A Role for the Protease-sensitive Loop Region of Shiga-like Toxin 1 in the Retrotranslocation of Its A1 Domain from the Endoplasmic Reticulum Lumen*

Shiga-like toxin I (Slt-I) is a ribosome-inactivating protein that undergoes retrograde transport to the endoplasmic reticulum to exert its cytotoxic effect on eukaryotic cells. Its catalytically active A1 domain subsequently migrates from the endoplasmic reticulum (ER) lumen to the cytoplasm. To study this final retrotranslocation event, a suicide assay was developed based on the cytoplasmic expression and ER-targeting of the cytotoxic Slt-I A1 fragment in Saccharomyces cerevisiae. Expression of the Slt-I A1 domain (residues 1–251) with and without an ER-targeting sequence was lethal to the host and demonstrated that this domain can efficiently migrate from the ER compartment to the cytosol. Deletion analyses revealed that residues 1–239 represent the minimal A1 segment displaying full enzymatic activity. This fragment, however, accumulates in the ER lumen when directed to this compartment. The addition of residues 240–251 restores the translocation property of the A1 chain in yeast. However, single mutations within this region do not significantly alter this function in the context of the 251-residue long A1 domain or affect the toxicity of the resulting Slt-I variants toward Vero cells in the context of the holotoxin. Since this mechanism of retrotranslocation is common to other protein toxins lacking a peptide motif similar in sequence to residues 240–251, the present results suggest that the ER export mechanism may involve the recognition of a more universal structural element, such as a misfolded or altered peptide domain localized at the C terminus of the A1 chain (residues 240–251) rather than a unique ER export signal sequence.

Shiga-like toxin I (Slt-I) 1 is a member of a class of ER-routing protein toxins (ERTs) that undergo retrograde traffic to the ER of cells before relocating their enzymatic domain to the cytosol (1)(2)(3). Slt-I is composed of a catalytic A subunit non-covalently associated with a pentamer of B subunits responsible for binding to the glycolipid receptor, CD77 (also known as globotriao-sylceramide or Gb3) (4 -6). Receptor binding is followed by clathrin-mediated endocytosis and retrograde transport of the toxin through the Golgi apparatus en route to the ER lumen (3,7). While migrating through the secretory pathway, a proteasesensitive loop (residues 242-261) located in the C-terminal region of the A chain is cleaved, dividing the A chain into a B pentamer-associated A 2 domain (residues 252-293) and an enzymatic A 1 domain (residues 1-251) (8). The A 1 domain remains associated with the A 2 /B subunits complex by virtue of a disulfide bond between cysteines 242 and 261. This disulfide bond is ultimately reduced in the ER lumen, liberating the enzymatic A 1 domain that is subsequently retrotranslocated to the cytosol. The N-glycosidase activity of the A 1 fragment depurinates a single adenosine residue at position 4324 in the 28 S rRNA inhibiting protein synthesis and subsequently leading to cell death (9).
The mechanism by which the catalytic domain of Slt-I and other ERTs is exported from the ER lumen to the cytosol remains poorly defined. Proteolysis of the Slt-I A chain to an A 1 fragment is a required event for toxicity to occur in mammalian cells (10). Proteolysis exposes a hydrophobic peptide present in the A chain of Slt-I and other ERTs, an event that may facilitate its interaction with ER-resident proteins or the ER membrane (11)(12)(13). However, studying the role of this region in retrotranslocation in isolation of other processes of retrograde traffic and processing remains challenging in the context of cell-based assays.
