A MicroRNA Network Controls Legionella pneumophila Replication in Human Macrophages via LGALS8 and MX1.

Cases of Legionella pneumophila pneumonia occur worldwide, with potentially fatal outcome. When causing human disease, Legionella injects a plethora of virulence factors to reprogram macrophages to circumvent immune defense and create a replication niche. By analyzing Legionella-induced changes in miRNA expression and genomewide chromatin modifications in primary human macrophages, we identified a cell-autonomous immune network restricting Legionella growth. This network comprises three miRNAs governing expression of the cytosolic RNA receptor DDX58/RIG-I, the tumor suppressor TP53, the antibacterial effector LGALS8, and MX1, which has been described as an antiviral factor. Our findings for the first time link TP53, LGALS8, DDX58, and MX1 in one miRNA-regulated network and integrate them into a functional node in the defense against L. pneumophila.

(Part of this work is included in the doctoral thesis of C. E. Herkt at Philipps University Marburg, Marburg, Germany, in 2018.)

RESULTS
miRNAs are differentially expressed in blood-derived macrophages after L. pneumophila infection and alter bacterial replication. In order to characterize the miRNA-mediated regulatory network upon Legionella infection, human blood-derived macrophages (BDMs) were infected with L. pneumophila at a multiplicity of infection (MOI) of 0.25 for 24 and 48 h, and miRNA expression changes were determined by means of small RNA sequencing with respect to control samples collected at both time points (deposited under GEO accession number GSE125559). Comparative analyses revealed that the expression of 85 miRNAs changed significantly as a function of infection and time (Fig. 1A). Most miRNAs were found to be differentially regulated 48 h after treatment in comparison to both control and 24-h expression values, while only four miRNAs were found to be significantly deregulated after 24 h compared to control (see Table S1 in the supplemental material). Among the 38 upregulated miRNAs, miR-146a and miR-155 were identified, which are known to play a crucial role in the inflammatory TLR response after bacterial infection (28)(29)(30). Further upregulated miR-NAs that have not yet been characterized in the context of infection include, e.g., miR-3196 or miR-4284. Of the 47 downregulated miRNAs, miR-653, miR-548a, and miR-671 displayed the highest fold changes, but since no described function in the immune response for these miRNAs are described in the literature, they were not further analyzed.
Among the significantly downregulated miRNAs miR-221 and miR-125b were of particular interest, since they have been implicated in posttranscriptional control of monocytic immune tolerance to bacterial components (31). These miRNAs were validated by quantitative PCR (qPCR) in BDMs from six independent donors ( Fig. 1B and D) and in the THP-1 cell line ( Fig. 1C and E), which yielded comparable results. Additional testing of cytokine secretion and gene expression (see Fig. S2 in the supplemental material) supported the notion of THP-1 cells as a suitable model for BDMs in further experiments to study L. pneumophila-regulated miRNAs in macrophages.
The promoter regions of miR-155 and miR-146a (upregulated) and of miR-221 and miR-125b (downregulated) were investigated using our own previously published H4 acetylation chromatin immunoprecipitation (ChIP) sequencing data (GEO accession number GSE80214) from Legionella-infected BDMs (32). The promoter regions of miR-155 and miR-146a revealed an increased H4 acetylation upon L. pneumophila infection at an MOI of 10 for 1 h, with linear fold changes of 1.9 and 1.3, respectively, while miR-221 and miR-125b, mainly the paralog hsa-miR-125b-2 on chromosome 21, showed a decreased H4 acetylation at their promoters with linear fold changes of 0.77 and 0.54, respectively ( Fig. 2A).
As these results suggested transcriptional regulation of miRNA expression in L. pneumophila infection, as opposed to changes in miRNA processing, the pri-miRNA expression levels of miR-146a, miR-155, miR-125b, and miR-221 were analyzed by qPCR. pri-mir-16 served as an endogenous control (33,34). Upregulation of pri-mir-155 and -146a ( Fig. 2B and C), as well as downregulation of pri-miR-221 and -125b ( Fig. 2D and E), was observed after infection with L. pneumophila, which led us to conclude that regulation of these miRNAs upon infection occurs at the transcriptional level. In summary, our miRNA profiling data show, among several key immune-associated miRNAs, the downregulation of immune-tolerance-associated miRNAs miR-221 and miR-125b during Legionella infection.
