then occur under the same conditions that trigger membrane insertion of PA63 and subsequent EF/LF translocation (10-14). The finding that the action of Lys6DTA was competed by a small polycationic peptide is consistent with this model. The events in translocation that follow membrane insertion by PA63 and the putative release of a polycation- or LFN-tagged protein remain uncharacterized, but it has been speculated that the tagged protein may cross the membrane in

Cell



an unfolded (or partially folded) form by means of the aqueous channel formed by PA63 (23). If the polycationic tag binds to PA63, the site of interaction must be distinct from the LF/EF binding site, since free LFN did not block the activity of polycationic-tagged DTA. This finding also indicates that interaction with the LF/EF site is not an obligatory step in PA63-mediated translocation. Internalization of the polycation-tagged protein may not require direct binding to PA63, however. An alternative mechanism for internalization of the polycation-tagged protein would involve binding of the tag to other components of the cell surface (e.g., phospholipids or high molecular weight acidic glycoproteins) that are co-endocytosed with PA63, allowing the tagged protein to be delivered to the same endocytic compartment as PA63 and to cross to the cytosol by means of PA63. Regardless of the mechanism by which polycation-tagged DTA is translocated, the results presented may aid in developing a general system for delivering heterologous proteins into the cytosol of mammalian cells. Development of a general translocation system has a number of potential medical and research applications, including: therapy for certain genetic diseases by protein complementation; antigen presentation to elicit specific major histocompatibility complex class I-restricted immune responses and clonal expansion of the relevant CD8+ cytoplasmic T lymphocytes; modulation of the activity of cytoplasmic target proteins; conditional expression of a protein's biological activity; and introduction of a protein that has been modified in vitro (i.e., phosphorylated, radiolabeled, isoprenylated, epitope-tagged, mutated). Two general classes of toxins have been used to deliver heterologous proteins to the cytosol: the pore forming toxins and the intracellularly acting toxins. The pore forming toxins, such as the a-toxin from Staphylococcus aureus and streptolysin 0, permeabilize the plasma membrane, permitting both entry and release of macromolecules by diffusion (28-30). The advantage of these toxins is that the protein to be delivered does not need to be physically linked to a component of the translocation machinery. They have the major disadvantage, however, of inducing leakage of cellular contents into the external medium, thereby perturbing cellular function (28-30). The intracellularly acting toxins, such as diphtheria and anthrax toxins, do not induce cell leakage, but do require that the heterologous protein be linked to the toxin, a step generally necessitating genetic manipulation. This requirement and the potential instability of resulting fusion proteins have been a barrier to fully investigating the feasibility of this approach. Nonetheless, studies by Olsnes and coworkers of the properties of heterologous proteins fused to DT have yielded valuable information (31-32). Binary toxins, such as anthrax toxin, may have significant advantages over other classes of intracellularly acting toxins in cytoplasmic delivery of heterologous proteins. For a singlechain toxin in this category such as DT to be used, the toxin must contain a lesion that eliminates its own enzymic activity, and the heterologous protein must replace or be fused to the enzymic domain. Because these toxins are multidomain entities, there are significant risks that incorporation of the heterologous protein will affect folding, stability, and/or expression. Similar problems are inherent in the use of nonbinary oligomeric toxins such as cholera toxin, in which a modified A subunit incorporating the heterologous protein must assemble with the other subunits of the toxin before the protein is released from the bacteria (33). For the binary toxins, in which the A and B moieties associate on the surface of a mammalian cell, the difficulties of incorporating a heterologous protein may be minimized. In the case of LF, we know that the enzymic domain(s) of the A chain may be entirely deleted, leaving the LFN domain as the only region required for translocation (19-21, 23). This domain, which is stable and soluble, functions when fused to

8441

either end of a protein and apparently only needs to retain its affinity for PA63 to mediate translocation. The criteria for the fusion of heterologous proteins to LFN have not yet been defined, but, on the basis of studies with diphtheria toxin, they may include the absence of disulfide bridges within the A moiety (32). A second advantage of the anthrax toxin system is that, because the cloned A and B moieties are normally synthesized as entirely separate nontoxic proteins in different bacterial strains, one avoids the regulatory restrictions that hamper recombinant DNA studies with DT and certain other toxins. Our findings raise the possibility of substituting small polycationic peptides for LFN in designing PA-dependent delivery systems. Although the peptides we tested were less efficient than LFN, they may prove to be advantageous in other respects. Because of their strong charge, the polycationic peptides are almost certainly highly hydrated and lack secondary structure, thus eliminating any requirement for their own folding. The likelihood that such peptides will interfere with folding of fused heterologous proteins is unknown, but is likely to be minimal for similar reasons. Appropriate expression vectors with polylinker sites would facilitate fusion of a tract of lysine residues to the amino terminus of candidate heterologous proteins. Further studies are needed to determine whether the polycationic tracts retain PA-dependent translocation activity if incorporated at other locations than the amino terminus. Tagging DTA with a polycationic peptide permitted us to use cation-exchange chromatography to purify it, and this may prove to be a general advantage in expediting purification of such fusion proteins. These properties make polycationic tag-mediated, PA-dependent translocation an attractive area for further research. We acknowledge Dr. Jimmy Ballard for helpful discussions. This work was supported by National Institutes of Health Grants AI-22021 and AI-22848 (R.J.C.), and Postdoctoral Fellowship Award NIH8469