Yeast represents an attractive model for analyzing how the enzymatic A 1 domain of Slt-I is able to escape from the lumen in the ER and kill the host cell. We report the development of a suicide assay in yeast where the fully active enzymatic A 1 fragment of Slt-I is rapidly targeted to the ER before it can accumulate in the cytosol. The ability of the Slt-I A 1 fragment to subsequently retrotranslocate to the cytosol from the ER is lethal to the host and was the basis of a screen to define a region of the toxin involved in this relocation process. This yeast assay was used to identify a C-terminal region of the Slt-I A 1 fragment involved in retrotranslocation.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-pRSATT was created by amplifying two regions of the GAL1 promoter from pRS854 and cutting them into the multiple cloning site of pRS316. The first region was created with a sense primer (STATA2, GAG AGA GAA TTC tac gct taa ctg ctc att gc) complementary to the sequence upstream of the GAL4 repeats and an antisense primer (ATATA2, GAG AGA GGA TCC GAG AGA GCA TGC gtt aat aga tca aaa atc atc gct tcg ctg) complementary to the sequence around the TATA element of the GAL1 promoter. This construct resulted in a product with an upstream EcoRI cloning site (underlined) and a downstream BamHI site (with a nested SphI site, both underlined), which was cloned into pRS316 between EcoRI and BamHI (creating an intermediate vector). The second region was created with a * This work was supported by funds from the Ontario Cancer Research Network, the Canadian Breast Cancer Research Alliance, and the United States Department of Defense Breast Cancer Research Program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Ontario Cancer Inst., 610 University Ave., Toronto, Ontario M5G 2M9, Canada. Tel.: 416-946-2967; Fax: 416-946-6529; E-mail: gariepy@uhnres.utoronto.ca. 1 The abbreviations used are: Slt-I, Shiga-like toxin I; DETOX, catalytically inactive Slt-I A chain; ER, endoplasmic reticulum; ERAD, endoplasmic reticulum-associated degradation; ERT, endoplasmic reticulum-routing toxin; PBS, phosphate-buffered saline. sense primer (SsphI, GAG AGA GCA TGC gta aat gca aaa act gca taa cca c) complementary to the TATA region of the GAL1 promoter and an antisense primer (AbamHI, GAG AGA GGA TCC ggg gtt ttt tct cct tga cg) complementary to the Ϫ1 region of the GAL1 promoter. The resulting product incorporated an upstream SphI site and a downstream BamHI site (both underlined) and was subsequently cloned into the intermediate vector containing the first fragment between SphI and BamHI. The final vector pRSATT-ER was created by cloning an ERtargeting sequence into the BamHI site of pRSATT. The sequence was created by gene synthesis from two complementary primers (fER, GAG AGA AGA TCT atg atg aag aaa aac aat gcg tta gca cta gcg ctt gcc ctt gcg cta gca ctg gct ttg gcc ctg gcc and rER, GAG AGA GGA TCC cgc gtt ggc ggt gcc tag cgc aag tgc aag cgc tag tgc taa cgc taa ggc cag ggc caa agc cag) with BglII and BamHI sites, respectively (both underlined). The primers were annealed and extended in a PCR reaction and subcloned into pBLUESCRIPT. The insert was then cut out with BglII and BamHI and cloned into the BamHI site of pRSATT. DNA cloning was carried out in the Escherichia coli strain DH5 [F Ϫ .80dlacZ M15 (lacZYA-argF) U169 deoR recA1 endA1 hsdR17 (r K Ϫ , m k ϩ ) supE44 Ϫ thi-1 gyrA96 relA1). The wild-type yeast strain W303a was provided by James Friesen (Banting and Best Department of Medical Research, University of Toronto, Ontario, Canada). RSY1293 and 1295 strains were obtained from Randy Schekman (Department of Cell and Developmental Biology, University of California at Berkeley).

Construction of Yeast Expression Vectors Encoding Slt-I A-
Cassettes encoding portions of the Slt-I A chain coding sequence were cloned into pRSATT and pRSATT-ER between BamHI and SacI. These cassettes were all generated by PCR using the sense primer P 1 in combination with the antisense primers (D 1 through D 13 ). The primers and corresponding products are shown in Table I. The overexpression plasmids (for the immunoprecipitation experiments) were created by cloning the wild-type Gal1 promoter into pRS416 between EcoRI and BamHI (creating pRS416Gal). The BglII-BamHI insert (coding for the ER-targeting sequence used to make pRSATT-ER) was then cloned into the BamHI site of pRS416Gal creating pRS416GalER. Cassettes coding for N-terminally myc-tagged and catalytically inactive (DT) A 1 variants were generated by PCR using the sense primer P 2 in combination with D 1 and D 6 (1-251 and 1-239 fragments, respectively) and cloned into pRS416Gal and pRS416GalER between BamHI and SacI. Thirty cycles of amplification (denature 94°C, 30 s; anneal 58°C, 30 s; extension 72°C, 1 min) were carried out with a PerkinElmer Life Sciences thermal cycler and Pwo polymerase (Roche Applied Science). The PCR products and vectors were cut with BamHI and SacI (Roche Applied Science) and cloned using T4 DNA ligase (New England Biolabs).
mRNA Isolation, Reverse Transcription-PCR, and Amplification-Total RNA was recovered from 10 7 yeast cells grown at 30°C in SCuracil medium made with 2% galactose. Cells were harvested in mid-log phase and RNA purified with the Qiagen RNeasy mini kit. RT-PCR was performed with the 1st Strand cDNA synthesis kit (Amersham Biosciences) according to manufacturer's instructions. PCR of cDNA was performed with primers P and D 6 for 40 cycles.
