Type III secretion systems: what is the translocator and what is translocated?. Technoreview

Over the past few years, we have seen much progress in understanding the function of type III secretion systems found in pathogenic Gram-negative bacteria ( Hueck, 1998). Perhaps the most important new concept to emerge was that type III secretion systems serve to deliver bacterial proteins into the plasma membrane and cytoplasm of eukaryotic host cells ( Cornelis and Wolf-Watz, 1997; see Fig. 1). Ever since this hypothesis was put forward, the challenge has been to test its generality experimentally. A major part of this effort went into establishing for each species which proteins act as translocators and which are translocated into the eukaryotic cytoplasm. This has turned out to be a difficult undertaking. Indeed (i) type III secretion systems are complex machines involving many gene products; (ii) insertion of the translocators and translocation of effectors appears to be host contact-dependent and often involves minute amounts of proteins; (iii) powerful adhesins are generally required for the proper operation of the system and, in some cases, there is no way to gently detach adhering bacteria from their target cell; and finally (iv) some bacterial species do not remain extracellular or intravacuolar during infection but instead gain access to host cytosol. These issues mean that the methods that had been developed for demonstrating the cellular translocation of other bacterial factors, for instance diffusible A/B type toxins, could not be applied. Instead, investigators sought (i) to determine genetically which among the secreted gene products are required for the translocation of the others, but not for their secretion, and (ii) to define the localization of each of the secreted proteins in infected cells. The assumptions were that the translocator(s) must be inserted in the host membrane while the effector proteins should be found in host cytosol. Working model . That many Yersinia Yop proteins are indeed translocated into host cell cytoplasm was firmly established using the CyaA fusion protein approach, where N-terminal portions of proteins putatively translocated are fused C-terminally to a part of the adenylate cyclase from Bordetella pertussis that displays calmodulin-dependent activity and thus serves as a reporter for access to host cytosol ( Sory and Cornelis, 1994) . This method also allowed the demonstration that, among the secreted proteins, only two gene products, YopB and YopD (which are both hydrophobic proteins), are required for translocation of the others. The value of this technique is that it respects the topology of the problem. However, it is rather cumbersome technically. Indeed, it requires the generation of fusion constructs and the method used for detection is an expensive commercial kit that takes several hours to develop and is not very sensitive (signal to noise ratio of less than 10:1). In addition, the hybrid constructs are often not under the original yop gene promoter but rather under the lac promoter on multicopy plasmids. This means that the fusion proteins are always present in the bacterial cytoplasm and translocated into the host cell in very high quantities. As a consequence, comparison of relative amounts of translocation between different strains carrying mutations within secretion machinery encoding genes is difficult. In addition, the stability to the hybrid proteins in host cytosol remains to be determined. Furthermore, it can not be applied to bacteria that become internalized into host cells via a macropinocytic-like ruffle and proceed to lyse their vacuole unless cytochalasin D is added to prevent the actin polymerization required for ruffle formation. Finally, it also has biological limitations, for instance it will not discriminate between a protein that is completely translocated into the host cell cytosol or one which is inserted into the host membrane with the terminus to which the cyclase moiety is attached facing the host cytosol. Could this have been recently observed in the EPEC system for an EspB–CyaA fusion protein which was shown to yield cyclase activity but only in a strain which also contained a wild-type copy of EspB ( Wolff et al., 1998 )? Indeed, in this paper EspB–CyaA was detected in a cytosolic fraction of cells but by immunofluorescence EspB appeared mostly at the bacteria–host membrane interface. Other groups have preferred to address the question of translocation using two cell biological localization techniques: fractionation and immunofluorescence microscopy of infected host cells. Both of these techniques have their own problems. Where translocation is examined by immunofluorescence, bacterial infections of tissue-cultured cells are performed for long periods (up to 3 h). In addition, for those bacteria that use their type III secretion