Listeria monocytogenes phosphatidylinositol-specific phospholipase C has evolved for virulence by greatly reduced activity on GPI anchors

Abstract

Listeria monocytogenes phosphatidylinositol-specific phospholipase C (PI-PLC) plays a critical role in escape of this human pathogen from host cell vacuoles. Unlike classical bacterial PI-PLCs, the L. monocytogenes enzyme has very weak activity on glycosylphosphatidylinositol (GPI)-anchored proteins. Previous crystal structure analysis has revealed that a small β-strand (Vb) is present in Bacillus cereus PI-PLC and is absent in the enzyme from L. monocytogenes. This Vb β-strand in B. cereus PI-PLC forms contacts with the glycan linker of GPI anchors, which presumably increases its activity on GPI-anchored proteins. In this study, we show that, of all known bacterial PI-PLCs, those from listeriae are the only ones that lack the β-strand. Expression by L. monocytogenes of B. cereus PI-PLC, which has strong activity on GPI-anchored proteins, inhibited bacterial escape from a vacuole and cell-to-cell spread, resulting in greatly reduced virulence in mice. Deletion of the Vb β-strand from B. cereus PI-PLC abolished its ability to cleave GPI-anchored proteins, decreased its inhibitory effects, and increased its virulence in mice. These results strongly suggest that L. monocytogenes PI-PLC has evolved as an important determinant of L. monocytogenes pathogenesis by absence of the Vb β-strand, thus leading to greatly reduced activity on GPI-anchored proteins.

Listeria monocytogenes is a Gram-positive facultative bacterial pathogen that has evolved for intracellular growth and virulence (1, 2). Subsequent to uptake by a host cell, the bacterium escapes from the phagosome, multiplies in the cytosol, and spreads directly to adjacent cells (3). Several virulence factors have been found to promote this intracellular adaptation. The expression of PrfA, a transcriptional activator of the virulence regulon, is thermoregulated through an mRNA secondary structure, leading to full activation of other virulence genes at 37°C (4). Listeriolysin O (LLO), a secreted pore-forming cytolysin, has specifically evolved for activity in a phagosome by having a low pH optimum (5). Two other virulence factors, a hexose phosphate transporter and a lipoate protein ligase, are specific adaptations for intracellular replication and survival (6, 7). Here, we report that Listeria monocytogenes phosphatidylinositol-specific phospholipase C (LmPI-PLC) has also evolved for intracellular growth and virulence by a specific structural modification that leads to greatly reduced activity on glycosylphosphatidylinositol (GPI)-anchored proteins.

LmPI-PLC (encoded by plcA) is comparable with Bacillus thuringiensis PI-PLC (BtPI-PLC) and Bacillus cereus PI-PLC (BcPI-PLC) in its ability to cleave PI; however, its activity on GPI-anchored proteins is much lower (8, 9). LmPI-PLC is an ortholog of BcPI-PLC, but their primary sequences are quite different, with 29% amino acid identity and 42% similarity. Despite this divergence, the tertiary structures of LmPI-PLC and BcPI-PLC are surprisingly similar. Both enzymes consist of a single domain with a barrel-like structure containing eight α-helices and eight β-strands (10, 11). The major structural difference between LmPI-PLC and BcPI-PLC is that the latter has an extra β-strand (Vb) in the C-terminal region (10). The Vb β-strand provides contacts for the glycan linker of GPI-anchored proteins, which presumably enhances its ability to cleave GPI anchors (12). Based on these structural comparisons, we extended our analysis to other bacterial PI-PLC-like sequences available from GenBank, and found that only the PI-PLCs produced by pathogenic Listeria species lack the Vb β-strand (Fig. 1). This analysis led us to consider the hypothesis that Listeria PI-PLC has evolved for intracellular growth and virulence by reducing its activity on GPI-anchored proteins through loss or absence of the Vb β-strand.

Fig. 1.