Steady-state Determination of Slt-I A 1 Levels in the Cytosol and ER-Twenty-five (25) ml cultures of W303a yeast cells transformed with pRS416, pRS416Gal mycDT1-251, pRS416GalER mycDT1-251, pRS416Gal mycDT1-239, and pRS416GalER mycDT1-239 were grown to an A 600 of 0.8. The cells were washed once in buffer 1 (100 mM Tris, pH 9.4, 10 mM dithiothreitol) and then resuspended in 1 ml of spheroplasting buffer (20 mM Tris pH 7.5, 1 M sorbitol, 2 g/ml zymolyase 100-T, in SC-uracil 2% galactose broth) and incubated for 1 h at 37°C. Spheroplasts were pelleted and resuspended in 1 ml of lysis buffer (20 mM HEPES, pH 6.8, 50 mM potassium acetate, 200 mM sorbitol, 2 mM EDTA, with protease inhibitors). Spheroplasts were lysed in a Dounce homogenizer on ice. The lysate was centrifuged for 5 min at 500 ϫ g to remove unbroken cells and nuclei. The supernatant was centrifuged for 1 h at 100,000 ϫ g in a TLA-100 ultracentrifuge to pellet ER membranes. The remaining cytosolic components in the supernatant were precipitated with trichloroacetic acid, washed in acetone and resuspended in 1 ml of RIPA buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.5% deoxycholate) prior to immunoprecipitation with a mouse anti-myc monoclonal antibody at 70 g/ml (9E10, kind gift from Bill Balch, The Scripps Research Institute). Briefly, the ER membrane pellet was washed once with lysis buffer and resuspended in 1 ml RIPA buffer. Immune complexes formed overnight at 4°C were captured on protein-G-Sepharose beads for 30 min at 4°C. The beads were washed once with PBS and then boiled in sample buffer for SDS-PAGE. A 1 constructs were detected by Western blot analysis using rabbit anti-Slt-I A chain polyclonal antibodies (Molecular Templates Inc., Toronto).
Yeast Transformation and Growth Properties (Spot Assay)-Yeast cells were transformed according to the lithium-acetate method (14). Five-ml overnight cultures were grown at 30°C of selective media supplemented with 2% glucose. These cultures were then pelleted and washed twice with PBS and resuspended in 1 ml of PBS. Samples were then diluted to an A 600 of 1.0 and serially diluted (10ϫ). Each dilution was then spotted (10 l) on SD-Ura plates supplemented with 2% galactose. The plates were incubated at 24, 30, or 37°C for 72 h.
Recombinant His 8 -tagged Protein Expression, Purification, and Rabbit Reticulocyte Lysate Assay-Constructs coding for the Slt-I A 1 fragments 1-238, 1-239, and 1-251 were cloned into the bacterial expression vector pET22b (Novagen) and the fragments expressed as N-terminal His-tag fusion proteins. Constructs coding for Slt-I holotoxins (wild-type and DETOX) were cloned into the pECHE10a vector (Molecular Templates Inc., Toronto, Ontario, Canada), expressed with a His 8 -tag at the N terminus of their A chain and purified using nickelnitrilotriacetic acid resin (Qiagen). The enzymatic activity of these molecules was tested in a rabbit reticulocyte lysate assay (TNT assay GAG AGA GAG CTC TCA CTA att cag tat taa tgc cac gct tcc AC SacI 241 TAG TGA D 5 GAG AGA GAG CTC TCA CTA cag tat taa tgc cac gct tcc AC SacI 240 TAG TGA D 6 GAG AGA GAG CTC TCA CTA tat taa tgc cac gct tcc cag AC SacI 239 TAG TGA D 7 GAG AGA GAG CTC TCA CTA taa tgc cac gct tcc cag AC SacI 238 TAG TGA D 8 GAG AGA GAG CTC TCA CTA agc tat taa tgc cac gct tcc cag aat tgc AC SacI L240A 240 D 9 GAG AGA GAG CTC TCA CTA att tat taa tgc cac gct tcc cag aat tgc AC SacI L240N 240 D 10 GAG AGA GAG CTC TCA CTA acg tat taa tgc cac gct tcc cag aat tgc AC SacI L240R 240 D 11 GAG AGA GAG CTC TCA CTA gtc tat taa tgc cac gct tcc cag aat tgc AC SacI L240D 240 D 12 GAG AGA GAG CTC TCA CTA tct ggc aac tcg cga tgc atg gtg atg aca att agc tat taa tgc cac gct tcc cag aat tgc kit, Promega) using the expression of luciferase as a measure of protein synthesis. The loss of relative light units is related to the ribosomeinactivating activity of Slt-I A chains. Cytotoxicity Assay on Vero Cells-The toxicity of wild-type Slt-I and the variant Slt-I (L240D) toward Vero cells were measured using the sulforhodamine B dye binding assay (15). Fifty-thousand Vero cells in 200 l of ␣-minimal essential medium were cultured in wells of 96-well plates and exposed to a range of toxin concentrations prepared in PBS for a 1-h period. The toxin-containing solutions were subsequently diluted with the appropriate medium containing fetal calf serum, and the treated cells were cultured for another 48 h. The medium was removed and the remaining adherent cells were fixed with ice-cold 10% trichloroacetic acid, air-dried, and stained with 0.4% sulforhodamine B (Molecular Probes, Eugene, OR) dissolved in 1% (v/v) acetic acid in water. The excess dye was washed away, and the remaining bound sulforhodamine B dye was extracted from the cells with 10 mM Tris base. The absorbance of the dye was read at 540 nm using a plate reader. Each point in the cytotoxicity curves represents the average of experiments performed in triplicate.
