The anti-apoptotic Coxiella burnetii effector protein AnkG is a strain specific virulence factor

The ability to inhibit host cell apoptosis is important for the intracellular replication of the obligate intracellular pathogen Coxiella burnetii, as it allows the completion of the lengthy bacterial replication cycle. Effector proteins injected into the host cell by the C. burnetii type IVB secretion system (T4BSS) are required for the inhibition of host cell apoptosis. AnkG is one of these anti-apoptotic effector proteins. The inhibitory effect of AnkG requires its nuclear localization, which depends on p32-dependent intracellular trafficking and importin-α1-mediated nuclear entry of AnkG. Here, we compared the sequences of ankG from 37 C. burnetii isolates and classified them in three groups based on the predicted protein size. The comparison of the three different groups allowed us to identify the first 28 amino acids as essential and sufficient for the anti-apoptotic activity of AnkG. Importantly, only the full-length protein from the first group is a bona fide effector protein injected into host cells during infection and has anti-apoptotic activity. Finally, using the Galleria mellonella infection model, we observed that AnkG from the first group has the ability to attenuate pathology during in vivo infection, as it allows survival of the larvae despite bacterial replication.

Coxiella burnetii is a zoonotic Gram-negative pathogen with an obligate intracellular lifestyle. The bacteria cause Q fever in humans. The acute form of Q fever is mainly characterized by a self-limiting flu-like illness and rarely by an interstitial pneumonia or hepatitis 1 . Acute Q fever is treatable with doxycycline, but many patients might also develop the post-Q fever fatigue syndrome 2 , which can last for up to 10 years 3 . At present, there is no evidence-based recommendation for treatment of post-Q fever fatigue syndrome. In addition, the infection may progress to chronic Q fever, which mainly manifests as a potentially fatal endocarditis 1 . Chronic Q fever develops months or years after infection, suggesting that C. burnetii persists silently within the host before massive bacterial replication leads to chronic Q fever. To control the infection, patients have to be treated for at least 18 months with doxycycline in combination with chloroquine 4 . Thus, a much more effective treatment for chronic Q fever would be highly desirable.
In general, humans become infected by inhaling C. burnetii-containing aerosols. Monocytes and macrophages are the primary target cells that take up C. burnetii by α v β 3 integrin-mediated phagocytosis 5,6 . The invasion into non-phagocytic cells seems to be at least partially dependent on the outer membrane protein A (OmpA) from C. burnetii 7 . After uptake, the C. burnetii-containing vacuole (CCV) matures by consecutive fusion events to a phagolysosome-like compartment [8][9][10] . Usually, phagocytosis of bacterial microorganisms leads to their destruction in the phagolysosome 11 . However, C. burnetii survives within the phagolysosome-like environment and requires these conditions for the generation of its progeny 12 . Cellular apoptosis is also important in the defense against intracellular pathogens 13 . As this is an anti-inflammatory mechanism it eliminates infected cells without inducing inflammation and tissue destruction 14 . Yet, C. burnetii efficiently inhibits host cell apoptosis 15,16 . Thus, inhibition of apoptosis and withstanding phagosomal maturation are thought to be important evasion The AnkG variants from the three groups respond differently to apoptosis induction. We previously demonstrated that the amino acids 1-69 of AnkG NM are required for anti-apoptotic activity 26 . Therefore, we analyzed whether AnkG Soyta (member of the second group), which contains all 69 amino acids of the first group, or AnkG F3 (member of the third group), where only the first 28 amino acids, except for amino acid 11, are identical to AnkG NM , would affect host cell survival (Fig. 2a). We ectopically expressed GFP or GFP-AnkG variants transiently in CHO cells. First, we demonstrated stable expression by immunoblot analysis (Fig. 2b). Next, we quantified the percentage of fragmented nuclei as readout for apoptosis after treatment with the apoptosis-inducer staurosporine. Our results demonstrated that AnkG of the first group (AnkG NM ) inhibited apoptosis, AnkG of the second group (AnkG Soyta ) did not affect apoptosis induction, and AnkG from the third group (AnkG F3 ) displayed apoptosis-promoting activity (Fig. 2c). Importantly, without staurosporineinduction, AnkG F3 did not display any pro-apoptotic activity (data not shown). As the anti-apoptotic activity of AnkG NM requires nuclear localization, which is mediated by binding to p32-and importin-α1 32,33 , we analyzed the interaction of the AnkG variants with the host cell proteins p32 and importin-α1. We hypothesized that AnkG Soyta and AnkG F3 bind both host cell proteins, as the binding of AnkG NM to importin-α1 was mapped to the amino acids surrounding residue 11, while the amino acids neighbouring residues 22 and 23 had been identified as the binding region for the interaction with p32 32,33 . Indeed, as shown in Figs. 2d and 2e, all three AnkG variants were able to bind to p32 and importin-α1. The interaction of the HA-AnkG variant with Flag-tagged importin-α1 is not mediated indirectly by the anti-FLAG beads, as shown by us before 33 . Next, we analyzed the subcellular localization of the AnkG variants, as nuclear localization of AnkG NM was demonstrated to be essential for its anti-apoptotic activity 32 . AnkG NM displays vesicular staining with close association with host cell mitochondria in healthy cells. After cellular stress, AnkG NM migrates to the host cell nucleus. In contrast, AnkG NM 1-69 was present within the host cell nucleus under all conditions 32 . Thus, we hypothesized that without cell stress induction the AnkG variants from the second and third groups might also localize to the nucleus, which was, in fact, the case (Fig. 2f). Taken together, although AnkG Soyta bound to p32 and importin-α1 and was translocated to the nucleus, it was unable to inhibit apoptosis. AnkG F3 was also found in the nucleus and bound to p32 and importin-α1, but displayed staurosporine-dependent pro-apoptotic activity in contrast to the antiapoptotic activity of AnkG NM .
