Abstract
Legionella pneumophila is the most common cause of the severe respiratory infection known as Legionnaires’ disease. L. pneumophila is typically a symbiont of free-living amoeba, and our understanding of the bacterial factors that determine human pathogenicity is limited. Here we carried out a population genomic study of 900 L. pneumophila isolates from human clinical and environmental samples to examine their genetic diversity, global distribution and the basis for human pathogenicity. We found that although some clones are more commonly associated with clinical infections, the capacity for human disease is representative of the breadth of species diversity. To investigate the bacterial genetic basis for human disease potential, we carried out a genome-wide association study that identified a single gene (lag-1), to be most strongly associated with clinical isolates. Molecular evolutionary analysis showed that lag-1, which encodes an O-acetyltransferase responsible for lipopolysaccharide modification, has been distributed horizontally across all major phylogenetic clades of L. pneumophila by frequent recent recombination events. Functional analysis revealed a correlation between the presence of a functional lag-1 gene and resistance to killing in human serum and bovine broncho-alveolar lavage. In addition, L. pneumophila strains that express lag-1 escaped complement-mediated phagocytosis by neutrophils. Importantly, we discovered that the expression of lag-1 confers the capacity to evade complement-mediated killing by inhibiting deposition of classical pathway molecules on the bacterial surface. In summary, our combined population and functional analyses identified L. pneumophila genetic traits linked to human disease and revealed the molecular basis for resistance to complement-mediated killing, a previously elusive trait of direct relevance to human disease pathogenicity.
Significance Legionella pneumophila is an environmental bacterium associated with a severe pneumonia known as Legionnaires’ disease. A small number of L. pneumophila clones are responsible for a large proportion of human infections suggesting they have enhanced pathogenicity. Here, we employed a large-scale population analysis to investigate the evolution of human pathogenicity and identified a single gene (lag-1) that was more frequently found in clinical isolates. Functional analysis revealed that the lag-1-encoded O-acetyltransferase, involved in modification of lipopolysaccharide, conferred resistance to the classical pathway of complement in human serum. These findings solve a long-standing mystery in the field regarding L. pneumophila resistance to serum killing, revealing a novel mechanism by which L. pneumophila may avoid immune defences during infection.
Introduction
Legionella pneumophila is a γ-proteobacterial species that parasitises free-living amoeba in freshwater environments (Hilbi et al., 2011). L. pneumophila hijacks the phagocytic process in amoebae and human alveolar macrophages by subverting host cellular mechanisms to promote intracellular replication (Strassmann and Shu, 2017). Legionella infections are a global public health concern presenting as either a severe pneumonia known as Legionnaires’ disease or Pontiac fever, a self-limiting flu-like syndrome (Herwaldt and Marra, 2018; Joseph et al., 2010; Lam et al., 2011; Wolter et al., 2016). Importantly, recent surveillance studies have indicated a steady increase of legionellosis incidences globally (Beauté and Network, 2017; Parr et al., 2015).
L. pneumophila infection in humans is considered to be the result of accidental environmental exposure and the selection for pathogenic traits among L. pneumophila is likely to be driven by co-selective pressures that exist in its natural habitat (O’Connor et al., 2011). The pivotal mechanism required for intracellular replication is the type IV secretion system (T4SS) that is conserved across all known members of the genus Legionella (Burstein et al., 2016). A very large repertoire of effector proteins in different combinations are encoded by Legionella species and can be secreted by this system to mediate critical host-pathogen interactions (Gomez-Valero et al., 2019). In addition, the ability to infect eukaryotic cells has evolved independently many times (Gomez-Valero et al., 2019). Despite this, less than half of all currently described Legionella species have been reported to cause human disease (Newton et al., 2010). Furthermore, there is an over-representation of a single serogroup (Sg-1) of L. pneumophila in human infections, which is responsible for more than 85% of all reported cases of legionellosis (Yu et al., 2002). Sg-1 strains can be further subdivided phenotypically using monoclonal antibodies (mAbs) that recognise various components of the lipopolysaccharide (LPS). The most prevalent mAb subtype in human infections is associated with an LPS O-acetyltransferase enzyme encoded by the lag-1 gene (lpg0777), which confers an LPS epitope recognised by the mAb 3/1 from the Dresden mAb panel (Ditommaso et al., 2014; Kozak-Muiznieks et al., 2014).