Circular Dichroism Experiments-CD spectra were recorded on an Aviv 62A DS circular dichroism spectrometer using a 0.5-cm path length rectangular cuvette with a 2-ml sample volume. Protein samples (16.9 M) of N-terminal His-tagged Slt-I A 1 fragments 1-238, 1-239, and 1-251 were prepared in sample buffer (25 mM sodium phosphate, pH 7.0, and 100 mM NaCl). Wavelength scans were recorded from 300 to 195 nm, with a 1-nm spectral bandwidth (1 nm between points) and an averaging time of 8 s. Triplicate spectra were recorded for each protein sample. 1 Fragment in Yeast-Two yeast expression vectors were constructed to study the retrotranslocation of the A 1 fragment of Shiga-like toxin 1 (Fig. 1A). The first vector, termed pRSATT, was constructed to express Slt-I A 1 chain variants directly into the cytosol under the control of an attenuated Gal1 promoter (16). The second vector, pRSATT-ER, was designed such that Slt-I A 1 variants would be expressed with an N-terminal ER-targeting sequence (17) to en-sure the concomitant synthesis and translocation of the nascent A 1 chains into the ER lumen. Folded A 1 mutant chains unable to retrotranslocate from the ER (or another compartment of the secretory pathway) back to the cytosol would result in a survival phenotype. A comparison of toxicity levels observed for cloned Slt-I A 1 variants expressed from pRSATT and pRSATT-ER was the basis of a method to delimit region(s) of the A 1 chain involved in retrotranslocation (Fig. 1B).

Expression of the Slt-I A
Identification of a Retrotranslocative C-terminal Peptide in the Slt-I A Chain-We hypothesized that distinct peptide domains mediated cytotoxicity and retrotranslocation and that a minimum cytotoxic fragment of the Slt-I A 1 fragment would not retain its ability to escape from the ER lumen. A series of cassettes coding for the wild-type Slt-I A 1 fragment (residues 1-251) and C-terminally truncated variants were created and cloned into pRSATT and pRSATT-ER. The expression of the 1-251 Slt-I A 1 fragment from pRSATT was lethal to the host ( Fig. 2A; fragment 1-251). Slt-I A 1 fragments truncated at their C terminus also remained lethal to the host until residue 239 was deleted ( Fig. 2A; fragments 1-238 and 1-239). The viability of yeast cells transformed with the pRSATT vector coding for residues 1-238 was indistinguishable from that observed in yeast transformed with the empty vector (pRSATT only) and in yeast expressing a catalytically inactive E167A/ R170A double mutant termed DETOX ( Fig. 2A). Wild-type Slt-I holotoxin and its detoxified variant as well as Slt-I A 1 fragments coding for either residues 1-238, 1-239, or the entire A chain were then separately expressed in bacteria, purified, and subsequently tested to confirm their ability to inactivate ribosomes in vitro (N-glycosidase activity) in a rabbit reticulocyte lysate (TNT) assay (Fig. 2C). As expected, the enzymatic activities of the A chain and of the 1-239 fragment were comparable with that of the wild-type Slt-I toxin, while the 1-238 fragment FIG. 1. A, yeast vector constructs used in this study. The pRSATT vector was used for the cytosolic expression of Slt-I A chain variants from an attenuated Gal1 promoter, while pRSATT-ER was designed to express Slt-I A chain constructs harboring the N-terminal signal sequence (SS) able to direct their localization to the ER lumen. B, schematic diagram of the yeast assay. Slt-I A variants are shown as rectangles with the cryptic hydrophobic domain hydrophobic domain depicted as a gray area and the HAT signal sequence required for ER-targeting as a black rectangle. Cloned toxin elements expressed in the cytosol lead to cell death. Targeting of Slt-I A variants to the ER lumen will also lead to cell death unless a mutation (M) in a putative retrotranslocative domain prevents its retrotranslocation. and the DETOX form of the toxin had no detectable catalytic activity (Fig. 2C). Genes coding for Slt-I A 1 1-251 and 1-239 were then cloned into pRSATT-ER to determine whether the deleted C-terminal region (residues 240 -251) was essential for routing the A 1 fragment from the ER lumen to the cytosol. Both Slt-I A 1 1-251 and 1-239 were equally toxic when expressed into the cytosol of yeast cells (Fig. 2B), while only the 1-251 construct remained toxic to cells when expressed with an Nterminal signal sequence (Fig. 2B, fragment ER 1-251). A survival phenotype was observed from the ER-routed A 1 1-239 variant implying that the truncated protein was unable to relocate back to the cytosol from the ER lumen or any other compartments along the secretory pathway or was simply secreted by the yeast cells ( Fig. 2B; fragment ER 1-239).