Characterization of AnkG Soyta activity. AnkG Soyta represents the second group and forms a 10.4 kDa protein. While the first 83 amino acids are identical to AnkG NM , the residues 84 -92 differ. Although AnkG Soyta bound to p32 and importin-α1 and displayed nuclear localization, it was unable to inhibit apoptosis. To investigate the underlying reason(s), we first constructed truncations of AnkG Soyta by stepwise removal of amino acids from its C-terminus (Fig. 3a). We ectopically expressed GFP or GFP-AnkG Soyta variants transiently in CHO cells and demonstrated stable expression by immunoblot analysis (Fig. 3b). Next, we measured apoptosis after stimulation with staurosporine by quantifying nuclear fragmentation, which was visualized by DAPI staining. As shown in Fig. 3c, stepwise removal of amino acids from the C-terminal region in AnkG Soyta restored its antiapoptotic activity. This suggests that the C-terminal region in AnkG Soyta disturbs the activity of the protein. We hypothesized that this might have been caused by the length of the protein. To address this experimentally, we constructed a truncated AnkG NM 1-92 to clarify whether the length of the protein might be the cause for the lost activity of AnkG Soyta , which also contains 92 amino acids. As shown in Fig. 3d, all constructs were ectopically expressed following transient transfection of CHO cells. AnkG NM 1-92 inhibited staurosporine-induced apoptosis as efficiently as full-length AnkG (Fig. 3e), indicating that the reduced length is not the cause for the missing anti-apoptotic activity of AnkG Soyta . These results suggest that the amino acids 84-92 of AnkG Soyta might influence the folding or intra-or intermolecular binding capacity of the protein.
Characterization of AnkG F3 activity. AnkG F3 represents the third group and forms a 6 kDa protein.
While the first 28 amino acids, except for amino acid 11, are identical to AnkG NM , the amino acids 29 -51 are different. AnkG F3 bound to p32 and importin-α1 and translocated into the nucleus, but it showed pro-apoptotic www.nature.com/scientificreports/ activity after apoptosis-induction. To shed light on the underlying reason, we created different AnkG truncations (Fig. 4a). We constructed a truncated variant AnkG F3 1-28 to clarify whether the first 28 amino acids, which are identical between AnkG F3 and AnkG NM , are sufficient for anti-apoptotic activity. Additionally, we constructed AnkG NM 1-51 to clarify whether the length of the protein might result in a change of its activity. First, we ectopically expressed GFP or the GFP-AnkG variants transiently in CHO cells and demonstrated stable expression by immunoblot analysis (Fig. 4b). As shown in Fig. 4c, while GFP-AnkG F3 significantly increases staurosporine-induced apoptosis, GFP-AnkG F3 1-28 inhibited apoptosis similarly to GFP-AnkG NM . This demonstrates that AnkG F3 contains the anti-apoptotic domain, which comprises the first 28 amino acids. Expression of GFP-AnkG NM 1-51 also resulted in inhibition of apoptosis, indicating that the length of AnkG F3 , which contains 51 amino acids, cannot explain its pro-apoptotic activity. These findings indicate that the amino acids 29-51 of AnkG F3 negatively affect the anti-apoptotic activity of the N-terminal region.
To determine whether the amino acids 1-28 are necessary and sufficient for the anti-apoptotic activity, we ectopically expressed GFP, GFP-AnkG NM , GFP-AnkG NM 1-28 or GFP-AnkG  transiently in CHO cells and determined cell death after staurosporine treatment. The expression of GFP-AnkG NM 1-28 significantly inhibited staurosporine-induced apoptosis. In contrast, the expression of GFP-AnkG NM 29-338 did not protect the cell from apoptosis (Fig. 4d). These data suggest that the amino acids 1-28 are necessary and sufficient for the antiapoptotic activity of AnkG.
