Listeria Pathogenesis and Molecular Virulence Determinants

Clinical FeaturesThe clinical signs of L. monocytogenesinfection are very similar in all susceptible hosts. Two basic forms of presentation can be distinguished: perinatal listeriosis and listeriosis in the adult patient. In both instances, the predominant clinical forms correspond to disseminated infection or to local infection in the central nervous system (CNS). Listeriosis is usually a very severe disease—in fact, one of the most deadly bacterial infections currently known—with a mean mortality rate in humans of 20 to 30% or higher despite early antibiotic treatment (422, 423,554, 594).

Fetomaternal and neonatal listeriosis.This form of presentation mainly results from invasion of the fetus via the placenta and develops as chorioamnionitis. Its consequence is abortion, usually from 5 months of gestation onwards, or the birth of a baby or stillborn fetus with generalized infection, a clinical syndrome known as granulomatosis infantiseptica and characterized by the presence of pyogranulomatous microabscesses disseminated over the body and a high mortality (336a) (Fig. 1). The infection is usually asymptomatic in the mother or may present as a mild flu-like syndrome with chills, fatigue, headache, and muscular and joint pain about 2 to 14 days before miscarriage. Less frequently (10 to 15% of perinatal cases), late neonatal listeriosis is observed. It generally occurs 1 to 8 weeks postpartum and involves a febrile syndrome accompanied by meningitis and, in some cases, gastroenteritis and pneumonia. It presumably results from aspiration of contaminated maternal exudates during delivery, although hospital-acquired cases involving horizontal transmission by fomites or medical personnel in neonatal units have been reported (177, 554, 628). The mortality of late-onset neonatal listeriosis is lower (10 to 20%), but like early-onset neonatal listeriosis, it may have sequelae such as hydrocephalus or psychomotor retardation (173). L. monocytogenes is one of the three principal causes of bacterial meningitis in neonates (391, 554, 594, 645).

• Open in new tab
• Download powerpoint

Fig. 1.

Clinical and pathological characteristics ofListeria infection in animals and humans. (A to D) Neuromeningeal listeriosis in sheep. (A) Typical aspect of circling disease, first described by Gill in 1933 in New Zealand (215,216) and the most characteristic clinical manifestation of listeriosis in ruminants; the syndrome is characterized by involuntary torticollis and walking aimlessly in circles as a result of brainstem lesions. (B) In a further step of the infectious process, animals lie on the ground with evident signs of uncoordination (paddling movements) and cranial nerve paralysis (strabismus, salivation, etc.). (C) Section of the medulla oblongata of a sheep with listerial rhombencephalitis, showing clear inflammatory lesions in the brain tissue. (D) Microscopic preparation of the brainstem showing extensive parenchymal inflammatory infiltration and typical perivascular cuffing (one is indicated by an arrowhead) (panels C and D copyright M. Domingo, Barcelona, Spain). (E and F) Stillborn with generalized L. monocytogenesinfection, a clinical condition also known as granulomatosis infantiseptica and characterized by the presence of military-disseminated pyogranulomatous lesions on the body surface (E) and internal organs (F, liver of the fetus in E) and a high mortality. (G and H) Extensive pyogranulomatous hepatitis in a lamb experimentally infected with L. ivanovii, with multiple necrotic foci in the liver surface (G) and the parenchyma (H, hematoxylin/eosin-stained cut from the liver in G, showing two subcapsular pyogranulomes [magnification, ×60]).

PathogenesisThe pathophysiology of Listeria infection in humans and animals is still poorly understood. Most of the available information is derived from interpretation of epidemiological, clinical, and histopathological findings and observations made in experimental infections in animals, particularly in the murine model. As contaminated food is the major source of infection in both epidemic and sporadic cases (176, 512), the gastrointestinal tract is thought to be the primary site of entry of pathogenicListeria organisms into the host. The clinical course of infection usually begins about 20 h after the ingestion of heavily contaminated food in cases of gastroenteritis (118), whereas the incubation period for the invasive illness is generally much longer, around 20 to 30 days (386, 542). Similar incubation periods have been reported in animals for both gastroenteric and invasive disease (138, 219, 671).

Ingestion of L. monocytogenes is likely to be a very common event, given the ubiquitous distribution of these bacteria and the high frequency of contamination of raw and industrially processed foods. However, the incidence of human listeriosis is very low, normally around 2 to 8 sporadic cases annually per million population in Europe and the United States (176, 301, 554, 649). Higher incidence rates have been reported for sporadic listeriosis (for example, 10.9 cases per million population per year in Barcelona, Spain [476], and 14.7 in France [234]), but these are exceptional. In epidemic situations, the incidence in the target population increases by a factor of 3 to 10 (46, 386, 587,599). Thus, L. monocytogenes seems to have a lower pathogenic potential than other food-borne pathogens. This is consistent with the relatively high 50% lethal dose (LD₅₀) values reported for mice infected experimentally by the oral (10⁹ [18, 485]) or parenteral (10⁵ to 10⁶ [18, 333, 553]) routes. The minimum dose required to cause clinical infection in humans has not been determined, but the large numbers of L. monocytogenes bacteria detected in foods responsible for epidemic and sporadic cases of listeriosis (typically 10⁶) suggest that it is high. However, these data should be interpreted with caution, given the long incubation period of invasive listeriosis and the time normally elapsed between diagnosis and analysis of the food eaten, during which Listeria organisms can have multiplied in the patient's refrigerator. It therefore cannot be excluded that low doses may be able to cause infection, at least in immunosuppressed persons, old people, and pregnant women. Indeed, levels of contamination as low as 10² to 10⁴ L. monocytogenes cells per g of food have been associated with listeriosis in humans (176, 425). Obviously, the infectious dose may vary depending upon the pathogenicity and virulence of the L. monocytogenes strain involved and the host risk factors.

Entry and colonization of host tissues.

(i) Crossing the intestinal barrier.Before reaching the intestine, the ingested Listeria organisms must withstand the adverse environment of the stomach (Fig. 2). Oral infective doses are lower for cimetidine-treated experimental animals than for untreated animals (586a), and the use of antacids and H2-blocking agents has been reported to be a risk factor for listeriosis (286, 592). This indicates that gastric acidity may destroy a significant number of the Listeriaorganisms ingested with contaminated food.

• Open in new tab
• Download powerpoint

Fig. 2.

Schematic representation of the pathophysiology ofListeria infection. See text for details.

There is controversy concerning the point of entry and the mechanism of intestinal translocation used by L. monocytogenes. In an early study by Racz et al. with guinea pigs infected intragastrically with 10¹⁰ L. monocytogenes, detailed histological analyses revealed that all the animals developed enteritis (526). In the initial stages, bacteria were detected mostly in the absorptive epithelial cells of the apical area of the villi, whereas in later phases most were inside macrophages of the stroma of the villi, suggesting that L. monocytogenespenetrates the host by invading the intestinal epithelium (526). This is consistent with the observation thatL. monocytogenes is able to penetrate the apical surface of polarized, differentiated human enterocyte-like Caco-2 cells with a brush border (322) (see Fig. 3B and C). In other studies using mice inoculated per os with 10⁸ to 10⁹bacteria, no invasion of the intestinal villous epithelium was observed; instead, there was colonization of the Peyer's patches (400, 410, 411), suggesting that L. monocytogenes uses the M-cell epithelium as an entry portal, as reported for other bacterial pathogens (623). Direct evidence that L. monocytogenes may indeed penetrate the host via the M cells overlying the Peyer's patches has been provided by a recent study using a murine ligated-loop model and scanning electron microscopy (308). However, listerial penetration at this level appears not to be very efficient, because only small numbers of bacteria are observed in association with the follicle-associated epithelium overlying Peyer's patches (411).

Intestinal translocation of pathogenic listeriae occurs without the formation of gross macroscopic or histological lesions in the gut of mice (411). This suggests that an epithelial phase involving bacterial multiplication in the intestinal mucosa is not required by L. monocytogenes for systemic infection. Indeed, a recent study using a rat ileal loop model of intestinal infection (523) has shown that Listeria organisms are translocated to deep organs very rapidly (within a few minutes), demonstrating that crossing of the intestinal barrier occurs in the absence of prior intraepithelial replication. This study also showed that defined mutants defective in virulence factors known to be involved in epithelial cell invasion (InlA and InlB), intracellular survival (Hly), and cell-to-cell spread (ActA) (see below), and even the nonpathogenic species L. innocua, translocated at the same rate as wild-type L. monocytogenes. Translocation was dose dependent, and the presence of Peyer's patches in ligated loops did not affect the rate of translocation, levels of uptake being similarly low for Peyer's patches and villous intestine (50 to 250 bacteria per cm² of tissue after inoculation of the loop with 10⁹ bacteria) (523). These findings favor the view that listerial translocation is a passive, nonspecific process similar to that observed for other bacteria (36), at least in mice and rats. However, L. monocytogenes proliferated in the intestinal wall, whereas L. innocua did not. The preferential site for bacterial replication was the Peyer's patches, and the essential listerial virulence factor Hly (hemolysin) was indispensable for this process, showing that L. monocytogenes establishes an active local infection in these lymphoid structures of the intestine. The foci of infection consisted of pyogranulomatous reactions in the subepithelial follicular tissue, with bacteria visible inside permissive mononuclear cells, presumably nonactivated resident macrophages and dendritic cells (522a, 523). Thus, antigen presentation events probably already take place in the intestine during the early phases of host colonization, and this may play an important role in the acquired resistance to subsequent reinfection that develops after primary oral exposure to L. monocytogenes (394, 400). A recent study has shown that invasion of intestinal epithelial cells byL. monocytogenes induces the activation of NF-κB and the subsequent upregulation of interleukin-15 (IL-15), and that at early stages of an oral infection with L. monocytogenes in mice and rats, intestinal intraepithelial lymphocytes become activated and produce Th1-type cytokines. This suggests that the interaction between intestinal epithelial and lymphoid cells may be relevant in the local host defence against listerial infection in the gut (706).

Experimental observations made with the mouse and rat models of intestinal translocation do not, however, explain how L. monocytogenes causes enteritis. The association of gastroenteric symptoms with fever is consistent with invasive intestinal disease, as observed by Racz in the guinea pig model (526). PathogenicListeria organisms pass directly from cell to cell by a mechanism involving host cell actin polymerization (see below). Therefore, regardless of the mechanism of entry used, the bacteria that penetrate the intestinal wall might then invade neighboring enterocytes by basolateral spread, leading to enteritis. This is consistent with in vitro experimental data showing that L. monocytogenes enters polarized Caco-2 cells predominantly via the basolateral surface (196). Gross intestinal lesions develop in experimental animals only if large oral doses of L. monocytogenes are given (400, 523). Similarly, episodes of listerial gastroenteritis in humans occur in the form of outbreaks with very short incubation periods and high attack rates among immunocompetent adults (118, 577), consistent with the ingestion of a very high dose of bacteria (as high as 2.9 × 10¹¹, as estimated for one of these oubreaks, caused by the consumption of heavily contaminated chocolate milk [118]). Thus, intestinal invasion and the ensuing febrile gastroenteritis syndrome probably result from extensive exposure of the intestine to pathogenicListeria organisms. However, the possibility that someL. monocytogenes strains have a greater enteropathogenic potential and cause intestinal damage at a lower dose cannot be ruled out.

(ii) Multiplication in the liver.TheListeria organisms that cross the intestinal barrier are carried by the lymph or blood to the mesenteric lymph nodes, the spleen, and the liver (411, 523) (Fig. 2). As stated above, this initial step of host tissue colonization by L. monocytogenesis rapid. The unusually long incubation period required by L. monocytogenes for the development of symptomatic systemic infection after oral exposure in relation to that for other food-borne pathogens is therefore puzzling and indicates that listerial colonization of host tissues involves a silent, subclinical phase, many of the events and underlying mechanisms of which are unknown.

Experimental infections of mice via the intravenous route have shown that L. monocytogenes bacteria are rapidly cleared from the bloodstream by resident macrophages in the spleen and liver (99,112a, 405). Most (90%) of the bacterial load accumulates in the liver, presumably captured by the Kupffer cells that line the sinusoids. These resident macrophages kill most of the ingested bacteria, as shown by in vivo depletion experiments (161), resulting in a decrease in the size of the viable bacterial population in the liver during the first 6 h after infection. Kupffer cells are believed to initiate the development of antilisterial immunity by inducing the antigen-dependent proliferation of T lymphocytes and the secretion of cytokines (243). Not all Listeriacells are destroyed by tissue macrophages, and the surviving bacteria start to grow, increasing in numbers for 2 to 5 days in mouse organs (93, 99, 128, 181, 381, 408, 448).

The principal site of bacterial multiplication in the liver is the hepatocyte (100, 112a, 197, 241, 241a, 563). This finding has led to the dismissal of the long-held idea that the major host niche for the parasitic life of L. monocytogenes is the macrophage population. There are two possible ways for L. monocytogenes to gain access to the liver parenchyma after its intestinal translocation and carriage by the portal or arterial bloodstream: via Kupffer cells, by cell to cell spread, or by the direct invasion of hepatocytes from the Disse space after crossing the fenestrated endothelial barrier lining the sinusoids. L. monocytogenes has been shown to efficiently invade hepatocytes in vitro (149, 701).

Electron microscopy of hepatic tissue from infected mice suggests thatL. monocytogenes goes through the complete intracellular infectious cycle in hepatocytes (197, 621), including actin-based intercellular spread (see below). Direct passage from hepatocyte to hepatocyte would lead to the formation of infectious foci in which L. monocytogenes disseminates through the liver parenchyma without coming into contact with the humoral effectors of the immune system. This may explain why antibodies play no major role in anti-Listeria immunity (516).

During the early steps of liver colonization, polymorphonuclear neutrophils are recruited at the sites of infection, forming discrete microabscesses. Neutrophils have been shown to play an important role in controlling the acute phase of Listeria infection (561) and in mediating the destruction ofListeria-infected hepatocytes in vivo (99). Hepatocytes respond to Listeria infection by releasing neutrophil chemoattractants and exhibiting an increase in adhesion to neutrophils, resulting in microabscess formation (560). They also respond by initiating an apoptosis program that may be critical for removing Listeria-infected cells from the liver tissue at early stages of infection (448a, 560). Using transgenic mice expressing a degradation-resistant IκB transgene, it has been shown recently that NF-κB activation in mouse hepatocytes is essential to clear L. monocytogenes from the liver, possibly by coordinating local innate immune response (368). Two to four days after infection, neutrophils are gradually replaced by blood-derived mononuclear cells together with lymphocytes to form the characteristic granulomas (283, 408) (Fig. 1G and H). These granulomas are the histomorphological correlate of cell-mediated immunity and presumably act as true physical barriers that confine the infectious foci, impeding further bacterial dissemination by direct cell-to-cell passage (516).

Between days 5 and 7 postinfection, L. monocytogenesbacteria start to disappear from mouse organs until their complete clearance as a result of gamma interferon (IFN-γ)-mediated macrophage activation and the induction of an acquired immune response primarily mediated by CD8⁺ lymphocytes, which together destroyListeria-infected cells (241a, 268, 328, 441). These cytotoxic T lymphocytes (CTL) are directed against listerial epitopes present in secreted virulence-associated proteins, such as Hly (59, 71, 266, 281, 625, 677), the metalloprotease Mpl (79), and the surface protein p60 (203, 204, 267,624), and possibly also in somatic housekeeping proteins such as superoxide dismutase (280). Protection against L. monocytogenes also involves a vigorous Th1-biased CD4⁺T-cell response (125, 203) and innate mechanisms such as the activation of IFN-γ-producing natural killer (NK) cells in response to IL-12 and tumor necrosis factor alpha (TNF-α) secretion by infected macrophages (125, 166, 268, 290). Hepatocytes may contribute to protection by becoming less permissive to intracellular proliferation of Listeria organisms upon exposure to IFN-γ (244), especially if costimulated with other cytokines (646).

The above course of events is accelerated in immune animals, resulting in rapid elimination of L. monocytogenes from the liver (Fig. 2). This is probably the most common outcome of L. monocytogenes infection in humans and animals in normal conditions, given the potentially high frequency of exposure to the pathogen via contaminated food and the relatively rare occurrence of clinical disease. Indeed, T lymphocytes directed againstListeria antigens are commonly found in healthy individuals (457), probably due to chronic stimulation of the immune system by L. monocytogenes antigens that are presumably continuously delivered to the immune system via food. A recent study has shown that the nonpathogenic species L. innocua can enhance a previously established L. monocytogenes-specific T-cell memory via recognition of cross-reactive p60 epitopes shared by the two species (204). L. innocua is commonly found in food (573), and as stated above, it is translocated to the deep organs of mice as efficiently as L. monocytogenes. Therefore, repeated contact with L. innocua may boost protective immunity against pathogenic Listeria spp., possibly explaining in part the rare occurrence of listeriosis in the general population.

If the infection is not controlled by an adequate immune response in the liver, as may occur in immunocompromised individuals, unlimited proliferation of L. monocytogenes in the liver parenchyma may result in the release of bacteria into the circulation (Fig. 2).L. monocytogenes is a multisystemic pathogen that can infect a wide range of host tissues, as indicated by its capacity to cause septicemia involving multiple organs and by the variety of potential sites of localized Listeria infection (see above). However, the principal clinical forms of listeriosis clearly show that L. monocytogenes has a pathogenic tropism towards the gravid uterus and the CNS (Fig. 2).

(iii) Colonization of gravid uterus and fetus.Abortion and stillbirth due to Listeria spp. have been reproduced experimentally by intravenous, oral, and respiratory inoculation in naturally susceptible gestating animal hosts, such as sheep, cattle, rabbits, and guinea pigs, as well in pregnant mice and rats (1, 238, 239, 329, 450, 487, 488, 586). This shows that L. monocytogenes gains access to the fetus by hematogenous penetration of the placental barrier (Fig. 2). In pregnant mice, the blood-borne bacteria first invade the decidua basalis and then progress to the placental villi, where they cause diffuse inflammatory infiltration and necrosis (1, 535, 536). Macrophages appear to be excluded from the murine placenta, neutrophils acting as the main antilisterial effector cell population (251a). Using homozygous mutant mice, it has been shown recently that colony-stimulating factor-1 is required for the recruitment of neutrophils to the infectious foci in the decidua basalis. This occurs via induction of neutrophil chemoattractant synthesis by the trophoblast (251a). In humans, placental infection is characterized by numerous microabscesses and focal necrotizing villitis (502, 582). Colonization of the trophoblast layer followed by translocation across the endothelial barrier would enable the bacteria to reach the fetal bloodstream, leading to generalized infection and subsequent death of the fetus in utero or to premature birth of a severely infected neonate with miliary pyogranulomatous lesions (the above-mentioned granulomatosis infantiseptica) (Fig. 1E and F).

