SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Legionella pneumophila infects alveolar macrophages and protozoa through establishment of an intracellular replication niche. This process is mediated by bacterial effectors translocated into the host cell via the Icm/Dot type IV secretion system. Most of the effectors identified so far are unique to L. pneumophila; however, some of the effectors are homologous to eukaryotic proteins. We performed a distribution analysis of many known L. pneumophila effectors and found that several of them, mostly eukaryotic homologous proteins, are present in different Legionella species. In-depth analysis of LegS2, a L. pneumophila homologue of the highly conserved eukaryotic enzyme sphingosine-1-phosphate lyase (SPL), revealed that it was most likely acquired from a protozoan organism early during Legionella evolution. The LegS2 protein was found to translocate into host cells using a C-terminal translocation domain absent in its eukaryotic homologues. LegS2 was found to complement the sphingosine-sensitive phenotype of a Saccharomyces serevisia SPL-null mutant and this complementation depended on evolutionary conserved residues in the LegS2 catalytic domain. Interestingly, unlike the eukaryotic SPL that localizes to the endoplasmic reticulum, LegS2 was found to be targeted mainly to host cell mitochondria. Collectively, our results demonstrate the remarkable adaptations of a eukaryotic protein to the L. pneumophila pathogenesis system.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Legionella pneumophila, the major causative agent of Legionnaires' disease is a facultative intracellular pathogen that multiplies within and kills human macrophages and free living amoeba (McDade et al., 1977; Rowbotham, 1980). The infection occurs when aerosolized L. pneumophila are inhaled from man-made or natural freshwater reservoirs harbouring the bacteria. Upon internalization of L. pneumophila, the Legionella-containing phagosome actively avoids fusion with lysosomes and recruits mitochondria and rough endoplasmic reticulum (RER) to establish a replication niche (Horwitz and Silverstein, 1980; Horwitz, 1983; Tilney et al., 2001). The formation of this specialized compartment is orchestrated by the Icm/Dot Type IVb secretion system that translocates bacterial protein substrates into the host cell (Segal et al., 1998; Vogel et al., 1998; Segal et al., 2005).

Over the past few years a large number of proteins and protein families have been identified as substrates of the Icm/Dot secretion system (Ninio and Roy, 2007; Altman and Segal, 2008; de Felipe et al., 2008; Habyarimana et al., 2008; Kubori et al., 2008; Pan et al., 2008; Zusman et al., 2008). Some of these proteins were shown to be involved in disruption of host cell trafficking and recruitment of the RER vesicles to the Legionella-containing phagosome (Nagai et al., 2002; Shohdy et al., 2005; Machner and Isberg, 2006; Murata et al., 2006; Liu and Luo, 2007; Pan et al., 2008; Shin et al., 2008), while others either have no assigned function (Luo and Isberg, 2004; Ninio et al., 2005; Zusman et al., 2007; Altman and Segal, 2008; Habyarimana et al., 2008; Zusman et al., 2008) or play a role in other aspects of Legionella virulence (Chen et al., 2004; Laguna et al., 2006; Kubori et al., 2008). Most of the studied effectors have been found to be dispensable for intracellular growth, implying that some of them might have functional homologues or their role during infection of the particular host examined is dispensable (Ninio and Roy, 2007).

Most of the Icm/Dot effectors are unique to L. pneumophila and have no known homologues in other organisms, whereas some of them have been shown to posses a significant degree of homology to eukaryotic proteins (Nagai et al., 2002; Chen et al., 2004; de Felipe et al., 2005; Bruggemann et al., 2006; de Felipe et al., 2008). The genomes of all four sequenced L. pneumophila strains contain many genes predicted to encode eukaryotic homologous proteins or eukaryotic protein domains and some of them have already been implicated in having a role in L. pneumophila pathogenesis (Nagai et al., 2002; Habyarimana et al., 2008; Kubori et al., 2008; Pan et al., 2008). The evolutionary origin of the eukaryotic homologues in the Legionella genome is currently unclear, but it is likely that many of these genes originate from different protozoa, which are the Legionella environmental hosts.

One of these eukaryotic homologous genes, legS2, encodes a homologue of sphingosine 1-phosphate lyase (SPL), an enzyme, highly conserved among all eukaryotes (Saba et al., 1997; Ikeda et al., 2004). SPL is responsible for the irreversible degradation of sphingosine 1-phosphate (S1P) to phosphoethanolamine and hexadecanal (Van Veldhoven and Mannaerts, 1993). S1P is an endogenous sphingolipid metabolite that influences a wide range of physiological functions such as cell survival, proliferation, migration, differentiation and many others (Bandhuvula and Saba, 2007).

In this work we demonstrate that the L. pneumophila eukaryotic homologue of SPL, LegS2, is evolutionarily conserved among several Legionella species and it was most likely acquired by Legionella from a protozoan host. Moreover, we found that LegS2 retained the function of its eukaryotic homologue SPL, yet it is being targeted to the host mitochondria as opposed to ER localization of the eukaryotic SPL.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Identification of conserved Icm/Dot effectors in different Legionella species

The Legionella genome was demonstrated to be highly plastic and even within different L. pneumophila strains there are considerable differences in organization and content of effector-encoding genes (Cazalet et al., 2004; 2008; Chien et al., 2004; Zusman et al., 2008), thus identification of conserved effectors in different Legionella species might indicate for their possible importance during infection. To address this question, we used low stringency hybridizations to examine the distribution of 23 L. pneumophila effector-encoding genes (Ninio and Roy, 2007; Zusman et al., 2007; de Felipe et al., 2008; Shin et al., 2008) among 18 different Legionella species. The analysis indicated that effector distribution is highly variable among different Legionella species (Fig. 1). Twelve of the examined effector-encoding genes were found to be present only in L. pneumophila, and even then bioinformatics analyses of the available genomic sequences of the four L. pneumophila strains revealed variations in the distribution of these genes (data not shown). However, the other 11 effector-encoding genes examined were found to be present in a number of different Legionella species (Fig. 1); interestingly, they mostly included genes encoding eukaryotic homologous proteins. Despite the presence of these effector-encoding genes in many species, their distribution often appeared random and in most cases it did not correlate with the Legionella phylogenetic tree. Moreover, several genes were found to be present in various Legionella species but were absent from L. pneumophila sg-5 (Fig. 1). The distribution of the genes encoding eukaryotic homologous proteins among the different Legionella species raises the possibility that these genes were acquired through massive horizontal gene transfer (HGT) and were lost during the course of evolution.

