Targeting of Yersinia Yop proteins into the cytosol of HeLa cells: one-step translocation of YopE across bacterial and eukaryotic membranes is dependent on SycE chaperone

Authors

  • Vincent T. Lee,

    1. Department of Microbiology and Immunology, UCLA School of Medicine, 10833 Le Conte Avenue, University of California, Los Angeles, CA 90095, USA.
    Search for more papers by this author
  • Deborah M. Anderson,

    1. Department of Microbiology and Immunology, UCLA School of Medicine, 10833 Le Conte Avenue, University of California, Los Angeles, CA 90095, USA.
    Search for more papers by this author
  • Olaf Schneewind

    1. Department of Microbiology and Immunology, UCLA School of Medicine, 10833 Le Conte Avenue, University of California, Los Angeles, CA 90095, USA.
    Search for more papers by this author

Olaf Schneewind E-mail olafs@ucla.edu; Tel. (310) 206 0997; Fax (310) 267 0173.

Abstract

Pathogenic Yersiniae adhere to and kill macrophages by targeting some of their Yop proteins into the eukaryotic cytosol. There is debate about whether YopE targeting proceeds as a direct translocation of polypeptide between cells or in two distinct steps, each requiring specific signals for YopE secretion across the bacterial envelope and for translocation into the eukaryotic cytosol. Here, we used the selective solubilization of the eukaryotic plasma membrane with digitonin to measure Yop targeting during Yersinia infections of HeLa cells. YopE, YopH, YopM and YopN were found in the eukaryotic cytosol but not in the extracellular medium. When bound to SycE chaperone in the Yersinia cytoplasm, YopE residues 1–100 are necessary and sufficient for the targeting of hybrid neomycin phosphotransferase. Electron microscopic analysis failed to detect an extracellular intermediate of YopE targeting, suggesting a one-step translocation mechanism.

Introduction

Upon entering their human host, most bacteria are phagocytosed and killed by the cellular immune system. Three pathogenic Yersinia species, Y. pestis, Y. pseudotuberculosis and Y. enterocolitica, have devised a mechanism that, upon direct contact, kills host cells and allows these microbes to multiply within lymphoid tissues and to establish disease (Straley et al., 1993a; Cornelis and Wolf-Watz, 1997). Contact with eukaryotic cells provides a signal for Yersinia to target cytotoxic proteins, Yops (Yersinia out proteins), into the host cells by a type III mechanism (Rosqvist et al., 1994; Petterson et al., 1996). Once within the eukaryotic cytosol, Yop proteins interfere with signal transduction and cytoskeletal rearrangement events, thereby allowing Yersinia to evade the infected host's defence (Bliska et al., 1991; Galyov et al., 1993; Straley et al., 1993b).

In the absence of eukaryotic cells, type III secretion can also be induced by low calcium concentration and temperature shift to 37°C, which results in the secretion of 14 different Yop proteins into the surrounding medium (Michiels et al., 1990). (Herein, secretion is defined as type III export of Yops into the extracellular medium, whereas targeting refers to the localization of Yops into the eukaryotic cytosol.) The signal sufficient for the secretion of reporter fusions in low calcium-induced cultures has been mapped to the first 15 codons of Yops (Michiels and Cornelis, 1991; Sory et al., 1995; Schesser et al., 1996). This signal is probably encoded within yop mRNA, because frameshift mutations that completely alter its protein sequence promote secretion of the fused reporter proteins (Anderson and Schneewind, 1997). A second independent type III export pathway has been revealed for YopE (Cheng et al., 1997). Mutant YopE with a defective secretion signal can be exported in a manner absolutely dependent on the presence of SycE chaperone. SycE is a small homodimeric protein that binds to residues 15–100 of YopE in the Yersinia cytoplasm (Wattiau and Cornelis, 1993; Woestyn et al., 1996), and this interaction is also sufficient to initiate YopE into the secretory pathway (Cheng et al., 1997).