The depression of cell-mediated immunity during pregnancy (686) presumably plays an important role in the development of listeriosis. In laboratory rodents, pregnancy reduces resistance to L. monocytogenes (58, 399) and significantly prolongs the course of primary infection in the liver (1). The T-cell-dependent elimination of L. monocytogenes from the organs of pregnant mice is delayed in late pregnancy. This correlates with a failure in the production of IFN-γ and results in listerial invasion of placental and fetal tissues (463). Abnormally high physiological levels of estrogenic hormones, such as those that occur during late pregnancy, may account for the disruption of T-cell-mediated resistance to infection, as treatment of mice with steroids inhibits the proliferative response toL. monocytogenes of splenic T cells and increases susceptibility to Listeria infection (524). This correlates with a decrease in the production of IL-2 (524), a cytokine that stimulates resistance toListeria spp. in mice (255). The intracellular killing function of macrophages has also been shown to be decreased by β-estradiol, which is associated with inhibition of both IL-12 and TNF-α secretion in vivo (578). Local depression in the cellular immune response in the placenta, which physiologically is important for preventing rejection of the fetus, may also contribute to the higher susceptibility to uterine infection by L. monocytogenes (398, 535, 536). It is not known why mothers suffering Listeria-induced miscarriage never develop CNS infection or overt septicemic disease.

(iv) Invasion of the brain.In humans, CNS infection by Listeria spp. presents primarily in the form of meningitis. This meningitis, however, is often associated with the presence of infectious foci in the brain parenchyma, especially in the brain stem (391, 473), suggesting L. monocytogenes has a tropism for nerve tissue. The neurotropism and special predilection of L. monocytogenes for the rhombencephalon are shown most clearly in ruminants, in which listerial CNS infection, in contrast to the situation in humans, develops mainly as primary encephalitis. In these animals, infectious foci are restricted to the pons, medulla oblongata, and spinal cord (Fig. 1C). Although there is inflammatory lymphocyte or mononuclear cell infiltration of the meninges, this condition occurs as an extension of the brain process, and macroscopic lesions may not even be evident or may be restricted to basal areas, midbrain, and cerebellum. Unilateral cranial nerve paralysis is a characteristic of listerial rhombencephalitis in ruminants, leading to the well-known circling disease syndrome (92, 103, 215, 216, 312, 534, 666, 693) (Fig. 1A and B). In humans, primary nonmeningeal brain infection is seldom observed. However, as in ruminants, it develops as cerebritis involving the rhombencephalon (15, 20, 69, 390, 391, 473).

Brain lesions in listerial meningoencephalitis are typical and very similar in humans and animals. They consist of perivascular cuffs of inflammatory infiltrates composed of mononuclear cells and scattered neutrophils and lymphocytes (Fig. 1D). Bacteria are generally absent from these perivascular areas of inflammation. Parenchymal microabscesses and foci of necrosis and malacia are also typically present (Fig. 1D). Bacteria are relatively abundant in these lesions, within phagocytes or free in the brain parenchyma around the necrotic areas (92, 103, 312, 315, 412, 489). Depletion experiments in mice using a neutrophil-specific monoclonal antibody have shown that neutrophils play a critical role in eliminating L. monocytogenes from infectious foci in the brain (389a). Less commonly, bacteria are observed within neurons in both natural (412, 489) and experimentally induced (588) infections. This is consistent with in vitro data showing that the invasion of cultured neurons is a relatively rare event (151, 505). However, neurons are efficiently invaded in vitro by direct cell-to-cell spread from infected macrophages or microgial cells (151). A recent study (146) has shown that L. monocytogenes can efficiently invade sensory neurons of rat dorsal root ganglia but not hippocampal neurons, suggesting that there may be differences in the susceptibility to infection among different types of neurons. The relevance of neuronal invasion to the pathogenesis of Listeria encephalitis is subject to speculation, as discussed below.

The mechanisms by which L. monocytogenes infects the CNS are still largely unknown. Numerous attempts have been made to reproduce listerial meningoencephalitis in susceptible animal hosts and laboratory rodents using various routes of inoculation, with variable success. Intravenous delivery of the inoculum does not usually result in CNS infection in sheep (103). This, and the usual absence of visceral involvement in natural cases of ovine listerial encephalitis, seems to argue against a hematogenous route of infection. The topographic distribution of the lesions in sheep, strictly confined to the brainstem, and the frequent unilateral involvement of the trigeminal nerve, with neuritis affecting distal portions of a single or a few fasciculi in only one of the nerve branches, led to the hypothesis that L. monocytogenes invades the brain by centripetal migration along cranial nerves (17, 92, 661). To test this hypothesis, Asahi and coworkers (17) scarified oral, nasal, and labial mucosae of mice with L. monocytogenes. Most mice developed neurological signs, such as torticollis, rolling movements, and paralysis of the hindlegs, 5 days after infection. These signs correlated with the presence of mononuclear and round cell infiltrations along the trigeminal nerve, in the corresponding nerve ganglion, and in the medulla oblongata, suggesting retrograde bacterial ascension along the cranial nerve to the brainstem. In another experiment, 4 of 11 goats inoculated in the lip developed listerial encephalitis 17 to 28 days after inoculation, with microscopic lesions identical to those found in natural cases of listeriosis, with ganglioneuritis of the trigeminal nerve (17). In spontaneous cases of ovine listerial encephalits, bacteria have been observed in linear arrays in individual nerve fibers, inside axons, and with actin tails (91, 92, 489), suggesting active intra-axonal movement. Experiments using a double-chamber cell culture system and dorsal root ganglia neurons provided evidence that L. monocytogenes can infect axons and spread along them towards the nerve cell body (146).

Experimental evidence has been obtained that intra-axonal bacterial transport may indeed occur in vivo. Otter and Blakemore (489) developed a mouse model in which L. monocytogenes was injected distally into the sciatic nerve. Paralysis of the leg into which the bacteria were injected occurred 7 to 12 days later, and examination of the spinal cord revealed lesions very similar to those found in the brains of naturally affected sheep. Bacteria were seen in these lesions in neutrophils and, in some cases, in axons (489). The feasibility of ascending intra-axonal transport via the termini of the trigeminal nerve was tested by experimentally injecting L. monocytogenes directly into the dental pulp of sheep through a hole drilled in the tooth crown (22). Six of 21 sheep developed neurological signs 20 to 40 days after challenge, and most animals showed histological encephalitis and trigeminal ganglioneuritis on the side of inoculation. These observations favor the view that L. monocytogenes may directly invade exposed sensory terminal ramifications of the cranial nerve in the mouth and reach the brain by centripetal spread. They also demonstrate that cases of mild CNS invasion by Listeriaorganisms may occur without overt clinical signs. This is important because it provides one possible explanation for the relatively small number of cases with neurological manifestations that are observed among animals exposed to contaminated silage in a herd affected by listerial encephalitis (and, by extension, to humans exposed toListeria-contaminated food). However, in another study in which brains from natural cases of sheep with meningoencephalitis were analyzed, inflammatory infiltration was observed only in the intracranial portion of the trigeminal tract, always associated with meningitis adjacent to proximal parts of cranial nerves. The authors were of the opinion that cranial nerve invasion was only secondary to a hematogenous brain infection and occurred by centrifugal spread from the brainstem (103). The above-mentioned experiments with the two-chamber model (146) suggested that both retrograde and anterograde transport of L. monocytogenes may occur inside infected neurons.

Although intravenous inoculation only occasionally gives positive results, intracarotid inoculation consistently results in encephalitis in sheep (103, 314, 486). This supports the notion thatL. monocytogtenes may indeed reach the brain via the blood. Strong evidence that meningoencephalitis in ruminants has a vascular basis is provided by the nature of the brain lesions themselves, with extensive perivascular infiltrates and mononuclear microabscesses and necrotic foci developing in close proximity to brain capillaries (103) (Fig. 1D). In contrast to intravenous inoculation, intracarotid delivery provides the inoculum direct access to the vascular system of the brain, enabling bacteria to bypass the clearance mechanisms of organs such as the spleen and liver. In an immunocompetent host, these clearance mechanisms should be sufficient to eliminate rapidly a single-dose intravenous inoculum, preventing the bacteria from invading the brain. Intracarotid inoculation in sheep causes CNS lesions involving the choroid plexus and the ependyma of the cerebral ventricles and aqueduct, resulting in primary meningitis and secondary lesions in the adjoining cerebral tissues (17). These brain structures are only rarely affected in natural cases of listerial meningoencephalitis in ruminants (103). The reason for these differences between natural and experimental lesion patterns probably resides in the concentration of bacteria in the blood, as this is an important factor affecting the mode by which bacterial pathogens gain access to the CNS (659, 660). This concentration would reach an extraordinarily high peak in the brain after intracarotid delivery, and high-grade bacteremia is more likely to cause meningitis due to massive bacterial penetration across the epithelium lining the choroid plexus, which has an exceptionally high rate of blood flow. In contrast, low-grade bacteremia, such as that resulting from the release of bacteria into the blood from infectious foci located in distant organs (e.g., the liver), tends to be associated with multiple foci in the brain parenchyma due to individual penetrations along the extensive microvascular endothelial bed (660).

Experimental infections in small laboratory animals also provide evidence for a hematogenous route of CNS invasion for L. monocytogenes. Intravenous inoculation in mice results in meningoencephalitis in a significant proportion of animals that survive the initial systemic phase of infection. Lesions include choroiditis, meningitis, and the characteristic perivascular cuffing and parenchymal pyogranulomatous foci in the brainstem (31, 103). No brain invasion is detected in mice challenged with low infective doses that induce only transient bacteremia, even if there is significant bacterial multiplication in the spleen and liver. However, brain invasion is achieved with high infective doses causing severe systemic disease and prolonged bacteremia (31). This suggests that the persistence of large numbers of bacteria in the blood is essential for the induction of meningoencephalitis by L. monocytogenes. This conclusion is consistent with a recent study by Blanot and coworkers (47), who adapted a model of otitis media in gerbils to study listerial CNS invasion. With a low infective dose of L. monocytogenes (10³/ear), these authors induced persistent low-level bacteremia (10²bacteria/ml) which was associated with the development of rhombencephalitis in gerbils mimicking that observed during human listerial meningoencephalitis, with leptomeningitis, parenchymal foci, and perivascular sheaths. The number of bacteria in the liver and spleen of gerbils decreased by day 5 postinoculation, but growth in brain tissue was unrestricted (47), suggesting that nerve tissue is a privileged niche for the multiplication of L. monocytogenes. CNS invasion was also regularly observed in mice after oral or subcutaneous inoculation (9, 17, 522). These infection routes imply systemic involvement and, clearly, also lymphohematogenous dissemination of bacteria to the brain.

In experiments with orally or subcutaneously infected mice, parenchymal brain lesions are irregularly observed, but marked inflammatory lesions in the meninges and ventricular system (choroid plexus, lateral, third and fourth ventricles, and aqueduct) are consistently recorded (9, 522). This striking affinity of L. monocytogenes for mouse ventricular structures was also noted after direct intracerebral inoculation (588), which prevents primary blood-borne exposure of the ependyma via the choroid plexus. In contrast, no involvement of the ventricular system was noticed in the gerbil model (47). These observations suggest that there may be anatomical and physiological differences in the brains of the various animal species that affect L. monocytogenes neuropathogenesis. Indeed, this may account for the differential characteristics of listerial CNS infection in humans and ruminants. However, it cannot be excluded that the peculiar characteristics of listerial meningoencephalitis in ruminants result from the fact that an alternative primary nonvascular mechanism of CNS invasion, which may perfectly coexist with hematogenous dissemination, has specifically evolved in these animals. Ruminants eat plant material, the physical characteristics of which may lead to small breaches of the oral mucosa. If animals are fed contaminated silage, these small wounds are repeatedly exposed to L. monocytogenes during rumination, favoring invasion of the trigeminal nerve terminals and subsequent intraneural, direct spread to the brain stem.

Bacterial pathogens such as H. influenzae, N. meningitidis, S. pneumoniae, and Streptococcus suis cross the blood-brain barrier (BBB) primarily via the choroid plexus. This enables them to reach the CSF and create purulent meningitis by spreading through the subarachnoid space (525, 659, 660,697). Although in human listerial infections of the CNS, purulent meningitis also occurs and bacteria may also be detected in the CSF, it is clear that L. monocytogenes, unlike other meningitis-causing bacterial pathogens, tends to affect brain parenchymal tissue. This may reflect a tropism for the microvascular endothelium, in particular that lining the capillary bed of the rhombencephalon, resulting in significant or preferential (in ruminants) crossing of the BBB at this point. Direct uptake by endothelial cells of bacteria circulating free in the blood is one possible mechanism by which L. monocytogenes may cross the microvascular BBB. Evidence for this mechanism is provided by an electron microscopic study of human brainstem tissue, in which bacteria have been observed within endothelial cells or adhering to the luminal face of the microvascular endothelium (335). This mechanism is also consistent with in vitro data showing that L. monocytogenes is able to invade cultured human brain microvascular endothelial cells (BMEC) (245, 246). The listerial surface protein InlB, of the internalin family (see below), has been shown to be required for invasion of endothelial cells in vitro (246,498), indicating that specific bacterial molecules are actively involved in the interaction with the BBB. L. monocytogenesefficiently replicates for long periods of time within brain microvascular cells without causing any evident damage, creating heavily infected foci from which bacteria spread to neighboring cells by actin-based motility (246). In this way, L. monocytogenes may reach and disseminate easily into the protected spaces of the CNS.

L. monocytogenes infection induces a potent response in cultured endothelial cells which is accompanied by the upregulation of endothelial adhesion molecules (P- and E-selectin, ICAM-1, and VCAM-1) (see below), leading to an increase in the binding of polymorphonuclear leukocytes to infected endothelial cells (152, 247, 330, 600,618). Heterologous plaque assays using infected macrophages as vectors have demonstrated that L. monocytogenes can spread efficiently from macrophages into endothelial cells (157,246). Therefore, the recruitment of phagocytes to the infected endothelium, primarily a host defense response, may also be used byL. monocytogenes, in addition to direct invasion, to cross the BBB. Evidence for a significant role in vivo for such a Trojan horse mechanism of CNS invasion, presumably involving direct bacterial cell-to-cell spread from infected phagocytes carried by the blood after adhesion to endothelial cells, has been provided recently by Drevets (154). This author showed that phagocyte-associatedL. monocytogenes accounted for 30% of the total bacterial population circulating in the bloodstream after intravenous inoculation. These entrapped bacteria spread to endothelial cells in vitro, and Listeria-infected peripheral blood leukocytes successfully established brain colonization in vivo (154). It is unknown whether trafficking of infected phagocytes across the BBB, as part of the inflammatory influx induced by endothelial and underlying brain tissue infection, occurs in vivo during CNS invasion by L. monocytogenes.

InternalizationThe cycle begins with adhesion to the surface of the eukaryote cell and subsequent penetration of the bacterium into the host cell. The invasion of nonphagocytic cells involves a zipper-type mechanism, in that the bacterium gradually sinks into diplike structures of the host cell surface until it is finally engulfed. During this process, the target cell membrane closely surrounds the bacterial cell and does not form the spectacular local processes or membrane ruffles characteristic of invasion by Salmonella andShigella spp. (147, 322, 358, 435, 642) (Fig.3B and C). The structures, mechanisms, and signal transduction cascades involved in the interaction between the bacterium and the cell during phagocytosis are only beginning to be elucidated.

The multisystemic nature of listerial infection indicates that L. monocytogenes probably recognizes a number of different eukaryotic receptors. The C3bi and C1q complement receptors have been reported to be involved in L. monocytogenes uptake by phagocytic cells (10, 115, 155, 156). L. monocytogenes is also efficiently internalized in the absence of serum, indicating that nonopsonic receptor-ligand interactions are also involved in host cell recognition by Listeria spp. (509). The eukaryote cell receptors used by Listeria spp. include the transmembrane glycoprotein E-cadherin (435), the C1q complement fraction receptor (61), the Met receptor for hepatocyte growth factor (HGF) (615a), and components of the extracellular matrix (ECM) such as heparan sulfate proteoglycans (HSPG) (12) and fibronectin (217). The macrophage scavenger receptor was shown to bind L. monocytogenes lipoteichoic acids and hence may also be involved inListeria-macrophage interactions (160, 240). The bacterial ligands identified to date are all surface proteins, such as the internalins InlA and InlB, the actin-polymerizing protein ActA, and p60 (see below). Recently, several putative adhesion factors ofL. monocytogenes have been identified by screening transposon mutants for defective attachment to eukaryotic cells. Among these putative adhesins are the surface protein Ami, an autolysin with a C-terminal cell wall-anchoring domain similar to InlB (60,429, 445) (see below), and Lap, a 104-kDa surface protein involved in attachment to Caco-2 cells (581a). Also recently, a 24.6-kDa fibronectin-binding protein has been identified inL. monocytogenes by screening a genomic library inEscherichia coli for fibronectin binding (215). Bacterium-host cell interactions involving lectins (482) may also be involved in L. monocytogenes adhesion to eukaryotic cells (174). Indirect evidence has been obtained for an L. monocytogenes adhesin containing α-d-galactose, involved in uptake by mouse CB1 dendritic cells (253) and human HepG22 hepatocarcinoma cells (114) upon recognition of a carbohydrate receptor at the eukaryotic cell surface. The presence in L. monocytogenes of lectin-like ligands mediating binding to d-glucosamine,l-fucosylamine, andp-aminophenyl-α-d-mannopyranoside moieties has also been indirectly demonstrated (110).

Intracellular Proliferation and Intercellular SpreadDuring invasion, Listeria bacteria become engulfed within a phagocytic vacuole (194) (Fig. 3D). Little is known about the characteristics of the Listeria-containing vacuolar compartment, but the vacuoles become acidified soon after uptake (25). There is evidence that L. monocytogenes ensures its intravacuolar viability by preventing phagosome maturation to the phagolysosomal stage (11). Thirty minutes after entry, the bacteria begin to disrupt the phagosome membrane (194), and within 2 h, about 50% of the intracellular bacterial population is free in the cytoplasm (655) (Fig. 3E). This membrane disruption step is essential for listerial intracellular survival and proliferation (220) and is mediated by the hemolysin in combination with phospholipases (see below).