Recently, it was shown that a very limited number (n=5) of L. pneumophila sequence types (ST)s are responsible for almost half of all human infections and that these clones have undergone very recent emergence and expansion (David et al., 2016). In addition, it was shown that recombination between the dominant STs has led to sharing of alleles that may be beneficial for human pathogenicity (David et al., 2016). However, the genetic basis for the enhanced human pathogenic potential of these STs is unknown. Early studies identified the capacity for some L. pneumophila strains to resist killing in human serum, a phenotype that may correlate with increased virulence (Plouffe et al., 1985). However, the molecular mechanism of strain-dependent serum resistance by L. pneumophila has remained elusive for over 30 years.
Here, we have carried out a population genomic analysis of a diverse dataset of 900 whole genome sequences of L. pneumophila isolates from both clinical and environmental sources to investigate the diversity of genotypes associated with human disease. Further, we employ genome-wide association analyses to identify genetic traits associated with human pathogenic strains revealing lag-1 to be the most strongly associated determinant of human pathogenic potential in L. pneumophila. We demonstrate that acquisition of lag-1 has occurred widely across the species by recent horizontal gene transfer and recombination, leading to strains that have enhanced resistance to serum killing and neutrophil phagocytosis. Importantly, functional analysis reveals that lag-1 confers the capacity to evade human complement-mediated killing, suggesting a key role in the early stages of pathogenesis.
Results and Discussion
The potential for human infection is distributed across the L. pneumophila species phylogeny
In order to examine the diversity of L. pneumophila associated with human clinical infection in comparison to those from environmental sources, we carried out whole genome sequencing of 400 clinical and environmental L. pneumophila subsp. pneumophila isolates from an archived collection (1984 to 2015) of Legionella species held at the Scottish Haemophilus, Legionella, Meningococcus & Pneumococcus Reference Laboratory, Scottish Microbiology Reference Laboratory, Glasgow (SHLMPRL) as detailed in the SI Methods section. The dataset included 166 clinical isolates of infected patients, primarily from sputum and bronchoalveolar lavages (BAL) and 239 environmental isolates of L. pneumophila subsp. pneumophila from water sources such as cooling towers and plumbing systems sent to the reference laboratory for routine testing and surveillance. To place the SHLMPRL isolates into context of the known global diversity of L. pneumophila subsp. pneumophila isolates, we included 500 assembled whole genomes that were available in the public database (NCBI Genbank). We constructed a Maximum-Likelihood phylogenetic tree from 139,142 core genome single nucleotide polymorphisms (SNP) which indicates the segregation of the L. pneumophila subsp. pneumophila population into seven major clades (Fig. 1) each supported by a minimum of one and a maximum of two sub-clusters defined by Bayesian Analysis of Population Structure (BAPS) analysis (Supplemental Fig. 1). Isolates from the SHLMPRL collection were distributed among all major clades indicating that the isolates are representative of the global species diversity (Fig. 1). Although recombination has played an important role in diversification of the species, complicating the accurate reconstruction of the phylogeny (David et al., 2017), the major phylogenetic clusters are consistent with those identified in previous population studies employing fewer isolates (Qin et al., 2016; Underwood et al., 2013). Recombination is more likely between phylogenetically related (and genetically similar) isolates which may amplify the phylogenetic signal that defines these major groups (Skippington and Ragan, 2012). Of the 166 clinical isolates, 80 belonged to the five most common sequence types (ST) implicated in human infections (ST1, 23, 36/37, 47 and 62). The remaining 86 clinical isolates were from STs distributed across all seven major phylogenetic clusters (Fig. 1). These data indicate that although only 5 STs are responsible for almost half of all infections, the other half of human infection potential comes from diverse genetic backgrounds distributed across the species. Of note, five clusters from distinct outbreaks of Legionnaires’ disease in Scotland between 1985 and 2012 originated from different clades (Fig. 1). Similarly, the 239 environmental isolates are also distributed across all major phylogenetic groups (Fig. 1). Notably, 34% (81 of 239) of environmental isolates belong to one of the five clinically dominant STs suggesting that at least a third of all environmental L. pneumophila in Scotland have human pathogenic potential. Overall, although some STs appear to have higher human pathogenicity, our findings support the understanding that the capacity for causing human disease is widely distributed across diverse L. pneumophila genetic backgrounds.