Analyzing transcription of the ankG gene from the three different groups. The ankG gene from the first group is transcribed and translated as a full-length protein. The ankG alleles from the second and third groups contain a premature stop codon leading to the translation of a 10 and 6 kDa protein, respectively. To determine whether the ankG variants from the second and third groups encode additional proteins containing AnkG amino acid sequences of three C. burnetii isolates -each representatives for its group. The first group contains the reference strain Nine Mile II (AnkG NM ) and nineteen additional strains expressing a 338 amino acid protein (blue). The second group includes thirteen isolates and is represented by C. burnetii Soyta (AnkG Soyta ) expressing a 92 amino acid protein. AnkG Soyta is identical in the first 83 N-terminal amino acids (red), but harbors 9 different amino acids at the C-terminus (yellow) compared to AnkG NM . The third group contains four isolates and is represented by C. burnetii strain F3 (AnkG F3 ) expressing a 51 amino acid protein. AnkG F3 has an amino acid exchange at position 11 (isoleucine to leucine) and is otherwise identical to AnkG NM in the first 28 N-terminal amino acids (green), but contains 23 different amino acids at the C-terminus (yellow). The asterisk (*) at the C-terminus of the amino acid sequence representing the second and third group indicates premature truncation. www.nature.com/scientificreports/ www.nature.com/scientificreports/ the C-terminal translocation signal, we inspected the ankG sequences for the presence of additional CDS within the original full-length transcript. Indeed, ankG contains three further start codons, which are in-frame with the stop codon at the C-terminus. These start codons are located at slightly varying base pair positions in the three ankG alleles near bp 337, 532 and 598 in the respective alleles ( Fig. 5a). In order to determine whether the C-terminal fragments are transcribed, we analyzed the mRNA from C. burnetii isolates belonging to the three different groups. As representative for the first group, we chose Nine Mile I (NMI) and for the second and third groups 19/34 and Z3574-1/92, respectively. First, we had to establish axenic growth of these isolates, as axenic media does not support growth of all C. burnetii strains 35 . While growth of NMI in ACCM-2 and ACCM-D has been shown, growing 19/34 and Z3574-1/92 in axenic media has never been attempted. The strain 19/34 grew similarly well in ACCM-2 as NMI. However, Z3574-1/92 was only able to grow in ACCM-D media. As C. burnetii has different growth kinetics in ACCM-2 and ACCM-D media 36 , which might influence transcriptional activity, we always used NMI grown in the respective media as control. Thus, we isolated RNA from 19/34 and NMI grown in ACCM-2 and RNA from Z3574-1/92 and NMI grown in ACCM-D. After treatment with DNase, we confirmed the lack of DNA contamination by performing a no reverse transcriptase control using specific primers for the dotA gene (Fig. 5b). Next, we reverse transcribed the mRNA into cDNA and determined whether the RNA fragment encoding the N-terminal residues of the three different ankG groups was transcribed. As shown in Fig. 5c, the base pairs 3-150 were transcribed in all three groups, suggesting that the 5´part of the mRNA can be transcribed and translated into the respective proteins. Next, we analyzed if the full-length ankG open reading frame of all three groups is transcribed into the corresponding mRNA. This is indeed the case for all three groups (Fig. 5d). Thus, the 3´ part of the ankG mRNA from the second and third groups is transcribed and might be translated into a protein containing the C-terminal translocation signal. Still, although these proteins might have the potential to be injected into the host cell via the T4BSS, they most likely lack anti-apoptotic activity due to missing amino acids 1-28.
Role of AnkG Soyta and AnkG F3 during a C. burnetii infection. During a C. burnetii infection AnkG NM is transported into the host cell nucleus, the site of its anti-apoptotic activity 33 . While we do not fully under- www.nature.com/scientificreports/ stand the mechanism by which an effector protein is recognized for delivery through the T4BSS, we know that the C-terminal end of the effector serves as the translocation signal 37 . In contrast to AnkG NM , AnkG F3 and AnkG Soyta lack the C-terminal end and should therefore not be translocated through the T4BSS. However, it was demonstrated that AnkG NM , AnkG NM 1-69 as well as AnkG NM 70-338 are translocated by the Legionella pneumophila T4BSS 26 . This indicates that AnkG NM might contain two translocation signals, one in its C-terminal part and another one within the first 69 amino acids. Therefore, AnkG Soyta and/or AnkG F3 might also be recognized as an effector protein and translocated into the host cell by the T4BSS. To determine if this is the case, we transformed C. burnetii with plasmids containing 3xFlag-tagged AnkG F3 or AnkG Soyta . Single C. burnetii pFlag-AnkG F3 or pFlag-AnkG Soyta clones were isolated and transgene expression was analyzed (data not shown). Mouse embryonic fibroblasts (MEFs) were infected with C. burnetii pFlag-AnkG NM , C. burnetii pFlag-AnkG F3 and C. burnetii pFlag-AnkG Soyta under inducing conditions and the localization of Flag-AnkG was analyzed by confocal imaging. As shown in Fig. 6a, Flag-AnkG NM was associated with the bacteria and localized within the host cell nucleus, confirming previous results 33 . This indicates that Flag-tagged AnkG NM can be translocated by C. burnetii into the host cell. In contrast, Flag-AnkG Soyta and Flag-AnkG F3 were only associated with the bacteria and not present in the host cell. This indicates that neither AnkG Soyta nor AnkG F3 are T4BSS effector proteins.