Once in the cytosol, bacteria multiply (Fig. 3E), with a doubling time of approximately 1 h (194, 518), i.e., approximately three times slower than in rich medium. In L. monocytogenes, intracellular growth does not involve the upregulation of known stress proteins (261) (see below), and auxotrophic mutants requiring aromatic amino acids or purines for growth in minimal medium grow inside cells at rates similar to those of the wild-type parent strain (413). This indicates that the cytoplasmic compartment is permissive for Listeria proliferation. However, three metabolic genes (purH, purD, andpyrE, involved in purine and pyrimidine biosynthesis) and an arginine ABC transporter (arpJ) have been shown to be induced within cells (336). Mutation of these genes had no major effect on intracellular proliferation and virulence (336), which can mean that certain nutrients are present in the host cell cytoplasm at concentrations that, although not limiting, are sufficiently low that L. monocytogenes has to upregulate certain metabolic genes to grow efficiently within cells. Recent experimental evidence indicates that pathogenicListeria spp. may exploit hexose phosphates from the host cell cytoplasm for efficient intracellular growth (543) (see below). Evidence has also been recently presented that antimicrobial peptides from the host cell cytosol, such as ubiquicidin found in macrophages (283a), may play a role in restraining intracellular proliferation of L. monocytogenes. Nothing is known about the listerial mechanisms counteracting these inhibitory peptides.

Intracytoplasmic bacteria are immediately surrounded by a cloud of a fine, fuzzy, fibrillar material composed of actin filaments, which later (around 2 h postinfection) rearranges to form an actin tail of up to 40 μm at one of the poles of the bacterium (see Fig. 6B). This actin tail consists of two populations of cross-linked actin filaments, one formed by relatively long, axially arranged actin bundles and the other by shorter, randomly arranged filaments. The polar assembly of actin filaments propels the bacterial cell in the cytoplasm with a mean speed of 0.3 μm/s, in a fashion reminiscent of the “action-reaction” principle that governs rocket movement. Moving Listeria cells leave an evanescent trail of F-actin undergoing depolymerization at the distal end, the steady-state length of which is proportional to the speed of bacterial movement. This movement is random, so some bacteria eventually reach the cell periphery, come into contact with the membrane, and push it, leading to the formation of finger-like protrusions with a bacterium at the tip (Fig. 3F and G). These pseudopods penetrate uninfected neighboring cells and are in turn engulfed by phagocytosis, resulting in the formation of a secondary phagosome delimited by a double membrane (Fig.3H), with the inner membrane originating from the donor cell (Fig. 3G). Bacteria escape rapidly (within 5 min) from the newly formed vacuole by dissolving its double membrane, reach the cytoplasm, and initiate a new round of intracellular proliferation and direct intercellular spread (117, 453, 549, 601, 652, 654-656) (Fig. 3A). In tissue culture, this infectious cycle is reflected in the formation of plaques in the cell monolayer (641, 672). Actin-based intracytoplasmic movement and cell-to-cell spread are mediated by the listerial surface protein ActA (see below).

HemolysinThe hemolysin gene, hly, was the first virulence determinant to be identified and sequenced in Listeria spp. Subsequent characterization of the hly locus led to discovery of the chromosomal virulence gene cluster in which most of the genetic determinants required for the intracellular life cycle of pathogenic Listeria spp. reside (see below). Thehly product, Hly, was also the first virulence factor for which a precise role in the pathogenesis of Listeriainfection was demonstrated. Hly is a key virulence factor essential for pathogenicity, having a vital role not only in intracellular parasitism but also in several other functions in the interaction of listeriae with their vertebrate host.

Characterization and role in escape from phagosome.In 1941, Harvey and Faber (269) demonstrated for the first time the production of a soluble hemolysin by L. monocytogenes. During the 1960s and 1970s, various researchers tried to purify this hemolysin and characterize its biochemical and toxic properties (218, 303, 305, 475, 620, 683). Jenkins et al. (303) were the first to provide evidence that theListeria hemolysin is similar in function and antigenicity to streptolysin O (SLO) from Streptococcus pyogenes. Kingdon and Sword (334) showed that the hemolysin was inhibited by cholesterol, that its optimum pH was below 7, and that it had cytotoxic properties in phagocytic cells. They were also the first to suggest that the hemolysin might be involved in the disruption of phagosomal membranes (15, 334). In 1985, Vicente et al. (676) reported the molecular cloning of a chromosomal fragment from L. monocytogenes that conferred hemolytic activity on E. coli. These authors were the first to use molecular genetics in the study of listerial virulence factors. Finally, in 1987, Geoffroy et al. (207) provided the first unambiguous evidence that the hemolysin of L. monocytogenesis an SLO-related cytolysin belonging to the family of cholesterol-dependent, pore-forming toxins (CDTX). This toxin was given the name listeriolysin O (LLO), and one of its key characteristics was determined, its low optimum pH (5.5) and the narrow pH range at which it is active (4.5 to 6.5) (207). Later, the 58-kDa CDTX hemolysin of L. ivanovii, ivanolysin O (ILO), was purified and characterized (669), and it was also shown that the weakly hemolytic but nonpathogenic species L. seeligeri produces, albeit in small amounts, an LLO-related CDTX (208, 376).

The strong correlation between hemolytic activity and pathogenicity in the genus Listeria (the pathogenic species L. monocytogenes and L. ivanovii are hemolytic [see Fig.5], whereas the nonpathogenic species are not; the exceptional case ofL. seeligeri will be discussed below) (250, 605,626) and the observation that the spontaneous loss of hemolysin production results in avirulence (288) led Gaillard et al. (195) and various other groups to generate isogenic hemolysin mutants of L. monocytogenes by transposon mutagenesis (108,323, 518). These mutants were much less virulent in mice (increase of >4 log units in the LD₅₀). Spontaneous revertants from which the transposon was lost recovered both hemolytic activity and pathogenicity, and reintroduction of the hlygene in trans into the nonhemolytic mutants restored virulence to wild-type levels (108, 195, 323, 518). Genetic analysis of the point of insertion for one of these mutants led, in 1987, to the identification and characterization of the hemolysin gene (432, 436). Cell culture studies of the effects of hly inactivation showed that hemolysin is required for the survival and proliferation of L. monocytogenes within macrophages and nonprofessional phagocytes (194, 359, 518). Electron microscopy of L. monocytogenes-infected cultured epithelial cells revealed thathly mutants remained within the vacuolar compartment, whereas hemolytic bacteria were free in the cytoplasm (194). The heterologous expression of hly inBacillus subtilis conferred on this nonpathogenic bacterium the ability to escape from the phagocytic vacuole and to proliferate within cells (although it did not become pathogenic, indicating that virulence factors other than LLO were required for infection) (45). These studies demonstrated the key role played by LLO in the intracellular infectious cycle of pathogenicListeria spp., as a mediator of phagosome membrane disruption. LLO not only mediates lysis of the primary phagosomes formed after the uptake of extracellular bacteria but is also required for the efficient escape of L. monocytogenes from the double-membrane vacuole that forms upon cell-to-cell spread (202) (Fig. 3D and H). The pores or membrane lesions caused by LLO (Fig. 4) probably facilitate the access of Listeria phospholipases to their substrates (see following section), leading to total dissolution of the physical barrier that delimits the phagosomal compartment.

• Open in new tab
• Download powerpoint

Fig. 4.

Pore-forming activity of LLO. Sheep erythrocyte ghost after treatment with purified LLO, showing the ring-shaped oligomeric structures of the toxin attached to the membrane. (Bar, 100 nm.) (Reproduced from reference 300 with permission of the publisher).

Sequencing of the genes for LLO (436), ILO, and theL. seeligeri hemolysin (seeligerilysin [LSO]) (257) revealed a high degree of similarity at both the protein (86 to 91%) and nucleotide (76 to 78%) levels and confirmed that the deduced proteins are members of the CDTX family of cytolysins. This family of toxins has 23 recognized members from different gram-positive genera, all of which are very similar in structure (40 to 70% amino acid sequence similarity) and show antigenic cross-reactivity (7, 8, 633). Portnoy et al. (519) demonstrated that, despite their structural and functional similarity, not all LLO-related toxins can mediate escape from phagosomes. These authors expressed the genes encoding SLO and perfringolysin O (PFO, the CDTX from Clostridium perfringens) in B. subtilis and observed that only PFO facilitated intracellular growth. However, unlike LLO, PFO damaged host cells (519). The pfo gene was also unable to complement an hly mutation in L. monocytogenesfor virulence in mice. The resulting strain was hemolytic and able to escape from phagosomes, but PFO again damaged the host cell, preventing the intracellular proliferation of the bacteria (310). This indicates that LLO has evolved structural features to prevent cytoxicity which are not present in PFO from an extracellular pathogen. By selecting for pfo-expressing L. monocytogenesmutants capable of normal intracellular growth, the same research group identified PFO mutations capable of mediating lysis of the vacuole but devoid of toxic effects on the infected host cell (311). One type of mutation decreased the cytosolic half-life of PFO (see below). Another type of mutation, very interestingly, modified the pH optimum of PFO, making it similar to that of LLO, with low levels of activity at pH 7.4 (a pH similar to that of the eukaryotic cell cytoplasm). This low optimal pH, similar to that within phagosomes, is thought to provide the basis for compartmentalized activity of LLO in an acidified vacuole and for protection against the potential deleterious effect on cell membranes of leakage of the toxin into the cytoplasm. This is probably a key element in the adaptation of L. monocytogenes to intracellular parasitism. Direct evidence for a link between the low optimum pH and compartment-specific pore-forming activity of LLO in macrophage phagosomes has been provided by a recent study using the membrane-nonpermeating fluorophore 8-hydroxypyrene-1,3,6-trisulfonic acid (25). It was shown that L. monocytogenes-containing phagosomes acidify rapidly after bacterial uptake. This was followed by an increase in pH and dye release from the vacuole. Loss of the fluorophore was prevented by lysosomotropic agents, such as ammonium chloride and bafilomycin A1 (25). Bafilomycin A1 was also shown to block L. monocytogenes escape from the primary vacuole in epithelial cells (102). These experiments indicate that a lowering in pH results in activation of LLO present in the phagosome lumen, leading to membrane destabilization, pH equilibration, and consequent deactivation of LLO.

Recent experimental evidence has shown that a PEST-like sequence present at the N terminus of LLO and absent in PFO is responsible for the rapid degradation of the listerial toxin in the host cell cytosol (127a). PEST-like sequences are thought to target eukaryotic proteins for phosphorylation and degradation, and deletion of this sequence in L. monocytogenes LLO led to increased cytotoxicity and lower virulence. When the sequence was introduced in PFO and the chimeric toxin was expressed in L. monocytogenes, bacteria were less cytotoxic than those expressing wild-type PFO and were able to replicate intracellularly. These data suggest that adoption of a PEST motif in LLO is another strategy used by L. monocytogenes, together with the low pH optimum, to restrict the activity of the toxin to the host cell vacuole, thereby preserving the intracellular niche for bacterial proliferation (127a). A recent report has also shown that L. monocytogenes mutants expressing a PEST-deleted hlyallele, although fully hemolytic, are strongly impaired in the ability to escape from the phagocytic vacuole. This suggests that the PEST motif in LLO may also play a specific role in the disruption of the phagosomal membrane, e.g., by serving as a ligand for targeting to that membrane other factors required for its efficient disruption (381a).

PhospholipasesPathogenic Listeria spp. produce three different enzymes with phospholipase C (PLC) activity that are involved in virulence. Two, PlcA and PlcB, are present in L. monocytogenes and L. ivanovii; the third, SmcL, is specific to L. ivanovii.

Characterization.Production of phospholipase activity by L. monocytogenes was first described in 1962 by Fuzi and Pillis (191), who reported opacity reactions in egg yolk agar that correlated in intensity with the hemolytic activity of the strains tested (see Fig. 12). This activity was detected in hemolytic fractions of L. monocytogenes culture supernatants and was at first wrongly identified as being due to the hemolysin itself (303, 304, 332). However, the application of careful fractionation procedures led to separation of the hemolytic and phospholipase activities (305, 620). The L. monocytogenes phospholipase was characterized as a secreted phosphatidylcholine (PC) cholinephosphohydrolase (PC-PLC;EC 3.1.4.3) similar to that responsible for the lecithinase activities ofBacillus cereus and C. perfringens(375). The purified PC-PLC/lecithinase is a zinc-dependent enzyme with an apparent molecular mass of 29 kDa, active over a wide pH range (5.5 to 8.0), and with a broad substrate spectrum (206,223). On lipid micellar dispersions, this PLC rapidly hydrolyzes PC, phosphatidylethanolamine (PE), and phosphatidylserine (PS). It also, to a lesser extent, hydrolyzes sphingomyelin (SM) but does not or only very weakly hydrolyzes phosphatidylinositol (PI) (206,223). However, PI is rapidly hydrolyzed in erythrocyte ghosts and in PI-PC-cholesterol liposomes, leading to the formation of inositol-l-phosphate (223). The purified enzyme is atoxic for mice; it is weakly hemolytic in guinea pig, horse, and human erythrocytes but not in sheep erythrocytes (206).

Lecithinase activity, like hemolytic activity, is an easily recognizable phenotypic marker closely associated with pathogenicity in the genus Listeria (528) (see Fig. 12). As forhly, transposon mutagenesis was used to identify the lecithinase gene. Two types of lecithinase-deficient, attenuated-virulence mutants were isolated. In the first type, the transposon was inserted into mpl (532), which encodes a zinc-metalloprotease (140), and the first gene of a 5.7-kb operon located immediately downstream from hly(434) (see Fig. 8). The second type of lecithinase-deficient mutant resulted from transposon insertion into the actA gene, located downstream from mpl in the operon and encoding the actin nucleator ActA (672) (see following section). Cloning of the DNA region targeted by the transposons in this mutant led to identification of the structural gene encoding the nonspecific listerial PLC, plcB. This gene occupies the third position in the 5.7-kb operon, called the lecithinase operon (see Fig. 8). The sequence of the plcBproduct, PlcB, is very similar to those of the PC-PLC precursor ofB. cereus (38.7% identity over 253 amino acids) and the α-toxin precursor from C. perfringens (22.4% identity over 223 amino acids) (672). Western immunoblotting of PlcB in the L. monocytogenes culture supernatant detects two antigenically related bands of 33 and 29 to 30 kDa (206, 470,532). This indicates that PlcB, like the B. cereusand C. perfringens PC-PLCs (309, 657), is secreted as an inactive proenzyme that is processed in the extracellular medium by proteolytic cleavage. The secreted PlcB proenzyme contains 264 amino acids, including a 26-residue N-terminal propeptide of 2.9 kDa, which is cleaved to give the faster-migrating mature, active enzyme. PlcB must be secreted in an inactive form to prevent bacterial membrane damage due to degradation of its phospholipids. The lecithinase-deficient mpl insertion mutant described above produced only the larger PlcB band corresponding to the inactive proenzyme (532), and complementation of this mutant with mpl restored the lecithinase phenotype and production of the active 29- to 30-kDa PlcB form (520). Thus, Mpl, which is encoded by the first gene of the lecithinase operon and, like PlcB, is also a zinc-dependent metalloenzyme, is involved in processing of the PC-PLC/lecithinase into its mature and active form in the L. monocytogenes culture supernatant.

The plcA gene, which encodes a PI-specific PLC, was identified in L. monocytogenes earlier than plcBby chromosome walking in the region upstream from hly(82, 377, 431) (see Fig. 8). Its secreted product, PlcA, is a 33-kDa polypeptide similar to the PI-PLCs from Bacillus thuringiensis, B. cereus, and Staphylococcus aureus(around 30% primary sequence identity) and eukaryotic PI-PLCs, such as that produced by Trypanosoma brucei (431). The purified PlcA enzyme is highly specific for PI, with a pH optimum between 5.5 and 6.5 in Triton X-100-mixed micelles, suggesting that, like LLO, it would be active in acidified phagocytic vesicles. No hydrolyzing activity was observed with PC, PS, PE, PI 4-phosphate, or PI 4,5-bisphosphate (222). The enzyme is able to hydrolyze glycosyl PI (GPI)-anchored eukaryotic membrane proteins (222,431), although much less efficiently than the orthologous enzyme from B. thuringiensis (200).

The animal pathogen L. ivanovii produces an additional PLC that is specific for SM. This SMase was purified from L. ivanovii culture supernatant after selective removal of ILO by sequestration with cholesterol and shown to contribute to the hemolytic activity of this species (669). In contrast to L. monocytogenes, which is weakly hemolytic, L. ivanoviiproduces a strong bizonal lytic effect on sheep blood agar, with an inner halo of complete hemolysis surrounded by a ring of incomplete hemolysis similar to that for β-toxin-producing S. aureusand B. cereus (546, 607, 669) (Fig.5). The outer ring of incomplete hemolysis is completely lysed if erythrocytes are exposed to a soluble product released by the actinomycete Rhodococcus equi(546, 626, 667, 668) (a cholesterol oxidase [467]), giving rise to a characteristic shovel-shaped patch of synergistic hemolysis (Fig. 5). This CAMP-like hemolytic reaction is routinely used for the identification of L. ivanovii (605, 606, 668). The gene encoding theL. ivanovii SMase, smcL, has recently been identified and sequenced. Its predicted protein product, SmcL, is highly similar (around 50% identity) to the known bacterial SMases from S. aureus (β-toxin), B. cereus, andLeptospira interrogans (226). Gene disruption and complementation studies have demonstrated that smcL is the determinant responsible for the differential hemolytic properties of L. ivanovii (225, 226). SmcL is functionally very similar to the S. aureus β-toxin in terms of membrane-damaging activity, as revealed by the similar ability of the two molecules to elicit shovel-shaped CAMP-like reactions with R. equi and the lytic selectivity for SM-rich membranes (both enzymes lyse sheep erythrocytes, in which SM accounts for 51% of total membrane phospholipids, but not horse erythrocytes, in which the SM content is only 13.5%) (Fig. 5). However, SmcL does not cause the hot-cold lysis typical of β-toxin (unpublished data), indicating that, despite their structural and functional similarities, there are differences in the mechanisms underlying membrane damage between these two homologous bacterial SMases.

• Open in new tab
• Download powerpoint

Fig. 5.

