Infection of cultured human endothelial cells by Legionella pneumophila.

Legionella pneumophila is a gram-negative pathogen that causes a severe pneumonia known as Legionnaires' disease. Here, we demonstrate for the first time that L. pneumophila infects and grows within cultured human endothelial cells. Endothelial infection may contribute to lung damage observed during Legionnaires' disease and to systemic spread of this organism.


Introduction
Legionella pneumophila causes an estimated 2-15% of community acquired pneumonia requiring hospitalization [1]. L. pneumophila is a gram-negative pathogen whose primary ecological niche appears to be as a parasite of protozoa. However, humans and other animals may become secondarily infected after inhaling or aspirating organisms.
Fascinatingly, the cell biology of infection of protozoa and humans cells appears remarkably similar. On entering both macrophages [2] and protozoa [3,4], L. pneumophila inhibits phagolysosome fusion [5] and multiplies in a compartment having properties of endoplasmic reticulum [6]. Escape from phagolysosome fusion, intracellular multiplication, and lysis of the host cell relies on a type IV secretion system encoded by the dot/icm gene complex [7][8][9].
Resident alveolar macrophages appear to be a Trojan horse for human infection [10]. L. pneumophila infects these cells and continues to grow within macrophages recruited during the subsequent inflammatory response. Infection may then spread to alveolar epithelial cells that line the airways, supported by observations of growth in primary [11] and transformed type I [12] and II pneumocytes [13], and growth of the related L. dumoffii in alveolar epithelial cells in vivo [14].
Bacterial dissemination does not stop at the alveolar wall. In both human autopsy studies [15] and animal models [16], L. pneumophila spreads to multiple organs. This suggests that the organism may infect and breach the remaining anatomic barrier in the lungpulmonary blood vessels -as a potential initiating event in systemic spread of the organism via the blood stream. Here, we begin to address this hypothesis through in vitro study of the interaction of Legionella pneumophila with primary human endothelial cells.

Infection of endothelial cells
We first tested whether Legionella pneumophila strain Lp01 [17] invades into primary human umbilical vein endothelial cells (HUVEC). Using a gentamicin protection assay, we assessed whether bacteria were protected from gentamicin [18], a polar antibiotic that penetrates slowly into eukaryotic cells and does not kill internalized bacteria during the time frame of the assay. Using this procedure, we found that L. pneumophila, strain Lp01, reproducibly invades into endothelial cells (see Fig. 1) with approximately 0.05% of bacteria internalized during a 2 hour incubation. The ability of L. pneumophila to invade into endothelial cells was confirmed by fluorescent microscopy where internalized bacteria were distinguished by differential staining as described in detail below.
Of note, the invasion efficiency was roughly 100 fold lower than what was previously observed for primary macrophages [19], an expected result since endothelial cells are not professional phagocytes. We compared invasion efficiency to the L. pneumophila, isogenic dotI mutant, HL051C [20], which was previously shown to be defective in inhibition of phagolysosome fusion and intracellular replication in macrophages. Although a trend towards slightly lower invasion efficiency by HL051C was noted, this did not reach statistical significance (Fig. 1). Previously, invasion of a dot/icm mutant, but not wild type L. pneumophila was shown to be blocked by the PI-3 kinase inhibitor, wortmannin [21], implying that invasion of wild type and mutant bacteria occur through distinct pathways. We found the same to be true in endothelial cells (Fig. 1). Here, treatment with 100 nM wortmannin led to a 4.5 fold reduction in invasion of HL051C (P = 0.005) vs. an insignificant reduction of 1.5 fold for Lp01. Therefore, since wortmannin did not substantially block invasion of wild type bacteria, invasion of wild type and the dotI mutant appears to occur by distinct pathways as described previously for macrophages [21,22].