system to force their way into cells, cytochalasin D is added to the medium to prevent bacterial internalization. As type III secretion systems seem to be activated by contact, prolonged host contact can lead to extensive release of translocator and effector proteins into the medium, presumably as a result of the inefficiency of the translocation process. To complicate matters further, secreted proteins can form large aggregates after release from certain bacteria (for instance Parsot et al., 1995 ). Over such long incubations however, the host cell is not passive but actively endocytoses its surrounding medium, even in the presence of cytochalasin D! Furthermore the cell will also be recycling its plasma membrane and endocytosing proteins inserted within it. Could it be that some of the putative translocated proteins seen in the host cells as 'punctate patterns' by immunofluorescence are actually not translocated into the host cytoplasm but instead present in the lumen or membrane of organelles of the endocytic pathway? In principle one would not expect to find putative effector proteins inside host cells in mutants of the potential translocators but in several species knockouts of the possible translocators leads to constitutive secretion (for instance see Williams and Straley, 1998) and such secreted proteins could be taken up by endocytosis. Another potential caveat is the following: when conventional immunofluorescence or even confocal microscopy is used the resolution of these methods is no greater than 400 nm and in the z-plane of confocal microscopy, it is substantially less ( Brelje et al., 1993 ). Thus, even if one were to use dual labelling experiments with a host plasma membrane marker to determine whether a particular bacterial protein was above or below it in the z-plane, there would be no way of rigorously assessing if the bacterial protein, despite its partial colocalization with the plasma membrane, was truly on the cytoplasmic or extracellular side of the 10 nm thick host membrane. One further method to approach this problem, which seems underused in cellular microbiology, is immunoelectron microscopy on thin frozen sections ( Griffiths, 1993). Indeed it has the potential of giving information at much higher resolution regarding the membrane or cytosolic localization of bacterial proteins within host cells. This method has heavy technical requirements, such as cryotome and transmission electron microscopy facilities and training, and may not be entirely appropriate for the problem because it usually requires that the antigen is locally abundant, as labelling efficiency is very poor. Indeed, the labelling efficiency has been estimated at 1–10% of the antigen, depending on the quality of the antibody, and this only on the very upper surface of the 50–100 nm thick section. Yet, there are cases in the current literature where type III secreted proteins have been reported to become clustered in the host membrane/cytosol zone below extracellular pathogens and this method might be informatively tried in such cases. Nevertheless, one must remember that because of the size of the colloidal gold-antibody sandwich (10–30 nm depending on the type of gold and antibody used) and the fact that it collapses randomly on the section as it is dried for observation, using this technique to assess if a particular membrane-associated protein is on the cytosolic or lumenal side of the membrane is not possible. The spacial resolution (but not the labelling efficiency) can be improved substantially using Fab fragments coupled to NanogoldTM ( Adams and Kilmartin, 1999). Subcellular fractionation has perhaps even more pitfalls. Many bacterial pathogens become tightly adherent to the host cell membrane during infection. Therefore, pure host plasma membrane fractions can be difficult to obtain. Some published protocols simply involve extracting infected cells with a mild detergent and recovering a detergent-soluble fraction containing cytosol and soluble membrane proteins and an insoluble fraction that contains the non-extracted membrane proteins, the intact bacteria and cytoskeletal elements. Other published work has used more sophisticated methods where infected cells are lysed by sonication or freeze/thawing to release the cytosolic contents while supposedly keeping the nuclei and bacteria intact. These preparations are then centrifuged at high speed to yield a 'cytosolic fraction' and the pellet is often extracted with various detergents that are thought not to permeabilize bacteria to yield a fraction containing 'soluble membrane proteins' and a pellet with intact bacteria, the cytoskeleton and non-extracted membrane proteins. Most experimenters checked, using bacterial cytosolic proteins as markers, that bacteria remained grossly intact during these procedures. That the extraction procedure worked is usually checked using