Multiple sequence alignment of PI-PLCs that are similar to LmPI-PLC and BcPI-PLC. The alignment of LmPI-PLC and BcPI-PLC is structure-based (10). The alignment of other sequences with respect to these two respectively was based on an automatic sequence alignment that was manually adjusted (13). β-strands are symbolized as arrows. Amino acids shown in red are assumed to form a Vb β-strand according to the BcPI-PLC sequence. Amino acids are numbered based on LmPI-PLC and BcPI-PLC, respectively.

Materials and Methods

Bacterial Culture. L. monocytogenes strains used in this study are listed in Table 1, which is published as supporting information on the PNAS web site. Unless otherwise specified, all bacterial strains were grown in brain-heart infusion (BHI) broth to stationary phase at 30°C. Selection and maintenance of strains harboring pAM401 was with chloramphenicol (10 μg/ml). Strains containing pKSV7 were selected and maintained with ampicillin (50 μg/ml) in Escherichia coli, and chloramphenicol (10 μg/ml) in L. monocytogenes. Bacterial growth in BHI was measured by changes in the OD₆₀₀.

Tissue Culture Cells and Growth Media. The mouse macrophage-like cell line J774 and murine L2 fibroblasts were grown in DMEM plus 7.5% FBS plus 1 mM l-glutamine at 37°C with 5% CO₂. For intracellular growth of L. monocytogenes, washed stationary phase bacterial cultures were used to infect J774 macrophages on 12-mm glass coverslips at 37°C. After 30 min, 50 μg/ml gentamicin was added to the culture medium to kill extracellular bacteria. Colony forming units (cfu) were determined by lysing host cells in sterile water and plating on BHI agar plates.

Construction of L. monocytogenes Mutants. Mutants were constructed by using either pAM401 for overexpression or pKSV7 for allelic exchange. Oligonucleotide primers used in this study are listed in Table 2, which is published as supporting information on the PNAS web site. All mutants were made through splicing by overlap extension; PCR and DNA sequences were confirmed by automated cycle sequencing. For overexpression, insertion of control and mutated genes into pAM401 was constructed by cloning with XbalI and SalI. All inserts are expressed under the plcA gene promoter with the LmPI-PLC original signal peptide sequence. Mutant constructs were electroporated into either DP-L1552 (AplcA) or WT, 10403S, and maintained with chloramphenicol (10 μg/ml). For allelic exchange, derivatives of shuttle vector pKSV7 were electroporated into competent L. monocytogenes and double crossover mutants were screened (14).

Escape from a Primary Vacuole. The escape of L. monocytogenes from the primary vacuole was determined by measuring the percentage of bacteria coated with actin filaments (stained with Alexa Fluor 568 phalloidin) in the cytosol (15). Briefly, J774 cells were plated and infected by using FITC-labeled bacteria for 90 min, and the ratio of escaped bacteria (red) over the total bacteria (green) was determined based on microscopy with appropriate filters.

Plaque Formation. Plaque formation assays using murine L2 fibroblasts were performed as described (16). Briefly, overnight, murine L2 fibroblasts cultured in six-well tissue culture plates were infected with L. monocytogenes for 60 min. The monolayer was then washed with PBS and covered with 2 ml of 1% DMEM-containing agarose with 10 μg/ml gentamicin. After incubation at 37°C for 3 days, a second overlay of agarose with Neutral Red was added to allow visualization of plaques. The diameters of plaques were measured and compared with that of the WT.

Mouse Infections. C57BL/6 female mice were housed in insulator cages and cared for in accordance with institutional animal care and use committee-approved protocols at the University of Pennsylvania School of Medicine animal facility. C57BL/6 mice, aged 6-10 weeks, were infected intravenously with 1 × 10⁴ cfu of L. monocytogenes in PBS. The infected mice were killed 3 days postinfection, and spleen and liver were harvested. Bacterial loads in spleen and liver were measured after homogenization in 1% Triton X-100/PBS, serial dilution, and plating on BHI agar plates at 37°C overnight.