Immune-responsive miRNAs facilitate intracellular Legionella replication. Previously, miR-221 and miR-125b, along with a third miRNA, miR-579, have been reported to confer immune tolerance of monocytic cells to bacterial components (31,35). We therefore hypothesized that the observed downregulation of this miRNA pool (see Fig. S1 in the supplemental material) might be involved in promoting antimicrobial defense. In a pilot set of experiments, the impact of miR-125b, miR-221, and miR-579 miRNAs Control MX1 and Bacterial Growth ® on the intracellular replication of L. pneumophila was determined. To this end, THP-1 cells were transfected with a pool of synthetic miR-125b, miR-221, and miR-579 precursors or inhibitors and were subsequently infected. Intracellular replication of L. pneumophila was monitored at 24, 48, and 72 h. Although treatment with the miRNA inhibitor pool resulted in reduced intracellular replication compared to a scramble control (scr), transfection of all three miRNA precursors resulted in a strong increase of intracellular replication. Individual transfection of a single miRNA precursor/inhibitor did not change replication of L. pneumophila (Fig. 3A). To further test assay validity, we verified Legionella replication in THP-1 cells with concomitant scramble RNA transfec-FIG 1 Differentially expressed miRNAs in blood-derived macrophages after L. pneumophila infection. BDMs from three different donors were infected with L. pneumophila at an MOI of 0.25 or left untreated as a control (ctr). Samples were taken 24 and 48 h postinfection. RNA was isolated and used for library preparation and enriched for small RNAs using a TruSeq small RNA kit. Sequencing was performed in a multiplexed run of 1 ϫ 51 cycle ϩ7 (index). (A) Heatmap of the miRNA log 2 fold change over the respective controls. The expression of distinct upregulated (B and C) and downregulated (D and E) miRNAs in BDMs and THP-1 cells after L. pneumophila infection was validated by qPCR using TaqMan assays and are displayed as log 2 fold changes. Boxes show the upper and lower quartiles with medians and whiskers indicating minimal and maximal values of five to nine independent biological replicates for dysregulated miRNAs. Paired t tests were performed. *, P Յ 0.05; **, P Յ 0.01; ***, P Յ .001; ****, P Յ 0.0001 (compared to the corresponding control). tion for up to 72 h (Fig. S3). These observations point to a concerted action of all three miRNAs in pathogen replication control in macrophages.
MX1 is involved in miRNA-dependent control of Legionella replication. To address the underlying mechanism, THP-1 cells were transfected with the miRNA precursor pool and additionally infected with L. pneumophila at an MOI of 0.5 for 24 and 48 h or left untreated as a control. As posttranscriptional regulation of coding genes commonly must also be evident on the protein level to be functional, the proteome patterns of these samples were comparatively profiled by mass spectrometry (MS) (Fig. 3B). This method comparatively profiles differences in protein levels based on isotope-labeled amino acid incorporation. Quantitative data were obtained for 1,670 proteins; of these, 113 proteins displayed reductions in protein level in response to transfection with the miRNA precursor pool in comparison to scr-transfected cells with (24 h ϩ L. pneumophila and 48 h ϩ L. pneumophila) or without infection (24 h -L. pneumophila) by L. pneumophila. The impact of the transfection with the miRNA precursor pool or in combination with L. pneumophila infection is illustrated in Fig. 3C with regard to these 113 proteins. Of note, miRNA precursor transfection alone or in combination with L. pneumophila infection triggered reductions in the level of MX dynamin-like GTPase 1 (MX1), as well as RPL13A, RPL14, and KRT2, in comparison to the respective controls. In addition, the levels of LGALS8 (galectin-8) were reduced following miRNA precursor pool transfection in cells infected with L. pneumophila for 48 h and in uninfected cells. To elucidate the miRNA-mediated influence on L. pneumophila replication, the focus was turned to MX1 and LGALS8, given their documented role in infection control.
To validate the results of the proteome profiling, intracellular MX1 was investigated by flow cytometry. After transfection with the miRNA precursor pool, we observed a reduced MX1 signal under all tested conditions ( Fig. 3D; see also the exemplary histogram in Fig. S4). Of note, mRNA levels of MX1 remained unaffected. In order to evaluate a potential direct effect of MX1 on infection, THP-1 cells or BDMs were FIG 2 Infection-related chromatin changes on miRNA promoters. The alterations of the acetylation pattern at the miRNA promoter regions of miR-155, miR-146a, miR-221, and miR-125b after L. pneumophila infection in BDMs were investigated. BDMs were infected with L. pneumophila for 1 h, and ChIP was performed using a pan-ac-H4-antibody (based on data published by Du Bois et al. [32]). (A) After sequencing, the enrichment of H4 acetylation at the miR promoters was analyzed. As confirmation for active sites in the promoter region, IP data for H3K4me3 and H3K27Ac provided by the Encode UCSC browser are shown. Horizontal dark green lines indicate the genomic locations of the promoter regions (2,000 bp) for the four analyzed miRNAs corresponding to given coordinates. Circles highlight the mature transcript of each miRNA. The expression of the pri-miRs (B, C, D, and E) in BDMs after L. pneumophila infection (MOI of 0.25 or 0.5) was examined by qPCR and is displayed as log 2 fold changes. Boxes show the upper and lower quartiles, with medians (when n ϭ 3 independent biological replicates) and whiskers indicating minimal and maximal values (when n ϭ 4 independent biological replicates). Paired t tests were performed. *, P Յ 0.05; **, P Յ 0.01; ***, P Յ 0.001 (compared to the corresponding control).