A maximum-likelihood phylogeny of 900 isolates based on 139,142 core genome SNPs divides the subspecies into seven major clades that are also supported by BAPS clustering. Isolates linked to five major outbreaks that have occurred in Scotland, UK between 1985 and 2012 originated from different lineages are indicated. In addition, isolates associated with sporadic cases of human infection have also emerged from diverse genetic backgrounds across all major clades. For reference, the position of five major disease associated clones (ST1, 23, 36/37, 47 and 62) and well-characterised reference genomes are indicated.
Genome-wide association analysis of L. pneumophila reveals genetic traits associated with human pathogenicity
The factors that contribute to the enhanced human pathogenic capacity of some L. pneumophila clones are unknown. Our strategy, to address this gap in understanding, involved sequencing of large numbers of genetically diverse environmental isolates, affording for the first time, a large-scale genome-wide association study (GWAS) of clinical vs environmental isolates to explore the bacterial genetic basis for human clinical disease. Initially, we performed systematic subsampling on our dataset to reduce the number of closely related isolates originating from the same type of source, either clinical or environmental, while retaining genetic diversity. This step removes over-represented endemic or epidemic clonal lineages in the dataset that arise from opportunistic, convenience sampling and also represents an additional control for a stratified bacterial population structure. From this reduced dataset (n=452), we used the programme SEER to identify a total of 1737 k-mers that were enriched (significantly associated, p < 0.05) among clinical isolates in comparison to environmental isolates. Mapping to the L. pneumophila Philadelphia 1 reference genome (Accession number: AE017354) revealed that 39% (n=673) of the k-mers aligned to a region of the genome spanning between loci lpg0748 and lpg078l representing an 18 kb cluster of genes involved in LPS biosynthesis and modification (Fig. 2) (Petzold et al., 2013). A total of 22 genes in this cluster each had at least one significantly enriched k-mer that mapped to it: namely, lpg0751, lpg0752, lpg0755, lpg0758, lpg0759, lpg0760, lpg0761, lpg0762, lpg0766, lpg0767, lpg0768, lpg0769, lpg0771, lpg0772, lpg0773, lpg0774, lpg0775, lpg0777 (lag-1), lpg0778, lpg0779, lpg0780 and lpg0781 (Fig. 2, Supplemental Table 2). Two additional conservative SEER analyses using more stringently subsampled datasets resulted in a smaller number of significantly enriched k-mers (205 and 61, respectively), that converge on a single gene within the LPS cluster (lpg0777, lag-1) that encodes an O-acetyltransferase involved in modification of the O-antigen of the L. pneumophila Sg-1 LPS (Supplemental Fig. S2). To corroborate the initial findings of the k-mer based SEER approach, we employed a different GWAS method (SCOARY) that examined the distribution of orthologous genes using the pan-genome pipeline ROARY (Brynildsrud et al., 2016; Page et al., 2015). Consistent with SEER, this approach also indicated that the LPS biosynthesis genes were enriched among clinical isolates with lag-1 exhibiting the strongest statistical support over other Sg-1-associated LPS genes (Supplemental Table S2). The corrected (Benjamini-Hochberg) p-value for lag-1 is several orders of magnitude lower (9.74E-11) than other Sg-1 LPS genes that were above the significance threshold (lpg0779: 2.35E-05, lpg0780: 2.35E-05, lpl0815/lpg0774: 3.84E-05 and lpg0767: 6.92E-05). As mentioned, lag-1 has previously been reported to be prevalent among clinical isolates (Edelstein and Edelstein, 1993; Lück et al., 2001; Lück et al., 2002; Lüneberg et al., 1998; N. Whitfield and S. Swanson, 2006). However, our large pangenome-wide analysis of the L. pneumophila species indicates that of all 11198 accessory genes, lag-1 has the strongest association with clinical isolates, suggesting a pivotal role in human disease.
Manhattan plot showing the genomic position of k-mers significantly associated with clinical isolates. The LPS biosynthesis and modification locus (822,150bp–855,010bp: lpg0751-lpg0781) in the Philadelphia 1 reference genome (AE017354) is represented and genes with significant k-mer associations coloured in red, with lag-1 in blue.