Detection of PI-PLC Activity on PI. Enzymatic activity on PI by PI-PLC was detected by using ALOA Listeria agar plates (Microbiology International, Frederick, MD). The size of an opaque halo surrounding a colony reflects the activity on PI. A quantitative method using [³H-inositol]PI was done as described (9).

Detection of PI-PLC Activity on GPI-Anchored Protein. Splenocytes were harvested from C57BL/6 mice. Cleavage of the GPI-anchored protein Thy1.2 on CD4⁺ and CD8⁺ T cells was measured by FACS. Ten microliters of overnight bacterial supernatant was mixed with 1 × 10⁶ splenocytes in a well of a 96-well plate and incubated at 37°C for 1 h. Cells were then stained in PBS, 1% BSA with mAb to CD4, CD8, and Thy1.2. After washing, cells were analyzed with a FACSCalibur (Becton Dickinson), and data were analyzed by using flowjo 3.7 (Tree-Star, Ashland, OR). If the activity was weak, supernatants were concentrated 50-fold by using a Microcon centrifugal filter device (Millipore) and the incubation time was extended to 2 h.

Results

L. monocytogenes Expressing a PI-PLC with Strong Activity on GPI-Anchored Proteins Exerts Inhibitory Effects on Virulence. To test the hypothesis that gain of GPI-anchor cleavage would result in reduced virulence, we examined the effect of expressing BcPI-PLC with canonical activity on GPI anchors on intracellular growth and virulence of L. monocytogenes. We constructed an L. monocytogenes strain expressing mature BcPI-PLC from the chromosome in place of LmPI-PLC under the plcA endogenous promoter, using the original LmPI-PLC signal peptide (strain HG-L1001). In vitro, BcPI-PLC secreted by strain HG-L1001 exhibited the expected activities on PI and GPI-anchored proteins (Fig. 2). We then tested the efficiency of strain HG-L1001 in escape from the primary vacuole and its ability to spread from cell to cell using the J774 cell line and the fibroblast L2 cell line, respectively. We found that, despite its apparent equivalent activity on PI (Fig. 2 A and B), HG-L1001 performed like an LmPI-PLC deletion strain, DP-L1552 (ΔplcA), with decreased ability to escape from the host primary vacuole, compared with the WT strain 10403S (Fig. 3A). Furthermore, HG-L1001 showed significant impairment in cell-to-cell spread compared with both the WT and the plcA deletion strain (Fig. 3B and Table 1). The plaques formed by HG-L1001 were 65% the size of WT bacteria compared with 90% of WT for the plcA deletion. These results indicate that BcPI-PLC does not complement LmPI-PLC in escape from a primary vacuole and exerts an inhibitory effect on cell-to-cell spread. These results did not result from reduced growth, because HG-L1001 exhibited normal growth both in BHI broth and in cells of the murine macrophage-like cell line J774 (Fig. 7, which is published as supporting information on the PNAS web site).

Fig. 2.

PI-PLC activity on PI and GPI-anchored proteins. (A) Activity on PI. The indicated strains were patched on an ALOA plate before incubation at 37°C overnight. The halo around the blue bacterial lawn indicates PI-PLC activity. (B) Activity of culture supernatants on PI using a quantitative [³H-inositol]PI method. (C) Activity on Thy1.2, a GPI-anchored protein. Shown is FACS analysis of surface expression levels of Thy1.2 on T cells after PI-PLC treatment. Yellow, pure BtPI-PLC (5.4 μg/ml, positive control); red, supernatant of HG-L1001 (BcPI-PLC); green, supernatant of 10403S (LmPI-PLC).

Fig. 3.