The expected miRNA mode of action prompted us to investigate the 3= untranslated region (UTR) of MX1 for binding sites of the miRNA pool (miR-125b, miR-221, and miR-579). In silico analysis by TargetScan Human 7.1 revealed no significant binding sites for any miRNA. Manual search for seed-compatible regions in the 3= UTR was equally unsuccessful. Qiagen Ingenuity Pathway Analysis (IPA) yielded a complex network of predicted and published targets of each miRNA. Filters were set to only include experimentally observed or high-confidence predicted miRNA/mRNA interaction partners. The newly found miRNA targets were interconnected using all available data sources in IPA with restriction to experimentally validated or high-confidence predicted interactions in humans (Fig. 5). Positive regulators of MX1 (blue) were prioritized over negative regulators (red), since miRNAs usually downregulate their targets. This filter strategy left 4 of 13 possible candidates for MX1 regulation mediated by the miRNA pool, two of which (gamma interferon [IFNG] and interferon ␣2 [IFNA2]) were discarded due to very low expression in THP-1 cells. The two remaining candidates were DExD/H-box helicase 58 (DDX58), also known as RIG-I, and tumor protein P53 (TP53). Intriguingly, the nucleic acid sensor DDX58 is a known regulator of host defense against Legionella infection (36). In silico analysis (TargetScan Human 7.1) of the DDX58 3= UTR revealed one binding site for miR-579 and two binding sites for miR-221 ( Fig. S5). Transfection of miR-221 alone and of the miRNA precursor pool resulted in decreased DDX58 mRNA expression. In comparison, transfection of miR-579 precursor alone did not result in reduced DDX58 mRNA expression (Fig. 6A). Analysis of miR-221 and miR-579 targeting of the DDX58 3= UTR (DDX58 WT ) by a luciferase reporter assay revealed significant functional interaction at 72 h after cotransfection of miR-221, miR-579, or the miRNA precursor pool (Fig. 6B). Binding specificity was tested by mutation of the putative miRNA binding sites (DDX58 mut ) and was positively ascertained for miR-221. The specificity of the miR-579 binding to the putative target site could not be confirmed and was hence disregarded.
Upon small interfering RNA (siRNA)-mediated DDX58 knockdown, MX1 protein levels were at 75% compared to scramble transfected cells as determined by flow cytometry (Fig. 6C). siDDX58-treated BDMs were infected with L. pneumophila at an MOI of 0.5 to assess replication at 24, 48, and 72 h postinfection. Successful knockdown was verified via qPCR ( Fig. 6D) and resulted in an up to 150% replication of L. pneumophila in BDMs (Fig. 6E), which was also shown in differentiated THP-1 cells (Fig. S6).
TP53, the second candidate to link miRNA precursor transfection to MX1, has previously been validated as a target of miRNA-125b by several studies (37)(38)(39). As expected, transfection of miR-125b precursor alone or the miRNA precursor pool reduced TP53 expression in THP-1 cells (Fig. 7A). Accordingly, transfection of miR-125b or the miRNA-pool with the partial 3= UTR of TP53 (TP53 WT ) containing the predicted miRNA binding sites resulted in a significant decrease of relative luciferase activity compared to scramble transfected cells in a luciferase-based reporter assay (Fig. 7B), which could be reversed by targeted mutation of the putative binding sites (TP53 mut ). CFU assays following efficient siRNA-mediated knockdown of TP53 in BDMs (Fig. 7C) revealed 157% replication at 72 h postinfection compared to the scramble control (Fig. 7D). The effect of TP53 on MX1 has already been described (40).  LGALS8 is targeted by miRNA-579 and influences L. pneumophila replication. As shown, all three miRNAs were necessary to mediate replication of L. pneumophila in macrophages, but our model did not yet include a role for miR-579. Since the protein LGALS8 is known to be involved in antibacterial defense (26,27,41) and displayed decreased levels upon miRNA transfection (Fig. 3C), LGALS8 was further investigated. First, the influence of miR-579 on LGALS8 mRNA was assessed. The individual transfection of miR-579 precursor resulted in significantly reduced levels of LGALS8, while transfection with the miRNA pool, which carries reduced per-miRNA concentrations compared to individual transfection, showed no influence (Fig. 8A). Accordingly, we could demonstrate a reduction in LGALS8 protein levels upon miR-579 transfection by flow cytometry (Fig. S7).