Sg-1 LPS genes were found in high frequency throughout the population, and in each major lineage indicating species-wide gene transfer. The Sg-1 LPS cluster is also found in other subspecies of L. pneumophila that can infect humans, such as subspecies fraseri and pascullei but has not been reported in other species of Legionella (George et al., 2016; Kozak-Muiznieks et al., 2016) Here, we observed that lag-1 can be associated with different combinations of Sg-1 LPS genes and still express the expected lag-1 phenotype represented by the mAb 3/1-specific epitope (Dresden mAb scheme) (Supplemental Table S1) (Helbig et al., 2002; Helbig et al., 1995). However, genetic instability and phase variation has been reported to affect the mAb phenotype between closely-related strains (Amemura-Maekawa et al., 2012; Bernander et al., 2003). One mechanism of phase variation is the excision of a 30 kb genetic element that results in a change in mAb specificity manifested by a loss of virulence in guinea pigs and loss of resistance to complement (Kooistra et al., 2002; Luneberg et al., 2001).
Recombination has mediated the dissemination of three dominant lag-1 alleles across the L. pneumophila species
It has been reported since the 1980s that isolates expressing the epitope recognized by the mAb 3/1 are more frequently associated with isolates from community-acquired and travel-associated infections (Harrison et al., 2007; Helbig et al., 2002; Kozak et al., 2009). However, the pathogenic basis for this association remains a mystery. In order to further investigate the role of the clinically-associated gene lag-1, we examined its diversity and distribution across the 900 L. pneumophila genomes employed in the current study. In total, three major allelic variants of lag-1 that had been previously identified to be representative of reference strains, Philadelphia, Arizona, and Corby, respectively, were identified (Kozak et al., 2009). In our data set, variant 1 (Philadelphia), was present in 195 (22%), variant 2 (Arizona) in 195 (22%) and variant 3 (Corby) in 178 (20%) of the 900 isolates examined (Fig. 3A). Each variant was distributed across the phylogeny, with variant 1 found in 3 major clades (1, 4 and 7) and variants 2 and 3 identified among isolates of all 7 major clades (Fig. 3A). Of note, clade 2 has the lowest frequency (12%) of isolates encoding the lag-1 gene and is also characterized by an under-representation of clinical isolates (31%). In addition to the three major lag-1 alleles, we identified a relatively small number (n=40) of derived minor allelic variants that differ by <1% nucleotide identity from any of the 3 major variants (Supplemental Table S3). Of these, only 10 are predicted to encode for full-length proteins suggesting most are likely to be non-functional pseudogenes. The limited number of allelic variants of lag-1 and their broad distribution across the species phylogeny indicates frequent horizontal dissemination of recently acquired lag-1 alleles driven by a strong selection pressure for lag-1 function. The closest homolog of lag-1 in the NCBI non-redundant protein database shares only 45% amino acid sequence identity and encodes a putative acyltransferase present in an environmental species of the genus Pseudomonas. The selective advantage for L. pneumophila to maintain lag-1 in the environment is still unclear. It has been proposed that the increased hydrophobicity of LPS when acetylated by O-acetyltransferase may enhance L. pneumophila survival in amoebae vacuolar compartments (Fernandez-Moreira et al., 2006).
A) Core-genome SNP-based phylogenetic tree indicating the distribution of the 3 major allelic variants of lag-1 across Sg-1 strains. Variant 1 (Philadelphia, Red), Variant 2 (Arizona, Green) and Variant 3 (Corby, Blue). Each major clade is associated with at least two different lag-1 alleles. Minor alleles that are within >99% nucleotide sequence identity to a major variant are shown with lighter colours. Isolates not encoding lag-1 are coloured grey. B) A neighbour-net phylogenetic network indicating recombination of Sg-1 genes including lag-1. The network was drawn using uncorrected P-distances with the equal angle method in Splitstree from a concatenated alignment of 10 conserved Sg-1 genes that were significantly associated with clinical isolates (9405 positions). Orthologs of 10 LPS cluster genes were extracted from 16 phylogenetically representative isolates coding 3 different variants of lag-1, 7 isolates missing lag-1 and 9 reference genomes (Lens, HL06041035, Philadelphia-1 [ST36], Paris [ST1], Corby, Alcoy, Pontiac [ST62], Lorraine [ST47], ST23). C) Presence of lag-1 correlates with resistance to killing in human plasma. L. pneumophila isolates containing the allelic Variant 1, 2, and 3 of lag-1 (depicted in red, green, and blue, respectively) or do not encode lag-1 (depicted in grey) were incubated with human plasma (coloured bars and dots) or heat inactivated plasma (open bars and dots) for 1 hour at 37°C. Each coloured or open dot per bar represents the average count of an individual plasma donor.