Effect of BcPI-PLC expression on L. monocytogenes intracellular growth and virulence. (A) Phagosomal escape. Results are expressed as the mean ± standard deviations for three cell preparations. (B) Cell-to-cell spread of L. monocytogenes in murine L2 fibroblasts, as indicated by plaque size. The data represent the means ± standard deviations of three experiments. HG-L1008 (BcPI-PLC ΔplcB) produced no detectable plaques. (C and D) Mouse infections. Bacterial loads were measured in spleen and liver 3 days after i.v. injection of 6-week-old C57BL/6 mice. The data are representative of three experiments. Shown are 10403S (WT, ▪), DP-L1552 (ΔplcA, ▴), HG-L1001 (expression of BcPI-PLC on the chromosome, ▾), HG-L1006 (expression of BcPI-PLC ΔVb on the chromosome, ♦). Statistical analyses were performed with Student's t test. ^***, P < 0.001.

A previous study has shown that deletion of both plcA and plcB, a gene encoding a phosphatidylcholine-preferring phospholipase C (PC-PLC), resulted in a more dramatic phenotype in the plaque assay for cell-to-cell spread (15). To verify the inhibitory effect of BcPI-PLC, we generated an L. monocytogenes strain expressing BcPI-PLC on the chromosome in place of LmPI-PLC in a strain that had a plcB deletion (strain HG-L1008, BcPI-PLC, ΔplcB). HG-L1008 made no detectable plaques on L2 cells whereas its parental strain DP-L1936 (ΔplcA,ΔplcB) and DP-L1935 (ΔplcB) generated plaque sizes on average 36% and 63% of that of WT, respectively (15) (Fig. 3B and Table 1), confirming that BcPI-PLC strongly inhibits L. monocytogenes cell-to-cell spread.

To determine whether expression of BcPI-PLC has an inhibitory effect on intracellular growth and virulence in an animal model, we infected C57BL/6 mice intravenously with strain HG-L1001 and determined bacterial numbers in the spleen and liver. The expression of BcPI-PLC resulted in bacterial loads in the spleen and liver that were 3 logs less than that of the WT compared with only a 0.5-1 log decrease for a strain carrying a plcA deletion (Fig. 3 C and D). These results demonstrate that expression of BcPI-PLC with greater activity on GPI anchors has a dramatic inhibitory effect on the ability of L. monocytogenes to grow during the early stages of a murine infection.

Gain of Activity on GPI-Anchored Proteins by Insertion of Vb β-Strand from BcPI-PLC into LmPI-PLC. Because BcPI-PLC has activity on PI similar to that of LmPI-PLC (9) (Fig. 2 A and B), we suggest that the much greater ability of BcPI-PLC to cleave GPI anchors is the cause of the significantly decreased virulence of L. monocytogenes expressing this enzyme. To confirm that the Vb β-strand is critical for activity on GPI-anchored proteins, we attempted to increase LmPI-PLC's ability to cleave GPI-anchored proteins. We inserted the BcPI-PLC Vb β-strand region into LmPI-PLC and generated an L. monocytogenes strain (HG-L1007) overexpressing the altered enzyme on a plasmid (pAM401) in strain DP-L1552 (ΔplcA, no LmPI-PLC expression) (Fig. 8A, which is published as supporting information on the PNAS web site). We found that HG-L1007 exhibited very low activity on PI as compared with that of the parental strain DP-L1559 (DP-L1552+pAM401::plcA) (Fig. 4A). However, it gained measurable activity on GPI anchors when the supernatant was concentrated ≈50-fold (Fig. 4B). This result indicates that the Vb β-strand in BcPI-PLC is a structural requirement for GPI-anchor hydrolysis.

Fig. 4.

Gain of GPI-anchored protein cleavage activity upon insertion of the BcPI-PLC Vb β-strand into LmPI-PLC. Expression of mutant LmPI-PLC was on a plasmid in the strain DP-L1552 (ΔplcA) background. (A) PI-PLC activity on PI. The indicated strains were patched on an ALOA plate and incubated at 37°C for 2 days. The halo around the blue bacterial lawn indicates activity. (B) PI-PLC activity on Thy1.2, a GPI-anchored protein. Shown is FACS analysis of surface expression levels of Thy1.2 on T cells after PI-PLC treatment. Yellow, supernatant of HG-L1001 (BcPI-PLC); red, supernatant of HG-L1007 (LmPI-PLC with insertion of the BcPI-PLC Vb β-strand); green, supernatant of 10403S (LmPI-PLC). All supernatants were concentrated 50-fold by using Microcon filters.