known host membrane proteins. The data show clear differential fractionation of host and bacterial proteins so why do the results so often disagree? Could there be bacterial proteins present at the host cell surface or in endosomes that would also be released by detergent extraction? Some authors have addressed surface contamination by protease treatment of the host membrane prior to detergent solubilization. Yet, if a bacterial protein is inserted into the host membrane and displays extracellular loops, these may be cleaved by the protease causing the fragments to migrate at much smaller molecular weights in SDS–PAGE and/or to loose antigenicity, which would lead to the erroneous conclusion that the proteins are not positioned in the host cell membrane at all. A better way to make use of protease would be to check for complete protection of a known translocated protein during infection with a wild-type bacteria compared with a strain in which a gene encoding a putative translocator has been mutated. In addition, what can guarantee that bacterial proteins inserted into the host membrane will be extracted to a similar extent as a host membrane marker? Indeed membrane proteins are differentially extracted depending on the detergent used and on the detergent–lipid ratio attained in the experiment ( Helenius and Simons, 1975 and Helenius et al., 1979 ). Thus, none of the fractions, especially those obtained after detergent treatment, are truly representative of a cellular compartment. The critical numbers that allow safe interpretation of such data are the degree of contamination by the protein under study of one fraction with the next. Wachter et al. (1999) described one way of obtaining such figures. The use of detergents on host cells results in mixing of cellular compartments and thus introduces an additional variable that is very difficult to control for and would best be avoided when trying to determine the location of a molecule in the 'unscrambled' cell. But how? We need to revive or design techniques to obtain host cytosol and membrane fractions devoid of bacteria without resorting to detergents. One approach to the problem is to use sucrose gradient floatation for intact membrane isolation ( Blocker et al., 1999 ). The recent controversy among Yersinia laboratories is an example of the kind of problems encountered in the course of using the techniques described above. The Cornelis and Wolf-Watz laboratories claim that the two hydrophobic proteins YopB and YopD are required for translocation of effector Yops using the CyaA method, confocal microscopy and cell fractionation ( Sory and Cornelis, 1994; Rosqvist et al., 1994 , Persson et al., 1995 ; Håkanson et al., 1996a), and propose that they form a pore in the host membrane ( Håkanson et al., 1996b; Neyt and Cornelis, 1999). In addition, Wolf-Watz's laboratory finds YopD inside the host cell cytoplasm using confocal microscopy and cell fractionation ( Francis and Wolf-Watz, 1998). This does not necessarily contradict a role of YopD in translocation of the effectors but it does complicate the picture. Furthermore, Lee and Schneewind (1999) provide evidence using a digitonin extraction procedure and immunofluorescence that YopB and YopD are not in the host membrane or cytosol but secreted into the extracellular milieu. So, which of these possibilities are correct? Only the right choice of methods and controls, cross-checking of each method with the next and careful interpretation of results can lead to a clearer picture in the future. It is important to note that location alone can not be used to draw a firm line between the translocator and the effectors. For example, Tir, the translocated intimin receptor is found in the host membrane but not involved in translocating other EPEC Esps into the host cell ( DeVinney et al., 1999 ) whereas IpaC, which is a component of the pore formed by Shigella in the host membrane ( Blocker et al., 1999 ), has a C-terminal part that is presumably exposed to host cytosol and is involved in activating actin polymerization under the bacterium ( Tran Van Nhieu et al., 1999 ). Thus, a bacterial protein inserted into the host membrane may not necessarily be a translocator and a translocator could also have effector functions. A safer way to assess and understand the role of each protein is using functional assays. We need to start designing those too; for instance, by generating semipermeablized host cells to assay the function of purified putative type III effectors ( Tran Van Nhieu et al., 1999 ). A.B. wishes to thank the members of the Sansonetti laboratory, including Philippe himself and Kirsten Niebuhr, Dana Philipott and Michelle Rathman for helpful discussions on the subject of this review, and Mike Donnenberg, Tony Pugsley and Hans Wolf-Watz for critical reading of the manuscript and communication of unpublished results.