Horizontal transfer of the 18 kb LPS biosynthesis gene cluster between Philadelphia-1 (ST36) and Paris (ST1) has been reported previously (Cazalet et al., 2008). To investigate the potential role of recombination in the distribution of the LPS genes including lag-1, we carried out a split network analysis based on the concatenated alignment of 10 LPS biosynthesis genes that were present in 23 representative isolates from across the phylogeny encoding different lag-1 variants (Fig. 3B). This analysis revealed extensive reticulation consistent with recombination across the gene cluster and identified horizontal transfer of LPS genes between three major clinical sequence types, ST23, ST47 and ST62, respectively (Fig. 3B). This network analysis revealed that even highly similar LPS biosynthesis gene clusters, such as those found in isolates 4454, Corby and Alcoy or 4616 and 4718, encode different variants of lag-1, consistent with recent gene conversion of lag-1. We also identified three examples of closely related genomes (average nucleotide identity across the genome of > 99.8%) that exhibited a signature of homologous recombination affecting lag-1 and the surrounding genomic region (Supplemental Fig. S3). These findings expand on a previously proposed mechanism of lag-1 gene deletion by Kozak and colleagues (Kozak et al., 2009) and demonstrate that recombination can restore a disrupted or missing lag-1 gene (Supplemental Fig. S3A and B) or replace one functional lag-1 variant with another (Supplemental Fig. S3C).
A comparison of re-assortment rates across major phylogenetic lineages showed no significant differences between LPS and non-LPS genes across the L. pneumophila genome (p=0.6275), suggesting that recombination is active across the genome (Supplemental Fig. S4 and Supplemental Table S4). Taken together, these findings highlight the wide dissemination of lag-1 and other LPS genes by recombination.
lag-1 confers resistance to killing by human plasma and broncho-alveolar lavage
Previous studies have reported an epidemiological correlation between lag-1 or its mAb 3/1-recognised epitope and clinical disease (Edelstein and Edelstein, 1993; Lück et al., 2001; Lück et al., 2002; Lüneberg et al., 1998; N. Whitfield and S. Swanson, 2006) but an understanding of a role in pathogenesis has proved elusive. Of note, an early study described an increased ability of an endemic nosocomial L. pneumophila strain to survive complement killing when compared to an environmental strain collected from the same medical facility (Plouffe et al., 1985). To investigate this unexplained phenotype in the context of lag-1 genotype, we examined resistance to killing in human plasma for the aforementioned 23 L. pneumophila strains selected to represent the diversity of lag-1 genotypes from across the phylogeny. We observed that all strains lacking a lag-1 gene were susceptible to killing in plasma, whereas the majority (9 of 16) of isolates containing lag-1 exhibited enhanced resistance, independent of the lag-1 variant encoded (Fig. 3C).
Within our dataset, we identified a group of epidemiologically-related isolates from a single Scottish healthcare facility typed as ST5 (a single locus variant of ST1) which were predicted to vary with regard to lag-1 functionality. ST5 has ony been found in this location, to date. The earliest isolates obtained from a nosocomial outbreak in 1984/1985 (Macfarlane and Worboys, 2012) contained variant 3 (Corby) of the lag-1 gene and all were found to express the mAb 3/1 epitope (Supplemental Table S1). In contrast, 16 of 19 environmental isolates from the same healthcare facility 12 to 21 years later (1997 to 2006) contained multiple independent mutations in lag-1 predicted to disrupt functionality. Specifically, a nonsense mutation (L48*), an insertion of a transposase, and the acquisition of a deleterious substitution (H28L) (Fig. 4A), each correlated with a lack of reactivity with mAb 3/1 from the Dresden panel classification (Supplemental Table S1). Consistent with our previous findings (Fig. 3C), the presence of a functional LPS O-acetyltransferase in this epidemiological cluster also correlated with resistance to killing in human plasma, whereas isolates with a non-functional lag-1 were susceptible to killing, independent of the type of deactivating mutation (Fig. 4B). Of note, no clinical episodes of disease were identified to be caused by this epidemiological cluster after 1985, and all clinical isolates contained a functional lag-1 gene. The identification of multiple independent mutations associated with loss of lag-1 function suggests a selection pressure that drives the inactivation of the lag-1 gene and LPS O-acetylation in this environment. The trend of lag-1 gene loss or deactivation was not observed among longitudinal ST1 healthcare facility-associated isolates from a cluster sequenced in a recent study by David and colleagues (David et al., 2017). However, a study on starvation of L. pneumophila in ultrapure water showed that in a short-term period, the viable cell numbers of all mAb 3/1-positive strains decreased strongly compared to the other strains suggesting a negative selection for lag-1 function in some water environments (Schrammel et al., 2018). Overall, these data indicate that the presence of a functional lag-1 correlates with enhanced resistance to serum killing.