To support this assumption, we fused AnkG NM , AnkG Soyta and AnkG F3 to the calmodulin-dependent adenylate cyclase toxin (Cya). CHO cells were infected with C. burnetii producing either Cya-AnkG NM , Cya-AnkG Soyta or Cya-AnkG F3 and the amount of cyclic AMP (cAMP) was measured. As Cya catalyzes the production of cAMP its amount correlates with the translocation of Cya-tagged AnkG variants into the cytoplasm. A concentration of 2.5-fold more cAMP than the negative control (Cya alone) was regarded as proof for translocation of the protein.
Cya-AnkG NM was translocated into the host cell cytosol (Fig. 6b), supporting previous results 26 . In contrast, Cya-AnkG Soyta and Cya-AnkG F3 were not translocated into the host cell cytoplasm. This data supports the hypothesis that neither AnkG Soyta nor AnkG F3 are T4BSS effector proteins. www.nature.com/scientificreports/  www.nature.com/scientificreports/ AnkG nM attenuates pathology during C. burnetii infection in Galleria mellonella. AnkG NM is translocated into the host cell during infection and has anti-apoptotic activity. Whether AnkG is important for the C. burnetii infection in vitro and more importantly in vivo is unknown. Recently, the Galleria mellonella infection model was used to analyze C. burnetii infection in vivo Galleria mellonella is susceptible to infection by C. burnetii NMI and NMII 24,27,[38][39][40] , suggesting that the pathogenicity of C. burnetii is independent of the phase variation in this infection model. Death of the larvae occurred in a dose-dependent manner and depended on a functional T4BSS 38 . Furthermore, this in vivo infection model was successfully used to assess the contribution of the T4BSS effector CvpB/Cig2 during infection 24,27 . Therefore, we used the Galleria mellonella model to determine the role of AnkG during an in vivo infection. To assess virulence, a lethal dose of 10 6 C. burnetii NMII wildtype (WT), C. burnetii NMII ΔdotA (ΔdotA), C. burnetii pFlag-AnkF (AnkF) or C. burnetii pFlag-AnkG (AnkG) per larvae was injected in the upper right proleg and larval survival was monitored for 7 days. Larvae injected with PBS solution served as a control for the injection stress, the ΔdotA mutant for the activity of the T4BSS and AnkF-expressing C. burnetii for the overexpression of an effector protein. 90% of the larvae receiving PBS solution survived, whereas 100% of the larvae infected with wildtype bacteria had died by day 7 post-injection. In contrast, the majority (~ 95%) of larvae infected with C. burnetii ΔdotA survived (Fig. 7a). This result confirms the importance of the T4BSS for C. burnetii virulence. All larvae infected with C. burnetii pFlag-AnkG survived till day 6 post-injection while all larvae infected with C. burnetii pFlag-AnkF had died by this time point (Fig. 7a). To verify that the observed pro-survival phenotype of larvae infected with C. burnetii pFlag-AnkG is due to increased expression of AnkG, we performed qRT-PCR to determine the expression level of AnkG in C. burnetii pFlag-AnkG either induced with IPTG or not induced. As shown in Fig. 7b, induction with IPTG resulted in increased expression of AnkG. The expression level decreased over time, but at seven days post-induction it was still significantly higher than in non-induced bacteria. This supports the assumption that the increased translocation of overexpressed AnkG into the host cells mediates improved vitality of the C. burnetii infected larvae. To exclude that this phenotype is due to a reduced infection or/and replication ability of the bacteria expressing pFlag-AnkG, we determined the ability of C. burnetii to infect and replicate in the larvae by qRT-PCR and immunofluorescence. Only C. burnetii ΔdotA was unable to replicate in the larvae. In contrast C. burnetii WT, Flag-AnkG and Flag-AnkF replicated in the larvae around 100-fold from day 1 to 5 (Fig. 7c).