Targeting of Yops into the cytosol of eukaryotic cells was first observed by immunofluorescence microscopy: YopE, YopH and YpkA (YopO) were found in the cytosol of HeLa cells that had been infected with Y. pseudotuberculosis (Rosqvist et al., 1994; Persson et al., 1995; Hakansson et al., 1996a). In another experiment, Y. enterocolitica were manipulated to express Yop fusions to Bordetella pertussis adenylate cyclase (Cya), which resulted in an increase in cAMP in the eukaryotic cytosol (Sory and Cornelis, 1994; Sory et al., 1995). The increase in cAMP is thought to be a measure of Yop targeting, because Bordetella adenylate cyclase absolutely requires calmodulin in the eukaryotic cytosol for enzymatic activity. Fusions of YopE, YopH and YopM to adenylate cyclase increased cytosolic cAMP levels, whereas YopN fusions did not (Boland et al., 1996). Shorter N-terminal Yop segments (YopE 1–50, YopH 1–71 or YopM 1–130) also increased cAMP levels, but fusions to the first 15 codons of Yops did not have this effect (Sory et al., 1995). Equal amounts of hybrid Cya proteins were found in the extracellular medium and in eukaryotic cell lysates, suggesting that the extracellular Yops may represent an intermediate step in Yop targeting (Boland et al., 1996). A specific role for Syc proteins in Yop targeting has not been demonstrated hitherto (Frithz-Lindsten et al., 1995).

On the basis of these observations, Yop targeting was proposed to occur by a two-step mechanism (Sory et al., 1995). Yops may first be exported across the bacterial envelope via their secretion signal (codons 1–15). The translocation domain of YopE, YopH and YopM (residues 15–50, 15–71 15–130 respectively) may then target these polypeptides from the medium or the bacterial surface into the eukaryotic cytosol (Sory et al., 1995; Boland et al., 1996). Here, we have developed a fractionation assay for Y. enterocolitica-infected HeLa cells to measure Yop secretion and targeting. Yops were located in the eukaryotic cytosol, the extracellular medium or remained associated with the bacteria. We could not detect an extracellular intermediate for those Yops that were localized to the HeLa cytosol. Targeting depended on the binding of SycE chaperone to YopE residues 15–100 in the Yersinia cytoplasm. During the infection of HeLa cells, the signal for YopE export in low calcium-induced Yersiniae (codons 1–15) did not lead to secretion into the extracellular medium. We propose a novel targeting mechanism that directs Yop proteins from the bacterial cytoplasm directly into the cytosol of eukaryotic cells.

Results

Fractionation of HeLa cells infected with Y. enterocolitica

We have developed an experimental scheme to fractionate infected tissue cultures and separate the eukaryotic cytosol from adherent bacteria. Digitonin interacts specifically with cholesterol-containing membranes (Esparis-Ogando et al., 1994). We reasoned that this detergent might cause the selective solubilization of the eukaryotic plasma membrane but not of bacteria, which lack cholesterol. To test this, HeLa cells were infected with Y. enterocolitica and incubated until signs of Yop toxicity could be observed. The growth medium was removed and centrifuged to sediment non-adherent bacteria from the extracellular medium. HeLa cells and attached Yersiniae were scraped off the culture flasks, extracted with digitonin and centrifuged to separate the bacteria and other insoluble material from the supernatant, containing eukaryotic cytosol. As a control, a duplicate sample of infected HeLa culture was treated with SDS to solubilize all eukaryotic and bacterial membranes. Protein in all fractions was precipitated with chloroform/methanol and analysed by immunoblotting.

Digitonin treatment released farnesyl protein-transferase (FPT) (Reiss et al., 1990) from the cytosol of HeLa cells into the extract supernatant (Fig. 1A). In contrast, bacterial chloramphenicol acetyltransferase (CAT) (Alton and Vapnek, 1979) was not solubilized by digitonin and sedimented with the bacteria. However, treatment with SDS released both FPT and CAT into the supernatant. No FPT was found in the extracellular medium, indicating that the plasma membrane of HeLa cells remained intact during the infection with Y. enterocolitica. Thus, digitonin treatment selectively disrupted the eukaryotic plasma membrane but not bacterial cells and was used to examine the location of Yop proteins. YopE, YopH, YopM and YopN were found in the supernatant of digitonin-extracted HeLa cells but not in the extracellular medium. Three Yops were secreted into the extracellular medium: YopB, YopD and YopR. YopB and YopR were also observed in the pellet fraction of digitonin extracts, whereas YopD was present in all the fractions examined. YopQ sedimented with the adherent bacteria into the pellet fraction of digitonin extracts.

Figure 1.

