Legionella pneumophila bacteria were maintained on solid agar plates BCYE (buffered charcoal yeast extract). They were inoculated and grown in AYE rich media for 24 hours prior to infection [25], [26]. Dictyostelium discoideum wild type strain AX4 was maintained axenically, at 22°C, in shaken liquid culture (HL-5 media) or on solid SM agar plates in association with Klebsiella aerogenes as a food source [27].

Acanthamoeba castellanii (ATCC 30234) was maintained as adherent cells in PYG media in 75 cm² tissue culture flasks at 25°C [28]. Human U937 cells (ATCC CRL-1593.2) were maintained as non-adherent cells in RPMI 1640 media supplemented with 2 mM L-glutamine, 10 mM HEPES, 1 mM sodium pyruvate, 4.5 g/L glucose, 10% fetal bovine serum and 50 µg/ml gentamicin, in 75-cm² tissue culture flasks, at 37°C. Macrophage-like cells were induced from U937 cells by adding 10⁻⁸ M phorbol-12-myristate-13-acetate (PMA, Sigma). Cells were allowed to differentiate for 48 hours prior to use in infection experiments [29].

Exponentially growing D. discoideum cells were harvested from HL-5 shaken culture (2∼4×10⁶/ml). The cells were washed twice with phosphate buffered saline (PBS). Aliquots of 10⁷ cells, were suspended in 5 ml modified HL-5 media (without addition of glucose, pH adjusted to 6.9), distributed to 60-mm culture dishes, and incubated at 26°C overnight to allow the cells to attach and adjust to the new media [21]. L. pneumophila cells used for infection were harvested at a post exponential OD₆₀₀>3.0, because the cells at this growth stage have the maximum infectivity [30]. The bacteria were washed twice in PBS buffer to eliminate any soluble extracellular factors and resuspended in PBS at the concentration needed to achieve the desired multiplicity of infection (MOI). Infections were initiated by adding 100 µl of bacterial suspension into equilibrated D. discoideum cultures. In control experiments, green fluorescent protein expressing L. pneumophila cells were used to follow the uptake of the bacteria by the amoebae. At various times, amoebae were harvested and washed repeated to remove bacteria and then examined by fluorescence microscopy. At MOI = 10, >90 percent of the amoebae had at least one L. pneumophila inside after 4 hours, and 100 percent contained L. pneumophila after 12 hours.

The infections of Acanthamoeba castellanii and human U937 cells were conducted in a similar way. Adherent cells were tapped off the bottom of 75-cm² stock culture flasks, washed in PBS, and resuspended in appropriate infection media and allowed to reattach to the bottom of 60-mm culture dishes. The infection of A. castellanii was assayed at 25°C or 37°C with PYG media deprived of glucose or with non-nutrient Ac buffer, as described [28]. The infection of human U937 cells was conducted at 37°C in RPMI 1640 media without antibiotics.

D. discoideum cell viability during L. pneumophila infection was assessed from triplicate samples by diluting cells from the infection media and allowing them to recover on SM agar plates inoculated with Klebsiella aerogenes bacteria as a food source. D. discoideum colonies emanating from single cells (CFU) were scored after four days of incubation at 22°C. Total RNA from infected cells was prepared for each time-point by harvesting the culture from the entire culture dish and dissolving the cell pellet in 1 ml TRIZOL reagent (Invitrogen).

Various types of cellular stress were induced in D. discoideum AX4 cells that were prepared in the same way as for L. pneumophila infection, as described above. The chemicals (Sigma) were added directly into cell culture. Camptothecin and etoposide were applied at 100 µm; staurosporine was used at 1 µm; H₂O₂ was at 1 mM; G418 was used at 100 µg/ml; chloramphenicol was used at 200 µg/ml. All the treatments were carried out at 26°C, the same temperature used for L. pneumophila infection. Total RNA was collected at various times after the addition of the chemicals and examined for cleavage of LSU rRNA by northern blot analyses.