In agreement with this result, we observed that at day 5 post-infection, the majority of hemocytes infected with WT, Flag-AnkG-or Flag-AnkF-expressing C. burnetii contained replicative CCVs. This was not the case in hemocytes infected with C. burnetii ΔdotA (Fig. 7d). These results demonstrate that a functional T4BSS is essential for intracellular replication, as demonstrated before 38 . In addition, these data showed that C. burnetii expressing Flag-AnkG replicates in hemocytes. Next, the infection rate of hemocytes was quantified (Fig. 7e). All bacteria exhibited a similar percentage of infection at day 1 post-infection. At day 5 post-infection, the infection rate had increased when infected with WT, Flag-AnkG-or Flag-AnkF-expressing C. burnetii, but not when infected with the C. burnetii ΔdotA mutant. In the latter case, the rate of infection stayed nearly constant. Thus, C. burnetii expressing Flag-AnkG have no defect in the ability to infect hemocytes, to establish a replicative CCV and to spread. We concluded from these data that exogenous AnkG mediates increased survival of C. burnetii infected larvae. How AnkG boosts survival of the infected larvae is currently unknown, but it cannot be explained by attenuated bacterial fitness of Flag-AnkG-expressing C. burnetii. These data suggest that AnkG might be an important virulence factor. In order to start investigating the role of AnkG for pathogenesis in vivo, we infected larvae with either wild type C. burnetii or a ΔankG mutant and monitored their survival over 8 days.
Larvae infected with the ΔankG mutant were significantly attenuated in virulence, as demonstrated by improved survival rates (Fig. 7f). However, the underlying reason for this attenuation in virulence will be investigated in future studies. Thus, whether there is a direct or indirect link between the anti-apoptotic activity of AnkG, which might be responsible for the survival of larvae infected with Flag-AnkG-expressing C. burnetii, and the attenuation of the ΔankG mutant has to be unraveled.

Discussion
Comparison of genomes from different C. burnetii isolates pointed to a considerable heterogeneity in the repertoire of T4BSS effector proteins 41 . Noticeable variation was detected in the sequences of the C. burnetii T4BSS effector CaeA derived from 25 different C. burnetii isolates 31 . CaeA was classified in seven different genotypic groups. In the current study, we focused on another effector (AnkG) and analyzed the AnkG nucleotide and amino acid sequences from the same 25 C. burnetii isolates in addition to 12 further isolates. These 37 different isolates were divided into three groups according to their respective ankG sequences (Fig. 1a,b). To get a better impression of the heterogeneity of the ankG sequence, we compared the ankG sequences from 57 complete, scaffold and draft genomes of C. burnetii strains (Tab. S1). These strains encode nine alleles of the ankG gene. Genome groups I, IIb and III 42 express the wild type sequence of the Nine Mile reference strain (group 1 in this study). Genome group IIa encodes a truncated protein of 92 residues due to a 2 bp frameshift deletion mutation at codons L83N84 and corresponds to AnkG Soyta (group 2 in this study). Genome group IV representatives all contain the Ile to Leu exchange at residue 11 and a 1 bp frameshift insertion mutation at Gly29, leading to a truncated protein with 51 residues and correspond to AnkG F3 (group 3 in this study). Three subgroups (designated 3, 3b and 3c in Tab. S1) contain different additional mutations in the remaining 3′ section of ankG. Genome groups V and VI both encode a full-length protein with 2 either amino acid substitutions at residues 11 (Ile to Leu) and 72 (Gly to Glu) in genome group V, while genome group VI contains just the mutation at amino acid position 11. There are only three exceptions to this sequence/genome group correlation. The Guyana strain, assigned to genome group I, has a frameshift insertion mutation at codon Gly294, leading to a protein of 327 residues with 33 altered amino acids at its C-terminus. The Cb185 strain, classified as genome group IIa, has the wild type sequence otherwise www.nature.com/scientificreports/ present in the genome groups I, IIb and III. The strain Cb109, a member of genome group IIb, has a frameshift mutation at residue Asn287, resulting in a protein of 299 residues with 13 altered amino acids at its C-terminus. Interestingly, only one C. burnetii isolate analyzed in this study, encodes AnkG and CaeA, namely C. burnetii Nine Mile. Nine isolates encode only the effector protein AnkG and four isolates encode only the effector protein CaeA. Thus, the majority of the strains analyzed encode none or only one of the two anti-apoptotic effector proteins AnkG and CaeA. There seems to be no correlation between the presence of a functional AnkG and/or CaeA effector protein and the geographic distribution of the isolate and its host species. However, this interpretation needs to be taken with caution, as the majority of the isolates were collected from infected ruminants in Germany. Importantly, AnkG and CaeA are not the only C. burnetii effector proteins involved in the regulation of host cell viability. So far, two others have also been characterized in more detail: CaeB was shown to inhibit intrinsic www.nature.com/scientificreports/ apoptosis 23 and IcaA was described to prevent pyroptosis 21 . However, it can be expected that other members of the so far identified ~ 150 effector proteins might also inhibit host cell death 43 . In line with this assumption, a large-scale transposon mutagenesis coupled to a high-content multi-phenotypic screening in the laboratory of Matteo Bonazzi identified transposon mutants that exhibit a cytotoxic phenotype 7 . This suggests that the genes affected might also contribute to preventing host cell death. Under certain conditions C. burnetii also induces apoptosis 44 or pyroptosis 45 , which might be due to the activity of pro-apoptotic effector proteins that still await identification and characterization. Therefore, the balance between bacterial anti-and pro-apoptotic effector proteins will determine the overall impact of C. burnetii on the host cell. In any case, the absence of AnkG and/ or CaeA or the expression and translocation of the C-terminal part of AnkG might result in altered pathogenesis of C. burnetii. Interestingly, the three different groups of AnkG exhibited different cellular activities (Fig. 2c), although they all bound to the host cell proteins p32 (Fig. 2d) and importin-α1 (Fig. 2e), which was shown to be a prerequisite for AnkG-mediated apoptosis inhibition. Additionally, ectopically expressed AnkG Soyta and AnkG F3 are localized within the host cell nucleus, the site of protein activity (Fig. 2f). Furthermore, AnkG NM , AnkG Soyta and AnkG F3 contain the same first 28 amino acids, which were shown to be necessary and sufficient for anti-apoptotic activity (Fig. 4c, d). The reason for the different cellular activity of the three AnkG variants might be altered protein folding or the inability to interact with themselves or other proteins 46,47 . However, further research is required to identify the underlying reason(s) for the different activities.