A) Schematic representation of the natural lag-1 deactivating mutations. In blue - functional lag-1, with dots - lag-1 with His28Leu mutation, with lines - lag-1 truncated by an insertion sequence in aa 251, in grey - lag-1 truncated by a premature stop codon. B) Isolates with lag-1 mutations show increased susceptibility to killing in human plasma. Closely related isolates from the same hospital-associated cluster were incubated with human plasma (coloured and patterned bars and dots) or heat inactivated plasma (open bars and dots) for 1 hour at 37°C. Each dot represent an average count of an individual plasma donor. One-way ANOVA, Tukey’s multiple comparisons test.
To investigate the role of lag-1 in mediating resistance to serum killing, we introduced the lag-1 gene encoded on an expression plasmid (pMMB207) into two L. pneumophila mAb 3/1 epitope-negative strains. We introduced the three lag-1 variants into the strain 4681 that contains a non-functional lag-1 gene due to the insertion of a transposase, and the lag-1 variant 1 into the lag-1 negative strain 4312 which contains a 1 bp deletion resulting in a frameshift at position 156. Introduction of the plasmid resulted in lag-1 expression and LPS 8-O-acetylation as confirmed by flow cytometric detection of the mAb 3/1 epitope (Fig. 5A). Strikingly, complementation of the strains with lag-1 resulted in resistance to killing in human serum independently of the lag-1 variant (Fig. 5B). Associated with the resistance to serum killing, we observed a decrease in human C3 deposition at the surface of the lag-1 expressing bacteria when compared to the isogenic lag-1 negative strain, at levels similar to the wild-type lag-1 positive isolate 3656 (Supplemental Fig. S5). Of note, it was previously reported that a wild-type lag-1 positive isolate and a spontaneously derived lag-1 mutant demonstrated equivocal levels of resistance to serum killing (Lück et al., 2001). We speculate that the discordance with our findings could be due to phase-variable expression of LPS (Lüneberg et al., 1998) or differences in transcriptional levels of lag-1 gene that have been previously observed depending on culture conditions (Faucher et al., 2011).
A) Introduction of any lag-1 gene variant leads to mAB 3/1 epitope expression. Detection of mAb 3/1 epitope by flow cytometry in WT and isogenic mutants of L. pneumophila isolates expressing the lag-1 variant 1, 2 or 3. B) lag-1 expression confers serum complement resistance to L. pneumophila strains with non-functional lag-1 gene. Isolates were incubated with human serum or heat inactivated serum for 1 hour at 37°C. Each point represents an average of triplicate CFU counts of a single sera donor. Bars represent mean+SEM. One-way ANOVA, Tukey’s multiple comparisons test, *p<0.05, ****p<0.0001. Comparasion to Empty plasmid isogenic strain represented on the top of the bars. C) mAb 3/1 negative L. pneumophila strains are susceptible to bovine BAL complement killing. Isolates were incubated with concentrated bovine BAL or heat inactivated BAL for 1 hour at 37°C. Each point represents a replicate. Bars represent mean+SEM. One-way ANOVA, Tukey’s multiple comparisons test, *p<0.05, ****p<0.0001. Comparasion to Empty plasmid isogenic strain represented on the top of the bars. D) lag-1 confers resistance to human neutrophil phagocytosis. WT lag-1 positive and negative strains expressing DsRed fluorescent protein were pre-incubated with 10% human serum 15 min prior incubation with human neutrophils for 30 min. Phagocytosis was evaluated by measuring neutrophils DsRed fluoresce by flow cytometry. Data representative of two independent experiments. L. pneumophila strains expressing lag-1 variant 1, 2 or 3 are represented in red, green or blue, respectively. Strains that do not express a functional lag-1 gene are represented in gray. Negative controls are represented in black. Bars represent mean+SEM. One-way ANOVA, Dunnett’s multiple comparisons test to non-infected control, *p<0.05, ****p<0.0001. HI - heat inactivated, BAL – bronchoalveolar lavage, NI – non-infected.