Hemolytic activities of pathogenic Listeriaspp. Left panels, spontaneous hemolysis of L. monocytogenes(L. mo.) and L. ivanovii (L. iv.) on sheep blood agar; compare the narrow ring of β-hemolysis produced by L. monocytogenes (here exceptionally evident) with the striking multizonal lytic activity of L. ivanovii, which is due to the production of an additional cytolytic factor (the sphingomyelinase SmcL). Bacteria were cultured at 37°C for 36 h and then left at 4°C for 48 h. Right panels, synergistic (CAMP-like) hemolysis ofL. ivanovii and β-toxin (sphingomyelinase)-producingS. aureus (S. au.) with R. equi (vertical streak); L. ivanovii and S. aureus elicit identical shovel-shaped patches of synergistic hemolysis, reflecting functional relatedness of the sphingomyelinases C produced by these two bacteria (see text and reference 226 for details). The CAMP-like effect results from the total lysis of the halo of sphingomyelinase-damaged erythrocytes exposed to a cholesterol oxidase from R. equi (467). As shown in the two upper right panels, the L. ivanovii sphingomyelinase is active towards sheep erythrocytes (which contain 51% sphingomyelin in the membrane) but not towards horse erythrocytes (13.5% sphingomyelin only). Note that the inner zone of spontaneous hemolysis that surroundsL. ivanovii colonies, due to the activity of the hemolysin (ivanolysin O), is not influenced by the source of erythrocytes. Bacteria were cultured at 37°C for 24 h.

ActAA great deal of attention has been paid in recent years to the study of the actin-based intracellular motility of pathogenicListeria spp. (27, 87, 105, 105a, 147, 366a, 401, 402,515, 581, 630, 636). No classical motor proteins such as myosin have been found to be involved in bacterial intracellular motility, and the actin tail of moving Listeria organisms is functionally similar to the poorly understood actin polymerization process observed in lamellipodia, responsible for the locomotion of eukaryotic cells. Therefore, interest in listerial actin-based motility has transcended microbial pathogenicity, as it provides cell biologists with a relatively simple model for dissecting the molecular mechanisms underlying actin cytoskeleton assembly and function. This model has the enormous advantage that a single bacterial protein is sufficient to promote the actin recruitment and polymerization events responsible for intracellular movement. This protein is ActA (142, 339), the product of the actA gene (672).

The central role of ActA in listerial intracellular motility and virulence was initially revealed by the unusual phenotype of an isogenic actA mutant of L. monocytogenes in infected tissue culture cells (339). After entry, these bacteria were capable of escaping into the host cytoplasm, but could not move intracellularly and accumulated as microcolonies in the perinuclear area of the cell (Fig. 6). The mutant was unable to recruit actin, as shown by staining with fluorescein isothiocyanate-labeled phalloidin, a fungal toxin that binds to F-actin and paralyzes the actin cytoskeleton. In addition, theactA mutant was highly attenuated in the mouse infection model (142, 339).

• Open in new tab
• Download powerpoint

• Open in new tab
• Download powerpoint

Fig. 6.

Actin-based intracellular motility. (A) Schematic structure and functional motifs of ActA (see text for details; amino acids numbered according to references 339 and 672). (B) Cos-1 cells infected with L. monocytogenes and labeled with FITC-phalloidin (actin stain, green) and rhodamine-conjugated anti-Listeria antibody (red) at 4 h postinfection; moving bacteria with actin tails are visible. (C and D) Giemsa stain of Caco-2 cells infected with L. monocytogenes wild-type (C) and ΔactA mutant (D) at 4 h postinfection; ΔactA mutant bacteria do not move intracellularly and do not spread to neighboring cells, growing in the cytoplasm as a microcolony close to the nucleus.

The actA gene encodes a secreted 639-amino-acid protein with a 610-residue mature form that is attached to the bacterial cell wall via its C-terminal region (Fig. 6). Several lines of experimental evidence have demonstrated the essential role of ActA in actin-based bacterial motility. Transfection experiments in epithelial cells showed that ActA is targeted to the mitochondrial surface and that these organelles recruit actin filaments in a way similar toactA-expressing bacteria (513). Microfilament-associated proteins such as α-actinin and fimbrin were also recruited to the ActA-decorated mitochondria in transfected cells. Thus, ActA, whether anchored to the bacterial or mitochondrial membrane, can induce the formation of an actin cytoskeleton. Although transfection experiments conclusively demonstrated that ActA alone was essential for actin filament recruitment, they did not show how this protein generates intracellular bacterial motility. This was achieved by heterologous expression of the actA gene in L. innocua and S. pneumoniae, bacteria that are otherwise incapable of actin-based motility. This demonstrated that ActA itself mediates actin polymerization-driven bacterial movement (340,631).

Mature ActA from L. monocytogenes can be arbitrarily divided into three distinct domains: (i) N-terminal domain (amino acids 1 to 234), which is rich in cationic residues; (ii) central region of proline-rich repeats (amino acids 235 to 394); and (iii) C-terminal domain (amino acids 395 to 610), with a highly hydrophobic stretch spanning residues 585 to 606 that anchors the protein to the bacterial surface (Fig. 6). Early structure-function studies with actAdeletion constructs expressed in transfected cells (189,513) or in a Listeria background and tested inXenopus egg extracts (365, 409) permitted an initial attribution of specific functions to these domains of the ActA protein. The N-terminal region plays an essential role, as it contains all the information necessary to initiate F-actin assembly and bacterial movement (Fig. 6). A domain rich in positively charged residues (amino acids 129 to 153) in this region was identified as being strictly required for actin assembly. Amino acids 117 to 121 were shown to be critical for filament elongation, and deletion of amino acids 21 to 97 led to discontinuous actin tail formation, suggesting a role in maintenance of the dynamics of actin-based motility for this domain. Interestingly, a synthetic peptide corresponding to part of this domain (residues 33 to 74) was shown to interact with F-actin in vitro (365). In the central region, a domain of proline-rich repeats (amino acids 265 to 396) is required for binding of the focal contact proteins vasodilator-stimulated phosphoprotein (VASP) and Mena (see below). Listeria mutants lacking this region of ActA are still able to recruit actin to the bacterial surface and to move intracellularly, albeit at significantly lower speeds. This indicates that while the VASP and Mena proteins are not essential for the initiation of actin assembly on the bacterial surface, they are important for the efficiency of the process (366, 472,632). There is significant structural and functional similarity between the C-terminal half of ActA (including the proline-rich repeat region) and zyxin (224), a eukaryotic protein associated with focal adhesions and actin stress fibers that may be considered a eukaryotic homolog of the listerial surface protein.

The mechanism by which ActA induces actin polymerization and bacterial motility is now beginning to be understood. ActA is not detected in the actin tail (341, 471), so clearly the listerial surface protein is not necessary for stabilization of the actin meshwork left behind by moving Listeria cells. In the proximal region of this tail, F-actin filaments tend to be unidirectionally polarized, with barbed ends oriented towards the bacterial cell surface, indicating that actin nucleation is occurring at this point (655,656). This suggested that ActA could act as an actin nucleator. However, globular (G)-actin monomer-binding activity could not be unequivocally demonstrated in ActA until very recently (96,708), so the question arose whether ActA must be modified in vivo in the mammalian cell to have adequate actin-binding function (e.g., by phosphorylation [70]), or whether another host molecule interacts with ActA and cooperates with it to trigger actin nucleation. Systematic searches for such a host factor among known actin cytoskeleton-binding, cross-linking, and regulatory proteins revealed that many, including α-actinin, tropomyosin, talin, vinculin, fimbrin (plastin), filamin, villin, gelsolin, ezrin/radixin, cofilin, profilin, coronin, Rac, CapZ, Arp3, and VASP, colocalize withListeria-associated actin filaments (86, 105, 117,126, 139, 338, 361, 409, 514, 564, 650, 653).

Profilin and VASP were found exclusively at the point of actin assembly, in close association with the bacterial cell (88,653), suggesting that they are actively involved in initiation of the Listeria-induced actin polymerization process. Profilin is a G-actin-binding protein, and its function in actin assembly is to bring ATP-actin to the barbed end of the filament in a complex that lowers the critical concentration for polymerization. Profilin interacts with polyproline (503), and poly-l-proline beads have been used to deplete profilin from cell extracts, resulting in the abolition of actin tail formation and bacterial motility (653). As profilin colocalizes with ActA at the site of actin polymerization and ActA has four polyproline motifs, it was thought that profilin was involved in actin nucleation by recruiting G-actin to the site of ActA expression on the bacterial surface. However, profilin was shown to associate only with bacteria in cell extracts or within infected cells, and not with ActA-expressing bacteria grown in culture medium, indicating that it requires an additional host factor to interact with ActA (653). VASP, in contrast, was shown to interact directly with ActA in vitro (88). VASP and its homolog Mena, which is also present between the bacterial surface and the growing actin tail (210), normally colocalize with actin filaments in focal contacts and in dynamic membrane processes (515, 538). VASP and Mena belong to the Drosophila Enabled (Ena)/VASP family of proteins, known to be involved in neural development. They are organized into three domains, an EVH1 (Ena-VASP homology 1) domain, a proline-rich central domain, and an EVH2 C-terminal domain. VASP has been shown to act as a ligand for profilin, vinculin, and zyxin (88, 210, 319, 537). Thus, by binding to ActA and profilin, VASP and Mena may establish a direct connection between intracellular Listeria cells and the cytoskeletal components of the host cell (210, 367). The EVH1 domain of the Ena/VASP-like proteins has been shown to bind directly to the proline-rich motif FPPPP, repeated three to four times in ActA, depending on the L. monocytogenes strain. This proline-rich ActA sequence is a novel polyproline motif that mimics the natural interaction between Ena/VASP proteins with their endogenous ligands in the focal adhesion proteins zyxin and vinculin. VASP has also been shown to interact with filamentous (F)-actin via its C-terminal EVH2 domain (85, 367, 472). These data strongly suggest that VASP may act as a bridge between polymerization-competent profilin/actin complexes and ActA and between ActA and the actin tail itself, thereby favoring assembly of the actin network that is used for bacterial propulsion across the cytosol. However, profilin is not essential for listerial actin-based motility (409), and the same is true for Ena/VASP-like proteins because, as indicated above, deletion of their ligand motif on ActA does not prevent (although it does impair) listerial actin-based motility.

A major clue to the mechanism by which ActA induces actin polymerization came from recent experiments showing that the N-terminal 264 amino acids of the protein interact with the Arp2/3 complex, resulting in a dramatic increase in actin nucelation activity (690, 691). First identified in Acanthamoeba, the Arp2/3 complex, which includes two actin-related proteins, Arp2 and Arp3, caps pointed ends, cross-links actin filaments, and has weak actin nucleation activity (403, 456). This complex was shown to initiate actin assembly on the surface of Listeriacells in a reconstituted cell-free system (689). The Arp2/3 complex is activated by proteins of the Wiskott-Aldrich syndrome protein (WASP)/Scar family, which connect the cytoskeleton with the signaling pathways involved in the dynamic remodeling of the actin framework (404, 562, 700, 704). It has been demonstrated that the p21-Arc subunit of the Arp2/3 complex binds to the C-terminal domain of Scar1 and the related WASP. Strikingly, the N-terminal 264 amino acids of ActA, the C-terminal p21-Arc-binding fragment of Scar1, and the C-terminal fragment of WASP are similar in sequence, and all increase the nucleation activity of the purified Arp2/3 complex (404). Indeed, C-terminal fragments of Scar1 that bind Arp2/3 completely block actin tail formation and Listeriamotility in cell extracts and in cells overexpressing the corresponding Scar1 constructs, although Listeria organisms are able to initiate actin cloud formation. Motility is restored by adding purified Arp2/3 complex (419). Thus, Arp2/3 is clearly an essential host cell factor for actin-based motility, and the major role of ActA is probably to regulate the activity of this multiprotein complex. Recent studies have demonstrated that the N-terminal region of ActA plays a critical role in binding and stimulating the nucleation activity of the Arp2/3 complex (514a, 626a, 708). Two domains in this ActA region (residues 56 to 76 and 92 to 109), with sequence similarity to WASP homology 2 domains, bind to actin monomers (708) (Fig. 6). A third N-terminal domain (residues 115 to 150, in particular the positively charged motif 117-KKRRK-121), with sequence similarity to the Arp2/3 homology region of WASP family proteins, binds Arp2/3 and competes for this binding with the the WASP family proteins N-WASP and Scar1 (514a, 708) (Fig. 6). It thus appears that ActA uses a mechanism similar to that of the endogenous WASP family proteins to promote Arp2/3-dependent actin nucleation.

However, using pure proteins in solution with assembled F-actin in the presence of ATP (the hydrolysis of which releases free energy for actin assembly), Loisel et al. (389) demonstrated recently that Arp2/3 in combination with VASP or WASP is not sufficient for actin-based motility. Wild-type movement was only reconstituted upon addition of capping protein and actin depolymerizing factor (ADF or cofilin), previously shown to be required for the depolymerization of actin tails (564) and to increase the rate of propulsion of L. monocytogenes in highly diluted, ADF-limited cell extract (86). In accordance with these observations, a cofilin homology sequence similar to that in WASP family proteins has been identified in the N-terminal fragment of ActA and shown to be critical for stimulating actin nucleation with the Arp2/3 complex in vitro and for actin-based motility in cells (626a). These data indicate that adequate turnover and the maintenance of high levels of ATP-G-actin and profilin-actin are essential for elongation of the actin tail and motility (389).

Gelsolin, a protein that caps barbed ends and severs actin filaments and which concentrates behind motile bacteria, was also shown to increase the length of actin tails and to support higher intracellular velocities. However, unlike ADF, gelsolin is not essential forListeria motility (361). Loisel et al. (389) also confirmed that neither profilin nor VASP is essential for actin-based motility, although their absence causes the rate of bacterial movement to decrease 3- to 10-fold. Similarly, the actin filament cross-linking protein α-actinin, which is uniformly distributed over the actin tail (117), is not required for movement, but in its absence there is an apparent lack of rigidity in the actin tail and a greater tendency for bacteria to drift in solution (389). These findings emphasize that while relatively few proteins are sufficient to build the core of the actin motor ofL. monocytogenes, many other host cell proteins, some with as-yet-unknown functions, may be involved in the actin cytoskeletal processes that lead to ActA-mediated listerial intracellular motility. Among these proteins is LaXp180 (506), recently identified using the yeast two-hybrid system aproach and shown to colocalize with intracellular Listeria cells. LaXp180 is a putative binding partner of the phosphoprotein stathmin, involved in the regulation of microtubule dynamics. By immunofluorescence it was demonstrated that stathmin also colocalizes with intracellular ActA-expressing L. monocytogenes (506). The role in infection of these newly identified proteins that are recruited to the listerial cell surface remains to be elucidated.

The motile force for intracellular bacterial movement is thought to be provided by the continuous deposition of actin monomers at the interface between the bacterial cell surface and the growing tail of tangled actin filaments immobilized in the cytosol that serves as a support structure for propulsion. To be effective, this motion system requires that actin deposition occur at only one pole of the bacterial cell, pushing it forward. This is indeed the case, as shown by the fact that the actin tail extends only from one end of the bacterium. This unidirectional actin polymerization appears to be related to the asymmetrical distribution of ActA on the bacterial cell. ActA accumulates at one of the bacterial poles and is totally absent from the other (the newly formed pole after bacterial division), which corresponds to the leading edge of movement opposite the place where actin tail formation occurs (341, 656). The basis of this polar ActA expression is not fully understood, but it clearly depends on the bacterial cell cycle and septum formation. Recent experiments with S. pneumoniae cells and polystyrene beads hemispherically decorated with purified ActA have provided convincing evidence that asymmetric distribution of ActA is indeed sufficient for directional actin assembly and movement (81, 631).

Because actin-based intracellular motility and intercellular spread are crucial for Listeria virulence, it is not surprising that the second pathogenic species of the genus, L. ivanovii, also possesses these features (320). The actAgene from L. ivanovii, i-actA, encodes a protein larger than ActA from L. monocytogenes. The similarity between the two proteins is also limited (34% identity) (230,348). However, ActA and i-ActA have a similar overall structure (charged N-terminal region, central region of polyproline motifs, and C-terminal membrane anchor region), and the functional domains required for actin recruitment and VASP binding are also conserved (209). The two proteins do indeed have the same function, and i-actA restores actin tail formation in anactA mutant of L. monocytogenes(230). Remarkably, the ActA proteins are the only polypeptides encoded by the central virulence gene cluster that have the same function but a high degree of sequence divergence betweenL. monocytogenes and L. ivanovii (see below). Differences in the interactions of the two species with the eukaryote host may have resulted in different selection pressures for conservation of actA genes in L. monocytogenesand L. ivanovii. Alternatively, actin-based motility may only require a set of conserved functional modules correctly exposed in space, the core protein carrying these modules tolerating significant deviations and even the insertion of other species-specific functional motifs. L. seeligeri has also been shown to containactA-homologous sequences (232). However, this species cannot induce actin polymerization in mammalian cells even if it is manipulated so that it can escape from the phagosomes of infected cells (321). The avirulence of this strain may have led to the mutational corruption of a protein with a function that is no longer required.

Two other intracellular bacteria, Shigella flexneri andRickettsia spp., and vaccinia virus have similar mechanisms of intracellular movement and cell-to-cell spread by continuous polar actin assembly (38, 190, 231, 278, 389). Bacterially induced actin polymerization has also been studied in great detail inS. flexneri, and interestingly, the bacterial protein involved in this process, IcsA, has no sequence similarity to ActA. IcsA also has a polar distribution on the bacterial surface and induces polar actin deposition but uses a different mechanism of actin nucleation, involving activation of the Cdc42 effector WASP to form a ternary IcsA-WASP-Arp2/3 complex (105a, 164). Thus, it appears that ActA and IcsA have evolved convergently, with both subverting the signaling pathway that controls actin cytoskeletal dynamics but by different strategies, accounting for their structural differences.

Recently, evidence has been presented that ActA may also be involved in the entry of L. monocytogenes into eukaryotic cells, probably by recognition of an HSPG receptor (12). Indeed, putative heparan sulfate (HS)-binding motifs, typically consisting of clusters of positively charged amino acids with interdispersed hydrophobic residues, have been identified in the N-terminal region of ActA. One such motif is very similar to the HS ligand of thePlasmodium circumsporozoite protein, and treatment of target cells with a synthetic peptide containing this malarial HS-binding motif significantly inhibited listerial attachment. In addition, by using CHO cells deficient in HS synthesis, it was shown that the absence of ActA reduced the binding and uptake of L. monocytogenes only in cells containing this glycosaminoglycan. A family of HSPGs, the syndecans, are bound to cell membranes via a membrane-spanning core protein, colocalize with actin bundles at the eukaryotic cell surface, and are thought to act as bridges between the ECM and the cortical cytoskeleton (530, 710). It therefore seems plausible that the interaction of ActA with an HSPG receptor induces transmembrane signaling events leading to cytoskeletal rearrangements and phagocytosis. Definitive confirmation that ActA plays a major role in L. monocytogenes internalization by host cells has recently been obtained by using an in-frame ΔactA mutant. Loss of ActA production in this mutant resulted in a decrease in invasiveness of up to a factor of 60 to 70 in epithelial cells, which was fully recovered upon complementation with the actA gene (Suárez et al., submitted for publication).