image

Figure 1. The distribution of L. pneumophila effector-encoding genes among different Legionella species does not correlate with the Legionella phylogenetic tree. The Legionella phylogenetic tree was reconstructed using the Mip protein sequences from each of the indicated species as described in the Experimental procedures section. Coxiella burnetii Mip protein was used as an outgroup. On the right, a summary of the low stringency hybridization results is presented for the genes indicated above each column: legLC4, ralF, legS2, legK3, ylfA, sidB, sidA, sdbA, ceg23, lirB and legK1. The following genes resulted with a hybridization signal only within L. pneumophila: lidA, lepA, wipA, legC5, sidC, sidE, sidD, sidG, vipA, legU2/lubX, legL5 and ceg25 (marked with asterisk ‘*’ in the left column). Plus ‘+’ indicates a positive hybridization signal.

Download figure to PowerPoint

If in fact these genes were acquired through HGT, it is clear that incorporation of such eukaryotic genes into a bacterial pathogenesis system would require many adaptations such as acquisition of a promoter and a translocation signal. To try and follow these adaptations, we decided to focus our study on LegS2, a protein highly homologous to the eukaryotic SPL (36% identity and 52% similarity to Tetrahymena thermophila). The known function of SPL in eukaryotic cells and the wide distribution of LegS2 among distant Legionella species, enabled in-depth study of the evolution, function and adjustments of this effector to the Icm/Dot pathogenesis system, as described below.

L. pneumophila LegS2 originates from a protozoan host

To explore the evolution of LegS2 in the Legionella genus we cloned the genes encoding LegS2 homologues from two distantly related Legionella species: L. jamestowniensis and L. dumoffii, which localize to different branches of the Legionella evolutionary tree (Fig. 1). Cloning and sequencing of these genes revealed that L. jamestowniensis and L. dumoffii encode a protein highly homologous to L. pneumophila LegS2 (68%/65% identity and 80%/79% similarity respectively). To determine whether legS2 was acquired though HGT more than once or, rather, through a single acquisition at an early stage of the Legionella evolution and then differentially lost from many species, we analysed several parameters such as G+C content, codon usage and G+C content of the wobbling third codon of the legS2 genes from the three Legionella species. The analyses revealed that in all three species the above parameters did not differ from other genes in the genomes of each of the species examined (data not shown). These findings strongly favour a single acquisition of the LegS2-encoding gene rather than several more recent acquisitions, which probably would have resulted in pronounced differences in the above mentioned parameters. In addition, we reconstructed the SPL phylogenetic tree based on SPL consensus sequences (Fig. 2), and this analysis further supports the single acquisition theory as the L. dumoffii LegS2 was found to be evolutionarily closer to the L. pneumophila LegS2 in comparison to the L. jamestowniensis LegS2, like in the acceptable Legionella phylogenetic tree (Ratcliff et al., 1998). Moreover, the reconstruction of the SPL phylogenetic tree clearly indicated that LegS2 was acquired from a eukaryotic organism as several other bacteria that were found to harbour a SPL homologous gene, such as Symbiobacterium thermophilum and Myxococcus xanthus, distributed to other branches of the SPL evolutionary tree (Fig. 2). Finally and most importantly, the Legionella LegS2 was found to form a monophyletic group with the protozoan SPL, implying that LegS2 was in fact acquired (directly or indirectly) from a protozoan host. It is important to note that the analysis was based on the alignment of 500 of SPL homologous proteins (data not shown) and the presented tree was reconstructed using a representative group of sequences.

image

Figure 2. The L. pneumophila legS2 gene was most likely acquired from protozoa. Rooted phylogenetic tree of the SPL proteins was reconstructed based on the consensus sequence from SPL homologous proteins obtained through PSI-BLAST. The presented SPL consensus sequence phylogenetic tree was reconstructed using the phyml program (for details see the Experimental procedures section). Glutamate decarboxylase catalytic domain from Francisella philomiragia was used as outgroup, as this enzyme is homologous to SPL, but performs a different function. Numbers indicate bootstrap values. Dashed circle marks monophyletic group of protozoa and Legionella.

Download figure to PowerPoint

Overall, the presented data supports the notion that the legS2 gene was acquired by Legionella on a single occasion early during Legionella evolution probably from a protozoan host and was then differentially lost from different Legionella species.

LegS2 is expressed and translocated into host cells via the Icm/Dot secretion system

As Legionella legS2 gene was most likely acquired from a protozoan organism, we were interested to determine its involvement in the L. pneumophila pathogenesis. For this purpose we examined the legS2 expression, translocation and requirement for intracellular growth. Using a lacZ fusion, the legS2 gene was found to be expressed in L. pneumophila and its expression levels were higher when examined at stationary phase in comparison to exponential phase (Fig. 3A). In addition, alignment of the upstream regulatory regions of the legS2 gene from L. pneumophila, L. dumoffii and L. jamestowniensis revealed a conserved sequence similar to −10 promoter elements previously found in many effector regulatory regions (Zusman et al., 2007; Altman and Segal, 2008) (Fig. 3A, upper panel). Substitution of the first TA nucleotides of this sequence caused a complete abolishment of the legS2::lacZ expression both at the exponential and at stationary growth phases (Fig. 3A), indicating that the legS2 gene is expressed from a regular prokaryotic promoter.

image

Figure 3. L. pneumophila LegS2 is expressed and translocated via the Icm/Dot secretion system. A. Upper panel: Alignment of the legS2 regulatory region from L. pneumophila (Lpn), L. jamestowniensis (Ljm) and L. dumoffii (Ldm). The putative −10 promoter is in bold and the two nucleotides mutated are underlined. Lower panel: Expression levels of legS2::lacZ wild-type fusion and the legS2::lacZ-mut fusion containing a mutation in the putative −10 promoter element were measured at exponential (white) or stationary (grey) growth phases. Results were obtained from at least three independent experiments; error bars represent standard deviation. B. Translocation of the CyaA::LegS2 fusion with its C-terminus exposed was examined from the JR32 wild-type strain and ΔicmT deletion mutant strain (GS3011). ‘Vec’ indicates vector containing only the CyaA protein in the wild-type strain (JR32). The results represent the means of cAMP levels per well obtained from at least three independent experiments; error bars indicate standard deviations. Protein levels were assessed by Western blot using α-CyaA antibody and are shown below each bar. C and D. Intracellular growth analysis of the legS2 knock-out mutant in either A. castelanii (C) or HL-60-derived human macrophages (D). Shown are wild-type L. pneumophila JR32 (solid diamonds); ΔicmT (GS3011) (solid boxes); ΔlegS2::Km (ED1001) (open circles). The error bars represent standard deviations.