In view of the extreme toxicity of ribosome-inactivating proteins (18) such as Slt-I, yeast vectors harboring a strongly attenuated Gal1 promoter were used in this study to minimize the expression of Slt-I A 1 and thus avoid overwhelming the ER import pathway. As a result of this experimental constraint, cytotoxic A 1 variants could not be detected by conventional biochemical techniques (Western blot, immunoprecipitation, microscopy) when expressed from the attenuated promoter. Three sets of experiments were then performed to confirm that the Slt-I A 1 fragment is translocated into the ER lumen and that the survival of yeast cells expressing the ER-targeted 1-239 fragment was not due to the lack of toxin production.
First, the expression of messenger RNAs was confirmed using RT PCR for all A 1 mutants targeted to the endoplasmic reticulum and leading to a survival phenotype in yeast (Fig. 3A).
Second, catalytically inactive forms (E167A/R170A) of the A 1 chain variants 1-251 and 1-239 were constructed and overexpressed in yeast cells to monitor their distribution within cellular compartments. More specifically, myc-tagged versions of the 1-251 and 1-239 fragments were cloned into a high copy number vector (pRS416) and expressed under the control of the wild-type Gal1 promoter, with and without the ER signal sequence. W303a yeast cells were then transformed with these vectors and grown in galactose to induce the expression of the toxin variants. The yeast cells were fractionated into cytosolic and ER fractions, and the myc-tagged protein was immunoprecipitated from each fraction. Immunoprecipitated material was then separated by SDS-PAGE and Western blots were performed using anti-Slt-I A chain polyclonal antibodies. As expected, 1-251 and 1-239 toxin fragments were recovered from the cytosol but not from ER fractions of yeast cells expressing the 1-251 and 1-239 toxin fragments lacking the ER signal sequence (ϳ29-kDa band; Fig. 3B). Similarly, toxin was only recovered from the cytosol of yeast cells expressing the 1-251 fragment fused to the ER signal sequence (ϳ33-kDa band; Fig.  3B). In this case, a larger molecular mass band (ϳ33 kDa band; 1-251 fragment with ER signal sequence) is observed in the cytosol as expected from the cytoplasmic overexpression of this construct as well as a dominant cleaved A 1 chain band (ϳ29 kDa band) corresponding to the mass of fragment 1-251 lacking the ER signal sequence. These results suggest that the A 1 chain has been directed into the ER and then rerouted to the cytosol after removal of the signal sequence. Finally, the fragment expressing residues 1-239 fused to the signal-cleaved ER signal sequence was predominantly found in the ER fraction of yeast cells (ϳ29-kDa band) suggesting that it accumulated there after translocation and signal cleavage. Predictably, some of the overexpressed ER signal-containing version of the 1-239 fragment (ϳ33 kDa band) was also observed in its unprocessed form the cytosolic fraction.
Last, a genetic approach was devised to confirm that the survival phenotype of yeast expressing the ER-targeted 1-239 fragment was due to ER import and not due to a lack of enzymatic activity. Our strategy was based on the use of a cold-sensitive yeast strain, which expresses a mutant form of Sec61p that conditionally limits ER import. The strain RSY1295 is non-viable at 17°C but propagates in the temperature range from 23 to 37°C where the disruption in ER import is not so severe as to prevent cell growth (19,20). Transforming RSY1295 with the vector expressing ER-targeted 1-239 (ER 1-239; Fig. 3C) and growing this particular strain at a temperature causing a partial blockage of ER import should lead to a reduction in yeast viability due to the accumulation of newly synthesized toxic 1-239 fragments directly into the cytosol. At the permissive temperature for ER import (30°C), RSY1295 displayed the expected sensitivity to cytosolically expressed DETOX, 1-251 and 1-239, and ER-targeted 1-251 and 1-239 ( Fig. 3C; DETOX, 1-251, 1-239, ER 1-251, and ER 1-239) as was seen for the isogenic wild-type RSY1293 strain transformed with the same vectors. However, at 37°C, RSY1295 was 2 orders of magnitude more sensitive to the ER-targeted 1-239 than RSY1293 (Fig. 3C, ER 1-239). The same effect was observed, albeit to a lesser extent, at 24°C (Fig. 3C) indicating that the toxic A 1 fragment 1-239 with the N-terminal ERtargeting sequence was produced but poorly shuttled to the ER lumen, thereby resulting in ribosome inactivation. Similar results were obtained for all non-retrotranslocative A 1 mutants expressed in the cold-sensitive RSY1295 yeast strain (data not shown).
Taken together, these results suggest that A 1 chains fused to the ER signal sequence are produced as enzymatically active molecules and targeted to the ER lumen. Furthermore, the data indicate that the 1-251 fragment of the Slt-I A chain efficiently retrotranslocates from the ER, while the 1-239 fragment accumulates there.