. Localization of Yop proteins during HeLa cell infections of Y. enterocolitica. Media (Med.) were decanted from HeLa cells and centrifuged to sediment non-adherent bacteria (P, pellet) and separate them from secreted Yops in the supernatant (S). HeLa cells and attached Yersiniae were extracted with digitonin (Dig.) to solubilize the HeLa plasma membrane specifically and release Yop proteins targeted into the eukaryotic cytosol. Unlysed bacteria, eukaryotic membranes and organelles were sedimented by centrifugation (P) and separated from the HeLa cytosol in the supernatant (S). As a control for the solubilization of all membranes, infected HeLa cells were extracted with SDS followed by centrifugation. Proteins were precipitated with chloroform/methanol, separated on SDS–PAGE, electroblotted and immunostained with specific antisera. Infections were with either (A) wild-type Y. enterocolitica W22703 or (B) yopN1 mutant strain VTL1. Yersinia Yop proteins, SycE and CAT as well as HeLa cell cytosolic FPT and nuclear TFIID were analysed. For wild-type Yersinia, 36% (± 6%) of YopE, 77% (± 9%) of YopH, 77% (± 21%) of YopM and 7% (± 10%) of YopN were soluble after digitonin extraction. Standard deviations (in brackets) were calculated from at least four independent measurements.

These results suggested that YopE, YopH, YopM and YopN were located in the eukaryotic cytosol, whereas other Yops were either secreted (YopB and YopR) or remained associated with the bacteria (YopQ). Yop targeting might be accomplished only by Yersinia that have attached to HeLa cells but not by non-adherent bacteria (Rosqvist et al., 1994; Petterson et al., 1996). To measure Yersinia adherence to HeLa cells, we counted bacteria in the extracellular medium and those present in detergent extracts by dilution and colony formation. Most bacteria (95%) were attached to HeLa cells during tissue culture infection. CAT, a constitutively expressed protein, was present in adherent as well as non-adherent bacteria (93% in the digitonin pellet and 7% in the media pellet). In contrast, SycE, the secretion chaperone of YopE in the cytoplasm of Yersinia, was found only in adherent bacteria that sedimented after digitonin extraction. Similar results were observed for all other Yops, with the notable exception of YopB and YopD (Fig. 1A). These data suggested that non-adherent bacteria synthesized YopB and YopD, whereas the attachment of Yersinia to HeLa cells provided a signal for the expression of other Yops.

The targeting signal of YopE

Previous work has suggested that the N-terminal 50 and 73 residues of YopE and YopH are sufficient to promote the targeting of adenylate cyclase fusions (Sory et al., 1995). However, about half of all adenylate cyclase fusion protein was located in the extracellular medium (Boland et al., 1996). Because this report was in conflict with our measurements of YopE and YopH localization, we examined the targeting of neomycin phosphotransferase (NPT) (Reiss et al., 1984) fusions to YopE with our fractionation assay (Fig. 2A). Hybrid proteins containing either full-length YopE or residues 1–100 of YopE fused to NPT were found in the supernatant of digitonin-extracted HeLa cells (34% and 81% respectively), indicating that they had been targeted in a manner similar to wild-type YopE (Fig. 2B). In contrast, fusions containing YopE residues 1–50, 1–15 or NPT alone remained in the pellet of digitonin extracts (0% targeting for all three hybrids). None of the hybrid YopE–NPT proteins were secreted into the extracellular medium.

Figure 2.

. The targeting signal of YopE. A. Translational fusions of YopE to neomycin phosphotransferase (NPT) were analysed for their targeting into the eukaryotic cytosol. B. Wild-type Y. enterocolitica strain W22703 (sycE +, yopN +). C. sycE1 mutant LC2 (sycE , yopN +). D. yopN1 mutant VTL1 (sycE +, yopN ) expressing YopE–NPT translational fusions infected HeLa tissue cultures for 3 h before digitonin fractionation as described in the legend to Fig. 1. Targeting was measured as the percentage amount of digitonin-solubilized YopE–NPT fusion: (B) YopE–NPT 34%, YopE1–100–NPT 81%, YopE1–50–NPT 0%, YopE1–15–NPT 0% and NPT 0%; (C) no targeting (0%) was observed for all YopE–NPT fusions in sycE1 strain LC2; (D) the yopN1 mutant strain VTL1 secreted YopE–NPT fusions into the surrounding medium. As a control, YopE1–15,+1–NPT (pDA72) harbouring a mutant secretion signal as well as NPT alone (pDA35) remained in the cytoplasm.