Download figure to PowerPoint

To find out whether LegS2 is a part of the L. pneumophila pathogenesis system, we constructed a CyaA fusion and determined that LegS2 is a translocated substrate of the Icm/Dot secretion system (Fig. 3B). Very recently, LegS2 was also demonstrated to translocate into host cells using a TEM1 (β-lactamase) fusion (de Felipe et al., 2008). To explore the involvement of LegS2 in intracellular growth, we constructed a legS2 knock-out mutant and examined it for intracellular growth in Acanthamoeba castelanii and HL-60-derived human macrophages. As was previously shown for many L. pneumophila effectors (Ninio and Roy, 2007), no intracellular growth defect was observed (Fig. 3C and D).

Collectively, these results indicate that LegS2 is expressed and translocated into host cells, but its absence does not impair L. pneumophila intracellular growth in the examined hosts.

L. pneumophila LegS2 contains a C-terminal translocation signal

Alignment of Legionella LegS2 proteins with the protozoan SPL proteins revealed an extension of about 30 amino acids at the C-terminal end of the Legionella LegS2 (Fig. 4A). Previous works on Icm/Dot effectors indicated that the signal for effector translocation was located at their C-terminal end (Nagai et al., 2005; Kubori et al., 2008). We were interested to examine whether LegS2 relies on this 30-amino-acid extension at the C-terminal end of the protein for its effective translocation into host cells. For this purpose we constructed a CyaA–LegS2 fusion missing the 30 C-terminal amino acids (LegS2-short) and a fusion containing only these 30 C-terminal amino acids directly fused to the CyaA reporter (Lpn30). Deletion of the extension resulted with an almost twofold decrease in cAMP levels (Fig. 4B), implying that this region is involved in promoting LegS2 translocation into host cells. Translocation of the CyaA reporter directly fused to these 30 C-terminal amino acids resulted in an almost 10-fold increase in translocation when compared with the CyaA protein alone, this translocation was found to be dependent on a functional Icm/Dot system (Fig. 4B), and it was about twofold less efficient in comparison to the full-length LegS2. Taken together, these results indicate that the 30 C-terminal amino acids of LegS2 are sufficient for translocation via the Icm/Dot secretion system.

image

Figure 4. The 30 C-terminal amino acids of LegS2 comprise an Icm/Dot translocation domain. A. Protein sequence alignment of the LegS2 C-terminus from L. pneumophila (Lpn), L. jamestowniensis (Ljm) and L. dumoffii (Ldm), as well as SPL from Dictyostelium discoideum (Dds), Entamoeba histolytica (Eht) and T. thermophila (Ttm), was performed using ClustalW. The arrow indicates the beginning of the C-terminal translocation domain. B. Translocation of the full-length LegS2 fusion (LegS2), LegS2 missing its C-terminal 30 amino acids (LegS2-short), the C-terminal extensions of the L. pneumophila (Lpn30), L. dumoffii (Ldm30) and L. jamestowniensis (Ljm35) LegS2 proteins (as indicated by the arrow) and the CyaA-vector control (Vec.) were examined in the wild-type strain JR32 (WT – grey), the ΔicmT knock-out mutant (ΔT – white), the ΔicmS knock-out mutant (ΔS), the ΔicmW knock-out mutant (ΔW) and the ΔicmS–icmW double deletion mutant (ΔWS) strains. The results represent the means of cAMP levels per well obtained from at least three independent experiments; error bars represent standard deviations. Protein levels were assessed by Western blot using α-CyaA antibody and are shown below each bar.

Download figure to PowerPoint

The LegS2 protein was found to be highly conserved between L. pneumophila, L. dumoffii and L. jamestowniensis; however, their C-terminal extension is very dissimilar (Fig. 4A). Previous works revealed that the Icm/Dot secretion system is conserved among different Legionella species (Segal et al., 2005), moreover a recent work also identified the first Coxiella burnetti effectors using L. pneumophila as a surrogate host (Pan et al., 2008). Nevertheless, no effector proteins from Legionella species other than L. pneumophila have been identified up to date. To determine if LegS2 serves as an effector protein also in L. dumoffii and L. jamestowniensis, the C-terminal extension of LegS2 from these two bacteria was fused to the CyaA protein. Both these fusions were found to efficiently translocate into host cells in an Icm/Dot-dependent manner (Fig. 4B). It is important to emphasize that this C-terminal extension, which is present in the Legionella LegS2 proteins (Fig. 4A), is missing from the eukaryotic SPL proteins, thus implying that it was most likely incorporated to the Legionella LegS2 after its acquisition as part of LegS2 adaptations to the Icm/Dot pathogenesis system.

Previously, it was reported that the IcmS–IcmW cytosolic complex is involved in translocation of the Icm/Dot substrates into host cells (Ninio et al., 2005; de Felipe et al., 2008). However, not all of the Icm/Dot translocated substrates were found to require the IcmS–IcmW complex, among them another eukaryotic homologous effector RalF (Cambronne and Roy, 2007). Thus, we were interested to determine whether this cytosolic complex is involved in LegS2 translocation. For this purpose we examined the translocation of LegS2-CyaA fusion from the ΔicmW, ΔicmS and ΔicmSΔicmW knock-out mutants. The translocation of LegS2 was reduced in these mutants in comparison to the wild-type strain (Fig. 4B) and the translocation efficiency was similar between the three mutants examined. This result indicates that LegS2 translocation depends on a functional IcmS–IcmW complex and that the presence of only one component of this complex is insufficient for effective translocation. Interestingly, the translocation of the LegS2-short fusion was dependent on a functional Icm/Dot system, but it was found not to be affected by the IcmS–IcmW cytosolic complex (Fig. 4B), what might indicate for another Icm/Dot secretion signal in the LegS2 effector.