Although it is a major surface-exposed virulence factor, ActA is not presented by either MHC class I or class II molecules. This is only achieved if ActA is manipulated so that it is delivered in soluble form to the cytosolic MHC class I-processing compartment. However, the CD8⁺ CTLs so elicited against the listerial surface protein are not protective in vivo (120).

InternalinsInternalins are the protein products of a family of virulence-associated genes found in pathogenic Listeria spp. The first members of this family to be characterized, InlA and InlB, encoded by the inlAB operon, were identified in L. monocytogenes by screening a bank of transposon-induced mutants for impaired invasiveness in Caco-2 cell monolayers. InlA was shown to function as an invasin, mediating bacterial internalization by these normally nonphagocytic epithelial cells, and was therefore named internalin (193). A large number of internalin homologs have since been identified in both L. monocytogenes andL. ivanovii (150, 168, 169, 527) (Fig.7). An element common to all internalins is a leucine-rich repeat (LRR) domain consisting of a tandem repeat arrangement of an amino acid sequence with leucine residues in fixed positions (317). The typical LRR unit of internalins consists of 22 amino acids, with leucine or isoleucine residues at positions 3, 6, 9, 11, 16, 19, and 22 (— —L— —L— —L—L— —N—I— —I/L— —L). This sequence forms a novel, right-handed helix designated parallel β-helix, with a turn after every LRR unit, first identified in the pectate lyase of Erwinia chrysanthemi(277, 705). The LRR domain is thought to be involved in specific protein-protein interactions. LRR domain proteins are numerous among eukaryotes, in which they are involved in a variety of functions, such as cell adhesion, as is the case for decorin, fibromodulin, and the glycoprotein Ibg, or receptor-ligand interactions and signal transduction pathways, as is the case for Trk, CD14, and the adenylate cyclase of Saccharomyces cerevisiae (337). LRR proteins are less frequent among prokaryotes and most are virulence factors, like YopM of Yersinia pestis and Yersinia pseudotuberculosis, IpaH of Shigella flexneri, and the filamentous hemagglutinin of Bordetella pertussis (55,264, 407). The occurrence in pathogenic Listeria spp. of a multigene family of LRR-containing proteins is unusual among prokaryotes.

• Open in new tab
• Download powerpoint

Fig. 7.

Structure of the members of the internalin multigene family from L. monocytogenes and L. ivanovii(reproduced with modifications from reference 358 with permission of the publisher with data from Dominguez-Bernal et al. [unpublished data]). S, signal peptide; B, B-repeats; C, Csa domain repeats C-repeat; D, D-repeats. See text for details.

The internalins best characterized in terms of structure and function are InlA and InlB. InlA comprises 800 amino acids, which can be subdivided into two functional regions. The N-terminal half of InlA contains, in addition to a signal peptide (a feature common to all internalins), 15 LRR units (A repeats), whereas the C-terminal half contains three longer repeat sequences (B repeats) and a cell wall anchor comprising the sorting motif LPXTG followed by a hydrophobic membrane-spanning region of approximately 20 amino acids and a few positively charged amino acids (Fig. 7). This distal LPTXG motif, like similar motifs in other surface proteins covalently linked to the peptidoglycan in gram-positive bacteria, has been shown to be responsible for the attachment of InlA to the bacterial cell envelope in a process mediated by the enzyme sortase (105b, 135, 179, 371,465, 466, 590). InlB comprises 630 amino acids with seven LRR units in the N-terminal part (149). The LPXTG motif and the hydrophobic tail are absent from the C-terminal part of InlB. Instead, InlB has a 232-amino-acid region consisting of tandemly arranged repeats about 80 amino acids long, each starting with the sequence GW. This domain, known as the cell surface anchor (Csa) domain, is responsible for the attachment of InlB to the bacterial surface (60). The Csa domain mediates a novel mechanism of protein-bacterial surface association in which InlB is partially buried in the cell wall, interacts with the cytoplasmic membrane, and uses lipoteichoic acid as a ligand (313). This type of association appears to be relatively weak, as significant amounts of InlB are found in the supernatant of L. monocytogenescultures and InlB is released from the bacterial surface upon incubation in buffers with high Tris-HCl concentrations (60,455). Another surface protein of L. monocytogenes, the putative autolysin/adhesin Ami, also associates with the bacterial cell wall via a Csa domain (313).

inlAB mutants are severely impaired in host cell invasion, and both InlA and InlB have been shown to be required for entry ofL. monocytogenes into normally susceptible nonphagocytic cells in vitro (143, 149, 193, 196, 197, 242, 246, 385). The invasion processes mediated by InlA and InlB are apparently similar from a mechanistic point of view (of the zipper type [417,642]), but the two proteins follow different signaling pathways to achieve entry (see below). InlA and InlB are sufficient to trigger internalization by appropriate host cells, as shown by their ability to confer invasiveness on normally noninvasive bacteria, such as L. innocua and Streptococcus faecalis, and to induce the internalization of latex beads coated with the purified proteins (62, 63, 373, 455).

Functional analysis of InlA has shown that the N-terminal region encompassing the LRR domain and the interrepeat region is necessary and sufficient for bacterial entry into permissive cells (373). Monoclonal antibodies directed against the LRR domain block InlA-mediated entry, indicating that this structure is essential in the entry process (435). The host cell receptor for InlA is E-cadherin, a calcium-dependent intercellular adhesion glycoprotein composed of five extracellular domains and a cytoplasmic tail (205, 703). This receptor was identified by affinity chromatography on a column with covalently coupled InlA onto which a Caco-2 cell extract was loaded. Definitive evidence that E-cadherin is a principal receptor for InlA was obtained by transfection experiments using fibroblasts nonpermissive for invasion by L. monocytogenes; expression of the gene of the chicken homolog of E-cadherin, L-CAM, made these fibroblasts fully susceptible to invasion by recombinant L. innocua expressing theinlA gene but not to invasion by wild-type L. innocua or by a ΔinlA mutant of L. monocytogenes (435). E-cadherin accumulates at the gap junctions between intestinal epithelial cells but is also present over the entire basolateral face of these cells, consistent with experimental data showing that L. monocytogenespreferentially penetrates polarized intestinal Caco-2 cells via their basolateral surface (196, 650). E-cadherin is also present on the surface of hepatocytes, dendritic cells, brain microvascular endothelial cells, and the epithelial cells lining the choroid plexus and placental chorionic villi, all of which are potential targets during Listeria infection. The cadherin specific to the nervous system, N-cadherin, does not promote InlA-mediated entry (435), which correlates with the observation that direct invasion of neurons by L. monocytogenes is not efficient (151). However, certain sensory neurons seem to be permissive for InlA-mediated internalization of L. monocytogenes because they may express E-cadherin (146). InlA presumably interacts, via its LRR region, with the first extracellular domain of E-cadherin (372). This interaction leads to bacterial adherence, but entry is mediated by the intracytoplasmic domain of E-cadherin, which presumably leads to actin cytoskeleton rearrangement via α- and β-catenins (372a).

Upon contact with susceptible eukaryotic cells, InlB induces membrane ruffling and tyrosine phosphorylation of several proteins, including Gabl. Cbl, and Shc, implicated in membrane localization and activation of the phosphatidylinositol 3-kinase (PI3K) p85/p110. This PI3K, known to be involved in the control of the actin cytoskeleton, is stimulated by InlB, resulting in activation of PLC-γ1 and in increased levels of phosphoinositide second messengers. Inhibition studies have shown that p85/p110 PI3K does not participate in InlA-mediated uptake but is indispensable for InlB-mediated phagocytosis (45a, 296, 297). Met, a receptor tyrosine kinase which physiologically serves as ligand for hepatocyte growth factor (HGF), has been recently identified as a signaling receptor for InlB. Interaction with InlB tyrosine phosphorylates the 145-kDa subunit of Met, a binding partner of PI3K. Met-deficient cell lines are unable to mediate InlB-dependent entry, suggesting a major role for this receptor in L. monocytogenes invasion (615a). Using affinity chromatography, a second receptor for InlB has been identified in gC1q-R, the cellular ligand of the globular part of the C1q complement fraction. Pretreatment of target cells with soluble C1q or anti-gC1q-R antibodies impaired InlB-mediated entry, and transfection of the gC1q-R gene in nonpermissive cells promoted internalization of InlB-coated beads (61). However, gC1q-R lacks a transmembrane domain, suggesting that it may function as a coreceptor rather than a signaling receptor. As for InlA, the LRR region of InlB is sufficient for entry into mammalian cells (62, 297). The X-ray crystal structure of the InlB LRR domain has been recently determined: it consists of an elongated structure resembling a bowed tube with an extensive solvent-exposed surface area. Three regions were identified as potential candidates for mediating protein-protein interaction: two were in the concave face of the tube, consisting, respectively, of clusters of hydrophobic and negatively charged amino acids; the other was at the tip and contained two calcium ions, which may serve as bridging molecules in the interaction with host cell receptors (412a).

Evidence indicates that InlA and InlB have different cell specificities and may therefore play an important role in cell and host tropism. Dramsi et al. (149) analyzed in detail the effect of in-frame inlA and inlB deletions and corresponding gene complementations in an internalin double mutant onL. monocytogenes entry into various cell types. They showed that InlA but not InlB mediates entry into human Caco-2 intestinal cells and that InlB is necessary for invasion of murine TIB73 hepatocytes, whereas InlA is not. However, in the human hepatocyte cell line HepG-2, both InlA and InlB were required for internalization. More recently, the entry of L. monocytogenes into HUVECs and brain microvascular endothelial cells has been shown to be dependent on InlB but not on InlA (157, 246, 247, 498). The available data indicate that the role of InlA in invasion is restricted to E-cadherin-expressing cells, whereas InlB mediates entry into a wider range of cell types from various animal species, including certain epithelial and fibroblast cell lines, such as Vero, HEp-2, HeLa, CHO, L2, and S180 cells, as well as hepatocytes and endothelial cells (150, 246, 435, 455, 498). InlA has also been shown to contribute to at least 50% of serum-independent phagocytosis by macrophages (584), indicating that at least one macrophage receptor recognizes this listerial surface protein.

A major breakthrough in our understanding of InlA function has been the recent discovery by Lecuit et al. (372) that a Pro residue at position 16 in the first extracellular domain of E-cadherin is critical for this interaction with InlA. The presence of a Pro residue at position 16 of E-cadherin correlated with permissivity to invasion of the host cell. Thus, cells expressing mouse and rat E-cadherin, both of which have a Glu residue at this position, do not allow InlA-dependent entry, whereas cells expressing human E-cadherin, L-CAM, and guinea pig E-cadherin, with Pro16, are totally permissive to invasion by InlA-expressing bacteria. A Pro16→Glu substitution in human E-cadherin abolished interaction with InlA, whereas replacement of the Glu16 residue with a Pro in mouse E-cadherin resulted in a gain of function (372). These findings are important in that they are the first to identify a possible molecular basis for host specificity in a bacterial pathogen. They also make necessary reevaluation of the available experimental information on the role of InlA and InlB in virulence, in particular, data obtained from in vivo studies in mice and rats, in which no clear pathophysiological role of the inlAB determinant could be identified despite its proven involvement in cell invasion in vitro. In such in vivo studies, for example, the inlAB locus has been shown to affect only very slightly the persistence of L. monocytogenes in mouse liver and not to be essential for the invasion of hepatocytes in vivo (14, 149, 150, 197, 242, 385). Similarly, no role for theinlAB-encoded internalins was detected in intestinal invasion and translocation in mouse and rat ligated ileal loop models (119, 523) or in the development of murine cerebral listeriosis (589). Despite the different cell specificities of InlA and InlB, a double ΔinlAB mutant was 10 times less invasive than a single inlA mutant in Caco-2 cells, indicating that the two internalins synergize to optimize invasiveness. Such synergistic activity of InlA and InlB is also evident in human hepatocytes, with ΔinlA, ΔinlB, and ΔinlAB mutations reducing the rate of entry by factors of 2 to 5, 10, and 100, respectively (149). However, the heterologous expression of inlB did not confer on L. innocua the ability to invade murine hepatocytes (149). It is therefore possible that the absence of InlA recognition by E-cadherin in mouse tissues impairs the targetting ofL. monocytogenes to certain critical cell types in vivo, thereby preventing the appropriate interaction of InlB with host cell surface receptors.

The study of Lecuit et al. (372) raises important questions about the usefulness of the widely used mouse model, at least for those aspects in which InlA and InlB are potentially involved. Indeed, as described above, oral infection with L. monocytogenes leads to invasion of the intestinal epithelium in guinea pigs, which have the permissive type of E-cadherin, but not in mice and rats, which express a Listeria-nonpermissive isoform of the intercellular adhesion molecule.

The observation that InlB alone is not sufficient to mediate entry ofL. innocua into murine hepatocytes whereas complementation with the inlB gene restores full invasiveness to a ΔinlABmutant (149) clearly shows that other L. monocytogenes-specific products are involved in the interaction with host cells. This is also indicated by the significant level of residual invasiveness of inlAB mutants both in vitro (143, 150, 157, 196, 197, 247, 253, 385, 435) and, especially, in vivo (14, 149, 196, 242, 385, 589). It seems very likely that other members of the Inl repertoire play a role similar to that of InlA and InlB in the interaction with the eukaryotic cell surface. In addition to the inlAB operon, six other genes encoding putative surface-associated internalins (inlC2, inlD, inlE, inlF, inlG, and inlH) have been identified in L. monocytogenes. These internalin genes are distributed in three loci, two of which contain three internalin genes each and correspond to a single internalin islet that has presumably evolved differently in two isolates of L. monocytogenes serovar 1/2a (150, 527) (see Fig. 7 and below). All these newly identified internalins have signal peptides and LRR domains with various numbers of repeat units in their N-terminal halves and, like InlA, an LPXTG motif at their C-terminal ends, suggesting that they are covalently bound to the bacterial cell wall. Deletion of theinlGHE gene cluster did not affect entry into a number of mammalian cell lines, suggesting that these internalins, unlike InlA and InlB, are not involved in invasion. However, counts of the ΔinlGHE mutant in the liver and spleen were 3 and 2.5 log₁₀ units below those of the wild-type parental strain, indicating that the missing internalins are somehow important for host tissue colonization in vivo (527). Deletion of theinlGHE-homologous inlC2DE gene cluster and of the additional inlF gene identified in the other strain also had no effect on entry into various cell types; however, here the absence of these internalin loci did not affect the bacterial counts in mouse organs (150). This, together with evidence indicating that internalin gene homologs are also present in the nonpathogenic speciesL. innocua (193), suggests that members of this multigene family may have biological functions not exclusively related to virulence.

All the above-mentioned internalins are large, cell wall-associated proteins which, until very recently (see below), had been found exclusively in L. monocytogenes. Their genes are preferentially expressed extracellularly and, with the exception of theinlAB operon, are independent of the listerial virulence regulator PrfA (75, 527). A new group of LRR proteins has recently been identified in both L. monocytogenes andL. ivanovii. Structurally, they are very similar to the large internalins but lack the C-terminal (B) repeat region, the LPXTG motif, and the membrane anchor (Fig. 7). The internalin proteins of this class are therefore generally smaller (around 30 kDa versus the 80 and 71 kDa of InlA and InlB, respectively) and are released in soluble form into the culture supernatant. InlC (167), also called IrpA (143), is the prototype of this group of small, secreted internalins (167) and is the only secreted internalin detected to date in L. monocytogenes. Unlike the genes encoding the large, cell wall-associated internalins, all the genes encoding these secreted internalins are PrfA dependent. Indeed, this property was fundamental to the discovery of this subfamily of internalins: InlC was identified in the culture supernatant as a major 30-kDa protein that was overproduced in L. monocytogenesafter complementation with the prfA gene in multicopy, which results in the artificial upregulation of PrfA-dependent proteins (143, 167, 384). In L. ivanovii, five PrfA-dependent secreted internalins (i-InlC, i-InlD, i-InlE, i-InlF, and i-InlG) have been reported (168, 169, 350), three of which (i-InlE, i-InlF, and i-InlG) are encoded in a large virulence gene cluster containing additional genes encoding secreted internalins (225; Dominguez-Bernal et al., unpublished; see also below). Internalization and intracellular growth were found to be similar for an L. monocytogenes inlC deletion mutant and the parental wild-type strain when tested in several cell types in vitro. TheinlC mutation, however, resulted in significantly lower virulence in mice (decrease in 1 log₁₀ in the bacterial load in organs) after infection by the intravenous or oral route (143, 167). A significant loss of virulence in mice has also been observed for the Δi-inlE and Δi-inlF mutants of L. ivanovii (169). Like most PrfA-dependent genes (see below), inlC is preferentially expressed in the cytoplasm, especially at late stages of infection, when bacteria are in the process of active intercellular spread (75, 167). This suggests that InlC may be involved in the dissemination of infection. However, cell-to-cell spread was found not to be affected in an inlC mutant (143, 167,246). No intracellular target has yet been identified for InlC, and the role in virulence of the small, secreted internalins remains unknown.

Other Virulence FactorsThe listerial proteins reviewed so far are true virulence factors because they primarily accomplish tasks specifically required for parasitization of the vertebrate host and have evolved exclusively in pathogenic Listeria species. In addition to these virulence factors, other listerial products have been identified that also contribute to survival within the host. Although some of them have a major influence on the outcome of the host-parasite interaction, their participation in pathogenesis is more indirect, as they are probably involved in general houskeeping functions also necessary for saprophytic life. This does not exclude the possibility that some of these bacterial products may have evolved specific characteristics and differ in certain aspects from their counterparts in nonpathogenicListeria spp. as a result of adaption to the lifestyle of a facultative intracellular pathogen.

Central Virulencee Gene ClusterSix of the virulence factors responsible for key steps of L. monocytogenes intracellular parasitism (prfA, plcA, hly, mpl, actA, and plcB) are physically linked in a 9-kb chromosomal island formerly known as the hly or PrfA-dependent virulence gene cluster (88a, 350) and now referred to as LIPI-1 (for Listeria pathogenicity island 1) (670a). Figure 8 shows the physical and transcriptional organization of this “intracellular life” gene cassette. The hly monocistron, encoding LLO, occupies the central position in this locus. Downstream fromhly and transcribed in the same orientation is the lecithinase operon, comprising the mpl, actA, andplcB genes (434, 672). These genes are transcribed either as a single 5.7-kb mRNA under control of thempl promoter or as shorter transcripts, one specific formpl and the other starting at a promoter immediately upstream from actA and covering the actA andplcB genes and three additional small open reading frames (ORFs). The plcA-prfA operon is located upstream fromhly and is transcribed in the opposite orientation, either as a bicistronic mRNA of 2.1 kb or in monocistronic transcripts of 1.2 and 0.8 to 0.9 kb, respectively (187, 188, 433).