SycE is required for YopE targeting

To discern whether SycE binding to YopE polypeptide was required for targeting, we used Y. enterocolitica LC2 (sycE1 ), which lacks the cytoplasmic SycE chaperone (Cheng et al., 1997). This strain secretes YopE into the medium via its secretion signal (codons 1–15) under low calcium conditions. However, when tested during HeLa cell infections, YopE, YopE–NPT and YopE1–100–NPT remained in the pellet of digitonin extracts, indicating that neither targeting nor secretion had occurred (0% for each of the three strains) (Fig. 2C). The defect in the sycE1 strain was specific for the targeting of YopE–NPT fusions, as YopM and YopH were still located in the HeLa cytosol (data not shown). The targeting defect of strain LC2 was caused by the mutation carried by the sycE1 allele, because it could be complemented by a plasmid-encoded wild-type allele (L. W. Cheng and O. Schneewind, data not shown).

The presence of YopE–NPT fusions in the cytosol of HeLa cells was visualized with anti-NPT staining and immunofluorescence confocal laser microscopy (Fig. 3). To define the eukaryotic cytosol, the HeLa cell plasma membrane was stained with Texas red-labelled wheatgerm agglutinin. Targeting of YopE–NPT and YopE1–100–NPT could be detected as Oregon green staining of the HeLa cytosol. YopE1–50–NPT, YopE1–15–NPT or NPT alone yielded the same amount of background staining as a control culture infected with Yersiniae that did not express NPT. During infection with the sycE1 mutant Y. enterocolitica LC2 strain, no fluorescent staining was observed for either YopE–NPT or YopE1–100–NPT. Together, these results demonstrate that SycE binding to YopE substrate is absolutely required for targeting of the polypeptide into the HeLa cell cytosol.

Figure 3.

. Immunofluorescent detection of YopE–NPT fusions by confocal laser microscopy. HeLa cells were infected with Y. enterocolitica strains expressing YopE–NPT fusion proteins, and targeting was detected by indirect immunofluorescence using anti-NPT and Oregon green-conjugated secondary antibody. The HeLa plasma membrane was stained with Texas red-conjugated wheatgerm agglutinin. Infections of Y. enterocolitica W22703 harbouring (A) pDA36 (YopE–NPT), (B) pDA44 (YopE1–100–NPT), (C) pDA45 (YopE1–50–NPT), (D) pDA46 (YopE1–15–NPT), (E) pDA72 (YopE1–15,+1–NPT), (F) sycE1 strain LC2 expressing YopE–NPT (pDA36), (G) W22703 pDA35 (NPT) and (H) W22703 no plasmid control.

The secretion signal (codons 1–15) and YopE targeting

To examine the role of the secretion signal encoded within the first 15 codons of yopE in targeting, we analysed YopE–NPT mutants in which this signal had been either mutated or replaced. Hybrid proteins were examined for their ability to serve as substrates in both the targeting reaction as well as the low calcium-induced secretion of either wild-type or sycE1 mutant Yersiniae (Fig. 4). Replacement of the first 15 codons of YopE with those of two non-secreted gene products, E. coli chloramphenicol acetyltransferase (cat ) (Takeshita et al., 1987) and β-galactosidase (lacZ ) (Kalnins et al., 1983), did not affect secretion in low calcium-induced wild-type Yersiniae (Fig. 4). As reported previously, the YopE+1–NPT fusion protein is also secreted by wild-type Yersiniae (Cheng et al., 1997). The SycE dependence of YopE–NPT secretion was tested in Y. enterocolitica LC2. Low calcium-induced secretion was absolutely dependent on the binding of SycE to YopE residues 15–100, because the three hybrid proteins remained in the cytoplasm of the sycE1 mutant strain (Fig. 4). When tested during the infection of HeLa cells, all three mutant YopE–NPT fusions sedimented with the bacteria after digitonin extraction (Fig. 4). Thus, the binding of SycE to YopE residues 15–100 is not sufficient for the targeting reaction, and some property of the first 15 codons or amino acids of YopE must also be recognized by the type III machinery.

Figure 4.