LegS2 functions as a sphingosine 1-phosphate lyase

Several L. pneumophila effector proteins that posses eukaryotic homologous domains such as the Sec7 guanine nucleotide exchange factor domain in RalF (Nagai et al., 2002) and the U-box containing protein LubX (Kubori et al., 2008) have been demonstrated to retain the function of their eukaryotic homologues. As was previously indicated, LegS2 is highly homologous to the eukaryotic SPL, especially at the putative catalytic domain of the protein (Fig. 5A). In eukaryotes, SPL is responsible for the irreversible degradation of S1P to phosphoethanolamine and hexadecanal (Van Veldhoven and Mannaerts, 1993; Saba et al., 1997; Ikeda et al., 2004). The first SPL-encoding gene was cloned from Saccharomyces cerevisia (Saba et al., 1997), and later the first mammalian and Drosophila SPL were identified and characterized using the yeast system (Zhou and Saba, 1998; Herr et al., 2003). We decided to use a similar approach in order to determine whether L. pneumophila LegS2 functions similarly to the eukaryotic SPL. The S. cerevisiae SPL deletion mutant strain, Δdpl1, was shown to be sensitive to d-erythro-sphingosine, due to its inability to degrade sphingosine that accumulates in the cell and subsequently leads to growth arrest (Saba et al., 1997). To examine whether L. pneumophila LegS2 can complement this phenotype, S. cerevisiae Δdpl1 was transformed with plasmids expressing either S. cerevisiae dpl1 or L. pneumophila legS2 under the regulation of galactose promoter, and the strains were plated on minimal plates supplemented with d-erythro-sphingosine. The results of this experiment unequivocally demonstrated that L. pneumophila LegS2 was capable of reversing the sphingosine-sensitive phenotype of the yeast Δdpl1 strain as transformants expressing L. pneumophila LegS2 were resistant to d-erythro-sphingosine in the presence of galactose but not in the presence of glucose (Fig. 5B). As expected, the yeast Δdpl1 strain transformed with the empty vector remained sensitive to sphingosine (Fig. 5B).

image

Figure 5. LegS2 complements the sphingosine-sensitive phenotype of the S. cerevisiaeΔdpl1 mutant. A. Sequence alignment of the most conserved region of LegS2 and the eukaryotic SPL proteins. Arrows indicate conserved amino acid residues important for SPL function. The bracket indicates a conserved amino acid stretch essential for SPL catalytic activity. B. Functional complementation of the yeast Δdpl1 strain by L. pneumophila LegS2. Transformants were template-inoculated in fivefold dilutions onto YPD plate (control) and minimal plates containing 50 μM d-erythro-sphingosine with either glucose or galactose (indicated above each plate). Two different clones are shown for LegS2. LegS2mut represents a LegS2 mutant containing two point mutations C328A and K362A [indicated by arrow in (A)].

Download figure to PowerPoint

Site-directed mutagenesis studies of the human SPL protein have identified several residues that are crucial for its function, including a 20-amino-acid stretch (amino acids 344–364), which includes a predicted cofactor binding site and several conserved lysine residues, which are part of a potential binding site for the S1P substrate, one of these residues is K359 (K362 in L. pneumophila LegS2) (marked in Fig. 5A). Additionally, a conserved cysteine at a position 317 (C328 in L. pneumophila LegS2) (marked in Fig. 5A) is likely to be required for proper protein folding (Van Veldhoven et al., 2000). To further confirm the result that L. pneumophila LegS2 indeed relies on its SPL function for reversing the sphingosine-sensitive phenotype of the S. cerevisiae Δdpl1 strain, we introduced these two point mutations (K362A and C328A) into L. pneumophila LegS2. S. cerevisiae Δdpl1 transformant expressing LegS2 K362A/C328A under the regulation of galactose promoter remained sensitive to sphingosine (Fig. 5B), thus supporting the finding that LegS2 functions as SPL. The protein levels of both the wild-type and the mutated LegS2 were similar when grown on inducible plates containing galactose, as was determined by Western blot analysis (data now shown).

LegS2 is targeted to the host cell mitochondria

Eukaryotic SPL is an integral membrane protein, which is predominantly located at the ER with its large C-terminal domain facing the cytosol (Ikeda et al., 2004). Previous works on L. pneumophila effectors used ectopic expression to determine their cellular localization (Derre and Isberg, 2005; Pan et al., 2008). To determine whether LegS2 effector localizes to the same organelle as the eukaryotic SPL, LegS2 was N-terminally fused to the His6 tag and cloned under the CMV promoter in a mammalian expression vector. Co-transfection of His-tagged LegS2 with Sec61-GFP, an ER marker, into COS7 cells resulted in little to no colocalization of both proteins, pointing to the localization of the greater part of LegS2 to an organelle different from the ER (Fig. 6A). Additionally, the observed pattern of LegS2 subcellular localization in the transfected cells resembled distribution of mitochondria in mammalian cells. To examine the possibility that L. pneumophila LegS2 is targeted to the host cell mitochondria, the His6-tagged LegS2 was coexpressed with the mitochondrial marker (MitoRed) and clear colocalization was observed (Fig. 6B). Localization of the LegS2 K362A/C328A mutant was similar to the wild-type LegS2, implying that the protein cellular localization is independent of its catalytic activity (Fig. 6C). Similar results were also observed when His6-tagged LegS2 was coexpressed with the mitochondrial marker in HeLa cells (data not shown).

image

Figure 6. LegS2 is targeted to the host cell mitochondria. Examination of colocalization between LegS2 (His tagged – red) and the ER marker Sec61 (GFP – green) (A) shows no colocalization. Examination of wild-type (B) and mutated (C) LegS2 (His tagged – green) with the mitochondrial marker dsRed-Mito (red) shows clear colocalization (yellow) in the right panels. The images are representatives of the cell population coexpressing both specified proteins. The images were taken using confocal microscopy and processed using a deconvolution tool as described in the Experimental procedures section.