• Open in new tab
• Download powerpoint

Fig. 8.

Physical and transcriptional organization of the central virulence gene cluster (LIPI-1) of L. monocytogenes and structure of the locus in other Listeria spp. Genes belonging to LIPI-1 are in grey, and the flanking loci are in black. Dotted lines above L. monocytogenes LIPI-1 genes indicate known transcripts. LIPI-1 is inserted in a chromosomal region delimited by the prs and ldh genes, encoding the housekeeping enzymes phosphoribosyl-pyrophosphate synthase and lactate dehydrogenase, respectively. In the plcB-ldh intergenic region, two ORFs, orfA and orfB, are found in allListeria spp., indicating that the insertion point of LIPI-1 is between the prs and orfB loci. In theplcB-orfB intergenic region of L. monocytogenes, there are two small ORFs, orfX and orfZ(stippled), which delimit the putative excision point of LIPI-1 inL. innocua. In the plcB-orfB intergenic region ofL. ivanovii, two small ORFs, orfX (encoding a homolog of the orfX product from L. monocytogenes) and orfL (stippled), are also present, which, like those in L. monocytogenes, delimit the deletion point of LIPI-1 in the nonpathogenic species L. welshimeri. orfZ and orfL encode polypeptides with significant similarity to bacteriophage proteins (respectively Orf6 of the A511Listeria phage and a serine/threonine protein phosphatase from λ [88a, 670a]), suggesting that LIPI-1 was originally acquired by phage transduction (see text for details). The additional ORFs found in the L. seeligeri virulence gene cluster are hatched (the orfK product shows 39% similarity with the phosphatidylinositol phospholipase PlcA and dplcBis a truncated duplicate of the plcB gene). The figure has been constructed using data from references 88a, 232, 350, and672 and unpublished results from J.K. (L. seeligeri) and B.G.-Z. and J.-A.V.-B. (L. ivanovii). (Adapted from reference 670a with permission of the publisher.)

Phylogenetic analyses based on the sequence of the 16S and 23S rRNA have revealed that the genus Listeria comprises two evolutionary branches, one relatively distant, corresponding toL. grayi, and the other containing the remaining species. The latter are in turn divided into two sublines of descent, one comprising L. monocytogenes and the closely related nonpathogenic species L. innocua, and the other containingL. ivanovii and the nonpathogenic species L. seeligeri and L. welshimeri (98, 579) (Fig. 9). L. ivanovii, which has an intracellular life cycle similar to that of L. monocytogenes (320), carries a copy of LIPI-1 (232, 350, 670a) (Fig. 8). The genetic structures of LIPI-1 from L. monocytogenes and L. ivanovii are identical, but the corresponding DNA sequences exhibit only 73 to 78% similarity (350, 670a), a degree of divergence compatible with the genetic distance that separates the two species. One gene of LIPI-1, actA, is specially dissimilar, reflecting a divergent evolution in these two Listeria spp. This lack of similarity is also reflected at the amino acid level (34% identity) even though the corresponding ActA proteins have similar functions in actin-based motility (230, 348). LIPI-1 is absent from the genomes of the nonpathogenic Listeria spp. with one exception, L. seeligeri (88a, 350, 670a). Although in this species LIPI-1 genes are present in a structurally intact form (232, 257), they are not expressed correctly due to the insertion of a divergently transcribed ORF (orfE) between plcA and prfA, disrupting the positive autoregulatory loop crucial for adequate PrfA-dependent virulence gene activation (321, 350) (Fig. 8) (see below). The L. seeligeri LIPI-1 contains additional ORFs (orfCD) and duplications of the plcA and plcB genes (350, 670a). In L. monocytogenes, L. ivanovii, and L. seeligeri, LIPI-1 is stably inserted at the same chromosomal position (Fig. 8), has a DNA composition similar to that of the core genome, lacks any obvious traces of mobility factors (integrases, transposases, or insertion sequences [IS]), and its insertion point is devoid of typical integration signals (direct repeats or tRNA genes) (88a, 350, 670a). These data are consistent with LIPI-1's having been acquired a long time ago by a common listerial ancestor (350, 670a).

• Open in new tab
• Download powerpoint

Fig. 9.

Evolution of the central virulence gene cluster (LIPI-1) in the genus Listeria. The presence of LIPI-1 at the same chromosomal position in various species belonging to distinct evolutionary branches of the genus suggests that the virulence gene cluster was acquired before speciation (see text for details). Thick lines indicate the evolutionary pathway followed by LIPI-1; solid arrows indicate conservation of a functional LIPI-1 (in L. monocytogenes and L. ivanovii), whereas the dotted lines with an open arrowhead indicate functional corruption of the virulence gene cluster (in L. seeligeri). L. innocua and L. welshimeri presumably arose by excision of LIPI-1 in the corresponding evolutionary branches. The dendrogram is a schematic reconstruction of the phylogeny of Listeriaaccording to references 98 and 579. (Adapted from reference 670 with permission of the publisher.)

It is clear that LIPI-1 provides the host bacterium with access to a new niche, animal host tissues, which imposes selection pressures totally different from those encountered in the environment. The evolutionary history of the genus Listeria can thus be interpreted such that the acquisition of LIPI-1 by a clone of the primordial listerial ancestor was a crucial event that led to separation of the genus into two different phylogenetic branches, one corresponding to bacteria that remained saprophytic, resulting in the current species L. grayi, and the other to bacteria which, by virtue of the intracellular life cassette, became parasitic (350, 670a) (Fig. 9). It is reasonable to assume that the outcome of LIPI-1 in each of the two lines of descent of the parasitic branch also played a key role in the further diversification of the genus, as follows: (i) stabilization, in L. monocytogenesand L. ivanovii; (ii) early deletion (possibly when LIPI-1 still had its original mobility functions), in L. innocuaand L. welshimeri; and (iii) functional inactivation, inL. seeligeri (Fig. 9). The persistence of the LIPI-1 genes in intact form in L. seeligeri is intriguing. Because these genes no longer confer any virulence properties on the host bacterium, one would expect them to be eliminated or corrupted during evolution. One explanation for their conservation is that theprfA-inactivating insertion was a relatively recent event in the evolution of L. seeligeri, the additional ORFs present in LIPI-1 of this species representing subsequent gene insertions that were tolerated by the afunctional PAI. Alternatively, the gene cluster present in L. seeligeri may represent an ancestral form of LIPI-1 which is still evolving and which, in its present form, provides an advantage in a specific natural habitat other than the mammalian host but which also requires phagosome escape functions for survival (e.g., in protists that take up bacteria from soil) (350).

Two virulence determinants present in LIPI-1, hly andplcB, have homologs which are exclusively present in low-G+C gram-positive bacteria closely related to Listeria, such asBacillus and Clostridium. Homologs of two other LIPI-1 genes, plcA and the thermolysin-related zinc-metalloprotease gene mpl, are also common in this branch of gram-positive bacteria. This clearly suggests that the source of LIPI-1 genes is within the same phylogenetic division to which the genus Listeria belongs. PrfA, a member of the Crp/Fnr superfamily of transcription factors, is widespread in both gram-negative and gram-positive bacteria (221). ActA is the only LIPI-1-encoded protein for which no obvious structural homolog has been identified yet in prokaryotes. ActA carries functional domains present in eukaryotic proteins involved in the assembly of the actin cytoskeleton, such as vinculin, zyxin, and the WASP family of proteins (see above). It remains to be determined whether this reflects convergent evolution or indicates horizontal gene transfer between Listeria and eukaryotic cells. Whether LIPI-1 was acquired in a single recombinational event or whether it specifically evolved in the Listeria ancestor by a stepwise assembly process also remains unclear. Recent sequence analyses of theprs-ldh intergenic region of L. innocua andL. welshimeri have provided evidence that LIPI-1 was indeed once present in the chromosome of these species and that it became entirely excised, possibly in a single event, from the chromosome of these bacteria. Consistent with this, two small ORFs at the right end of LIPI-1 have been shown to exhibit significant similarity to bacteriophage or viral proteins, suggesting that the gene cluster (or at least part of it) was originally mobilizable by phage transduction (88a, 670a) (Fig. 8).

Internalin IslandsInternalins belong to a multigene family exclusive toListeria. Except for the inlC and inlFgenes of L. monocytogenes, the members of this family are always found in clusters of two to several genes, all oriented in the same direction (Fig. 10). The largest of the known internalin islands is LIPI-2, a chromosomal locus specific to L. ivanovii which contains 10 inl genes (see below). All inl genes have very similar sequences and encode highly homologous proteins with variable numbers of LRR repeats and, in the case of the large, surface-associated internalins, B and Csa (or GW) repeats. These repetitive DNA sequences are naturally prone to recombination, suggesting that the diversity of inl genes on the listerial chromosome is the result of gene duplications and intra- and intergenic rearrangements. There is evidence for such recombination events in inl genes. Comparison of the same internalin locus in two different L. monocytogenes strains revealed a different gene organization. The two strains shared only two structurally identical inl genes (inlG and E), and two internalin genes present in one isolate (inlC2 andinlD) (150) presumably recombined to generate a new internalin gene in the other (inlH) (527) (Fig. 10). In L. ivanovii, a tandem fusion of two small, secreted internalin genes in LIPI-2 has led to the generation of a hybrid gene encoding a larger, secreted internalin, i-inlG(Dominguez-Bernal et al., unpublished). (Fig. 7).

• Open in new tab
• Download powerpoint

Fig. 10.

Internalin islands and islets of L. monocytogenes and L. ivanovii. The scheme shows the different gene arrangements found for the same internalin locus in two isolates of L. monocytogenes, possibly resulting from rearrangements between inl genes (inlH presumably arose by recombination of the 5′-terminal part of inlC2 and the 3′-terminal part of inlD [350, 527]). See text for details.

The origin of the inl genes is intriguing. LRR domains are widespread in eukaryotic proteins but rare in bacterial proteins, suggesting that they are a relatively recent acquisition by prokaryotes. inl genes are also found in the nonpathogenic species L. innocua (193; EuropeanListeria Genome Consortium, unpublished). The presence of members of the internalin multigene family in Listeria spp. representing distinct phylogenetic branches of the genus suggests thatinl genes were already present in the common listerial ancestor and, consequently, that evolutionarily they are probably as old as LIPI-1. However, inl loci have a different distribution, gene organization, and location in the differentListeria spp. The inlG(C2D/H)H internalin islet of L. monocytogenes is present between two housekeeping genes (ascB and dapE) that are contiguous inL. ivanovii and other Listeria spp. (350). Also, inlC of L. monocytogenes is inserted alone, whereas its counterpart inL. ivanovii, i-inlC, is adjacent to another inlgene (i-inlD), and both are inserted at a different place of the chromosome (168). Another example is the large internalin island LIPI-2 of L. ivanovii, which is inserted between two housekeeping genes that are otherwise contiguous inL. monocytogenes. The analysis of the insertion points ofi-inlCD and LIPI-2 revealed that these two loci are associated with tRNA genes (168, 225). tRNA loci are commonly used as a target for integration of lysogenic phages and other mobile elements (539) and are frequently found at the insertion sites of PAIs in gram-negative bacteria (259,260). Consequently, inl genes have probably been not only transmitted vertically during evolution but also exchanged by horizontal transfer within the genus Listeria, possibly carried by phages, which in these bacteria are abundant, have interspecies infectivity, and are capable of DNA transduction (287, 338).

The conservation of such a repertoire of inl homologs in the genome of pathogenic Listeria spp. must necessarily reflect that each of the individual internalins plays a role in the biology of these bacteria. It has been suggested that variability in the exposed face of internalin LRR motifs is important to create a variety of specific protein-protein interactions (412b). Consistent with this notion, a direct involvement in receptor-facilitated invasion of permissive host cells has been demonstrated for the LRR domains of InlA and InlB (62, 373) and a role in cell and host tropism has been demonstrated for these internalins (149,372) (see above). In L. monocytogenes, genes encoding large, surface-anchored internalins are abundant (150, 193, 527), whereas there seem to be very few genes encoding small, secreted internalins (only one, inlC, has been identified to date) (167). In L. ivanovii, the reverse seems to be true, with many genes encoding small, secreted internalins and no genes encoding large, surface-associated internalins (168, 169,226; Dominguez-Bernal et al., unpublished). It is therefore possible that the different gene complements in these twoListeria spp. and the structural diversity of the internalins they encode facilitate the recognition of a variety of host receptors and are thus at least partly responsible for their different pathogenic characteristics.

In line with this assumption is the recent discovery in L. ivanovii of LIPI-2, a large (22 kb) gene cluster composed of many small, secreted internalin genes and the smcL gene, encoding the L. ivanovii SMase (225). LIPI-2 is specifically present in L. ivanovii within the genusListeria and may play a significant role in the pathogenic tropism of this species for ruminants. At least for the smcLproduct, there are indications for such a role in host tropism, as it selectively lyses sphingomyelin-rich membranes (e.g., those from sheep erythrocytes) (226) and promotes intracellular proliferation in bovine MDBK cells but not in canine MDCK cells (225). Most of the LIPI-2 inl genes are PrfA dependent, whereas smcL is absolutely independent of PrfA and is transcribed in the opposite orientation (226). This suggests that LIPI-2 was formed by the insertion of smcLinto a preexisting core of inl genes. ThesmcL-encoded enzyme is homologous to the S. aureus β-toxin and the B. cereus SMase (226), suggesting that the gene was acquired by L. ivanovii via horizontal transfer from either of these two low-G+C gram-positive bacteria. LIPI-2 has a DNA composition significantly different from that of the genome of L. ivanovii and is unstable, leading to a 17-kb deletion that affects the right part of the PAI. These characteristics, together with the expressional subordination of most of its genes to the LIPI-1-encoded PrfA protein, indicate that LIPI-2 is a more recent development that LIPI-1 in the evolutionary history of the genus Listeria. Spontaneous LIPI-2 deletion mutants are significantly impaired in virulence for mice and sheep and exhibit altered organ tropism in the latter animals (Dominguez-Bernal et al., unpublished).

PrfA, Master Regulator of VirulencePrfA is the only regulator identified to date inListeria spp. which is directly involved in the control of virulence gene expression. This protein is the main switch of a regulon including the majority of the known listerial virulence genes. Some of the members of the regulon, including LIPI-1 genes (withprfA itself; see below) and several genes of the subfamily of secreted internalins (e.g., inlC of L. monocytogenes and i-inlE of L. ivanovii), are tighly regulated by PrfA; others, such as the inlABoperon, are only partially regulated by PrfA (148, 167, 350, 385,433, 544). However, the expression of some listerial virulence genes is totally independent of PrfA, as is the case for the SMase gene of L. ivanovii, smcL (226), and theinlGHE internalin locus of L. monocytogenes(527). There is evidence that PrfA also negatively regulates some L. monocytogenes genes, such as the stress response mediator gene clpC (548) and the motility-associated genes motA (439) andflaA (544). This suggests that PrfA has a more global regulatory role in L. monocytogenes.

The deduced amino acid sequence of the PrfA protein shows similarities (20% overall identity) to that of the cyclic AMP (cAMP) receptor protein of E. coli (Crp, also known as CAP), on the basis of which the listerial regulator has been included in the Crp/Fnr family of bacterial transcription factors (347, 364). Crp/Fnr-like proteins have a role in virulence gene regulation in other mammalian and plant pathogens (221, 347, 694). Although most members of the Crp family of transcription factors are from gram-negative bacteria, Crp/Fnr-related transcriptional regulators have been also described in gram-positive bacteria (298). The PrfA homologs encoded by the prfA genes of L. ivanovii and L. seeligeri have a primary structure 80% identical to that of the L. monocytogenes virulence regulator protein. In L. ivanovii, PrfA also plays a central role in the coordinate regulation of virulence gene expression. InL. seeligeri, virulence genes are not expressed, but complementation with the plcA-prfA bicistron from L. monocytogenes activates transcription of the seeligerilysin O (slo) gene, leading to bacterial escape from the phagosome and intracellular bacterial multiplication (321). This suggests that the avirulence of this Listeria species is at least partly due to the insertion of the divergently transcribedorfE between the plcA and prfA genes (Fig. 8), resulting in a PrfA⁻ phenotype even ifprfA is intact and the corresponding PrfA protein is potentially active (see preceding section). Whether the complementedL. seeligeri strain is virulent has not yet been tested.

Two structural features that are important for Crp function are also present at the corresponding positions in PrfA (Fig. 11). One is a helix-turn-helix (HTH) motif in the C-terminal region (amino acids 171 to 191) which, as in Crp, mediates the interaction of the protein with specific DNA sequences in target promoters (612). These target sequences, called PrfA-boxes, are 14-bp palindromes centered on position −41.5 relative to the transcription start site in PrfA-dependent promoters (187, 437, 672) (Fig.11). DNA footprinting experiments have shown that PrfA binds in vitro to the PrfA-box (136). The second structural feature of Crp that is shared by PrfA is a series of short antiparallel β-strands delimited by Gly residues, which form a β-roll structure involving most of the N-terminal half of the protein (residues 19 to 99) (Fig. 11). In Crp, this structure forms the binding site for the activating cofactor, cAMP (344). However, cAMP is not detected and is not known to function as an intracellular messenger in gram-positive bacteria (291). Accordingly, most of the Crp residues involved in cAMP binding are not conserved in PrfA, and the addition of exogenous cAMP to L. monocytogenesdoes not lead to PrfA-mediated transcriptional activation (673). The role of this putative β-roll structure in PrfA is currently unknown. Two other functionally important regions of Crp, the D α-helix (amino acids 138 to 155), involved in transmission of the allosteric effect from the cAMP-binding N-terminal domain to the DNA-binding C-terminal domain, and activation region 1 (residues 156 to 164), involved in Crp-RNA polymerase interaction with class II promoters, are also particularly well conserved in PrfA (Fig. 11). PrfA does have some features that are unique among the members of the Crp-related family of transcriptional regulators, an additional HTH motif in the N-terminal region of unknown function, and a 25-amino-acid extension at its C terminus, which may form a leucine zipper (Fig. 11). Mutant forms of PrfA lacking this putative leucine zipper motif were found to be inactive (221).

• Open in new tab
• Download powerpoint

Fig. 11.