Deletion of the Vb β-Strand from BcPI-PLC Restores L. monocytogenes Virulence. To further test the hypothesis that PI-PLC with activity on GPI anchors is deleterious for L. monocytogenes virulence, we generated two constructs with changes in the BcPI-PLC Vb β-strand. One construct (pNVB) had an in-frame deletion of the Vb β-strand; the other (pWVB) had four threonines changed to alanines in the Vb β-strand, leading to a predicted loop structure in place of the Vb β-strand (Fig. 8 B and C). Both constructs were based on the plasmid pAM401 with its plcA endogenous promoter and signal sequence. We introduced these constructs into the L. monocytogenes DP-L1552 (ΔplcA) background. Strain HG-L1004 (with pNVB) exhibited decreased activity on PI but no activity on Thy1.2, a GPI-anchored protein, as predicted (Fig. 5 A and B). Strain HG-L1005 (with pWVB) exhibited full activity on PI but decreased activity on GPI-anchored proteins (Fig. 5 A and B). These data further support the prediction that the Vb β-strand in BcPI-PLC is critical for activity on GPI-anchored proteins (12) (Fig. 8D). To test the effect of the two constructs on cell-to-cell spread, we used the plaque assay with L2 fibroblasts. The plaque size of strain HG-L1004 (with pNVB) was restored to that of a strain expressing WT LmPI-PLC on a plasmid in a ΔplcA background (DP-L1559). In contrast, the plaque size of strain HG-L1005 (with pWVB) still exhibited the observed inhibitory effect even though the activity on GPI-anchored proteins was reduced (Fig. 5C and Table 1). When BcPI-PLC was expressed from a plasmid in the WT background, the resulting strain HG-L1003 produced plaques similar to those produced by HG-L1002 (Fig. 5C), indicating that BcPI-PLC had a dominant negative effect over LmPI-PLC for cell-to-cell spread. Overexpression of an inactive form of LmPI-PLC (H86A, HG-L1009) produced plaques of reduced size similar to those produced upon overexpression of WT LmPI-PLC (DP-L1559) (Table 1). Thus, reduced plaque sizes produced by HG-L1002 and HG-L1003 do not result from increased PI-PLC activity.

Fig. 5.

GPI-anchor cleavage activity and cell-to-cell spread of L. monocytogenes expressing WT and BcPI-PLC. (A) PI-PLC activity on PI. The indicated strains were patched on an ALOA plate before incubation at 37°C overnight. The halo around the blue bacterial lawn indicates activity. (B) PI-PLC activity on Thy1.2, a GPI-anchored protein. Shown is FACS analysis of surface expression levels of Thy1.2 on T cells after PI-PLC treatment. Orange, pure BtPI-PLC (5.4 μg/ml, positive control); red, supernatant of HG-L1002 (overexpression of BcPI-PLC); green, supernatant of HG-L1005 (overexpression of BcPI-PLC with the Vb changed to a loop); blue, supernatant of HG-L1004 (overexpression of BcPI-PLC ΔVb); pink, supernatant of DP-L1559 (overexpression of LmPI-PLC). (C) Cell-to-cell spread in murine L2 fibroblasts, as indicated by plaque size. The data represent the means ± standard deviations of three experiments. 10403S (WT), DP-L1552 (ΔplcA), DP-L1559, HG-L1002, HG-L1004, HG-L1005, HG-L1003 (overexpression of BcPI-PLC in WT background), DP-L 1511 (overexpression of LmPI-PLC in WT background). Except for DP-L1511 and HG-L1003, all other overexpressions were in the DP-L1552 (ΔplcA) background. Plasmid overexpression of plcA results in decreased plaque size as described by Camilli et al. (14). Statistical analyses were performed with Student's t test. ^***, P < 0.001.