Download figure to PowerPoint

To determine if LegS2 localization to the mitochondria also occurs during infection, we infected U937-derived human macrophages with L. pneumophila expressing either myc-tagged LegS2 or myc-tagged LegS2 K362A/C328A. Both wild-type LegS2 and the mutated LegS2 localized to host cell mitochondria during infection (Fig. 7), similarly to what was observed with the ectopically expressed LegS2 (Fig. 6). Additionally, acquired 3D images of LegS2 distribution during ectopic expression (Fig. 8A and B), as well as during infection (Fig. 8C), indicated that it surrounds the mitochondria from the outside and it seems to form patches on the mitochondria outer membrane during infection as well as when ectopically expressed (Fig. 8B and C).

image

Figure 7. LegS2 is targeted to host cell mitochondria during infection. (A) Non-infected U937-derived human macrophages stained with DAPI (blue), α-myc and Alexa-488 antibodies (green) and MitoTracker for mitochondria staining (red). U937-derived human macrophages were infected with wild-type L. pneumophila expressing myc-tagged LegS2 (B) or mutated LegS2 (C) stained with α-myc and Alexa-488 antibodies (green). Mitochondria were stained with MitoTracker (red) and cell nuclei were stained with DAPI (blue). Colocalization (yellow) of LegS2 with mitochondria visualized on the merge of images B and C (right panel). The images were taken using confocal microscopy and processed using a deconvolution tool as described in the Experimental procedures section.

Download figure to PowerPoint

image

Figure 8. High-resolution 3D analysis of the LegS2 localization to the mitochondria. A COS7 cell coexpressing His-tagged LegS2 (green) and the mitochondrial marker dsRed-Mito (red) were examined (A and B). Three-dimensional rendition of the same cell (A) with magnification on separated mitochondria [marked in (A)] depicting the exact localization of LegS2 on the mitochondrial surface (B). (C) 3D projection of the U937-derived human macrophages infected with L. pneumophila expressing Myc-tagged LegS2 (green). The mitochondria were stained with MitoTracker (red) and DNA with DAPI (blue). Arrow head marks the intracellular bacteria (C). Arrows indicated patches formed by LegS2 on the cytosolic side of mitochondria (B and C).

Download figure to PowerPoint

These results clearly indicate that LegS2 is targeted to the host cell mitochondria unlike the eukaryotic SPL (Ikeda et al., 2004), further emphasizing the power of bacterial evolution that is capable of changing cellular targeting of a protein as a part of its adaptation to the bacterial pathogenesis system.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

The L. pneumophila genome has been demonstrated to include a significant number of genes encoding proteins mainly found in eukaryotic organisms (Cazalet et al., 2004; Chen et al., 2004; de Felipe et al., 2005). Nearly 40 of these proteins have been shown to function as substrates of the Icm/Dot secretion system (Nagai et al., 2002; de Felipe et al., 2008; Pan et al., 2008), and effectors containing both eukaryotic protein-interacting domains, such as ankyrin repeat and coiled-coil domains, and catalytic domains, such as Sec7 and U-box, have been implicated in L. pneumophila pathogenesis (de Felipe et al., 2008; Habyarimana et al., 2008; Kubori et al., 2008; Pan et al., 2008). However, the question about the origin of these genes and the adaptations required for their integration into the L. pneumophila pathogenesis system remains open. As L. pneumophila has an intracellular lifestyle it is likely to assume that most of their acquisitions of genetic material from eukaryotic origin would arise from their hosts. Analysis of L. pneumophila LegS2, a protein highly homologous to the eukaryotic SPL, confirmed that the most likely origin of this gene is from one of its protozoan hosts. A similar result was also obtained for the L. pneumophila RalF protein when we performed analogous analysis for over 500 proteins containing a Sec7 domain and reconstructed its phylogenetic tree (Fig. S1). Although several Rickettsia species have been demonstrated to encode a RalF homologous protein (Nagai et al., 2002; Ogata et al., 2006) that clusters together with the Legionella RalF, the Sec7-containing proteins from both bacteria, cluster with the Sec7-containing proteins from protozoan organisms, thus implying that the original acquisition occurred from a protozoan organism, similarly to what was found for LegS2.

Once acquired by the bacteria a gene must undergo a number of changes in order to become a functional part of the pathogenesis system while retaining or altering its function. Eukaryotic genes have both exons and introns that prokaryotes are unable to handle. One possible way to overcome this obstacle is to acquire DNA by converting it from mRNA either in the host or in the bacterial cytosol. This scenario is not improbable as the Legionella genome encodes two putative reverse transcriptases (lpg1081 and lpg2070). Additionally, Legionella must be able to maintain newly acquired genes long enough for them to acquire regulatory elements, such as a promoter and a ribosomal binding site (this step might occur upon the gene entry into the bacterial genome, in case that the newly acquired DNA is integrated into a functional gene). As we showed here, analysis of upstream region of the legS2 gene in three Legionella species, L. pneumophila, L. dumoffii and L. jamestwoniensis, confirmed the presence of a functional bacterial −10 promoter element.

Finally, in order to become an effector of the Icm/Dot pathogenesis system, a protein must obtain a translocation signal, which will allow its translocation into the host cells. A number of previous works showed that the translocation signal of the Icm/Dot effectors is located at the C-terminal end of these proteins (Nagai et al., 2005; Kubori et al., 2008). Our results demonstrated not only that the signal is found at the C-terminal end of the protein, but that it is clearly an extension in comparison to eukaryotic proteins, an extension that was most likely incorporated into the gene after its acquisition. Moreover, this C-terminal region was found to be subjected to evolutionary changes as it becomes very dissimilar among the LegS2 proteins in the different species. To further explore the phenomenon of the variability in the C-terminal translocation signal in the eukaryotic homologous effectors we examined the effector RalF for this property. For this purpose we cloned the RalF-encoding gene from Legionella moravica and aligned it with the RalF proteins encoded by three different L. pneumophila strains. Similar to LegS2, the RalF proteins also appeared very dissimilar at the C-terminal end, even within the L. pneumophila strains (Fig. S2). This variability of the C-terminal translocation signal might indicate for the high flexibility of the Legionella translocation machinery that might allow the accumulation of many changes.