Second, the influence of miR-579 binding to the 3= UTR of LGALS8 was investigated with a luciferase assay. Cotransfection of miR-579 or the miRNA-pool with the wild-type version of the 3= UTR fragment of LGALS8 (LGALS8 WT ) revealed a significant decrease of relative luciferase activity (Fig. 8B). The transfection of miR-125b and miR-221 also led to a significantly reduced luciferase signal, but less efficiently compared to the miR-579 transfection. The transfection of the mutated version (LGALS8 mut ) with miR-579 yielded a significantly weaker decrease of relative luminescence compared to the wild-type 3= UTR. A significant decrease was detected if miR-125b or miR-221 were cotransfected with LGALS8 mut compared to LGALS8 WT . In summary, miR-579 reduced the LGALS8 3= UTR-dependent luciferase activity, and this effect was reversed by the mutation of the miRNA seed region, illustrating a direct binding of the miR-579 to the 3= UTR of LGALS8. In order to elaborate on the role of LGALS8 in the infection process, an siRNA knockdown of LGALS8 with subsequent CFU assay was performed. Transfected cells were infected with L. pneumophila at an MOI 0.5 for 24, 48, and 72 h. Successful LGALS8 knockdown in BDMs was verified via qPCR (Fig. 8C) and was constant over time FIG 5 MX1 is not directly targeted by the miRNA-pool. IPA was performed to investigate the connection between the miRNA pool (miR-125b, miR-221, and miR-579) and the MX1 protein. Filters were set to only include experimentally observed or high-confidence-predicted miRNA-mRNA interaction partners. Output was limited to 13 total candidates. The newly found miRNA targets were interconnected using all available data sources in IPA with restriction to experimentally validated or high-confidence-predicted interactions. Direct or indirect relationships between molecules are indicated by solid or dashed lines, respectively. Blue lines depict positive regulations, while red lines indicate negative regulations. Molecule classes (cytokine, enzyme, ion channel, ligand-dependent nuclease, phosphatase, transcription factor, other, or mature miRNAs) are indicated by distinct symbols.
irrespective of L. pneumophila infection. In BDMs, L. pneumophila replication was significantly increased in cells transfected with siLGALS8 (up to 150% at 72 h postinfection) (Fig. 8D). This increased L. pneumophila replication was also confirmed in differentiated THP-1 cells (Fig. S6). The major findings of this study are summarized as a schematic model in Fig. 9.

DISCUSSION
The present study describes for the first time the global miRNA regulation in human primary macrophages in response to L. pneumophila infection. Furthermore, it identifies a new miRNA-regulated cell-autonomous immune network containing LGALS8 and the antiviral factor MX1, which restricts L. pneumophila replication in human macrophages. We found this restriction to be mediated by the repressive action of miR-125 and miRNAs Control MX1 and Bacterial Growth ® miR-221 on MX1 and of miR-579 on LGALS8. Downregulation of these miRNAs during infection by decreased histone H4 acetylation might therefore contribute to enhanced bacterial clearing (Fig. 9).
We identified 85 differentially expressed miRNAs in primary human macrophages infected with L. pneumophila. Exemplarily, the miRNAs miR-146a, miR-155, miR-27a, and miR-125a were upregulated in a time-dependent manner, whereas miR-221, miR-222, miR-125b, miR-26a, and miR-29b were downregulated in response to L. pneumophila infection. We have shown before that miR-146a can facilitate Legionella replication inside the host cell via degradation of IRAK1 mRNA (42). In the present study, miR-125b, miR-221, and miR-579 were of particular interest. While miR-579, unlike miR-125b and miR-221, was only weakly detected in our data set, we considered all three miRNAs important because of their documented role in conferring macrophage tolerance to endotoxin (31,35) and because of their downregulation at 48 h postinfection versus control cells (see Table S1 in the supplemental material). Of note, we see a significant increase in Legionella CFU count upon miRNA overexpression at the same time point (Fig. 3).