The oligonucleotides used to study mitochondrial rRNAs are listed in (Table 1). The D. discoideum mitochondrial LSU rRNA (GenBank NC_000895) was amplified by reverse transcription polymerase chain reaction (RT-PCR) from a preparation of total RNA. In some experiments, RNA was purified from isolated mitochondria as described [31]. The antisense primer Dd3′-2863 was used to amplify the first strand cDNA, which then served as template for primers Dd5′-144 and Dd3′-2518 to amplify the entire mitochondrial LSU rRNA. The resulting product was digested with ClaI and HaeIII to generate four DNA fragments used as hybridization probes that correspond to segments of the rRNA. The approximate locations of the cleavage sites were determined by northern blot analysis using these four D. discoideum DNA probes and were estimated from the rRNA fragment lengths and the known length of the LSU rRNA. This information was used to design primers to map the cleavage sites precisely using the rapid amplification of cDNA ends (RACE) technique. The A. castellanii mitochondrial LSU rRNA (GenBank NC_001637) probe was amplified from its genomic DNA using primers Ac5′-73 and Ac3′-2679 as described previously. The human mitochondrial LSU rRNA (GenBank NC_001807) probe was amplified from human genomic DNA using primers Hs5′-18 and Hs3′-1545 as described.

Antisense primers Dd3′-2243 and Dd3′-2169 were used as gene specific primers (GSP) to run the 5′ RACE, modified from the manufacturer's protocol (Invitrogen). Briefly, the first strand cDNA was amplified by GSP1 Dd3′-2518 with Superscript III reverse transcriptase (Invitrogen) at 55°C, followed by capping the 5′ end of the cDNA with dCTP using terminal deoxynucleotidyl transferase. An anchor primer with a polyG tail (Invitrogen) and a nested GSP2 Dd3′-2243 were used to PCR amplify the capped cDNA. A universal anchor primer without polyG tail (Invitrogen) and the third GSP3 Dd3′-2169 were used to run nested PCR to confirm the presence of the required cDNA species. The PCR products were TOPO cloned into TA vectors and sequenced. The rRNA primary cleavage sites and the 3′ side secondary cleavage sites were determined by aligning the cDNA sequences next to polyC tails to the original rRNA sequences. After the primary cleavage site was determined, another set of primers GSP1 Dd3′-1391, GSP2 Dd3′-1351 and GSP3 Dd3′-1271 were designed to the 5′ side of the primary cleavage site. The 5′ side secondary cleavage site was determined in a similar way that the other two cleavage sites were resolved.

The secondary structure of D. discoideum mitochondrial LSU rRNA was derived from comparative prediction modeling [32]. The high-resolution ribosome crystal structure of the halobacterium Haloarcula marismortui was used as a reference model to map D. discoideum mitochondrial LSU rRNA structure in three-dimensions, using PDB 1FFK coordinates [33], [34]. RasMol software was used to manipulate and render the three-dimensional structures [35].

In the course of carrying out transcriptional profiling of D. discoideum during infection with L. pneumophila we observed a decrease in the steady-state levels of a number of mRNAs that are related to mitochondrial function (data not shown). Given the reported association of L. pneumophila-containing phagosomes with the host cell mitochondria, we examined this finding more closely. We assessed the steady-state mRNA levels of four mitochondria-encoded genes, each of which code for one subunit of each of the four complexes of the electron transport chain [23]. These four genes; nad5 (NADH dehydrogenase subunit), atp6 (ATP synthase, F₀ subunit), cytB (cytochrome b), and cox3 (cytochrome c oxidase subunit) are transcribed from the mitochondrial genome as part of four distinct polycistronic primary transcripts [36]. After infection with the virulent L. pneumophila strain JR32, we observed two modes of mRNA decline, nad5 and atp6 levels decreased dramatically within the first 4 h of infection, while cytB and cox3 levels declined more gradually throughout the infection (Figure 1A). Exposure of cells to an avirulent L. pneumophila strain (icmT mutant) or Klebsiella aerogenes (a bacterium commonly used as a laboratory food source for D. discoideum) had no detectable effect on transcript levels. In these experiments, cell viability began to decrease rapidly after 8 h, so the decline in nad5 and atp6 mRNAs appears to precede cell killing.