Although the genes encoding eukaryotic homologous proteins undergo many changes during their adaptation to the pathogenesis system, they seem to retain their original function as was shown for RalF, LubX and here for LegS2 (Nagai et al., 2002; Kubori et al., 2008). However, despite their conserved functions, some of these proteins were relocated to cellular organelles different from their original locations in the eukaryotic cells. For example, RalF, contains a Sec7 homologous domain which is generally used for the recruitment of Arf1 to the Golgi apparatus (Casanova, 2007). The proteins containing this domain are usually transmembrane proteins localizing either to the Golgi apparatus or to the plasma membrane (Casanova, 2007). The L. pneumophila RalF was shown to be localized to the Legionella-containing vacuole within the host cell and to recruit Arf1 to this site, thus redirecting intracellular vesicular trafficking (Nagai et al., 2002). In our case, the eukaryotic SPL is a transmembrane ER-localized protein with its catalytic domain facing the cytosol (Ikeda et al., 2004) and we demonstrated here that the Legionella LegS2 was localized to the cell mitochondria.

Effectors targeting host cell mitochondria have been previously described for pathogenic bacteria utilizing a type III secretion system (Nougayrede and Donnenberg, 2004; Layton et al., 2005; Munkvold et al., 2008); however, to the best of our knowledge, LegS2 is the first type IV secretion system effector targeted to the host cell mitochondria. SPL is one of the enzymes involved in sphingolipids metabolism; however, it is currently unclear what is the exact contribution of LegS2 to the Legionella pathogenesis. It is possible that Legionella requires the degradation products of S1P one of which is phosphoethanolamine that serves as a precursor for phosphatidylethanolamine synthesis, which in turn can be converted to phosphatidylcholine (Vance and Vance, 2004). A recent study showed that Legionella requires phosphatydilcholine synthesis for its virulence (Conover et al., 2008), and it is possible that it also relies on the host sphingolipids pools for it. Degradation of S1P alters a wide range of cellular processes that result in altered gene expression and at times diminished cell survival (Bandhuvula and Saba, 2007). How alteration of these processes might be of benefit to Legionella intracellular survival is a matter for further studies.

Experimental procedures

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Bacterial strains, plasmids, primers and media

Legionella species and strains used in this work are listed in Table S1. Plasmids and primers used in this work are described in Tables S2 and S3 respectively. Bacterial media, plates and antibiotic concentrations were used as described before (Segal and Shuman, 1997).

Cloning of legS2 homologous genes from L. jamestowniensis and L. dumoffii

In order to determine the size of the DNA fragment where the legS2 gene is located, genomic DNA from L. dumoffii and L. jamestowniensis was digested with EcoRI and used for low stringency hybridization with the L. pneumophila legS2 probe. The legS2 gene was found to be located in a 6.5 kb and 4.5 kb DNA fragment of L. jamestowniensis and L. dumoffii respectively. These fragments were cloned into pUC18 and correct clones (pED-spl-jam1 and pED-spl-dum1) were detected by colony hybridization using the L. pneumophila legS2 probe (Segal and Shuman, 1997). The legS2 gene from both bacteria, as well as the adjacent regions, was sequenced. As pED-spl-dum1 contained only the legS2 homologue and its upstream region, the downstream region of the L. dumoffii legS2 homologue was obtained by inverse PCR. The L. dumoffii genomic DNA was digested with HindIII followed by self-ligation and PCR amplification. The 3.5 kb PCR product was cloned into pUC18 digested with HincII to generate pED-spl-dum2. GenBank accession numbers EU107519 and EU107520 were assigned to the legS2 regions from L. dumoffii and L. jamestowniensis respectively.

Cloning of the ralF gene from L. moravica

Legionella moravica genomic DNA was digested separately with either HindIII or AvaIII (sticky end compatible with PstI) and hybridizations were performed using L. pneumophila ralF as a probe. The fragments with positive hybridization signal were 2.5 kb and 1.5 kb for HindIII and AvaII digested DNA respectively. These fragments were cloned into pUC18 digested with either HindIII or PstI to generate pED-moRalFHind and pED-moRalFAva, respectively, and sequenced. GenBank accession number FJ394378 was assigned for the L. moravica ralF gene.

Cloning for L. pneumophila allelic exchange

Fragments of 1 kb from the upstream and the downstream regions of the LegS2 gene were amplified by PCR using genomic L. pneumophila DNA as a template and the primers Spl-AF and Spl-AR, and Spl-BF and Spl-BR respectively (Table S3). The resulting fragments were digested with either EcoRI and SalI or SalI and BamHI, respectively, cloned into pUC18 to generate pED-SPL-A and pED-SPL-B, respectively, and sequenced. These two plasmids were then digested with the respective restriction enzymes and cloned into pUC18 together with the kanamycin (Km) resistance cassette digested with SalI to generate pED-pd-SPL. The whole fragment containing the upstream region, the downstream region and the Km cassette between them was digested with PvuII and cloned into pLAW344 digested with EcoRV to form pED-dSPL. This plasmid was used for allelic exchange as previously described (Segal and Shuman, 1998). The resulting strain (ED1001) contained the first 38 amino acids of the SPL, as well as its 105 last amino acids and its internal 458 amino acids were substituted by the Km cassette. For generation of the icmS–icmW double mutant (ED400), the previously described plasmid pGS-Le-33-D2-Km (Segal and Shuman, 1997) was introduced into the icmW mutant strain GY141 (Zusman et al., 2003) for allelic exachange. All the knock-out mutations were confirmed by PCR.