Structure of PrfA (A) and its target DNA sequences in PrfA-regulated promoters (B). See text for details. In PrfA, amino acid coordinates correspond to the peptide sequence deduced from theprfA gene (position 1 corresponds to the Met residue of the first triplet), whereas in Crp they correspond to the actual amino acid position in the primary structure of the native protein, which lacks the residue encoded by the first triplet. Therefore, the amino acid numbering of Crp is shifted one position with respect to that of PrfA, and hence, position 144 of Crp, where the Crp* mutation Ala→Thr lies, aligns exactly with position 145 of PrfA, where the Gly→Ser substitution leading to the PrfA* mutant phenotype is located. In Crp, A to F designate the α-helical stretches of the protein and AR1 is activating region 1 (two other activating regions [AR2 and AR3] are embedded within the β-roll structure). In PrfA, structures specific for this protein are in light gray. (B) N indicates any nucleotide. (Adapted from references 66 and 221.)

Environmental Control of Virulence Gene ExpressionVirulence factors are primarily required to colonize animal tissues and to overcome the host defense mechanisms brought into play during infection. Therefore, their synthesis outside the host may represent a burden compromising the ability of L. monocytogenes to survive in the natural environment and limiting its potential for transmission. Evidence is accumulating that L. monocytogenes has evolved signal-sensing and signal-transducing mechanisms to exploit the diverse physicochemical environments it may encounter to adapt virulence gene expression to the particular needs of saprophytic versus parasitic lifestyles. Temperature is one such environmental variables. Thus, PrfA-dependent transcription is weak below 30°C (environmental temperature) but becomes induced at 37°C (body temperature of warm-blooded animals) (124, 378,544). The mechanism underlying this thermoregulation is unknown.

An increase in temperature is, however, not sufficient for maximal levels of transcriptional activation of prfA and the genes of the PrfA-dependent regulon to be reached in L. monocytogenes. Wild-type strains express PrfA-regulated genes very weakly in rich media (e.g., brain heart infusion [BHI]) at 37°C, but transcription of these genes is strongly activated if the bacteria are cultured in BHI supplemented with activated charcoal (543-545, 548). The presence of this absorbent does not affect bacterial growth, indicating that the effect on virulence gene transcription involves a signal-sensing regulatory mechanism rather than a nutritional stress response due to sequestration of essential nutrients (545). One possible explanation for the “charcoal effect” is that activated charcoal absorbs a substance that is present in BHI but not in infected tissues and that allowsL. monocytogenes to detect that it is growing outside an adequate host compartment. One such substance might be iron, because low free iron availability defines the environment of host tissues (387) and there is evidence that hly expression is activated in conditions of iron limitation (101, 113,207). The specific binding of PrfA to target sequences in thehly and actA promoters is also strongly inhibited under high-iron conditions (51). However, charcoal-treated BHI does not retain its virulence-activating properties after removal of the charcoal by filtration (Novella et al., unpublished data), suggesting that the mechanism responsible is not related to the sequestration of a component of the culture medium.

Although the mechanism of charcoal-mediated activation of virulence genes remains unknown, the charcoal effect clearly demonstrates that signals from the growth medium play a critical role in regulating the expression of the PrfA regulon. It also shows that L. monocytogenes needs to sense a suitable combination of environmental signals of physical and, possibly, chemical nature to upregulate the PrfA regulon. This may represent a fail-safe mechanism by which Listeria bacteria prevent the expression of virulence genes in situations in which they are not required even if the temperature rises above the critical activation threshold. An induction of prfA and PrfA-dependent virulence genes similar to that observed with charcoal takes place when L. monocytogenes bacteria growing in BHI are transferred to minimal essential medium (MEM) (54, 635). MEM does not support normal growth of L. monocytogenes (635), so it is unclear whether the virulence gene-activating effect is due to a sensory mechanism involving chemical signals, a stress response to starvation, or both.

L. monocytogenes has been shown to respond to various stresses, such as exposure to high temperature (42°C) and oxidants or entry into stationary phase, by increasing, via induction ofprfA, the expression of a set of genes includinghly (634, 635). The upstream mediator involved in the activation of prfA under stress conditions is unknown. Detailed transcriptional analysis of the genes of LIPI-1 at 42°C revealed selective induction of hly andplcA, transcription of the genes of the lecithinase operon (mpl, actA, and plcB) remaining at basal levels (Ripio et al., unpublished). This expression pattern is identical to that observed if bacteria are confined within a phagocytic vacuole (75). It is therefore possible that a transient stress response involving the transcriptional activation of some genes of the PrfA regulon (those requiring lysis of the phagocytic vacuole) takes place when Listeria cells are still confined within the acidified phagosome, during the early stages of cell infection after invasion (see below). A simple decrease in pH to between 5.0 and 6.0 (i.e., the pH of acidified phagosomes) reduces hlyexpression in vitro (30, 124), indicating that other growth-arresting stress conditions found within phagosomes may be required for the induction of hly and the synergistic phagosome-disrupting virulence determinant plcA. It is also possible that de novo synthesis of hly plays a lesser role in phagosome disruption than acid-induced release of preformed intracellular stocks of the toxin, as postulated for PlcB (416).

Although wild-type L. monocytogenes cells produce low to undetectable levels of virulence factors in BHI, they are as virulent in mice as prfA* strains (545), which overexpress virulence genes constitutively in vitro (see below). As described above, wild-type strains can strongly induce PrfA-dependent virulence genes, up to expression levels similar to those in constitutive prfA* mutants, if grown with charcoal or in MEM. These observations led to the suggestion that the high levels of virulence gene induction reached by wild-type strains in vitro upon exposure to charcoal or MEM may correspond to the levels of expression required in vivo by L. monocytogenes to survive and proliferate in infected host tissues (545). A recent study has provided support to this hypothesis by showing that, despite their differences in spontaneous levels of expression in vitro, there is no significant difference in intracellular virulence gene activation between a wild-type strain and a prfA* mutant (75). Indeed, several PrfA-dependent virulence genes, such as plcA (336), actA (54, 137,451), and inlC (167), and PrfA synthesis itself (540), have been shown to be highly induced intracellularly. Investigation of L. monocytogenesintracellular gene expression using the green fluorescent protein as a reporter revealed that the actA promoter is activated within 30 min to 1 h of infection. This activation depended upon the ability of L. monocytogenes to reach the host cytosol, as it was not observed in a Δhly mutant unable to escape from the phagocytic vacuole (185). All these findings indicate that the expression of prfA and PrfA-dependent genes is upregulated if L. monocytogenes senses activatory signals specific to the host cytosolic compartment.

Carbon Source Regulation of Virulence GenesFermentable carbohydrates cause a strong repression of virulence genes in L. monocytogenes (30, 67, 124, 444,543). This downregulation occurs only if the carbohydrate is added in amounts sufficient to promote bacterial growth, suggesting that the underlying regulatory mechanism is related to catabolite repression (CR) (444). However, an L. monocytogenes mutant lacking CcpA, a central mediator of CR in gram-positive bacteria, did not show any alteration in carbon source regulation of virulence genes (29). Sugar-mediated repression affects not only the divergently transcribed virulence genes, plcA and hly, which have overlapping promoters, but also plcB, belonging to another transcriptional unit with no immediate physical link to plcAand hly and the promoter of which (PactA) is strictly dependent on PrfA (50, 67, 544) (Fig. 8). This suggests that repression is mediated by a trans-acting pleiotropic factor and somehow involves PrfA. Fermentable sugars such as cellobiose (336, 444, 541) and glucose (unpublished data) reduce PrfA-dependent virulence gene expression without affecting the levels of PrfA protein, indicating that this regulation is mediated by a repressor mechanism that either reduces the activity of the PrfA protein or blocks the binding of PrfA to its target promoters. The repressor mechanism appears to be situated hierarchically over the positive control pathway that activates the PrfA virulence regulon (see below), because fermentable carbohydrates annulate the induction of virulence genes by charcoal treatment (67, 543). Control of the synthesis of virulence-associated products by carbon source has also been reported for other pathogenic bacteria, such as E. coli (STb heat stable enterotoxin) (80),Actinobacillus actinomycetemcomitans (leukotoxin) (449), and Pseudomonas aeruginosa (PlcH hemolytic phospholipase) (575). Such a CR-like regulatory mechanism therefore seems to be widespread among pathogenic bacteria, suggesting that it may be important for the coordinate control of bacterial virulence. The actual relevance of this regulatory mechanism to the modulation of virulence expression in vivo is presently unknown.

An interesting observation is that in strain NCTC 7973 of L. monocytogenes, virulence gene expression was downregulated by the β-glucosides cellobiose and arbutin but not by common sugars, such as glucose, mannose, or fructose (67, 500, 501). Although the behavior of NCTC 7973 is anomalous and there is evidence that this strain has accumulated several regulatory mutations (30, 67,543), it suggests that sugars may use two independent mechanisms to bring about virulence gene repression in L. monocytogenes, one responding to any carbon source and another specific for β-glucosides. This is also suggested by the recent characterization of a mutant derived from the wild-type strain EGD, in which a clear uncoupling between glucose- and β-glucoside-induced PrfA-dependent virulence factor expression was observed (67). In addition, PlcB production is inhibited in NCTC 7973 by cellobiose and arbutin, but not by salicin (67), indicating that repression by β-glucosides is a complex phenomenon in which various sugar-sensing mechanisms, with different substrate specificities, may be involved. Recently, a 4-kb operon specific toL. monocytogenes, bvrABC, encoding a putative β-glucoside-specific phosphoenolpyruvate-sugar phosphotransferase system similar to that of the bgl and arb operons of E. coli and Erwinia chrysanthemi and thebgl regulon of B. subtilis, has been reported to be involved in virulence gene repression by cellobiose and salicin (67). Transcription of the enzyme II permease gene,bvrB, was found to be induced by cellobiose and salicin but not arbutin, and a bvrAB knockout mutation abolished the repression exerted by cellobiose and salicin. β-Glucosides are unique to the plant kingdom and are presumably abundant in decaying vegetation. Thus, Bvr-mediated sensing of β-glucosides may allowL. monocytogenes to recognize its presence in a soil environment and consequently switch off virulence genes. It is presently unknown how the the bvr locus triggers repression of PrfA-dependent genes. Another locus possibly involved in regulation of hly expression by cellobiose has also recently been identified by transposon mutagenesis (292).

In contrast to glucose and other common fermentable carbohydrates, glucose-1-phosphate (G1P) (in general, any hexose phosphate) efficiently stimulates the growth of L. monocytogeneswithout causing virulence gene repression (543). Sugar phosphate utilization is not constitutive in L. monocytogenes; it is a strictly PrfA-dependent phenotype that is coexpressed with other virulence factors (e.g., Hly, ActA, and PlcB), suggesting a role in pathogenesis. Consistent with this, the capacity to grow on hexose phosphate as a sole carbon source is also present in the pathogenic species L. ivanovi (as a PrfA-dependent phenotype as well) but absent from nonpathogenic Listeriaspp. As previously mentioned, PrfA-dependent genes become selectively activated in the cytoplasm of the host cell. In this compartment, the sugar phosphates G1P and glucose 6-phosphate are available as exogenous carbon sources for bacterial growth as a result of glycogen mentabolism. A role of hexose phosphate uptake in Listeriaintracellular proliferation has recently been confirmed with the identification and mutagenesis of the PrfA-dependent genehpt, encoding a sugar phosphate transporter (Chico-Calero et al., submitted for publication). The lack of repression of virulence genes observed with hexose phosphates indicates that uptake of these sugars bypasses the regulatory circuit responsible for triggering the CR response.

Regulatory Mechanism of PrfAThe clearest evidence that PrfA is a functional homolog of Crp was provided by the characterization of prfA* mutants ofL. monocytogenes (544). These mutants were first identified by Ripio et al. (545) when testing a panel of L. monocytogenes strains for Hly and PlcB production as markers of PrfA-dependent virulence gene expression. Three different phenotypes were observed (Fig.12). Phenotype 1 was exhibited by most of the strains tested and was characterized by low to undetectable levels of virulence factor production in BHI and the capacity to substantially (by 10- to 25-fold) increase this production upon culture medium supplementation with charcoal. Phenotype 2 was exhibited by only a few strains and was characterized by spontaneously high levels of virulence factor production and by unresponsiveness to charcoal, the levels of virulence factor production already being maximal. The third phenotype was exhibited by only one strain (LO28) and was characterized by intermediate levels of spontaneous Hly and PlcB production on BHI and responsiveness to charcoal. It was concluded that phenotype 1 corresponded to the wild type of L. monocytogenes and that phenotypes 2 and 3 represented mutant or variant phenotypes (545) (Fig. 12).

• Open in new tab
• Download powerpoint

Fig. 12.

Hemolysin (Hly) and lecithinase (PlcB) phenotypes ofL. monocytogenes. (A and B) Sheep blood agar; (C) egg yolk agar. (A) Original plate from which prfA* mutants ofL. monocytogenes were first isolated and identified; the culture shows a mixture of colonies of the wild-type isolate P14 (weakly hemolytic) and its spontaneous prfA* derivative, P14-A (strongly hemolytic) (see text and references 544 and545 for details). (B and C) Hly and PlcB phenotypes of L. monocytogenes (streaks 1, 2, and 3) compared with those ofL. ivanovii (streak 4). 1, weak to undetectable hemolytic and lecithinase activities typical of the L. monocytogeneswild type (clinical strain P37); 2, strong hemolytic and lecithinase activities of prfA* mutant (strains P14-A and NCTC 7973); 3, intermediate variant phenotype found in strain L028, of unknown molecular basis (L028 has a wild-type prfA); 4, strong hemolytic and lecithinase activities by L. ivanovii(clinical isolate P26); all the strains of this species characteristically overexpress PrfA-dependent virulence genes to levels similar to those of prfA* mutants of L. monocytogenes.

All phenotype 2 variants were strains that had been kept under laboratory conditions for a long time. Some were well-known collection strains widely used in research on listerial virulence, such as NCTC 7973 (Fig. 12) and the EGD Mackaness strain used in the laboratory of one of us (P.B.), originally obtained from R. J. North at the Trudeau Institute and thereafter renamed EGD-A to prevent confusion with the wild-type EGD strain currently used as a reference strain by the European Listeria Genome Consortium (545). Phenotype 2 variants were probably introduced inadvertently into research as model strains because their conspicuous hyperhemolytic phenotype facilitated the selection of Hly⁻ mutants during early studies on the role of hemolysin in Listeriavirulence. Phenotype 2 variants arise spontaneously at a very low frequency from wild-type strains during subculture in the laboratory and may be induced by chemical mutagenesis (171, 326,545). The prfA genes in all phenotype 2 strains tested to date have a point mutation at codon 145, leading to an amino acid substitution (Gly→Ser) in the PrfA protein. Complementation studies have shown that this mutation is responsible for the constitutive overexpression of virulence genes exhibited by phenotype 2 strains (544).

A PrfA protein with the Gly145Ser substitution renders the virulence regulon insensitive to a variety of environmental repressor signals, such as growth in rich medium, low temperature, and the presence of readily fermentable carbon sources (30, 543, 544). Gly-145 of PrfA aligns with Ala-144 of Crp (544, 673), a position at which any bulkier substitution (e.g., Ala144Thr) results in a mutant Crp* protein that no longer requires the cofactor cAMP to become transcriptionally active (344) (Fig. 11). Like Crp* mutant forms, the Gly145Ser PrfA protein has greater binding affinity for the specific target DNA sequence (673). The Gly145Ser mutation also increases the capacity of the PrfA-RNA polymerase complex to initiate transcription in vitro (Vega et al., unpublished data). From these observations, it has been suggested that PrfA functions via a cofactor-mediated allosteric transition mechanism similar to that of Crp and that the Gly145Ser mutation represents a cofactor-independent PrfA* form that is “frozen” in an active conformation (544,673).

The regulatory model currently proposed for PrfA is depicted in Fig.13. This model is based on two premises. The first is that PrfA has two functional states, inactive and active, and shifts from one to the other upon binding of a low-molecular-weight cofactor, not yet identified, that transduces the stimulatory environmental signals to the PrfA transcriptional machinery (544, 673). Evidence for the existence of this PrfA-activating cofactor has been obtained by the fractionation of bacterial extracts on sucrose gradients, in which a low-molecular-weight component seems to stimulate PrfA-target DNA complex formation (221). The second premise is that PrfA is always present, at least in minimal amounts, in the bacterial cytoplasm, which is also the case. prfA is expressed in two ways: (i) under the control of growth-regulated promoters in theplcA-prfA intergenic region, generating a monocistronic transcript from which PrfA is synthesized in low but detectable amounts even if bacteria are exposed to PrfA regulon-nonactivating conditions (186, 187, 433, 540, 541, 673); and (ii) under the control of a PrfA-dependent promoter in front of the plcA gene, which generates a plcA-prfA bicistronic transcript (83, 186, 187, 433) (Fig. 13; see also Fig. 8), creating an autoregulatory loop which is critical for the adequate activation ofprfA (378, 433, 541, 543, 544, 673). In the model proposed, the levels of virulence gene expression depend primarily on the functional status of the PrfA protein. This in turn controls the levels of prfA expression via positive feedback, resulting in strong PrfA-dependent transcriptional activation due to a rapid increase in PrfA in the active conformation, provided that the intracellular levels of the environmentally regulated stimulatory cofactor remain high (see the legend to Fig. 13). This positive-control model is highly versatile and makes possible a fine-tuned adaptive response to rapidly changing environmental conditions, such as would be encountered by L. monocytogenesduring its transition from free to parasitic life and during passage across the various physiological barriers, tissues, and compartments of the infected host. The model provides a coherent explanation of all the available experimental data on PrfA regulation and is also compatible with the possible existence of additional regulatory elements modulating the activity of PrfA or its interaction with the target DNA. These elements include the so-called PrfA-activating factor (Paf), a proteinaceous substance that may be a fraction of the RNA polymerase and that facilitates the formation of the PrfA/RNA polymerase/target DNA complex (50, 51, 136, 221), and the putative mediators of carbon source regulation of virulence gene expression (see above).

• Open in new tab
• Download powerpoint

Fig. 13.

Model for the mechanism of PrfA-mediated regulation. This model predicts that PrfA can undergo an allosteric transition between inactive (light gray) and active (dark gray) forms upon interaction with a low-molecular-weight hypothetical cofactor (see text for details). (A) Resting PrfA system. There is no cofactor, and the PrfA protein is synthesized at low, basal levels from the monocistronic transcripts generated from growth phase-regulated promoters in front of the prfA gene (small light gray arrow). (B) Activated PrfA system. If L. monocytogenes senses a suitable combination of environmental signals (i.e., 37°C and cytoplasmic environment), the intracellular concentration of the cofactor increases, leading to activation of the PrfA protein (a), which binds with increased affinity to the PrfA-boxes (black squares) in PrfA-regulated promoters (b); the transcriptionally active PrfA form causes the synthesis of more PrfA (in active conformation) via the positive autoregulatory loop generated by the PrfA-dependent bicistronic plcA-prfA transcript (c), thereby inducing the transcription of all PrfA-dependent genes (d) (dark gray arrows represent PrfA-dependent transcripts; the empty rectangle on the right represents any PrfA-dependent gene). The PrfA regulon remains switched on as long as the cofactor is present in the bacterial cytoplasm, but is rapidly switched off if the activating environmental signals cease and the concentration of the cofactor drops. (Reproduced from reference 673 with permission of the publisher.)