Delimiting the Retrotranslocative Peptide of the Slt-I A 1
Fragment-Progressively longer versions of the Slt-I A 1 fragment were cloned into pRSATT-ER and expressed in yeast to determine the minimal peptide segment within the region 240 -251 necessary to facilitate the retrotranslocation of the A 1 chain from the ER lumen to the cytosol. The addition of Leu 240 to the C terminus of 1-239 had a dramatic effect on retrotranslocation, restoring toxicity to nearly that of the wild type A 1 domain (Fig. 4A, ER 1-240 and ER 1-251). The additions of Asn 241 and Cys 242 appear to partly mask the retrotranslocative potential of Leu 240 , while the addition of the three consecutive histidines from residues 243 to 245 restored retrotranslocation to the same level observed for 1-251 (Fig. 4A, ER  1-241, ER 1-242, and ER 1-245). The toxicity of all lengths tested was indistinguishable from wild type when expressed from pRSATT (data not shown).

Substitution of Leucine for Aspartic Acid at Position 240 in the 1-240 and 1-251 Slt-I A 1 Fragment Results in a Reduction
in Retrotranslocative Potential-The addition of Leu 240 to the 1-239 fragment of the Slt-I A chain restored most of its retrotranslocative potential. To determine whether this effect was due to the presence of a specific amino acid at position 240 or simply a chain length effect, leucine 240 was replaced in the 1-240 Slt-I A 1 fragment with alanine, asparagine, arginine, and aspartic acid and tested for cytotoxicity when expressed and routed to the ER using the pRSATT-ER vector. The toxicity of these ER-targeted variants was reduced (Fig. 4B) in relation to the ER-targeted 1-240 and 1-251 fragments (Fig. 4B). Fulllength Slt-I A 1 mutants (1-251) were then constructed to determine whether mutations at position 240 were blocking retrotranslocation specifically because of their location at the C terminus of the peptide. Leucine at position 240 was replaced either with alanine, asparagine, arginine, or aspartic acid in the context of the full-length Slt-I A 1 (1-251), and the constructs were tested for toxicity when expressed from pRSATT-ER. The retrotranslocation defects conferred by mutations at position 240 in the 1-240 Slt-I A 1 fragment were restored when residues 241-251 were added back to this chain (Fig. 4C) suggesting that single amino acid substitutions were not sufficient to disrupt the retrotranslocation event in the context of the full A 1 domain. Yeast expressing Slt-I A 1 variants harboring substitutions at position 240 from the pRSATT vector exhibited the same growth properties as yeast expressing wildtype Slt-I A 1 indicating that all of the constructs created FIG. 3. A, RT-PCR and amplification of cDNA from total RNA. mRNA could be detected in total RNA preparations from yeast cells expressing ER-routed 1-239, L240D 1-240 , and L240D 1-251 as well as DETOX but not in yeast cells transformed with an empty vector (pRSATT). The PCR product in all cases is represented by the ϳ720-bp band. B, yeast cells overexpressing a catalytically inactive (E167A/R170A), myc-tagged 1-251 and 1-239 Slt-I A 1 fragment and ER-targeted 1-251 Slt-I A 1 fragment accumulate protein in the cytosol alone (ϳ29-kDa band). A 1 fragments harboring an ER-targeting sequence migrate with an apparent mass of ϳ33 kDa. Yeast cells overexpressing ER-targeted 1-239 Slt-I A 1 fragment accumulate protein both in the cytosol (uncleaved ϳ33-kDa fragment) and in the ER (cleaved ϳ29-kDa fragment). C, the production and routing of ER-targeted 1-239 Slt-I A 1 fragment were confirmed using a temperature-sensitive Sec61p mutant yeast strain (defect in ER import). The expression of ER-targeted Slt-I A 1 fragment 1-239 (ER 1-239) in the RSY1295 yeast strain is ϳ100-fold more toxic at 24 and 37°C (cytoplasmic accumulation) in comparison with its expression in the isogenic wild-type strain 1293. Slt-I A 1 1-251 constructs expressed in the cytosol (1-251) and ER-targeted (ER 1-251), as well as cytosolically expressed Slt-I A 1 1-239 (1-239) and inactive variant (DETOX), displayed comparable toxicities in the RSY1295 strain at 24, 30, and 37°C temperatures. 10-Fold serial dilutions of each yeast culture were spotted on SD-Ura-Leu plates supplemented with 2% galactose. retained full enzymatic activity (data not shown).
Finally, the impact of a single L240D mutation in retrotranslocation was further addressed in an assay involving the entire AB 5 Slt-I holotoxin. Wild-type and L240D Slt-I holotoxins were expressed, purified, and their cytotoxicity toward VERO cells monitored to verify if the L240D mutant was less toxic toward mammalian cells. Interestingly, both the wild-type and L240D holotoxins exhibited identical toxicity profiles toward VERO cells (Fig. 5A) (see "Discussion").