Construction of CyaA fusions

The LegS2 gene was amplified by PCR using spl-F-cyaA and spl-C-R primers (Table S3), digested with BamHI and PstI and cloned into pMMB-cyaA-C digested with the same enzymes to generate pED-cyaA-SPL. For construction of pED-cyaA-SPL-stop plasmid, a region encoding the first 571 amino acids of SPL (out of 601) was amplified by PCR using spl-F-cyaA and SPL-stop-R primers (Table S3), digested with BamHI and PstI and cloned into pMMB-cyaA-C vector digested with the same enzymes. For generation of pED-Lpn30, pED-Dum30 and pED-Jam30, regions encoding the 30 C-terminal amino acids (35 aa for L. jamestowniensis LegS2) of LegS2 protein from L. pneumophila, L. dumoffii and L. jamestowniensis were amplified by PCR using the following primers: spl-30C and spl-C-R, dum30-F and dum30-R, jam30-F and jam30-R respectively. The PCR products were cloned into pMMB-cyaA-C in frame to generate pED-lpn30, pED-dum30 and pED-jam30 respectively. CyaA translocation assay was performed as described elsewhere (Zusman et al., 2007).

Construction of spl::lacZ fusion

The promoterless lacZ vector pGS-lac-02 was used for cloning of the regulatory region of the legS2 gene. A 277 bp fragment containing the legS2 regulatory region was amplified by PCR using the SPL-lacZ-EI and SPL-lacZ-Bam primers (Table S3). The resulting fragment was digested with BamHI and EcoRI and cloned into the pGS-lac-02 vector to generate pGS-SPL-lacZ and sequenced. For mutagenesis of the putative −10 promoter element of the LegS2 gene, the overlap extension PCR method (Ho et al., 1989) was applied using the SPL-prom-mut-F and SPL-prom-mut-R primers and pGS-SPL-lacZ as a template, to generate pGS-SPL-mut-LacZ. The β-galactosidase assay was performed as described elsewhere (Zusman et al., 2007).

Cloning for yeast complementation

The dpl1 gene from S. cerevisiae BY4741 strain (MATa his3D1 leu2D0 met15D0 ura3D0), kindly provided by Professor M. Kupiec, Tel Aviv University, was amplified using DPL1compF and DPL1compR primers (Table S3), digested with BamHI and XhoI and cloned into pGREG523 to generate pED-yDpl1. For legS2, the gene was amplified using yelSPL-Bam and SPLhis-R primers (Table S3), digested with BamHI and XhoI and cloned into pGREG523 under the regulation of GAL promoter to generate pED-ySPL. To generate pED-ySPLmut containing two point mutations the PCR overlap extension approach was applied (Ho et al., 1989). For C328A mut3SPLAla-F and mut3SPLAla-R primers were used and for K360A SPL-mut-F and SPL-mut-R primers were used (Table S3). Briefly, first K360A mutated fragment was constructed using appropriate primers and was then used as a template for the generation of a double mutant fragment, which was then digested with BamHI and XhoI, cloned into pGREG523 to generate pED-ySPLmut.

Cloning of myc-tagged legS2

For cloning of myc-tagged legS2 and legS2 C328A/K360A into pMMB207 vector, the N-terminal region of legS2 was amplified by PCR using the primers Myc-F-XbaI and SPL-R-SacI (Table S3) and pED-ySPL as a template. The fragment obtained contains the 5′ region of legS2 fused to the 13myc tag, and this fragment was digested with XbaI and SacI. The wild-type 3′ region of LegS2 was digested from pED-cyaA-SPL with SacI and PstI. For the legS2 C328A/K360A the 3′ region was amplified by PCR using spl-F-cyaA and spl-C-R primers (Table S3), using pED-ySPLmut as a template, this fragment was digested with BamHI and PstI and cloned into pUC18 to generate pED-pSPLmut and the 3′ region of legS2 C328A/K360A was digested out from this plasmid with SacI and PstI. To generate pZT-207-myc-SPL pMMB207 was digested with XbaI and PstI and three-way ligation was performed using the two fragments described above. For generation of pZT-207-myc-SPLmut similar ligation was performed using the relevant fragment containing the 3′ region of legS2.

Cloning for ectopic expression

For cloning of legS2 into the mammalian pcDNA4 vector, legS2 was amplified by PCR using SPL-his-F and SPL-his-R primers (Table S3). The resulting fragment was digested with NdeI and XhoI and cloned into pET15b vector containing an in-frame N-terminal His6 tag to generate pED-hisSPL. The N-terminal His-tagged legS2 fragment was amplified using pMMB-His-EcoRI and SPL-R-SacI primers (Table S3) and digested with EcoRI and SacI, and the C-terminal fragment of legS2 was digested with SacI and PstI from pED-cyaA-SPL. These two fragments were cloned into pcDNA4 digested with EcoRI and PstI to generate pED-ectSPL, where the His6-LegS2 is under the regulation of the CMV promoter. To generate pED-ectSPLmut, the C-terminal fragment from pED-pSPLmut was used.

Low-stringency Southern hybridizations

Legionella chromosomal DNA was prepared from all the Legionella species listed in Table S1 as previously described (Segal and Ron, 1993). For the preparation of probes L. pneumophila JR32 genomic DNA was used as a template for PCR using the primers listed in Table S3. The hybridizations were performed as described elsewhere (Feldman and Segal, 2004).

Intracellular growth assays

Intracellular growth assays in A. castellanii and HL-60-derived human macrophages were performed as previously described (Segal and Shuman, 1999).

Bioinformatics analysis

All sequence homologies were determined by blast (Altschul et al., 1997). For reconstruction of Legionella genus phylogenetic tree Mip protein sequences from the relevant Legionella species were obtained from GenBank (Ratcliff et al., 1997; 1998). For reconstruction of phylogenetic tree of the sphingosine 1-phosphate lyase proteins, PSI-BLAST analysis was performed using default parameters in the non-redundant database (Altschul et al., 1997). Consensus sequences of the selected SPL proteins were aligned using web-available muscle program (Edgar, 2004), the tree was reconstructed using phyml program (Guindon et al., 2005) and visualized by the TreeView software. Consensus sequence of glutamate decarboxylase from Franscisella philomiragia was used as outgroup. For reconstruction of phylogenetic tree of Sec7 domain-containing proteins, 500 protein sequences were obtained through PSI-BLAST. Sequences of Sec7 domains from different proteins were aligned using web-available muscle algorithm, the tree was reconstructed using the phyml program (Guindon et al., 2005). Selected sequences were used for reconstruction of the presented tree. Distant Sec7 domain containing protein from Homo sapiens was used as outgroup. For calculation of G+C content and G+C content in the third codon of the coding regions in L. jamestowniensis and L. dumoffii, all available GenBank coding sequences from these two bacteria were obtained and arranged into one continuous coding sequence. The following website http://bioweb.pasteur.fr/seqanal/interfaces/geecee.html was used to calculate the values for G+C content. For G+C content of the third codon the http://bioweb.pasteur.fr/seqanal/interfaces/wobble.html website was used.