To better understand bacterial pathogenesis, appropriate infection models are essential. The Galleria mellonella infection model has been recently used to study pathogenic Gram-positive and Gram-negative bacteria 24,27,38,[48][49][50][51][52] . Importantly, several of these studies showed a good correlation between the G. mellonella and mammalian infection models. For example, the essential role of the type III and type IV secretion systems in virulence was also demonstrated in the G. mellonella infection model 38,49,51 . In order to examine the role of specific virulence factors in an in vivo model system, this factor might either be overexpressed or deleted. Here, we used C. burnetii NM overexpressing Flag-tagged AnkG 33 . Importantly, C. burnetii overexpressing Flag-tagged AnkG showed reduced killing of infected G. mellonella (Fig. 7a), although infection and replication were not reduced in comparison to wildtype bacteria (Fig. 7c, d). It can be only speculated how AnkG increased survival of the infected larvae; AnkG might adjust the host immune reaction, possibly by preventing destruction of hemocytes via its anti-apoptotic activity. Hemocytes are important phagocytic immune cells and depletion of circulating hemocytes upon infection correlates with G. mellonella mortality 53 . Similarly, the addition of AnkG to the repertoire of Legionella pneumophila effector proteins resulted in host tolerance to infection by preventing rapid pathogen-induced apoptosis in dendritic cells 26 .
Taken together, we have demonstrated that AnkG is an anti-apoptotic effector protein, which might mediate host tolerance to a C. burnetii infection in the G. mellonella model of infection. In addition C. burnetii lacking AnkG are attenuated in the G. mellonella infection model, indicating that AnkG might be an important virulence factor. Furthermore, we have further narrowed down the anti-apoptotic domain to amino acids 1 to 28 and showed that the effector function is not conserved in all C. burnetii isolates. Indeed, our analysis suggests that C. burnetii belonging to the genome group IIa and IV do not express a functional AnkG effector protein.

Methods
Reagents, cell and bacterial strains. Unless otherwise noted, chemicals were purchased from Sigma Aldrich as described in Schäfer et al. 33 . Complete Protease inhibitor cocktail mixture and Xtreme Gene 9 transfection reagent were purchased from Roche. Staurosporine was bought from Cell Signaling. Cell lines were cultured at 37 °C and 5% CO 2 in media containing 10% heat-inactivated fetal bovine serum (Biochrom). CHO (Chinese hamster ovary) fibroblasts were grown in minimal essential medium alpha medium (Invitrogen), HEK293T (human embryonic kidney) and MEF cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen). E. coli strains DH5α and BL21(DE3) were cultivated in Luria-Bertani (LB) broth supplemented with kanamycin or ampicillin if indicated. PCR, sequencing and sequence analysis. PCR, sequencing and sequence analysis were done similar to as described earlier in Bisle et al., 2016. Therefore, PCR primers (Table 2)   www.nature.com/scientificreports/ Table 2), purified with DyeEx 96 plates (Qiagen) and electrophoresed on a 3130XL Genetic Analyzer (Applied Biosystems). Sequence analysis and polymorphism evaluation was performed using Lasergene 9.0 (DNASTAR, Madison, WI) and MEGA5 54 . Nuclear fragmentation assay. Nuclear fragmentation assays were performed as described previously 23,31 .

Isolation of genomic
In brief, CHO cells were plated on coverslips in 24-well dishes at a density of 2.5 × 10 4 cells per well. After overnight incubation, cells were transfected with the plasmids indicated. Eighteen hours post-transfection, the cells were incubated with staurosporine (2 mg/ml) for 4 h at 37 °C and 5% CO 2 . The cells were fixed with 4% paraformaldehyde (Alfa Aeser) in PBS (Biochrom) for 20 min at room temperature, permeabilized with ice-cold methanol for 30 s, quenched with 50 mM NH 4 Cl (Roth) in PBS for 15 min at room temperature. The cells were mounted using ProLong Diamond with DAPI (ThermoFisher) to visualize the nucleus.