Replication of bacteria within endothelial cells
We next tested whether L. pneumophila replicates inside of endothelial cells. As shown in Fig. 2, we found that wild type Lp01 replicates to high levels in HUVEC with a two to three log growth increase, leading to lysis of host cells and release of bacteria (data not shown). In contrast, the dotI mutant, HL051C, failed to grow and was gradually eliminated. These results are consistent with previous observations in macrophages and suggest that growth is dependent on the dot/icm-dependent, type IV secretion system [17,23]. These observations were confirmed by fluorescent microscopy where large numbers of Lp01 were found intracellularly at later time points vs. the infrequent appearance of HL051C. The latter often appeared to have degraded nucleoids by DAPI staining. Similar growth curves were observed for other endothelial cell types tested including primary human lung microvascular and umbilical artery endothelial cells (data not shown). The ability to grow in the diverse types of endothelial cells surveyed is especially interesting as they are known to have diverse physiological properties [24]. Bacteria also grew within primary pulmonary artery endothelial cells of bovine origin (a species that also develops Legionnaires' disease [25]), suggesting that the ability to infect endothelial cells crosses species boundaries.

Altered phagosome properties
We next examined whether L. pneumophila inhibits fusion of bacterial phagosomes with lysosomes, as observed previously during infection of macrophages [5]. Using an immunofluorescence staining technique, we assessed the colocalization of internalized bacteria with a lysosomal marker, LAMP-1. At 1.5 hours after infection, only 6% of Lp01 colocalized with LAMP-1 vs. 57% of the dotI mutant (P,0.001, Fig. 3). Therefore, only the replication-competent Lp01, but not the replicationdefective dotI mutant, efficiently escaped from lysosomal fusion.
We next examined the morphogenesis of L. pneumophila containing vacuoles by transmission electron microscopy. Within 6 hours of infection by Lp01, most bacterial phagosomes were surrounded by ribosome studded endoplasmic reticulum (RER) (Fig. 4A,B), as observed previously in infected macrophages [26] and amoeba [27]. At 24 hours after infection, organisms had multiplied in vacuoles to largely fill the endothelial cell (Fig. 4C). In contrast, the dotI mutant was always contained in smooth vacuoles, usually as single bacteria, and without associated RER or ribosomes at all time points (Fig. 4D, E). Furthermore, mutant bacteria often appeared degraded and were occasionally found in vacuoles containing membranous debris (Fig. 4F). Both mutant and wild type organisms were found in membrane-bound phagosomes, never freely in the cytoplasm, although disruption of the vacuolar membrane was noted late in the infection cycle of wild type organisms. Taken together, intracellular parasitism of endothelial cells appears similar to observations in macrophages.

Discussion
Here, we for the first time identify the ability of Legionella pneumophila to invade and replicate inside diverse types of primary human endothelial cells, including endothelial cells obtained from the lung. We believe that this property may have special relevance for the pathogenesis of Legionnaires' disease. Previous studies have focused on the organism's ability to grow within macrophages and alveolar epithelial cells. However, the capillary network, formed by endothelial cells, also makes up a critical and potentially vulnerable part of the lung.
The biology of endothelial infection is consistent with what has been described for infection of macrophages, alveolar epithelial cells, and protozoa. The organism inhibits phagolysosome fusion and grows intracellularly in a process dependent on the dot/icm type IV secretion machinery. However, contact-dependent cytotoxicity observed during infection of macrophages and red blood cells [19,28] does not occur (i.e., as determined by LDH release assay [19] after pelleting bacteria onto an endothelial monolayer for 5 minutes at 800 g, data not shown). This may reflect a greater resistance to osmotic lysis, differences in the interaction with the bacterial type IV secretion machinery, or decreased invasion efficiency. Nevertheless, endothelial cell death still occurs after intracellular multiplication of bacteria. It is possible that endothelial cell infection and death in vivo leads to damage of pulmonary capillaries and allows bacteria access to the  Endothelial damage may also in part account for pathologies such as hemorrhage that have been observed during Legionnaires' disease [29,30] or adverse sequelae such as pulmonary fibrosis [31]. Here, the lung forms scar tissue rather than re-establishing surfaces capable of air exchange, leading to permanent disability after bacteriological cure [31]. Interestingly, patients with prosthetic heart valves appear to be at particular risk of developing endocarditis from Legionella [32][33][34]. Possibly, this relates to a susceptibility of endothelial cells overlying prosthetic material to infection with this organism.
Clearly, the extent and consequences of endothelial infection by L. pneumophila need to be examined using in vivo models. However, the vascular endothelium should now be considered a potential participant in the pathogenesis of Legionnaires' disease.