Transfection

COS7 cells were transfected with TransIt-LT1 (Mirus) transfection reagent according to manufacturer's instructions. Briefly, COS7 cells were grown in DMEM (Invitrogen) medium supplemented with 10% FBS and 2 mM Glutamine. A day prior to transfection cells were plated in 24-well plates containing 13 mm glass coverslips at a concentration of 5 × 104 cells per well. The next day the medium was replaced and the cells were transfected using 0.5 μg DNA per well. Following 18–20 h incubation cells were fixed and stained.

Infection of U937-derived human macrophages for fluorescence analysis

U937 were differentiated into human-like macrophages by addition of 10% normal human serum and 10 ng ml−1 of phorbol 12-myristate 13-acetate (TPA) (Sigma) at concentration of 4 × 105 cells per well on glass slides. Following 48-hour incubation, the cells were washed twice with RPMI supplemented with 2 mM glutamine and infected with the wild-type strain (JR32) expressing either the 13myc-tagged LegS2 or LegS2 C328A/K360A at multiplicity of infection of 5. Bacteria were allowed to adhere by centrifugation at 180 g for 5 min. Following 5 h incubation at 37°C medium was changed to serum-free medium containing 100 nM MitoTracker Red CMXRos (Invitrogen) and incubated at 37°C for 1 h. The cells were then fixed with 4% paraformaldehyde solution for 15 min and stained with purified monoclonal anti-c-myc (10E9) diluted 1:350 followed by staining with goat anti-mouse Alexa488 (1:250) and DAPI (1:1000).

Fluorescent staining

Cells were washed twice with PBS+ (PBSx1 containing 1 mM CaCl2 and 0.125 mM MgCl2) and fixed with 4% paraformaldehyde solution for 15 min. Coverslips were washed once with PBS+ followed by 5 min incubation in PBS++ (PBS+ containing 0.1% Tirton X-100 and 1% BSA). Coverslips were stained with appropriate antibodies diluted in PBS++ for 1 h, followed by two 5 min washes in PBS++ and stained with secondary antibodies diluted in the same buffer for 40 min. Following two 5 min washes with PBS+, the coverslips were mounted on glass slides using GelMount (Biomeda). Antibodies were diluted as follows: mouse anti-pentaHis (Qiagen) 1:25, followed by a 1:250 dilution of goat anti-mouse Alexa488 or goat anti-mouse Alexa647 (Molecular Probes), depending on the experiment. To label mitochondria, COS7 cells were transfected with pDs1Red-Mito (Clontech). To label the Endoplasmic Reticulum, COS7 cells were transfected with pSec61-GFP (a kind gift of Dr Gerardo Lederkremer Tel Aviv University).

Imaging

Images were acquired with a motorized spinning disk confocal (Yokogawa CSU-22 Confocal Head) microscope (Axiovert 200 M, Carl Zeiss MicroImaging) under control of SlideBook™ (Intelligent Imaging Innovations). Images were acquired with a 63× oil immersion objective (Plan Apochromat, NA 1.4). Illumination was with a 40 mW 473 nm and 10 mW 561 nm solid state lasers (Crystal lasers). A step size of 0.3 μm was used. For image presentation clarity, images were deconvoluted with either the Constrained Iterative or the Nearest Neighbour algorithms of the Slidebook™ software using a measured point spread function for each wavelength.

Sphingosine resistance assay

Saccharomyces cerevisiae BY4741dpl1Δ (MATa his3D1 leu2D0 met15D0 ura3D0 dpl1Δ::KanMx) strain was transformed with appropriate plasmids using standard lithium acetate protocol, and transformants were selected for histidine prototrophy. Sphingosine supplemented plates were prepared as previously described (Zhou and Saba, 1998). Specifically, minimal media containing either glucose or galactose was supplemented with 50 μM d-erythro-sphingosine (MP Biomedicals) and 0.0015% NP-40 (Sigma) as disperser. Strains of interest were grown in minimal liquid culture medium at 30°C, cell number was adjusted and a series of fivefold dilutions were made. The cultures were then template-inoculated onto YPD (Yeast extract/peptone/dextrose) control plate and minimal media plates containing 50 μM d-erythro-sphingosine with either glucose or galactose. Plates were incubated at 30°C for 2 days and visually assessed for differences in growth. Western blot was performed using anti-c-Myc antibody (Sigma) 1:2000.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

We would like to thank Adi Stern and Dr Tal Pupko for their help with the bioinformatics analyses, Yuval Mazor and Professor Martin Kupiec for their assistance with the yeast experiments and Olga Peker for her help with cell transfections.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Fig. S1. The L. pneumophila ralF gene most likely originated from a protozoan organism. Presented is a phylogenetic tree reconstructed from the alignment of Sec7 domains from GEF proteins. The tree was rooted using Arf6 GEF from Homo sapiens as outgroup. Protein sequences were obtained through PSI-BLAST, the tree was reconstructed using the PHYML program (for details see the Experimental procedures section). Numbers represent bootstrap values. Dashed circle marks monophyletic group of protozoa and Legionella.

Fig. S2. The translocation signal of RalF is dissimilar among different Legionella strains and species. Protein sequence alignment of the RalF C-terminus from L. pneumophila strains: Philadelphia-1 (Lpn), Paris (Lpp) and Lens (Lpl), and L. moravica (Lmr).

Table S1. Bacterial strains.

Table S2. Plasmids used in this study.

Table S3. Primers used in this study.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
CMI_1328_sm_Suplementary-Figures.pdf38KSupporting info item
CMI_1328_sm_Suplementary-Tables.pdf46KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.