Fragment 1-239 and Full-length Slt-I A 1 Domain (1-251) Adopt Similar Overall Structures in Solution-
The ER lumen hosts resident proteins and mechanisms that probe the folding and structural integrity of nascent proteins (21,22). Misfolded proteins are subsequently routed (retrotranslocated) to the cytosol for degradation (19,23). One of these pathways is typically exploited by antigen-presenting cells to degrade foreign viral proteins and to present a complex spectrum of peptide antigens to a limited set of major histocompatibility complex class I molecules in the ER lumen (24,25). In summary, ERresident mechanisms can recognize a broad range of altered or foreign proteins that may be perceived as misfolded elements (22). Such mechanisms are inherently less restrictive than the traditional concept of distinct receptors recognizing unique peptide signals. The Slt-I A 1 domain (1-251) as well as other ER-routed toxins may thus utilize such routing pathways, although none of these toxins share in common a domain homologous in sequence to residues 240 -251. If the absence of these residues blocks the ER export of fragment Slt-I A 1 (1-239) to the cytosol, one could conclude that the first 239 amino acids of Slt-I A 1 are not adopting a misfolded geometry in relation to the native Slt-I A 1 (1-251). Both segments Slt-I A 1 1-239 and 1-251 display comparable enzymatic activity (Fig. 2C) and thus must not differ significantly in terms of their overall structure. To confirm that these fragments do adopt comparable structures in solution, we expressed, purified, and recorded the circular dichroism spectrum of Slt-I A 1 1-238 (inactive A 1 chain; Fig. 2), 1-239, and 1-251 from 195 to 260 nm. All spectral profiles essentially overlap (Fig. 5B), suggesting that their overall secondary structures are similar. More precisely, neither the loss of catalytic activity nor the loss of retrotranslocation can be associated with a major conformational change. The addition of residues 240 -251 may thus restore the ER  (ER 1-251). The C-terminal sequence for each construct is shown on the right side of each panel. export event potentially through the presentation of a local misfolded C-terminal patch. This form of recognition would not depend on the preservation of a unique peptide motif and would explain the minimal impact of single mutations within residues 240 -251 of the 1-251 peptide domain. DISCUSSION ER-routed toxins, such as ricin, cholera toxin, and Shiga toxin, are endocytosed by eukaryotic cells and migrate in a retrograde fashion eventually reaching the ER lumen. Their catalytic domain must subsequently escape from this penultimate compartment and reach their cytosolic targets. The A 1 chain of Slt-I represents one such catalytic domain and was dissected in the present study to identify a C-terminal region serving a role in its retrotranslocation to the cytosol. More precisely, the Slt-I A 1 chain contains residues 1-251 of Shigalike toxin 1 A subunit. It harbors a hydrophobic segment within its C-terminal region (residues 224 -242; Fig. 6A), a region postulated to be involved in its ER export mechanism (12). Such hydrophobic domains have also been identified in the A chain of Shiga and Shiga-like toxins and in other ribosomeinactivating proteins such as abrin (12) and ricin (26) (Fig. 6B). To define a role for this region of the A 1 domain in retrotranslocation, S. cerevisiae cells were transformed with a suicide vector that either expressed the 251-residue-long cytotoxic A 1 domain in the cytosol of yeast cells or was directly routed to the ER lumen. This experimental approach essentially bypassed other routing events commonly associated with the intracellular trafficking of ER-routed toxins and directly addressed parameters associated with this final translocation event. As expected, the expression of either construct led to the eventual routing or deposition of this domain in the cytosol of yeast cells, resulting in the inactivation of ribosomes and in cell death (Fig.  2). The C terminus of the A 1 chain was subsequently truncated to delimit both the minimal catalytic domain and the retrotranslocative domain of the 251-residue-long A 1 domain of Slt-I. Our study revealed that the first 239 amino acids of the A 1 chain represented the minimal region displaying full catalytic activity when expressed in the cytosol of yeast cells. In contrast, the same fragment harboring an N-terminal ER-targeting sequence lacked the ability to migrate from the ER compartment to the cytoplasm and was consequently not toxic to yeast cells (Figs. 2 and 3).