. Substrate requirements for YopE secretion and targeting by the type III machinery. YopE–NPT fusion proteins were expressed in either wild-type Y. enterocolitica W22703 or sycE1 mutant strain LC2. Targeting of YopE–NPT fusions during the infection of HeLa cells was performed as described in the legend to Fig. 1. Digitonin-extracted cells were centrifuged to separate the supernatant (S), representing the HeLa cytosol, from the sedimented pellet (P) containing intact bacteria. Secretion was measured by low calcium- and temperature-induced type III export of YopE–NPT proteins into the culture medium (−Ca2+). Cultures were centrifuged, and sedimented Yersiniae (P) were separated from the supernatant (S) containing secreted Yops. Samples were precipitated with either chloroform/methanol (targeting) or TCA/acetone (secretion) and analysed by immunoblotting. Fusions were constructed by replacing the first 15 codons of yopE, specifying its secretion signal, with those of chloramphenicol acetyltransferase (cat )(pVL33), β-galactosidase (lacZ )(pVL35) and a + 1 frameshift mutation (pDA141) of the wild-type yopE secretion signal (yopE + 1 )(pDA139).

The YopE secretion signal in yopN mutant Yersiniae

Our result suggested that the secretion signal of YopE alone may not be functional during the infection of HeLa cells by Y. enterocolitica. In contrast to low calcium-induced secretion, in which all Yop proteins were secreted into the medium, the export of Yops during the infection of HeLa cells appeared to be regulated and directed to three distinct locations. yopN  mutants of Y. pseudotuberculosis or Y. enterocolitica are known to secrete large amounts of Yops into the extracellular medium even during tissue culture infections (Forsberg et al., 1991; Petterson et al., 1996; Boland et al., 1996). To examine a possible role for YopN in preventing the secretion of YopE, we constructed Y. enterocolitica strain VTL1 (yopN1 ). As expected, this mutant strain was temperature sensitive for growth and secreted Yops into the culture medium at 37°C even in the presence of calcium (data not shown). During the infection of HeLa tissue cultures, Y. enterocolitica VTL1 secreted all Yops into the extracellular medium (Fig. 1B). Some YopE, YopH and YopM could be found in the supernatant of digitonin extracts, suggesting that the yopN1 strain may still be able to promote Yop targeting. A plasmid-encoded wild-type yopN allele restored the fractionation pattern of Yops to that observed for Y. enterocolitica W22703, indicating that the phenotype of strain VTL1 was caused by the mutation carried by the yopN1 allele.

To test whether the secretion signal of YopE was functional in the yopN  mutant, the location of YopE–NPT fusions was examined during the infection of HeLa cells with Y. enterocolitica VTL1 (Fig. 2D). YopE–NPT, YopE1–100–NPT, YopE1–50–NPT and YopE1–15–NPT were secreted into the extracellular medium. As a control, NPT alone as well as YopE1–15,+1–NPT (pDA72), harbouring a defective secretion signal, sedimented with the yopN1 mutant bacteria. Thus, during the infection of HeLa cells the secretion signal located within the first 15 codons of YopE is functional in yopN  mutants but not in wild-type Yersiniae.

Electron microscopic detection of YopE–NPT fusions

Digitonin fractionation alone cannot distinguish between intrabacterial Yops and those that might be located on the surface of either bacterial or eukaryotic cells. For example, YopE–NPT fusions that sedimented after digitonin extraction (YopE1–50, YopE1–15) could be located either in the cytoplasm or on the surface of Yersiniae. If the latter were true, such a result would favour the two-step translocation model of Yops, whereas the cytoplasmic location of these NPT fusions would indicate that targeting occurred by a different mechanism. The location of YopE–NPT fusions was measured by electron microscopy. When infected with Y. enterocolitica expressing YopE–NPT, NPT-specific immunogold particles were detected in the cytosol of HeLa cells and in the bacterial cytoplasm but not on the surface of Yersiniae (Fig. 5 and Table 1[link]). YopE1–15–NPT was found in the bacterial cytoplasm but not on cell surfaces or in the HeLa cytosol. Similarly, when infected with the sycE1 mutant Y. enterocolitica LC2, immunogold staining of YopE–NPT was observed only in the bacterial cytoplasm but not in the cytosol of HeLa cells or on cell surfaces. Together, these data suggest that YopE–NPT fusions that were not targeted into the HeLa cytosol remained within the bacterial cytoplasm.

Table 1. . Immunoelectron microscopic detection of YopE–NPT fusion proteins. a. YopE–NPT fusion proteins were detected with anti-NPT followed by protein A–gold conjugate staining. Gold particles were counted and averaged per bacterium or μm2 of HeLa cytoplasm. Data were gathered from 25 bacteria.b. Fusion proteins were expressed from low-copy-number plasmids, either full-length YopE–NPT (pDA36) or YopE1–15–NPT (pDA46).Thumbnail image of
Figure 5.