Immunoblotting. As described in Schäfer et al. 33 , proteins were separated by SDS/PAGE and transferred to a PVDF membrane (Millipore). The membranes were probed with antibodies directed against GFP (Life Technologies) or HA (Covance Research Products). A chemiluminescence detection system (Thermo Scientific or Millipore) was used to visualize the proteins marked by HRP-conjugated secondary antibodies (Dianova).

Co-immunoprecipitation.
A modified protocol from Schäfer et al. 33  Determination of transcript variations of AnkG. RNA from different AnkG strains was isolated as described in "Quantification of AnkG Level". For DNase digestion and cDNA synthesis, the RNA was diluted to a 10 ng/µl final concentration. The following strains were used: NMI as member of the group one, 19/34 as member of the second group, and Z3574-1/92 as member of the third group (Table 1). To exclude DNA contamination, a no reverse transcriptase control was performed using 2 µl RNA as template in a PCR reaction with primers 760/761 (dotA). PCR for different ankG sections was performed using the DreamTaq DNA polymerase kit (ThermoScientific). Per sample, 25 µl of PCR reagents were used containing 18.8 µl of DNase-free H 2 O, 2.5 µl of DreamTaq Green Buffer, 0.5 µl of dNTPs, 0.5 µl of each Primer (10 µM), and 0.2 µl of polymerase. The amplification was run with the following parameters: 2 min at 95 °C; 30 cycles, each consisting of 30 s, 95 °C, 30 s at 58 °C, and 1 min 72 °C; final elongation of 10 min at 72 °C. Primers were used for the 5´, N-terminus encoding part of the mRNA (a817/ a818) and for the full length product (a817 and a822) ( Table 3).
Amplified ankG fragments were analyzed using a 2% agarose (Bio&Sell) gel containing 0.004% midori green (Nippon genetics). The gels were run for 30 min with 120 V and were analyzed using UV light in a gel documentation chamber.
Axenic cultivation of C. burnetii and infection of MEFs. The following experimental procedures were described before 33 . In brief, C. burnetii Nine Mile phase II (NMII), transformed with the respective plasmid, was propagated for 6 days in 75 cm 2 tissue culture flasks at 37 °C and 5% CO 2 , 2.5% O 2 in 30 ml acidified citrate cysteine medium (ACCM-2), which supports host cell-free (axenic) growth of C. burnetii 55  pFlag-AnkF were grown in 15 ml of axenic ACCM-2 with or without 3 µg/ml chloramphenicol in 25 cm 2 cell culture filter flasks at 37 °C, 95% CO 2 and 2.5% O 2 for 5 days. One day before infection, 2 mM IPTG was added to the C. burnetii pFlag-AnkG and C. burnetii pFlag-AnkF cultures to induce expression of 3xFlag-AnkG and 3xFlag-AnkF. At the day of infection, C. burnetii cultures were pelleted and resuspended in PBS at 5 × 10 4 /µl. Galleria mellonella larvae, purchased from TruLarv (UK), were randomized by size into groups of 10 larvae on wood chips. For infection, 20 µl of PBS or C. burnetii-containing solutions were injected into the uppermost right leg of the larvae. The larvae were incubated at RT for 7 days with survival monitoring every 24 h. The larvae were regarded as dead when they were not able to move or appeared black.
Quantification bacterial load/larvae. At day 1 and day 5, two infected Galleria mellonella larvae were frozen at -80 °C. On the same day all larvae were disrupted using a BeadRuptor and lysed with Proteinase K overnight. The next day genomic DNA of C. burnetii was isolated with the Qiagen DNeasy Blood and Tissue Kit and an RT-PCR was performed to determine the bacterial load. For the fold replication analysis, the genome equivalents of day 5 larvae were compared to the genome equivalents of day 1 larvae. The experiment was performed three times.
Immunofluorescence of infected hemocytes. Hemocytes from 3 Galleria mellonella larvae infected with either C. burnetii, C. burnetii ΔdotA, C. burnetii pFlag-AnkG or C. burnetii pFlag-AnkF were collected at 1 day and 5 days post infection 38 . Cells were seeded and centrifuged on poly-l-lysine-coated coverslips, washed twice with PBS, fixed with 4% paraformaldehyde (Alfa Aeser) in PBS (Biochrom) for 20 min at room temperature, permeabilized with 0.1% Triton-X 100, followed by quenching and blocking with 50 mM NH 4 Cl (Roth) in PBS/5% goat serum (Life Technologies) for 30 min at room temperature. The coverslips were incubated with primary antibodies directed against C. burnetii and actin (phalloidin labeled with Alexa Fluor 674) diluted in PBS/5% goat serum for 20 min at room temperature, washed three times with PBS and further incubated with secondary Alexa Fluor labelled antibodies Alexa 488 diluted in PBS/5% goat serum for 20 min. After three washes with PBS, the cells were mounted using ProLong Diamond with DAPI to visualize cell nuclei and bacterial DNA. Confocal fluorescence microscopy was performed using a Carl Zeiss LSM 700 Laser Scan Confocal Table 3. Plasmids used in this study. *Primer numbers are as in Table 4. www.nature.com/scientificreports/ Microscope. In Z-Stack images with 1 µm distance 100 randomized cells were counted and determined whether they were infected or not. The experiment was performed three times.