To test the effects of deletion of the Vb β-strand in vivo, we constructed an L. monocytogenes strain (HG-L1006) expressing BcPI-PLC without the Vb β-strand on the chromosome in place of LmPI-PLC as we did for HG-L1001. HG-L1006 showed decreased activity on PI and no detectable activity on Thy1.2 (Fig. 6 A and B). When HG-L1006 was tested for cell-to-cell spread, it seemed to have recovered its ability to form plaques similar to that of WT (Fig. 6C Table 1). Further, mouse infections confirmed this result, in that growth in both the spleen and liver was largely restored (Fig. 3 C and D). Strain HG-L1006, however, showed no improvement in its ability to escape from the primary vacuole of J774 cells (Fig. 3A). This result is in agreement with previous findings that PI-PLC activity is critical at this step (Fig. 6A). Thus, BcPI-PLC with a deletion of the Vb β-strand performed much like LmPI-PLC in L. monocytogenes during more prolonged infections. It is possible that this restoration of virulence resulted from instability in vivo of BcPI-PLC with the Vb strand deleted. If it were completely inactivated by the host, we would expect the strain expressing it to perform like the plcA deletion strain, which it does not (Fig. 3D).

Fig. 6.

Recovery of L. monocytogenes intracellular growth and virulence by deletion of the Vb β-strand from BcPI-PLC. (A) PI-PLC activity on PI. 10403S (WT), DP-L1552 (ΔplcA), HG-L1001 (BcPI-PLC), HG-L1006 (BcPI-PLC ΔVb). (B) PI-PLC activity on Thy1.2, a GPI-anchored protein. Shown is FACS analysis of surface expression levels of Thy1.2 on T cells after PI-PLC treatment. Red, supernatant of HG-L1001; green, supernatant of HG-L1006; blue, supernatant of DP-L1552. All supernatants were concentrated 50-fold by using Microcon filters. (C) Cell-to-cell spread in murine L2 fibroblasts, as indicated by plaque size. The data represent the means ± standard deviations of three experiments. Statistical analyses were performed with Student's t test. ^***, P < 0.001.

Discussion

Bacterial PI-PLCs have been identified in several Gram-positive bacteria including B. cereus (17), B. thuringiensis (18), Bacillus anthracis (19), Staphylococcus aureus (20, 21), Listeria ivanovii (22, 23), and L. monocytogenes (14, 22), but their precise role in pathogenesis remains unknown in most cases. Studies of L. monocytogenes infections in J774 cells have shown that diacylglycerol generated by hydrolysis of PI mobilizes PKC β-isoforms, which seem to play a role in escape from the phagosomes (24, 25). In this study, we demonstrated that the replacement of LmPI-PLC with BcPI-PLC inhibits L. monocytogenes intracellular growth and virulence most likely through the cleavage of GPI-anchored proteins. Our data suggest that LmPI-PLC has evolved as an essential determinant of L. monocytogenes pathogenesis by loss or absence of the Vb β-strand needed for efficient GPI-anchor hydrolysis. As predicted from a structural comparison, the absence of the Vb β-strand leads to greatly reduced activity on GPI anchors (8, 10). There are hundreds of diverse GPI-anchored proteins on eukaryotic cell surfaces, some of which are important for innate immunity and intracellular signaling (26). Many of these GPI-anchored proteins are concentrated in cholesterol- and sphingomyelin-rich domains in eukaryotic cell surfaces (26, 27). Intriguingly, the pore-forming activity of listeriolysin O (LLO) depends on the presence of cholesterol, and its activity is required for numerous physiological activities induced by PI-PLC in L. monocytogenes infections (28). It seems that GPI-anchored proteins are being spared by the LmPI-PLC. If so, learning which of these GPI-anchored proteins are critical for L. monocytogenes intracellular growth and virulence will further enrich our understanding of its adaptation to the host environment.