. Immunoelectron microscopic detection of YopE–NPT targeting. HeLa cells were infected with Y. enterocolitica W22703 (pDA36) as described in the legend to Fig. 1 and processed for transmission electron microscopy. Thin sections were incubated with anti-NPT and protein A–colloidal gold conjugate (9 nm) followed by staining with uranyl acetate and lead. See Table 1 for quantification of gold particles.

Discussion

Examination of the signals for the secretion of Yop proteins in low calcium-induced cultures leads to the identification of two distinct modes of type III secretion (Cheng et al., 1997). One pathway recognizes a signal encoded within the first 15 codons of all Yops examined so far (Sory et al., 1992; Schesser et al., 1996; Anderson and Schneewind, 1997). In most cases, this signal can be mutated by the insertion or deletion of nucleotides immediately after the start codon without loss of function (Anderson and Schneewind, 1997). This result, taken together with the fact that Yop proteins do not share either peptide sequence homology or common amino acids, suggests that a property of the mRNA may be recognized by the type III machinery. However, during HeLa cell infections, the YopE secretion signal alone did not lead to the secretion of fusion proteins into the extracellular medium. This result does not, of course, exclude the possibility that similar signals from other Yops may lead to secretion during infection.

The second mode of type III secretion requires the binding of SycE to residues 15–100 of YopE (Cheng et al., 1997). This interaction is also sufficient for the secretion of reporter fusions in low calcium-induced cells. Here, we report that SycE binding to YopE is absolutely necessary for YopE targeting into the HeLa cytosol. Forsberg and co-workers (Frithz-Lindsten et al., 1995) have investigated the SycE/YerA requirement by immunofluorescent microscopy and reported a low level targeting of YopE in a yerA mutant of Y. pseudotuberculosis. Our experiments compared the digitonin fractionation with immunofluorescent microscopy, and we found that the latter was less sensitive. Hence, even with the more sensitive digitonin fractionation technique, we cannot detect YopE in the HeLa cytosol during infection with sycE1 mutant Yersiniae and conclude that the binding of SycE is an absolute requirement for the targeting of YopE. SycE binding is not sufficient for YopE targeting, which also requires the first 15 codons or amino acids in addition to the SycE binding site (YopE residues 15–100). These results indicate that the substrate requirement for YopE targeting differs significantly from that identified for type III secretion by low calcium-induced Yersiniae.

Our data are in disagreement with those of Cornelis and co-workers (Boland et al., 1996), who reported that, during the infection of macrophages with Y. enterocolitica, Yop–Cya fusions are found in the tissue culture medium as well as in the eukaryotic cytosol. The YopE secretion signal (codons 1–15) was presumed to direct fusion proteins into the extracellular medium as an intermediary step in Yop targeting. This assumption led to the proposal of a two-step translocation mechanism for Yop targeting and to the conclusion of a modular organization of secretion and translocation signals within Yop proteins (Sory et al., 1995). We used digitonin fractionation of Yersinia-infected HeLa cells and report that YopE, H, M and YopN are localized to the eukaryotic cytoplasm but not to the extracellular medium. Another surprising finding of this experimental technique is the secretion of YopB, YopD and YopR into the extracellular medium. YopB and YopD are absolutely required for the targeting of other Yops, but it is not clear how this could be accomplished by their secretion into the medium. One caveat of our fractionation technique is that Yop proteins might be targeted into HeLa cells and later assume a digitonin-insoluble state, such as membrane integration or import into the eukaryotic nucleus. For example, the nuclear transcription factor TFIID (Sawadogo and Roeder, 1985) was not solubilized by digitonin extraction (Fig. 1). To uncover the precise location of those Yops that sediment with the bacteria during digitonin fractionation will require additional experimentation such as electron microscopy. YopQ (YopK) has been located by immunofluorescence microscopy to Y. pseudotuberculosis during HeLa cell infections (Holmstrom et al., 1997). However, its subcellular location is still unknown. Adenylate cyclase fusions to YopN did not yield an increase in cAMP during the infection of macrophages by Y. enterocolitica (Boland et al., 1996). In contrast, we report that YopN was solubilized in part by digitonin extraction. YopN–NPT fusions did not acquire such digitonin solubility, suggesting that the discrepancies between the two experimental approaches can be explained by the aberrant subcellular location of Yop fusion proteins (V. T. Lee, unpublished observation).