Classical pathway complement activity exists in healthy human broncho-alveolar lavage (BAL), despite the relatively low concentration of some complement proteins (Watford et al., 2000), and exposure to aerosolized LPS leads to a rapid increase of the level of these proteins in the lung of human volunteers (Bolger et al., 2007). The importance of this innate immune mechanism in lung health is supported by the observation that many patients with deficiencies in complement proteins or complement receptors have recurrent respiratory infections (Figueroa and Densen, 1991). Here, we examined the role of lag-1 in resistance to killing in bovine BAL. Similar to the impact of human serum, incubation of lag-1 negative L. pneumophila with concentrated bovine BAL resulted in significant bacterial killing compared with heat inactivated fluid (Fig. 5C), and complementation with each lag-1 variant conferred resistance to killing (Fig. 5C). Taken together, our data reveal a key role for lag-1 in conferring resistance to complement killing in both serum and bronchoalveolar fluid.
Presence of lag-1 correlates with resistance to neutrophil phagocytosis
The role of the complement system in the lung immune defences extends beyond the proteolytic cascade associated with bacterial lysis, as opsonization with complement C3b protein induces phagocytosis of the opsonized targets by neutrophils and macrophages (Heesterbeek et al., 2018). Accordingly, to test the effect of lag-1 expression on phagocytosis of L. pneumophila by neutrophils, human blood purified neutrophils were incubated with L. pneumophila fluorescent strains in the presence of non-immune human serum (Fig. 5D).
We selected L. pneumophila strains from different phylogenetic groups that express the three lag-1 variants, in addition to three lag-1 negative strains. We observed a considerably higher mean fluorescence intensity (MFI) in neutrophils infected with lag-1 negative strains compared to lag-1 positive strains, indicating reduced adherence or phagocytosis that correlated with lag-1 expression (Fig. 5 D; Fig. S6). Therefore, the ability of lag-1 positive L. pneumophila strains to escape complement deposition correlates with the capacity to escape internalization and killing by human phagocytes. These data are consistent with early L. pneumophila studies, where L. pneumophila strain Philadelphia 1 (lag-1 positive) was demonstrated to be resistant to complement and neutrophil killing in the absence of specific antibodies (Horwitz and Silverstein, 1981; Verbrugh et al., 1985). There is increasing evidence for the pivotal role of neutrophils in the resolution of L. pneumophila lung and either neutrophil depletion or blockage of recruitment to infected lungs renders mice susceptible to L. pneumophila (LeibundGut-Landmann et al., 2011; Mascarenhas et al., 2015; Tateda et al., 2001a; Tateda et al., 2001b; Ziltener et al., 2016). Taken together, our data indicate that L. pneumophila lag-1 mediated inhibition of complement deposition, disrupts the innate immune response via inhibition of complement-mediated lysis and blocking recognition by phagocytes and subsequent internalization.
Lag-1 confers resistance to killing by the classical complement pathway
We next investigated if a specific complement pathway was responsible for the serum killing of the Sg-1 mAb 3/1-negative L. pneumophila. As a similar inhibitory effect of lag-1 on bacterial killing was observed in both serum and plasma, we employed serum for these experiments to facilitate the use of commercially available depleted serum samples. EDTA can be used to inhibit the classical, lectin and alternative pathways via chelation of both Ca2+ and Mg2+, whereas chelation with EGTA/Mg2+ inhibits the Ca2+-dependent classical and lectin pathways only, leaving the alternative pathway unaffected (Fig. 6A). Killing assays in human serum in the presence of either EDTA or EGTA/Mg2+ resulted in complete abrogation of lag-1-negative strains susceptibility to complement, and no difference in cell viability compared to incubation in heat inactivated serum, suggesting the alternative complement pathway is insufficient to kill L. pneumophila (Fig. 6B). To distinguish the role of the classical from the lectin pathway, killing assays were performed in the presence of mannose that competes for the association of mannose binding lectin (MBL) with the bacterial surface and blocks the lectin complement pathway (Moller-Kristensen et al., 2006). Using this approach, there was no effect on bacteria viability, indicating that the lectin pathway is not required for L. pneumophila Sg-1 killing, consistent with the previous report that MBL polymorphisms are not associated with a higher risk for legionellosis (Herpers et al., 2009). The exclusion of the lectin pathway suggests the essential role of the classical pathway in the complement killing of L. pneumophila. To confirm this finding we depleted serum of C1q, required for the classical pathway, resulting in loss of serum-mediated killing (Fig. 6B). It is noteworthy that we also observed an increase in bacterial viability in the presence of Factor B-depleted serum (Fig. 6B), implying a possible role for the alternative pathway in amplification of classical pathway activation as previous described for Streptococcus pneumoniae (Brown et al., 2002). Previously, purified L. pneumophila LPS was reported to activate both classical and alternate pathways, primarily through the activation of the classical pathway dependent on natural IgM antibodies (Mintz et al., 1992). In another study, complement C1q protein was demonstrated to bind to the major outer membrane protein (MOMP) of L. pneumophila, activating complement in an antibody-independent way (Mintz et al., 1995).