Transcriptionally unrelated to miR-125b, the miR-125 family member miR-125a has been observed to be upregulated in response to infection with M. tuberculosis, L. monocytogenes (43,44), and also L. pneumophila, as we have previously shown (45). miR-221 and miR-222 are encoded as a cluster located on chromosome X and have been reported to be overexpressed in many types of cancer (46). In addition, it has been shown that miR-221 expression was selectively decreased in lipopolysaccharide (LPS)- treated macrophages, as well as in the lungs of LPS-challenged mice (47). These observations are in line with our observation of a downregulation of miR-125b and miR-221 in response to a Legionella infection, and we extended our findings from BDMs to THP-1 cells. Other studies also confirmed that the regulation of miRNAs such as miR-146a/b, miR-132, miR-155, miR-214, and miR-195a, was comparable in THP-1 and primary macrophages (48), supporting our conclusion that the miRNA profile in THP-1 LGALS8 is targeted by miR-579 and influences L. pneumophila replication. THP-1 cells were transfected with either the miR-579 mimic or the miRNA mimic pool (miR-125b, miR-221, and miR-579) at a final concentration of 30 nM. As a control, a scrambled precursor (scr) was transfected. At 24 h posttransfection, THP-1 cells were activated with PMA for another 24 h and then infected with L. pneumophila at an MOI of 0.5. (A) LGALS8 expression was examined with qPCR and is displayed as the log 2 fold change. Luciferase reporter assay analyses were performed in HEK-293T cells. The plasmid contained either the wild-type (LGALS8 WT ) or the mutated (LAGLS8 mut ) version of the 3= UTR of LGALS8. (B) Ratios of Renilla and firefly luciferase luminescence were normalized to the vector without insert. BDMs were transfected with an siRNA pool targeting LGALS8 (siLGALS8) or with a scrambled siRNA as a control (scr) and infected at an MOI of 0.5 or left untreated as a control. (C) Downregulation of LGALS8 expression with siRNA was verified by qPCR and is displayed as the log 2 -fold change. (D) CFU of Legionella were determined 24, 48, and 72 h postinfection after a knockdown of LGALS8. Boxes show the upper and lower quartiles, with medians (when n ϭ 3 independent biological replicates) and whiskers indicating minimal and maximal values (when n ϭ 4 or 5 independent biological replicates). Two-way ANOVA with Sidak's correction was performed. *, P Յ 0.05; **, P Յ .01; ***, P Յ 0.001; ****, P Յ 0.0001 (compared to the corresponding control). #, P Յ 0.05; ##, P Յ 0.01; ###P Յ 0.001 (compared to wild type) (B). ##, P Յ 0.01 (for global treatment effects) (D).

FIG 9
Scheme of MX1 and LGALS8 regulation by miRNA precursor pool transfection to influence L. pneumophila replication in macrophages. Precursor transfection of miR-125b reduces the expression of TP53, while DDX58 (RIG-I) is targeted by miR-221. Both targets further regulate the protein expression of MX1, which has an antibacterial effect on Legionella. Furthermore, miR-579 acts on LGALS8 to reduce its expression. Thus, reduction of MX1 and LGALS8 expression leads to an increased replication of Legionella in macrophages. In conclusion, these miRNAs are influencing the expression of their targets and have an impact on Legionella replication. miRNAs Control MX1 and Bacterial Growth ® cells seems to be largely comparable to that of primary BDMs. Since the response of differentiated THP-1 cells to a Legionella infection is similar to that of primary cells with respect to key inflammatory mediators, THP-1 cells were used as cell line model instead of BDMs when necessitated by experimental upscaling.
In order to understand the mechanism of miRNA regulation in response to L. pneumophila infection, transcriptional changes of miRNAs were correlated with a genomewide histone analysis of H4 acetylation after L. pneumophila infection of human primary macrophages. Acetylation of histones is one of the master regulators of gene expression in mammalian development and human disease (49) and can strongly impact inflammatory and host defense responses (50). We have previously shown that activating acetylation upon Legionella infection depends on recognition of flagellin by the host (51). The correlation of selected pri-miR expression with the acetylation level of the respective promoter region following infection with L. pneumophila supported the notion that the investigated miRNAs are regulated in macrophages in response to a L. pneumophila infection by chromatin-associated events. As shown previously, Legionella can influence the histone code of the host by secreting effector proteins to repress gene expression (52). We therefore aimed to establish whether the regulation pattern of these miRNAs is of beneficial or deleterious consequence for the host.