We propose a new model in which YopE is targeted directly from the bacterial cytoplasm into the cytosol of HeLa cells. Translocation across three membranes may be achieved by a type III secretion channel spanning the bacterial inner and outer membranes, which may be extended by other, hitherto unknown, proteins into the eukaryotic cytosol. Because the YopB and YopD proteins are absolutely required for targeting (Hakansson et al., 1996b), it is tempting to speculate that these polypeptides might fulfil such a role. The substrate requirements for YopE targeting are residues 1–100 bound to the SycE chaperone. Because none of the targeted Yops (YopE, YopH, YopM, YopN and YopO) display sequence homology, we think it is likely that the Syc proteins play an important role in substrate recognition. For example, a conserved C-terminal sequence element (Wattiau et al., 1996) or other features of these secretion chaperones could be recognized by the type III machinery. If chaperone delivery to the type III machinery were a universal feature of Yop targeting, one would predict the existence of other Syc proteins (for YopM, YopN and YopO) that have not yet been identified (Wattiau et al., 1994).

All Yop proteins require the type III machinery for their export from the bacterial cytoplasm (Allaoui et al., 1995). One explanation for the different locations of Yop proteins, i.e. the medium, HeLa cytosol or associated with the bacteria, would be that their structural genes are expressed at different times during tissue culture infection. Hence, Yop proteins that are synthesized once the Yersinia have docked on the surface of HeLa cells might be directed into the eukaryotic cytosol, whereas others that were expressed before attachment might be secreted. An alternative explanation for the different locations of Yop proteins would be that Yersinia switch the mode of substrate recognition for the type III machinery. For example, Yops might first be secreted into the medium by an mRNA-encoded signal. Later during infection, perhaps after docking of the bacteria on the surface of HeLa cells, type III export may occur only if Yops are properly delivered by their chaperones. It is equally plausible that Yersinia use both regulatory elements, gene expression and alternate modes of substrate recognition, to position their Yop proteins at different locations relative to the eukaryotic target cell. Our future experiments will aim to distinguish between these possibilities.

Experimental procedures

HeLa cell infections

Yersinia strains were grown in Luria broth at 26°C with 150 r.p.m. shaking, diluted 1:20 into fresh media and incubated for another 2 h (OD600 0.4). HeLa cells were grown in Dulbecco's minimal Eagle's medium (DMEM) supplemented with 10% FBS to confluency (2.5 × 107 cells per 75 cm2 flask). Cells were washed twice with 5 ml of PBS, incubated in 10 ml of DMEM for 30 min, infected with Y. enterocolitica [2.5 × 108 bacteria, multiplicity of infection (MOI) 10] and incubated for 3 h at 37°C, 5% CO2. Media were collected by decanting and placed on ice. HeLa cells attached to the flasks were lysed by the addition of 10 ml of PBS containing either 1% purified digitonin or 1% SDS and 10 mM EDTA. Detergent solutions were incubated for 20 min at room temperature with vigorous intermittent vortexing. Samples were sedimented by centrifugation at 20 000 g 15 min. A 6.6 ml portion of supernatant was transferred to a new tube and the remainder discarded. The sediment was suspended in 10 ml of 1% SDS in PBS and a 6.6 ml portion transferred to a new tube. Protein was precipitated with methanol/chloroform (Wessel and Flugge, 1984) and suspended in 400 μl of sample buffer (10% glycerol, 1% SDS, 0.1% bromophenol blue, 5.5 M urea, 2% β-mercaptoethanol, 36 mM Tris-HCl, pH 6.8). Proteins were separated on SDS–PAGE, electrotransferred onto PVDF membrane, immunoblotted with specific antiserum and identified as a chemiluminescent signal on X-ray film.

Immunofluorescence

HeLa cells (2 × 105) were grown in DMEM on cover slips in a 24-well plate for 48 h at 37°C, 5% CO2. Cells were washed twice with PBS, covered with 1 ml of DMEM and infected with 8 × 106 bacteria (MOI of 10) for 3 h. Samples were first washed with PBS and then fixed with 3.7% formaldehyde in PBS for 20 min. The reaction was quenched by washing in PBS and then adding 0.1 M glycine for 5 min. HeLa cells were permeabilized with 1% Triton in PBS for 30 min. Samples were blocked for non-specific staining with 5% non-fat milk, 0.05% Tween 20 in PBS for 15 min followed by incubation with anti-NPT (1:100 dilution) for 20 min. Samples were washed four times with PBS 0.05% Tween 20 for 5 min each and incubated with goat anti-rabbit IgG–Oregon 488 green conjugate as well as wheatgerm agglutinin–Texas red conjugate (Molecular Probes, both diluted 1:500) for 20 min. Samples were washed four times, dried for 1 h and viewed under a Leica confocal laser microscope.