A) Simplified schematic representation of the complement pathways and the inibitors used in this study. B) Inhibiton of both classical and alternative pathways confers L. pneumophila resistance to complement killing in human serum. 3956, 4451, 4681::Empty plasmid isolates were incubated with 90% heat inactivated serum, serum, serum with EDTA, EGTA/Mg2+ or mannose, Factor B- or C1q-depleted serum for a 1 hour at 37°C. Statistical analysis when compared to serum on top of the bars. One-way ANOVA, Tukey’s multiple comparisons test, **p<0.01 ****p < 0.0001
Concluding comments
The steady global increase in L. pneumophila infections is worrisome and studies that explore the evolutionary basis of the increased pathogenicity of some clones can provide insights into the nature of the public health threat posed. Here we have employed a combined large-scale population level study of clinical and environmental isolates, and a functional ex vivo analysis to reveal the basis for serum resistance in L. pneumophila. The identification of a specific modification of LPS that is required for resistance to classical complement mediated killing and inhibits phagocytosis by neutrophils could inform the design of novel therapeutic approaches that subvert the capacity of L. pneumophila to resist innate immune killing during severe human infection.
Materials and Methods
Bacterial strains and plasmids
Bacterial strains and plasmids employed in this study are described in SI Appendix. L. pneumophila cells were cultured in yeast extract broth (YEB) with constant shaking at 37 °C in the presence or absence of chloramphenicol (SI Appendix, SI Materials and Methods).
Gene cloning and complementation
The three lag-1 gene variants were cloned into the vector pMMB207 and transformed by electrophoresis into the L. pneumophila isolate 4681 or 4312. pSW001 plasmid, contitutevely expressing DsRed was transformed into L. pneumophila as described in SI Materials and Methods (Chen et al., 2006). Expression of the mAb 3/1 epitope and DsRed fluorescence was confirmed by flow cytometry (SI Appendix, SI Materials and Methods).
Whole genome sequence analysis
Sequence data are available on the European Nucleotide Archive, accession number PRJEB31628. For detailed methods please refer to SI Appendix, SI Materials and Methods.
Complement killing assays
L. pneumophila was incubated for 1 hour at 37°C in 90% normal human serum/plasma, heat inactivated serum/plasma, bovine concentrated BAL or heat inactivated BAL. Inhibition of complement pathways was carried out by adding 12.5 mM EDTA (Sigma Aldrich), 12.5 mM EGTA/Mg2+ (Sigma Aldrich) or 100 mM mannose (Acros Organics) to normal human serum or by using C1q- and Factor B-depleted serum (Pathway Diagnostics). (SI Appendix, SI Materials and Methods).
Neutrophil assay
L. pneumophila transformed with pSW001 plasmid was incubated with 10% normal human serum for 15 min prior to incubation with purified human blood neutrophils for 30 min at 37°C, 600 rpm. Bacterial internalization was evaluated by flow cytometry (SI Appendix, SI Materials and Methods).
Acknowledgements
This work was supported by funding to JRF from the Chief Scientists Office Scotland (Grant No. ETM/421) and the Wellcome Trust Collaborative award (Grant No. 201531/Z/16/Z). Computing resources were supported in part by MRC CLIMB (Grant Number: MR/L015080/1). We are grateful to Carmen Buchrieser for providing plasmid pMMB207. Thanks also to Dr Kenneth Baillie and Dr Sara Clohisey for organising the human blood donation study and the volunteers from the Roslin Institute who provided blood samples for the serum and plasma killing assays.