It has been previously established that miR-125b, miR-221, and miR-579 are instrumental in the process of endotoxin tolerance (31,35,53). Reversal of the observed downregulation of this set of miRNAs postinfection by transfection caused an increased replication of L. pneumophila, whereas enhanced downregulation of all three miRNAs led to a decreased replication of L. pneumophila. A combination of all three miRNAs was necessary for this effect. Upon confirmation of a probacterial role of miR-125b, miR-221, and miR-579, we attempted to identify downstream effector molecules. In a proteomics approach, MX1 protein was found to be downregulated after transfection of all three miRNA precursors. Given its established role in restricting viral replication (54)(55)(56)(57)(58)(59)(60)(61), it was further investigated. The human gene MX1 encodes a 78-kDa protein called MX1 or MxA (62,63) and belongs to and shares many properties of the dynamin superfamily of large GTPases (64). MX1 proteins appear to assemble into long filamentous structures at physiological salt concentrations and are detectable by electron microscopy. In the presence of guanosine nucleotides, the filaments rearrange into rings that tubulate negatively charged liposomes and associate with the smooth ER (65,66), hypothetically functioning in intracellular protein transport and sorting (67). The nonconstitutive expression of MX1 is tightly regulated by type I and II IFNs (68). The restrictive effect of type I IFNs on Legionella growth in epithelial-like cell lines and macrophages has already been documented (69)(70)(71)(72). Although it has been previously described that interferon-related genes, including MX1, are upregulated following transfection of random 23-mer double-stranded miRNA mimics (73), we did not observe an increase of MX1 expression in the present study. The antiviral activities of human MX1 are mediated by the direct interference with, for instance, synthesis and/or nuclear import of newly synthesized viral components (74,75). In our study, a knockdown of constitutive MX1 led to enhanced intracellular bacterial replication. It is known that L. pneumophila manipulates various key host regulatory pathways in order to establish an LCV. The LCV is decorated with many bacterial and host factors to prevent bacterial degradation and enable intracellular replication. Proteomics screens have identified several small and large GTPases that are associated with the LCV. It was demonstrated that these GTPases are functionally involved in the LCV formation, as well as in the Legionella replication process (76,77). The secreted Legionella effector protein Lpg1137 binds and cleaves syntaxin 17 during intracellular replication, leading to an altered activity of the large GTPase Drp1 (78). We extended this observation to MX1, since siRNA-or miRNA-mediated downregulation led to increased Legionella replication. Because MX1 could not be found in the proteome of LCVs within U937 macrophages infected with L. pneumophila strain AA100/130b (79), we hypothesize that MX1 is not directly attached to the LCV.
While we ruled out a direct effect of miR-125b, miR-221, and miR-579 on MX1 mRNA, Ingenuity Pathway Analysis revealed DDX58 and TP53 as possible functional targets, which was confirmed by luciferase assay and knockdown experiments. The gene DDX58 encodes the cytosolic RIG-I-like receptor (RLR) RIG-I, which recognizes cytosolic double-stranded RNA or single-stranded RNA containing 5=-triphosphate. An activation of RIG-I triggers a large-scale amplification of a signaling cascade leading to the expression of type I IFN-dependent genes and induction of proinflammatory cytokines (80). In our study, DDX58 knockdown led to reduced intracellular MX1 protein. Furthermore, the targeting of DDX58 by miR-221 was demonstrated. The involvement of miR-221 in the type I IFN response has been established to occur via the RIG-I/MDA5 pathway (81), and DDX58 could be the as-yet-missing miR-221-responsive molecule limiting the viral replication by inducing the type I IFN response. It has been shown that L. pneumophila secretes the factor SdhA, which suppresses the RIG-I/MDA5 pathway (36), suggesting the importance of this pathway to influence Legionella replication. It is possible that the direct inhibition of RIG-I by miR-221, as well as the inhibited activation of MX1 by RIG-I, is contributing to the replication effect observed in this study. TP53, the other established miRNA-responsive controller of MX1, is a direct target of miR-125b (37)(38)(39) and is a master regulator of the cellular mechanisms controlling responses to cellular stress such as DNA damage, aberrant oncogene activation, loss of normal cell-cell contacts, nutrient deprivation, and abnormal reactive oxygen species production. Importantly, it is also induced in response to viral infection as a downstream transcriptional target of type I IFN signaling (82). In addition, TP53 is also activated indirectly by type I IFN through other IFN-inducible proteins, such as promyelocytic leukemia protein, STAT1, or IFIX␣-1 (83)(84)(85). Several studies showed that TP53 has a role in controlling bacterial infection and that inhibition of TP53 may confer certain selective advantages to bacteria (86). An inhibition of TP53 activity was observed for infection with H. pylori (87,88), different Mycoplasma species (89), and Chlamydia trachomatis (90). Chlamydia and Legionella are intracellular bacteria, and they strictly rely on host resources. Since metabolic control of TP53 provides antibacterial protection, inhibition of TP53 seems to be important for bacterial survival and growth, as seen in this study. Of note, the association of TP53 and MX1 in IFN signaling has already been shown. An increased expression of TP53 is accompanied by enhanced expression of MX1, as well as RIG-I and IRF7, which strengthens its implication in the type I IFN signaling (40). In summary, counteracting the type I IFN response by DDX58 and TP53 knockdown led to an increased replication of L. pneumophila, which in our model is due to the downstream effector molecule MX1.