The lengthening of the chain (240 -251) within the context of the ER-targeted A 1 chain subsequently demonstrated that the addition of either leucine 240 alone or the entire segment 240 -251 achieved a similar effect in terms of restoring the ability of the A 1 chain to exit the ER lumen and kill yeast cells. A series of single mutations at position 240 revealed that the impact of changes at this site was not sufficient to prevent the retrotranslocation event in the presence of the remaining Cterminal peptide (241-251) within the framework of the complete A 1 chain (residues 1-251) (Fig. 4). This finding was further confirmed when the cytotoxic activity of the wild-type AB 5 toxin and a variant harboring a L240D mutation displayed identical toxicity profiles toward Vero cells (Fig. 5A). These results suggest that the nature of the translocation mechanism is not based on the recognition of a specific peptide sequence but rather on a potentially less stringent peptide element, possibly the exposure of a misfolded or hydrophobic sequence. This hypothesis is further supported by the fact that the sequence alignment of ER-routed toxins did not reveal the presence of a peptide motif similar to residues 240 -251 of Shigalike toxin 1 (Fig. 6B). An examination of the crystal structure of the Shiga holotoxin indicates that residues 236 -240 of its A chain (part of its hydrophobic domain, underlined in Fig. 6) are normally buried in the intact holotoxin (27). An analysis of the surface exposure of residues 236 -240 in the A 1 fragment further suggests that only leucine 240 would be partially exposed to an aqueous environment upon removing the A 2 and B subunit domains, such as in our yeast system (as calculated using the software program NACCESS). The 1-239 A 1 fragment (with no exposed hydrophobic residues in the C-terminal region) may be perceived as stable or folded in the secretory pathway and thus not a substrate for ER chaperones. The addition of hydrophobic leucine 240 may locally destabilize the A 1 fragment making it a substrate for the retrotranslocative machinery of the cell. The substitution of leucine 240 with a polar residue such as aspartic acid (L240D) may favor the solvation of this region of the A 1 fragment, a conformational event that is masked in the context of the 1-251 A 1 fragment in yeast (Fig. 4). The fact that the L240D mutation had no effect in the context of the Slt-I holotoxin is not surprising, since this residue is not exposed in the presence of the A 2 fragment and B subunits. The overlapping circular dichroism profiles (195-260 nm) observed for A 1 fragments 1-238, 1-239, and 1-251 indicate that all three fragments share a similar if not identical overall fold. The solvent exposure of residues 240 -251 in A 1 fragment 1-251 would thus be expected to only result in a local structural difference beyond residues 1-238 (Fig. 5B). This finding and the fact that both A 1 fragments 1-239 and 1-251 are equally active (Fig. 2C) would suggest that the peptide element (conformationally altered or misfolded peptide element) that triggers the local exposure of a hydrophobic patch in the A 1 fragment is localized to the region 240 -251. Further support for this concept comes from two additional sources. In the context of the complete Shiga-like toxin 1, the A chain is proteolytically cleaved by furin during cellular trafficking. The cleavage site (Arg 251 -Met 252 (8)) is situated in the A chain protease-sensitive loop (residues 242-261), a region constrained by a single disulfide bond linking Cys 242 to Cys 261 . Experiments aimed at altering or removing this furin cleavage site or at cleaving the A chain with other proteases (trypsin, calpain) have demonstrated that the processing of the A chain to an A 1 fragment dramatically increases the toxicity profile of the toxin but does not alter the catalytic activity of A chain in a cell free system (28,29). Since the A 1 fragment is only released from the holotoxin A 2 and B subunits in the ER lumen as the disulfide bond between Cys 242 and Cys 261 is finally reduced, the proteolysis results suggest that it is at the level of the retrotranslocation step that ER-resident proteins may preferentially recognizes the cleaved A 1 domain. Second, studies with cholera toxin have recently demonstrated that the cleavage of its protease-sensitive loop is required for the interaction of its A chain fragment with the ER-resident protein, protein disulfide isomerase (30,31). Protein disulfide isomerase is a chaperone, which recognizes a broad spectrum of misfolded proteins and prevents their aggregation by catalyzing the rearrangement of their disulfide linkages (32). Native proteins do not represent good protein disulfide isomerase substrates in comparison with unfolded proteins suggesting that the mechanism of substrate recognition is based on misfolded peptide elements (33,34).
It has been suggested that a large number of ERTs reach the cytosol by exploiting the ER-associated degradation system (ERAD) (35). A short misfolded peptide element exposed, in the context of their respective ERT, through processing could thus serve as a universal remodeling mechanism leading to the recognition of A 1 -like fragments by chaperones and ER-resident proteins linked to the ERAD system. One can envision a retrotranslocation model where the cleavage of the proteasesensitive loop within the A chain of Slt-I (Fig. 6) results in the exposure of a short peptide element located at the C terminus of the A 1 fragment. This peptide signature encompassing residues 240 -251 of the A 1 fragment triggers a local remodeling event leading to an interaction with ER-resident proteins involved in the recognition and retrotranslocation of misfolded proteins (ERAD pathway).
One cannot rule out the possibility that the C terminus of the A 1 fragment forms part of a ligand region recognized by an ER receptor distinct from chaperones. Such an interaction could act to retain the A 1 fragment in the ER for a sufficient period of time for retrotranslocation to occur. The disruption of this interaction may allow the A 1 chain to further migrate to other compartments of the secretory pathway making it unavailable for the ER retrotranslocation event. One may recall that the holotoxin Slt-I normally reaches the ER through the interaction of its B subunits with the glycolipid Gb3 (3). The A 1 chain is subsequently released from the B pentamer in the ER lumen. The ER-targeting feature is not associated with the free A 1 chain itself. We did observe a small amount of the signalcleaved 1-239 fragment in the cytosol of yeast cells overexpressing the ER-translocated 1-239 fragment (Fig. 3B). This observation is consistent with the idea that the 1-239 fragment can retrotranslocate to the ER but may not be retained long enough in that compartment for an efficient retrotranslocation event to take place.