Immunoelectron microscopy

HeLa cells were infected with Yersinia and incubated for 3 h at 37°C, 5% CO2 as described above. Cells were washed with 10 ml of PBS, scraped off the plate and fixed overnight at 4°C in 2% formaldehyde, 6% sucrose in PBS. Samples were washed twice with PBS, and the remaining formaldehyde was quenched with 0.01 M glycine in PBS for 10 min. The samples were dehydrated through a graded series of ethanol, placed in resin through a graded series of ethanol–LR white mixtures and baked overnight at 55°C to dry the resin. The embedded cells were cut with an ultramicrotome, and thin sections were collected on Formvar-coated nickel grids. Samples were immunostained at room temperature by floating the grids on a series of 50 μl droplets of different solutions: blocking in 50 mM HEPES, 0.3 M NaCl, 0.05% NaN3, 1% BSA, 0.01% cold water fish skin gelatine for 30 min, anti-NPT 1:5 in blocking solution for 1 h, seven washes with 50 mM HEPES, 0.3 M NaCl, 0.05% NaN3, protein A–colloidal gold conjugate (9 nm particles) 1:50 in blocking solution for 1 h and, finally, another seven washes. Samples were fixed in 2% glutaraldehyde for 10 min, stained first with 2% uranyl acetate for 10 min and then with Reynold's lead for 1 min. Grids were washed in water and viewed under a transmission electron microscope.

Antisera

Yops were precipitated from the supernatant of low calcium-induced Y. enterocolitica strain 8081 cultures (Portnoy et al., 1981) with ammonium sulphate (46%). Precipitated Yops were suspended in 6 M guanidine-HCl, 0.01 M phosphate buffer, 10 mM dithiothreitol (DTT) and separated by reverse-phase high-performance liquid chromatography (HPLC) on a C8 column (Anderson and Schneewind, 1997). YopE, H, M and YopN were purified in this manner. The coding sequences for YopB, D and YopR were PCR amplified with primers specifying abutted BamHI restriction sites for cloning into the pQE vectors (Qiagen). Histidine-tagged polypeptides were overexpressed in Escherichia coli and purified by affinity chromatography on Ni-NTA followed by separation on reverse-phase HPLC. Purified polypeptides were injected into rabbits for antibody production, whereas antisera against FPT (Signal Transductions), NPT (5→3′) and TFIID (Oncogene Research) were purchased.

Plasmids and strains

The pDA plasmids have been described elsewhere (Anderson and Schneewind, 1997). Plasmids pVL33 and pVL35 were generated by inserting annealed oligonucleotides specifying the codons 1–15 of either cat or lacZ between the NdeI and KpnI sites of pDA139. Plasmids were sequenced for confirmation and transformed into Yersinia strains. The Y. enterocolitica strains W22703 (Cornelis and Colson, 1975) and LC2 (sycE1 ) (Cheng et al., 1997) have been described previously. The yopN1 mutant strain has a stop codon followed by a nucleotide insertion and BamHI site inserted at codon four of the yopN gene (Forsberg et al., 1991). The yopN1 mutation was introduced by allele replacement following a standard protocol (Cheng et al., 1997). Procedures to measure Yop secretion by Yersinia strains have been reported previously (Anderson and Schneewind, 1997).

Acknowledgements

This material is based upon work supported under a National Science Foundation Graduate Fellowship to V.T.L. We thank Daniel Clemens, Jim Gober, Brigitta Sjostrand and Phoebe Stewart for their help with electron microscopy and confocal laser immunofluorescence microscopy, and Mailin Chu for generating the yopN1 mutant strain. D.A. was supported by the Microbial Pathogenesis NIH training grant AI07323. O.S. was supported by the Stein Oppenheimer Foundation, start up funds from the Department of Microbiology and Immunology as well as grants from the US Public Health Service AI38897 and AI33985.

Ancillary