In an attempt to find a mechanism behind the crucial, yet unexplained role of miR-579, pathway analysis revealed the putative targets DDX58 and LGALS8 that might be effectors of MX1 regulation. Galectin-8 (LGALS8) was detected by mass spectrometry to be downregulated following transfection of all three miRNA precursors. Expression of LGALS8 was significantly reduced following miR-579 precursor transfection, and this interaction was validated by luciferase assay, establishing for the first time a targeting of LGALS8 by miR-579. Luciferase assay results suggest that more, as-yet-unidentified miR-579 binding sites might exist. Galectins are glycan-binding, evolutionary conserved proteins and have pleiotropic roles in innate and adaptive immune responses (91,92). Galectins are localized in the cytosol and in the nucleus, and several of these galectins are also secreted (91). Within the cytosol, galectins prevent the formation of complex carbohydrates. Thus, they function as danger and pattern recognition receptors. For instance, galectin-3 accumulates on damaged bacterium-containing vesicles (93), notably including the LCV (94). Galectin-3 has been established as a marker for vacuole integrity during infection with S. enterica serovar Typhimurium, Shigella flexneri, and Trypanosoma cruzi (93,95). It was shown that the cytosolic galectin-8 also localizes to damaged vesicles and restricts Salmonella replication in HeLa cells (27). In addition, as shown in our study ( Fig. 8C and D), an siRNA-mediated knockdown of galectin-8 led to an increased replication of L. pneumophila. While galectin-8 was suggested to serve as a versatile receptor for vesicle damage (27), galectin-3 and galectin-8 have been demonstrated to support localization of antimicrobial guanylate binding proteins miRNAs Control MX1 and Bacterial Growth ® (GBPs) to bacterial vacuoles (96). GBPs, in turn, have a known role in the defense against Legionella (97,98). Thus, the absence of galectin-8 upon knockdown could explain the enhanced replication of Legionella (27). An association of galectin-8 with the LCVs of infected human macrophages at 6 h postinfection was previously validated, strengthening the putative role as a danger receptor (99).
In summary, L. pneumophila infection induced transcription-dependent downregulation of specific miRNAs. This leads to a derepression of MX1 and more efficient controlling of L. pneumophila replication, as suggested by miRNA overexpression, which increased bacterial replication. While the troika of miR-125b, miR-221, and miR-579 has already been described in the context of tolerance to bacterial components, the mechanism we propose here, centrally involving MX1, seems to be embedded in a different functional framework.
L. pneumophila culture. L. pneumophila strain Corby wild type or a gfp-expressing strain (kindly provided by K. Heuner, RKI, Berlin, Germany) was routinely grown on buffered charcoal-yeast extract (BCYE) agar plates at 37°C for 3 days. Infection was carried out as described before (42).
Cell culture and L. pneumophila infection. THP-1 cells (ATCC) and primary monocytes were cultured and differentiated as previously described (32). For all infection experiments, differentiated THP-1 cells or BDMs were infected with L. pneumophila strain Corby at an MOI of 0.25 or 0.5 or left uninfected as control. For SILAC (stable isotope labeling of amino acids in cell culture) experiments (100), THP-1 cells were cultivated in SILAC-DMEM with 2% (vol/vol) FCS lacking arginine and lysine. The medium was additionally supplemented with heavy-isotope-labeled [ 13 C]arginine and [ 13 C]lysine (Cambridge Isotope Laboratories, Tewksbury, MA), and THP-1 cells were cultured over five passages to reach a labeling efficiency of Ͼ97% for all proteins.
Transfection of macrophages. Lipofection of THP-1 cells or BDMs with ON-TARGETplus siRNA pools (Dharmacon, Lafayette, CO), targeting either MX1, DDX58, TP53, or LGALS8 mRNA was carried out according to the manufacturer's protocol (Lipofectamine 2000; Thermo Fisher Scientific, Waltham, MA). Transfection of THP-1 cells with miRNA mimics or inhibitors was performed with siPORT NeoFX (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. mirVana mimics and inhibitors and corresponding controls were purchased from Thermo Fisher Scientific.