Bacterial strains and infections
Legionella pneumophila strains were described previously [17]. Endothelial cells were purchased from Cambrex and passaged in EGM2-MV medium from the same manufacturer. To perform infections, bacteria were grow in AYE medium to early stationary phase, resuspended in EGM-2MV, and used to infect endothelial cells grown on collagen I (Vitrogen, Cohesion, Inc.) treated twenty-four well dishes.

Assessment of bacterial invasion
After infection at a bacterial to endothelial cell ratio of approximately 300:1, extracellular bacteria were killed by a 45 minute treatment with 100 mg/ml of gentamicin, and surviving intracellular bacteria enumerated after eukaryotic cell lysis with 0.1% saponin. Statistical comparisons were performed using JMP IN Vs. 5.1.2 (SAS Institute, Cary, NC) by the Wilcoxon Ranked Sum Test for continuous and Fisher's Exact Test for categorical variables with P,0.05 considered significant.

Assessment of bacterial replication
At multiple time points after infection, both extracellular and intracellular bacteria were enumerated by sampling extracellular medium and then intracellular compartments after saponin lysis. For bacterial replication experiments only, the infectious inoculum was harvested from aBCYE plates rather than from early stationary phase liquid cultures used for all other assays.

Endocytic trafficking assay
To assess fusion of L. pneumophila containing phagosomes with lysosomes, endothelial cells were plated on collagen I-treated, German glass coverslips (Bellco Glass) and infected at a ratio of approximately 300 bacteria per endothelial cell. Extracellular bacteria were stained prior to permeabilization with a rabbit anti-Legionella primary antibody (gift of Ralph Isberg, Tufts Medical School) followed by a Alexa-594-labeled goat anti-rabbit secondary antibody. After permeabilization with 0.2% TritonH-X 100 for 15 minutes, lysosomal compartments were stained with anti-LAMP-1 mouse monoclonal antibody (Santa Cruz Biotechnology, clone H5G11) [35] followed by Alexa-488-labeled goat anti-mouse secondary antibody (Molecular Probes). Bacterial nucleoids were then stained with 49,6-diamidino-2-phenylindole (DAPI). Intracellular bacteria were identified by the absence of extracellular antibody staining and positive staining with DAPI. Colocalization of intracellular bacteria (DAPI + , Alexa-594 -) and LAMP-1 (Alexa-488 + ) was assessed and quantified using a Nikon Eclipse E300 epifluorescence microscope and images capture using a Retiga 2000RV digital camera (QImaging, Surrey, British Columbia) and IPLab software 3.9.4r5 (BD Biosciences, Rockville, MD)  In this image, RER association is still evident (left side of vacuole), however, at this late stage in infection, vacuolar membrane appears partially disrupted. In contrast, dotI mutants do not associate with RER and are surrounded by a simple, smooth membrane, at times also containing membranous debris, e.g., (D) 6 hours and (E) 24 hours. Intact bacteria were far less numerous, while electron dense structures suggestive of partially degraded bacteria were often present, (F) 6 hours. Magnifications (in figure): A, 654,000; B, 654,000; C 619,000; D 682,000; E 6119,000; F 641,000. doi:10.1371/journal.pone.0002012.g004 using an inverted BEEMH capsule (Beem, Inc., Westchester, PA), and sectioned en face. Sections were stained with lead citrate and viewed with the digital image acquisition system on the Jeol (Tokyo, Japan) MEM-1011, transmission electron microscope.