We next tested whether the decrease in mitochondrial transcripts could result from any form of cellular stress by treating cells with several toxins known to kill cells in different ways. The pattern of transcript decrease observed with L. pneumophila infection was closely reproduced by treatment with hydrogen peroxide, but not by treatment with the aminoglycoside G418, or the topoisomerase II inhibitor etoposide (Figure 1B). The oxidative stress induced by hydrogen peroxide is known to lead to mitochondrial damage, so this suggests that L. pneumophila infection leads specifically to a rapid disruption of mitochondrial function by inducing oxidative stress, or conditions that resemble oxidative stress.

D. discoideum mitochondrial LSU rRNA is typically used as a normalization control in northern blots, but when we probed for the LSU rRNA it appeared to be cleaved during L. pneumophila infection. Using the LSU rRNA 5′ end as probe, smaller fragments appeared in RNA samples isolated from infected D. discoideum cultures, and these fragments accumulated as the infection progressed as judged by the increase in band intensities (Figure 2A). A larger fragment appeared to accumulate earlier in the course of infection than the smaller fragments, suggesting a specific and sequential cleavage, with a primary cleavage followed by secondary cleavage events. The cleavage did not occur when cells were exposed to avirulent strains L. pneumophila (Figure 2A; and see below). We also demonstrated by northern blot analyses that cytoplasmic rRNAs remained intact during infection by probing the same RNA samples that contained cleaved mitochondrial LSU rRNA (data not shown). The cleavage of the mitochondrial LSU rRNA also correlated with decreased survival of D. discoideum during infection (Figure 2B). The presence of detectable cleavage at 8 hours post infection, before any detectable loss in cell viability, suggests that LSU rRNA cleavage precedes cell killing. Contrary to the similar effect that virulent L. pneumophila and hydrogen peroxide had on mitochondrial mRNA levels, hydrogen peroxide treatment did not induced LSU rRNA cleavage in cells (Figure 2C).

It has been observed that virulent L. pneumophila infection induces host cell apoptosis in mammalian cells [37], [38], [39]. To begin to test whether the observed cleavage of mitochondrial LSU rRNA is a host cell response to any stress, or caused by L. pneumophila directly, we induced stress in D. discoideum cells by a variety of chemical treatments and assessed mitochondrial rRNA integrity on northern blots. Camptothecin and etoposide are apoptosis inducers that poison eukaryotic DNA topoisomerases I and II, respectively [40], [41]. Staurosporine is a potent protein kinase C inhibitor [42] while the antibiotics G418 and chloramphenicol inhibit eukaryotic cytoplasmic and mitochondrial ribosomal protein synthesis, respectively, and hydrogen peroxide induces oxidative stress in cells. These treatments caused D. discoideum cell death but none of them led to the mitochondrial LSU rRNA cleavage observed during L. pneumophila infection (Table 2). These results suggest that the observed rRNA cleavage is caused by L. pneumophila infection rather than a general host cell stress response.

All of the L. pneumophila pathogenic strains that have been sequenced so far carry type IV Dot/Icm and type II Lsp secretion systems that are believed to be required for pathogenesis [43], [44]. Most of the bacteria's pathogenic functions reported so far require the integrity of one or the other of these secretion systems [14], [16], [17], [45], [46]. We tested which of these two L. pneumophila secretion pathways are required for the rRNA cleavage to take place. The icmT and dotA genes reside within two different gene clusters that encode the core components of the type IV secretion machinery [13], [47], [48]. The lspF and lspG genes are clustered together and encode the core of the type II machinery, while the lspDE genes encode another component of the type II secretion machinery [43]. The type IV mutants, icmT and dotA, as well as the type II mutants lspF, lspG and lspDE were tested for their ability to induce rRNA cleavage. None of these mutants were observed to induce cleavage of the mitochondrial rRNA after mixing them with D. discoideum (Table 3). These results indicate that both the type IV and the type II secretion systems must be functional for D. discoideum mitochondrial LSU rRNA cleavage to occur.

We also tested the pathogenicity of these type IV and type II mutants on D. discoideum. As expected, none of these L. pneumophila mutants killed D. discoideum, as their parental wild type strains did, which corroborates the findings using other amoebae as hosts (Table 3). Since we could have observed some LSU rRNA cleavage without death of the Dictyostelium cells, but did not, a strict correlation exists between the pathogenesis of L. pneumophila and the cleavage of D. discoideum's mitochondrial LSU rRNA.