CFU assay. To analyze bacterial replication in transfected THP-1 cells or BDMs, cells were infected with L. pneumophila at an MOI of 0.5. Cells were lysed at indicated time points with 1% saponin, and diluted lysates were streaked onto BCYE agar plates. After 3 days of incubation at 37°C, L. pneumophila colonies were counted, and the bacterial load was calculated.
miRNA promoter prediction and differential promoter acetylation. The putative active transcription start sites (TSS) of miR-155, miR-146a, miR-221, and miR-125b in macrophages, and their genomic locations were determined using PROmiRNA software (101). The software uses a machine learning approach to predict the TSS of primary miRNA transcripts (pri-miRNAs) based on the signal from CAGE data from the FANTOM4 project libraries (102) and sequence features such as GC content and TATA box elements. We used epigenetic data, in particular H4Ac from our previous studies (32) to identify significant acetylation changes at miRNA promoters upon infection.
Illumina small RNA sequencing and bioinformatics analysis. The miRNome of L. pneumophilainfected BDMs was determined by Illumina TruSeq small RNA sequencing (Illumina, San Diego, CA). BDMs of three different donors were infected with L. pneumophila at an MOI of 0.25 for 24 and 48 h or left untreated as control. miRNAs with an absolute log 2 -fold change of Ն0.65 (corresponding to a linear fold change of 1.5) and an adjusted P value of Յ0.2 (Benjamini-Hochberg correction) in at least one comparison were considered significantly regulated.
SILAC. Quantitative proteome analysis of infected THP-1 cells was performed as described previously (103). Details of this analysis are described in the supplemental information (Table S2).
Western blot analysis. Cellular proteins were harvested by lysis and protein concentration was determined by BCA assay (Thermo Fisher Scientific). For protein separation, 10% SDS gels were used with 20 to 30 g of protein per lane. Proteins were transferred to a nitrocellulose membrane with the use of a tank blot for 1 h at 100 V. Primary antibody (␣-MX1 or ␣-tubulin-antibody) was added at a 1:1,000 dilution, followed by incubation for 2 h at room temperature on a tumbling shaker. HRP-conjugated secondary antibody (1:2,000) was added for additional 2 h at room temperature on a tumbling shaker. After removal of excess antibody by washing, protein signal was detected on a bioluminescence and chemiluminescence imager (INTAS Science Imaging Instruments GmbH, Göttingen, Germany). When required, quantification of signal was performed by densitometric analysis, using the LabImage 1D software (Kapelan Bio-Imaging GmbH, Germany).
Flow cytometry. After consecutive transfection, differentiation, and L. pneumophila infection of THP-1 cells, the cells were detached with trypsin-EDTA, fixed with 4% paraformaldehyde (PFA), and permeabilized with 0.5% Tween 20, followed by blocking with 0.5% (vol/vol) Tween with 10% FCS and Human BD Fc Block (BD Bioscience, San Jose, CA). Primary ␣-MX1 antibody, ␣-LGALS8 antibody and ␣-IgG isotype control antibody were added for 1 h. Then, secondary anti-rabbit Alexa 555-labeled antibody was added for 30 min at ambient temperature in the dark. After a washing step, the cells were analyzed on a Guava easyCyte flow cytometer (Merck Millipore). Data were analyzed using FlowJo version 7.6.5. In order to account for signal arising from unspecific antibody binding, only signal above the IgG control intensity was considered.
In silico pathway analyses. The IPA tool (104) was used to investigate the connection between the miRNA pool (miR-125b, miR-221, and miR-579) and the MX1 protein. Detailed methods are given in the supplemental material (Text S1).
Ethical statement. All blood donors gave informed written consent for use of their blood samples for scientific purposes.
Statistics. Statistic evaluation was carried out with GraphPad Prism (version 7). For data with multiple variables, two-way analysis of variance (ANOVA) was performed, followed by either Sidak's or Tukey's comparison test. For comparison of two data columns, a paired t test was used on log 2 -transformed data. For all tests, a Gaussian distribution was assumed, and the confidence interval was set to 95%. P values of Յ0.05 were considered significant. All statistical tests are specified in the respective figure legends.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.04 MB.

ACKNOWLEDGMENTS
We thank Nadine Siebert (Philipps University Marburg, Germany) for excellent technical assistance. We are grateful to Marie Schwerdtner and Isabelle Gollin (Philipps University Marburg, Marburg, Germany) for support with cloning the mutated 3= UTR of LGALS8 into psiCHECK2 vector. We thank Mirjam Feldkamp for the support with the small RNA sequencing.
Parts of this work were funded by Bundesministerium für Bildung und Forschung miRNAs Control MX1 and Bacterial Growth ®