Targeted DotA deletion in C. burnetii NMII. Targeted gene deletion of the NMII gene dotA was performed as described previously by homologous recombination with a suicide plasmid 56 . Briefly, 1 × 10 10 electrocompetent C. burnetii NMII were transformed with the plasmid harboring 2 kb 5′ and 3′ flanking region of genomic dotA CDS flanking a P Com1 -CAT cassette (BioRad Gene Pulser XCell, 18,000 V/cm 2 , 500 Ω, 25 µF in 0.1 cm gap cuvettes). Subsequently, transformed NMII were cultivated axenically in 6 ml 1xACCM-2 with 3 µg/ ml chloramphenicol and/or 350 µg/ml kanamycin at 37 °C and 2.5% O 2 atmosphere in T25 flasks for 6 days. Afterwards, cultures were passaged every 7 d for 3 weeks in 6 ml medium supplemented with chloramphenicol (CM) and kanamycin in T25 flasks. Identification of plasmid co-integrants was performed by colony PCR with primers as proposed 56 . Co-integrants were then subcultured in 3 ml medium supplemented with CM and 1% sucrose in 6-well plates for 4 d, followed by two passages for 6 d in 6 ml medium supplemented with CM in T25 flasks until late stationary phase. Identification of dotA knock-out (ΔdotA) strains was performed by colony PCR with specific primers 56 . Clonal isolation of ΔdotA strains was performed through plating on semi-solid agar plates.
burnetii grown in ACCM-D medium were electroporated with 10 µg pJB-CAT-Cya-AnkG Soyta , pJB-CAT-Cya-AnkG NM or pJB-CAT-Cya-AnkG F3 . Transformants were selected by culturing the bacteria in ACCM-D medium containing 3 µg/ml chloramphenicol for 5 days, plated onto ACCM-D/0.25% agarose plates supplemented with 3 µg/ml chloramphenicol. Single clones were picked after 7-10 days as previously described 56 . CHO cells were infected with the different C. burnetii mutants in a 24 well plate at an MOI of 200 for 3 days at 37 °C and 5% CO 2 . The translocation assay was performed by measuring the concentration of cAMP in cell lysates using the cAMP enzyme immunoassay (GE Healthcare) as previously described 19 . The threshold was set as 2.5-fold more cAMP than the controls (uninfected cells and cells infected with C. burnetii expressing CyaA alone) 25 .
For creation of the constructs pCMV-HA-AnkG Soyta and pCMV-HA-AnkG F3 , the genes were amplified from pEGFP-AnkG Soyta and pEGFP-AnkG F3 using the primers listed in Table 4, restricted with BglII and NotI, and ligated with likewise-restricted pCMV-HA.
The construct pEGFP-AnkG 1-28 was produced using the primers in Table 4. As template pEGFP-AnkG NM was used and AnkG 1-28 coupled to GFP was amplified and ligated in a pEGFP backbone lacking GFP. The restriction enzymes NdeI and BamHI were used for directed cloning.
Generation of a C. burnetii ΔankG mutant. C. burnetii Nine Mile phase II were electroporated with 10 µg pJC-CAT::ankG-5´3´-lysCA as previously described 58 . Co-integrants were selected by culturing the bacteria in ACCM-D media lacking lysine, but containing 2% sucrose for 4 days as previously described 57 . Surviving transformants were expanded in ACCM-D media lacking lysine for 7 days. After spreading the diluted culture on 0.25% ACCM-D agarose without lysine clonal ΔankG mutants were picked after 7 days of culture. The picked clones were expanded in ACCM-D media without lysine. Quantification of the ankG expression level. RNA was isolated using the TriFast Reagent (VWR) according to the manufacturer's instructions and treated with DNase (RNase-Free DNase Set, Qiagen). Therefore, 1,000 ng RNA were digested using 2 µl RDD Buffer and 1 µl DNase (ad 20 µl). cDNA synthesis was performed with SuperScriptII reverse transcriptase (Thermo Scientific) as recommend by the manufacturer using specific primers for ankG and dotA. Subsequent quantification by qRT-PCR was performed using the SybrGreen qPCR Mix (Thermo Scientific) with primers for ankG (65/66) and dotA (737/738) for normalization (Table 4).

Statistical analysis. An unpaired Student´s t-test was used for statistical analysis.
Received: 3 September 2019; Accepted: 26 August 2020