We attempted to identify possible L. pneumophila enzymes responsible for the rRNA cleavage by using a candidate gene approach. We inspected all predicted ribonuclease genes in the genomic sequence of L. pneumophila and excluded genes predicted to encode ribonucleases with defined functions (10). We found a single predicted T2 ribonuclease with an RNase core domain plus uncharacterized domains that we hypothesized might be involved in determining substrate specificity (GenBank AE017354). A deletion mutant of the gene encoding this protein, provided by the laboratory of Ralph Isberg, was assayed for LSU rRNA cleavage activity upon infection of D. discoideum. Mitochondrial LSU rRNA cleavage was observed, just as in wild type L. pneumophila infection, indicating that this protein is not responsible for the rRNA cleavage activity (data not shown).

We used RACE (Rapid Amplification of cDNA Ends) to map the specific cleavage sites on the rRNA [49]. In order to design the RACE primers we mapped the approximate locations of the cleavage sites by northern blot analyses using segments of RT-PCR-amplified mitochondrial LSU rRNA as probes (Figure 3A and 3B). Two sets of primers were designed for the RACE reactions that bordered the approximate locations of the cleavage sites. We carried out RACE reactions on samples of total RNA collected from cells that had been infected with virulent and avirulent L. pneumophila. Using two different RACE primers two fragments and the full-length copy of the rRNA were easily detectable, but only from cells infected with virulent L. pneumophila (Figure 3C). The RACE products representing sub-fragments of the copied and amplified rRNA were cloned into plasmids and the cleavage sites were then determined by sequencing the inserts (Figure 3C). The sequencing of the RACE products revealed the precise locations of one major and one minor cleavage site on the D. discoideum mitochondrial LSU rRNA (Figure 3A). With the major cleavage site mapped, RACE primers were designed that lay to the 5-prime side of that site and used to amplify the other minor cleavage fragment (Figure 3D).

The northern blot data indicates that the LSU rRNA is cleaved into discrete fragments and the unambiguous ends identified by sequencing each of the RACE products strongly suggests that the fragmentation results from cleavage at discrete sites. The major cleavage site mapped between nucleotides 1533 and 1534 (Figure 3A). A minor cleavage site on the 5′ side of the major site mapped between nucleotides 862 and 863 and another minor cleavage site on the 3′ side of the major site mapped between nucleotides 1878 and 1879. Because the eukaryotic mitochondrial ribosomes are thought to have shared a common ancestor with archaebacterial ribosomes, we mapped the cleavage sites onto the predicted secondary structure of the D. discoideum mitochondrial LSU rRNA based on the halobacteria Haloarcula marismortui rRNA structure [33], [50] (Figure 4A). When mapped onto the corresponding regions of the H. marismortui rRNA high-resolution crystal structure, the major cleavage site is predicted to be on the surface of the ribosome (Figure 4B). This is consistent with the expectation that the initial cleavage site is accessible on the intact ribosome similar to the sarcin/ricin loop, a distinct site that is subject to cleavage by those toxins (Figure 4B and 4C). The minor cleavage site at base 1878 is predicted to be in domain IV of the LSU rRNA (Figure 4A), where essential functional contacts are made between the ribosome's large and small subunits (Figure 4B and 4C). The minor cleavage site at base 862 and is predicted to be deep within the ribosomal structure (Figure 4C).

We tested whether LSU rRNA cleavage occurs in other cells infected with L. pneumophila. Acanthamoeba castellanii and human U937 cells are two widely used models of L. pneumophila infection [28], [29]. We infected these cells using established protocols and assayed for host cell viability and mitochondrial LSU rRNA integrity, but we did not observe any cleavage events during either of these infections (Figure 5). This result with human cells is not surprising since the entire structural domain where the primary cleavage site is located in D. discoideum LSU rRNA has been lost in human LSU rRNA [32], [51]. However, in A. castellanii, the major cleavage site and the minor site at base 1878 are conserved, whereas the minor cleavage site at base 832 is within a loop that is unique in D. discoideum [32].