Fine Regulation of Virulence DeterminantsA second level of regulation is provided by the differential response of PrfA-dependent promoters according to the structure of their PrfA-boxes (Fig. 11). Binding affinity to PrfA-boxes is affected by the number of nucleotide mismatches they carry, becoming weaker as the sequence diverges from the perfect inverted repeat. Promoters with perfectly symmetrical palindromes (e.g., Phly and PplcA) are more sensitive and respond more efficiently to activation by PrfA than those with nucleotide mismatches (e.g., Pmpl, PactA/plcB, and PinlA) (Fig.11), which require larger amounts of the regulatory protein for full activation (53, 72, 186, 187, 544, 613, 698). This complementary cis-acting regulatory mechanism appears to be critical for the temporal and spatial expression of virulence genes and consequently for the correct development of the intracellular infectious cycle of L. monocytogenes. Early work by Freitag et al. (186, 187) and Camilli et al. (83) using different combinations of mutants affected in the monocistronic or bicistronic expression of prfA demonstrated that transcription of prfA from its own promoters in theplcA-prfA intergenic region was sufficient to mediate escape from the phagocytic vacuole and subsequent intracellular proliferation, but not to mediate cell-to-cell spread, which required a functional autoregulatory loop. These observations suggested that small amounts of PrfA synthesized during the early stages of bacterium-cell interaction are sufficient to activate rapidly and fully the high-affinity promoters of the hly and plcA genes, the products of which are involved in phagosome disruption. In contrast, activation of the low-affinity actA promoter, which directs the expression of the actA and plcB determinants involved in cell-to-cell spread, requires higher levels of PrfA, as provided by induction of the bicistronic transcript in the cytoplasm. Consistent with this, a recent study comparing wild-type bacteria withhly and actA mutants (which are confined in the phagosome and host cell cytoplasm, respectively) has shown that the PrfA-dependent membrane-damaging determinants hly andplcA are predominantly expressed in the phagosomal compartment, whereas the actA and mpl genes are selectively induced in the cytoplasm of the infected host cell (75).

As described above, experimental evidence has recently been obtained that PrfA synthesis is indeed activated in the host cell cytoplasm, correlating with higher activity of the PrfA-dependent plcApromoter that directs the expression of the plcA-prfAbicistron (540). However, prfA* mutants, in which PrfA-dependent expression is constitutively activated, are highly cytotoxic for the host cell (unpublished data). This may result from excess, unrestricted production of toxic virulence factors such as Hly and the phospholipases. In other words, activation of the PrfA system in the cytoplasm may be deleterious and incompatible with intracellular parasitism unless modulated by additional regulatory mechanisms (especially when L. monocytogenes reaches high intracellular population densities after multiplication). prfA transcript levels are surprisingly low within host cells (75), suggesting that a high turnover of prfA mRNA may be a mechanism that helps limit PrfA synthesis within host cells. The reported short half-life of Hly (127a, 311, 677) and PlcB (415) in the host cell cytoplasm may also play an important role in posttranslationally regulating the intracellular levels of cytotoxic virulence factors. This is clearly suggested by the observation that, whereas the levels of actA transcription in the host cell cytoplasm are only three times higher than those ofhly, the amounts of the corresponding proteins are very different, with 70 times more ActA than Hly (451). Additional negative control mechanisms acting at the transcriptional level may also contribute to modulate prfA expression within the host cell. For example, uptake of glucose from the host cell cytoplasm may counteract, by carbon source-mediated repression (see above), the stimulatory effect of the intracellular environment on PrfA function. Freitag et al. (186, 187) reported that in the absence of PrfA, transcription from the two promoters immediately upstream from prfA is increased, and deletion of the −35 region of one of these two promoters (P2) (see Fig. 11 and 13), with a structure reminiscent of a PrfA-box, leads to higher levels ofprfA transcription. This suggests that, by binding to theplcA-prfA intergenic region, PrfA may short-circuit the autoamplification loop, thereby downregulating its own expression. One study did not confirm the existence of such a PrfA-mediated negative autoregulatory mechanism (613), but a more recent study has shown that the transcription of a reporter gene under the control of the P2prfA promoter is significantly repressed by PrfA (72).

Some L. monocytogenes virulence genes, although part of the PrfA regulon, can also be expressed independently of PrfA. This may provide additional expressional characteristics important for the specific role of individual virulence factors in pathogenesis. A clear example is the internalin operon, inlAB, which is expressed from three promoters, two of which are PrfA independent. The third promoter is PrfA dependent, but its PrfA-box has two substitutions, resulting in a low binding affinity and inefficient induction by the listerial virulence regulator (53, 136, 148, 613). This promoter configuration is obviously responsible for the peculiar expression characteristics of the inlAB operon, which, in contrast to other PrfA-regulated loci, is significantly expressed during extracellular growth in broth medium at 37°C and positively regulated by elevated iron concentrations (i.e., PrfA-downregulating conditions) but poorly expressed within the cell cytoplasm or in MEM (i.e., PrfA-activating conditions) (53, 75,101, 167). This is consistent with the role of the products of this operon, the internalins InlA and InlB, as invasins mediating internalization from the extracellular space and with no known involvement in intracellular proliferation.

The hly gene, which is primarily PrfA dependent, has an alternative means of expression via a weak PrfA-independent promoter (141). This explains why L. monocytogenes prfAdeletion mutants have weak hemolytic activity (141, 544) and low-level expression of hly is detectable in L. monocytogenes in conditions of PrfA downregulation, such as extracellular growth in broth culture. As discussed earlier (617,619), low concentrations of exogenous Hly released by extracellular bacteria may contribute to pathogenesis by triggering cell responses, and the PrfA-independent promoter of the hlygene may be important in this respect.

Signal Transduction Pathways Associated with Epithelial Cell InvasionLittle is known about the signaling cascades triggered byListeria organisms during entry into cells that are normally nonphagocytic. InlB has been shown to be an agonist of the lipid kinase p85/p110 (PI 3-kinase [PI3K], and the InlB-mediated uptake ofL. monocytogenes by Vero cells has been shown to be associated with activation of the PI3K (296, 297). In addition to InlB, this activation also requires tyrosine phosphorylation in the host cell, because uptake is inhibited by genistein treatment, which blocks tyrosine-specific protein kinases. The mechanism by which the PI-3 kinase mediates uptake is unknown. However, the products of the reaction catalyzed by PI3K, phosphoinositide-3,4-bisphosphate and phosphoinositide-3,4,5-triphosphate, may directly interfere with the actin cytoskeleton by uncapping the barbed ends of actin filaments, thereby affecting cytoskeleton dynamics. In addition to phosphoinositide-phosphates, leukotrienes are also thought to be involved in L. monocytogenes signaling during the invasion of epithelial cells. Nordihydroguaretic acid, an inhibitor of 5-lipooxygenase activity, which generates leukotrienes from arachidonic acid, also inhibits the invasion by L. monocytogenes of Hep-2 cells (430). However, it is not yet known at which step of the putative signaling cascade leukotrienes are involved, or whether the leukotriene signal is initiated by InlA- or InlB-receptor interactions.

The treatment of various types of cell with the protein kinase C (PKC) inhibitor genistein (or staurosporine) inhibits invasion by L. monocytogenes, suggesting that PKC activity is critical for regulating the cytoskeletal changes necessary for listerial uptake (253, 353, 648, 675). It has recently been reported (648) that the invasion of HeLa cells by L. monocytogenes activates not only ERK-1 and ERK-2, but also two other MAP kinases (p38 and JNK) and the MAP kinase kinase MEK-1. ERK-2 is also phosphorylated upon invasion by LLO-negative mutants and is a downstream target of MEK-1. These data suggest that MEK-1/ERK-2 activation is one step in the signaling cascade leading to L. monocytogenes uptake by host epithelial cells. ERK-2 activation is not sensitive to wortmannin treatment, showing that it is not a downstream event of PI3K activation. Obviously, the PI3K and MEK-1/ERK-2 pathways may form two components of converging or independent signal transduction systems required for L. monocytogenes invasion. Another protein tyrosine kinase, pp60^c-src, is clearly activated during the entry of L. monocytogenes into epithelial cells because specific inhibition of pp60^c-src by herbimycin blocksinlAB-dependent entry (664). Treatment of host cells with TcdB toxins led to a dramatic breakdown of the normal actin cytoskeleton but did not abrogate L. monocytogenes invasion (or actin-based motility), indicating that the small GTP-binding proteins of the Rho and Ras subfamilies are not involved in the entry process (162).

The L. monocytogenes invasion process has been found to be sensitive to treatment with cytochalasin D for all cell types tested, demonstrating the importance of microfilaments for internalization (194, 359). Several recent reports have demonstrated that microtubules also play some role in invasion. The uptake of L. monocytogenes by mouse dendritic cells, mouse bone marrow-derived macrophages, and human brain endothelial cells is sensitive to nocodazole treatment, which disrupts microtubules (246, 253,352).

Activation of NF-κB and Modulation of Host Gene Expression in MacrophagesDifferential gene expression in eukaryotic cells depends largely on modification of the transactivating activity of inducible transcription factors. NF-κB, which is involved in regulation of the expression of many immunologically important genes (622), is the prototype of a family of transcription factors. NF-κB DNA-binding activity in response to L. monocytogenesinfection has been studied in the macrophage-like cell line P388D₁ (272, 273), endothelial cells (330, 600), and epithelial cells (271). A rapid invasion-independent increase of the NF-κB RelA/p50 DNA-binding activity is observed in macrophages within 10 to 20 min of addition of the bacteria. Studies with in-frame deletion mutants have shown that NF-κB is induced with biphasic kinetics upon L. monocytogenes infection. The first transient induction of NF-κB requires only the adhesion of L. monocytogenes to the P388D₁ cells. This induction also occurs upon infection with mutants of L. monocytogenes that have lost one or more of the known virulence genes and involves lipoteichoic acid, a cell surface component of L. monocytogenes. A second, permanent induction of NF-κB is detectable after release of theListeria bacteria into the cytoplasm of the host cell. This occurs exclusively with virulent L. monocytogenes strains and requires the synthesis of the bacterial phospholipases PlcA and PlcB in the infected host cell. It has been speculated that DAG and CER, likely products of the two enzymes (see above), act as second messengers in the induction of persistent NF-κB activation.

The expression of cytokines upon treatment of mammalian cells withL. monocytogenes or L. monocytogenes-derived products has been analyzed in many different cell systems. Infections were performed with defined cell populations, such as mouse bone marrow-derived macrophages (BMM), macrophage-like cell lines, and complex mixtures of cells, like spleen cell preparations (356,357). We will briefly review some of the studies and focus on those in which defined cell types were used as hosts for L. monocytogenes.

Early studies by E. A. Havell (274-276) showed that murine fibroblasts infected with L. monocytogenes produce significant amounts of alpha/beta interferon (IFN-α/β) and IL-6 and that mouse alveolar macrophages respond to L. monocytogenesinfection by producing IL-6 and TNF-α. Differences in the induction of cytokine expression were assessed for macrophages infected with live and killed L. monocytogenes-infected macrophages. Live and killed bacteria differ in IL-1, IL-6, and TNF-α production (447, 709), whereas IL-12 is induced in vitro by live and killed Listeria cells (94).

Using various well-defined mutants of L. monocytogenes, it has been shown that viable bacteria rapidly induce IL-1α, IL-1β, IL-6, and TNF-α mRNAs in the mouse macrophage-like cell line P388D₁, whereas killed bacteria induce only IL-1β mRNA (355). Hly⁻ mutants are unable to induce IL-1α, IL-6, and TNF-α but do induce IL-1β mRNA. The infection of BMM with L. monocytogenes also induces the proinflammatory cytokines (132). However, an Hly⁻ L. monocytogenes mutant was found to induce the same types and amount of cytokines as the wild-type strain, indicating that intracellular growth is not necessary for the transcriptional induction of these cytokines in BMM.

Not all the bacterial products mediating the induction of cytokines upon L. monocytogenes infection have been identified. The bacterial cell wall polymer lipoteichoic acid has been shown to induce the expression of the proinflammatory cytokines IL-1α/β, IL-6, and TNF-α in BMM (353) and in human monocytes (40). Nishibori et al. (474) analyzed cytokine gene expression upon treatment of spleen cells with purified LLO. Total spleen cell preparations containing macrophages and NK cells responded rapidly with increases in the synthesis of IL-1α, IL-6, IL-10, IL-12, TNF-α, and IFN-γ mRNAs. By depleting the cell preparations of either NK cells or macrophages, the contribution and interaction of the two cell types to cytokine production were assessed. It was found that IL-1α, IL-12, and TNF-α were produced mainly by macrophages, whereas IFN-γ was clearly produced by NK cells stimulated by macrophages, probably via TNF-α and IL-12.

The synthesis of cytokine receptor mRNAs in BMM infected with L. monocytogenes seems to be regulated inversely to the corresponding cytokines (132). TNF receptor type I and IFN-γ receptor are transcriptionally downregulated following the infection of BMM withL. monocytogenes, whereas IL-1 receptor type II mRNA is unaffected. In contrast, TNF receptor type II expression is upregulated. The bacterial requirements for this downregulation of cytokine receptor expression are unknown, but the downregulation is specific for infection with L. monocytogenes and was not found upon infection with nonpathogenic L. innocua strains.

Phagocytosis of L. monocytogenes is required for transcriptional induction of the heat shock protein 70 (hsp70) gene in macrophages. The synthesis of LLO seems to accentuate the hsp70 response (598). In contrast, expression of the gene encoding MHC class II antigen I-Ab is downregulated upon phagocytosis, at a time when the bacteria are still located mainly in the phagosomal compartment (595). Again, LLO seems to be important for this host response. Another important host cell gene that is downregulated on infection of macrophages by L. monocytogenes encodes antigen H-2K, a major component of the class I MHC molecule. Unlike downregulation of the gene encoding antigen I-Ab from the class II MHC, this event requires not only LLO, but also the expression of one or more genes of the listerial lecithinase operon (595).

Host Responses during Infection of Endothelial CellsHost cell responses elicited in endothelial cells during infection or by treatment with listerial exotoxins include lipid second-messenger generation (617, 618), NF-κB activation (330,600), stimulation of cytokines and chemokines, and induction of cell adhesion molecule expression (152, 153, 247, 330, 351,699). HUVECs were shown to respond to L. monocytogenes infection or treatment with purified LLO by increasing phosphoinositide metabolism and inducing the production of the vasoactive inflammatory mediators platelet-activating factor and prostaglandin I₂. This LLO-triggered second-messenger induction is further stimulated by the phospholipase PlcA, which acts in synergy with LLO (see above). It has been speculated that the pore-forming activity of LLO might facilitate the uptake of PlcB by the HUVECs, resulting in DAG and inositol phosphate generation (617). Listerial phospholipase-mediated increased levels of CER were also found in HUVECs during L. monocytogenesinfection. This increase is followed by an elevation in the activity of the transcription factor NF-κB (152, 600). The cell adhesion molecules for which expression was analyzed in L. monocytogenes-infected HUVECs and BMECs include E-selectin, P-selectin, VCAM-1, and ICAM-1 (152, 153, 351, 699). P-selectin expression peaked 30 to 60 min after infection, whereas the induction of E- selectin, ICAM-1, and VCAM-1 started 4 to 8 h postinfection. Cell adhesion molecule expression is partly dependent on LLO expression (351). E-selectin expression and the concomitant induced adhesion of polymorphonuclear leukocytes to HUVECs is particularly dependent on PlcA and PlcB synthesis by L. monocytogenes. These data show that the phospholipases are important virulence factors in the L. monocytogenes-endothelial cell interaction, interfering with signal transduction pathways mediated by lipid second messengers and causing changes in gene expression (600). The expression of IL-6 and IL-8 is also induced during the infection of HUVECs, but no data are available concerning the bacterial requirements for the induction of cytokine and chemokine expression upon infection (247).

Apoptosis L. monocytogenes has recently been shown to induce apoptosis in hepatocytes (560), Caco-2 epithelial cells (662), and dendritic cells, in which it was also demonstrated that LLO is the bacterial component triggering this event (252). L. monocytogenes-infected hepatocytes undergo apoptosis in vitro and in the infected mouse. It has been suggested that hepatocyte apoptosis, which is linked to neutrophil recruitment, eliminates infected cells rapidly from the tissue. Thus, apoptosis would inhibit the spread of L. monocytogenes to the neighboring cells and simultaneously would make the intracellularListeria cells accessible to effector cells. In dendritic cells, which are important antigen-presenting cells, LLO-induced apoptosis should result in lower levels of antigen presentation to immune cells, thereby downregulating the immune response.

Figure 14 summarizes the information available on the signaling cascades involved in the host cell response to Listeria infection.

• Open in new tab
• Download powerpoint

Fig. 14.

Summary of modulated signal transduction pathways and host cell responses identified during L. monocytogenesinfection of murine bone marrow-derived macrophages, murine P388D₁ and J774 cell line macrophages, human Caco-2 and HeLa epithelial cells, mouse dendritic cells, and human umbilical vein endothelial cells. Abbreviations: aSMase, acidic sphingomyelinase; gC1q-R, complement C1q receptor; Hsp70, heat shock protein 70; Hsp90, heat shock protein 90; HSPG-R, HSPG receptor; ICAM-1, intercellular adhesion molecule 1; IFN-γR, IFN-γ receptor; IPx, inositolphosphates; LTA, lipoteichoic acid; MEK-1, mitogen-activated protein kinase kinase 1; Met, receptor tyrosine kinase for HGF; MKP-1, mitogen-activated protein kinase phosphatase 1; PAF, platelet-activating factor; PGI₂, prostaglandin I₂; PIP2, phosphatidylinositol-(3,4)-bisphosphate; PIP3, phosphatidylinositol-(3,4,5)-trisphosphate; PI-PLC, phosphatidylinositol-specific phospholipase C; SR, scavenger receptor; TNF-RI, TNF receptor type 1; VCAM-1, vascular cell adhesion molecule-1. Reproduced with modifications from reference 357 with permission of Elsevier Science.