(S.R.B.). 1. Gill, D. M. (1978) in Bacterial Toxins and Cell Membranes, eds. Jeljaszewicz, J. & Wadstrom, T. (Academic, New York), pp. 291-332. 2. Considine, R. & Simpson, L. (1991) Toxicon 29, 913-936. 3. Leppla, S. H. (1982) Proc. Natl. Acad. Sci. USA 79, 3162-3166. 4. Hanna, P. C., Kochi, S. & Collier, R. J. (1992) Mol. Bio. Cell 3, 1269-1277. 5. Hanna, P. C., Acosta, D. & Collier, R. J. (1993) Proc. Natl. Acad. Sci. USA 90, 10198-10201. 6. Hanna, P. C., Kruskal, B. A., Ezekowitz, R. A. B., Bloom, B. R. & Collier, R. J. (1994) Mol. Med. 1, 7-18. 7. Leppla, S. H. (1995) in Bacterial Toxins and Virulence Factors in Disease, eds. Moss, J., Iglewski, B., Vaughan, M. & Tu, A. T. (Dekker, New York), pp. 543-572. 8. Escuyer, V. & Collier, R. J. (1991) Infect. Immun. 59,3381-3386. 9. Klimpel, K R., Molloy, S. S., Thomas, G. & Leppla, S. H. (1992) Proc. Natl. Acad. Sci. USA 89, 10277-10281. 10. Friedlander, A. M. (1986) J. Biol. Chem. 261, 7123-7126. 11. Gordon, V. M., Leppla, S. H. & Hewlett, E. L. (1988) Infect. Immun. 56, 1066-1069. 12. Blaustein, R. O., Koehler, T. M., Collier, R. J. & Finkelstein, A. (1989) Proc. Natl. Acad. Sci. USA 86, 2209-2213. 13. Koehler, T. M. & Collier, R. J. (1991) Mol. Microbiol. 5, 15011506. 14. Milne, J. C. & Collier, R. J. (1993) Mol. Microbiol. 10, 647-653. 15. Milne, J. C., Furlong, D., Hanna, P. C., Wall, J. S. & Collier, R. J. (1994) J. Bio. Chem. 269, 20607-20612. 16. Bragg, T. S. & Robertson, D. L. (1989) Gene 81, 45-54. 17. Labruyere, E., Mock, M., Ladant, D., Michelson, S., Gilles, A.-M., Laoide, B. & Barzu, 0. (1990) Biochemistry 29, 49224928. 18. Quinn, C. P., Singh, Y., Klimpel, K. R. & Leppla, S. H. (1991) J. Biol. Chem. 266, 20124-20130. 19. Arora, N. & Leppla, S. H. (1993) J. Biol. Chem 268, 3334-3341. 20. Arora, N. & Leppla, S. H. (1994) Infect. Immun. 62, 4955-4961. 21. Arora, N., Williamson, L. C., Leppla, S. H. & Halpern, J. L. (1994) J. Biol. Chem. 269, 26165-26171.

8442

Cell Biology: Blanke et al.

22. Little, S. F., Leppla, S. H., Burnett, J. W. & Friedlander, A. M. (1994) Biochem. Biophys. Res. Commun. 199, 676-682. 23. Milne, J. C., Blanke, S. R., Hanna, P. C. & Collier, R. J. (1995) Mol. Microbiol. 15, 661-666. 24. Ausubel, F., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith, J. A. & Struhl, K. (1987) Current Protocols in Molecular Biology (Wiley, New York). 25. Studier, F. W. & Moffatt, B. A. (1986)J. Mol. Biol. 189, 113-130. 26. Blanke, S. R., Huang, K., Wilson, B. A., Papini, E., Covacci, A. & Collier, R. J. (1994) Biochemistry 33, 5155-5161. 27. Leppla, S. H., Friedlander, A. M. & Cora, E. M. (1988) in Bacterial Protein Toxins, eds. Fehrenbach, F. J., Alouf, J. E., Goebel, W., Jeljaszewicz, J., Jurgen, D. & Rappuoli, R. (Fischer, Stuttgart, Germany), pp. 111-112.

Proc. Natl. Acad. Sci. USA 93

(1996)

28. Bhakdi, S. & Tranum-Jensen, J. (1985) Biochem. Soc. Symp. 50, 221-233. 29. Bhakdi, S. & Tranum-Jensen, J. (1987) Rev. Phys. Biochem. Pharmacol. 107, 147-223. 30. Bhakdi, S. & Tranum-Jensen, J. (1991) Microbiol. Rev. 55, 733-751. 31. Wiedlocha, A., Madshus, I. H., Mach, H., Middaugh, C. R. & Olsnes, S. (1992) EMBOfJ. 11, 4835-4842. 32. Falnes, P. O., Choe, S., Madshus, I. H., Wilson, B. A. & Olsnes, S. (1994) J. Biol. Chem. 269, 8402-8407. 33. Holmes, R. K., Jobling, M. G. & Connell, T. D. (1995) in Bacterial Toxins and Virulence Factors in Disease, eds. Moss, J., Iglewski, B., Vaughan, M. & Tu, A. T. (Dekker, New York), pp. 225-256.