Phenolic Glycolipid Facilitates Mycobacterial Escape from Microbicidal Tissue-Resident Macrophages

1When M. tuberculosis (Mtb) is aerosolized into the lower lung, it first encounters lung-resident alveolar macrophages that patrol the air-lung epithelium interface. In the first few days post-infection, Mtb is found exclusively within alveolar macrophages. Thereafter, it traverses the lung epithelium to reside within other myeloid cells that have aggregated into granulomas. The difficulty of tracking the early fate of individual mycobacteria in traditional animal models has precluded elucidation of how mycobacteria move from alveolar macrophages into other cells and indeed how they survive these broadly microbicidal first responders.

2We have exploited the optical transparency of the zebrafish larva to study the early mycobacterium-phagocyte interactions by infecting Mycobacterium marinum (Mm), a close genetic relative of Mtb, into the zebrafish larval hindbrain ventricle, an epithelium-lined cavity. In this model, pathogenic mycobacteria manipulate host responses immediately upon infection so as to inhibit the recruitment of neutrophils and microbicidal monocytes, and instead recruit and infect mycobacterium-permissive myeloid cells. To avoid detection by microbicidal monocytes, mycobacteria mask exposed pathogen-associated molecular patterns (PAMPs) with the cell-surface phthiocerol dimycoceroserate (PDIM) lipid, thus preventing recognition of PAMPs by Toll-like receptors (TLRs). Mycobacteria thus inhibit monocyte signaling through TLRs, which would normally recruit prototypical microbicidal iNOS-expressing monocytes. In conjunction, pathogenic mycobacteria recruit growth-permissive monocytes using a PDIM-related surface lipid, phenolic glycolipid (PGL) that induces the host monocyte chemokine CCL2. CCL2 recruits mycobacterium growth-permissive monocytes through signaling via its cognate receptor CCR2. The recruitment of growth-permissive monocytes is critically important for the ability of mycobacteria to establish infection. PGL-deficient mycobacteria fail to recruit normal numbers of monocytes and their ability to establish infection is attenuated.

3However, mycobacteria still have to contend with resident macrophages that are thought to be the first phagocytes encountered during infection. Here we found that resident macrophages are default first-responders to invading bacteria, including mycobacteria, and phagocytosed them rapidly. These first-responding resident macrophages were microbicidal to virulent mycobacteria, and capable of eradicating infection unless the mycobacteria escaped into more permissive cells. We found that PGL rapidly induces the production of CCL2 in the resident macrophages via a Sting-associated pathway. CCL2 recruited CCR2+ monocytes to the close proximity of the infected resident macrophage. The bacteria then transferred from microbicidal resident macrophages into these monocytes, thus escaping into a growth-permissive niche to establish infection. Resident macrophage-mycobacterium interaction is thus possibly the earliest determinant of whether infection will be established or cleared, with PGL acting as a very early mycobacterial immune evasion determinant. Furthermore, our findings suggest that STING and CCL2 are host susceptibility factors that act at the very first steps of infection.

4When Mtb is aerosolized into mouse lung, it is found for the first few days exclusively within alveolar macrophages. In the zebrafish larva, directly posterior to the hindbrain ventricle infection site (Figure 1A), is the brain which, like most organs, has a population of resident macrophages. We asked whether these brain-resident macrophages or microglia, analogous to the resident macrophages of the mammalian lung, participated in the immune response to mycobacterial infection. In addition to their tissue-specific functions, tissue-resident macrophages, including those of the brain, play a central role in host defense against infection. Like lung-resident macrophages, brain-resident macrophages phagocytose Mtb and produce inflammatory cytokines in response to it.

5To distinguish between brain-resident macrophages and monocytes, we used the nuclear dye Hoechst 33342 that does not cross the blood brain barrier; injection of Hoechst 33342 into the caudal vein of zebrafish larvae labels cells, including myeloid cells, in the body but not in the brain. We injected Hoechst dye into the caudal vein and then injected wild-type Mm into the HBV 2 hr later (Figure 1A). Three hours following infection, recruited cells were identified as either brain-resident macrophages (Hoechst-negative) or peripheral monocytes (Hoechst-positive) (Figure 1A).

6Our prior work had shown that myeloid cell recruitment to the HBV was substantially dependent on bacterial PGL and host Ccl2-Ccr2. We asked whether recruitment of resident macrophages, monocytes, or both were dependent on these. Wild-type Mm recruited both resident macrophages and monocytes, whereas the PGL-deficient Mm strain (Δpks15) recruited resident macrophages but not monocytes (Figure 1B). Correspondingly in Ccr2-deficient animals, wild-type Mm recruited resident macrophages but not monocytes (Figure 1C). We asked whether resident macrophages in the zebrafish larvae arrived more rapidly to mycobacteria, similar to resident macrophages in the mammalian lung. A temporal analysis revealed that they were the first responders to infection and arrived independently of Ccr2 signaling (Figures 1D and 1E). In contrast, monocytes arrived later and in a Ccr2-dependent fashion (Figures 1D and 1E). Thus, similar to Mtb infection of the mammalian lung, Mm infection of the zebrafish HBV recruits both resident macrophages and peripheral monocytes. The two cell types appear to be recruited sequentially, and through distinct pathways—Ccr2-independent for resident macrophages and Ccr2-dependent for peripheral monocytes.

7We found that resident macrophages were also the first-responders in bacterial infections wherein overall myeloid cell recruitment is dependent on Toll-like receptor (TLR-MyD88) signaling rather than the CCL2-CCR2 axis, such as in the case of PDIM-deficient Mm (ΔmmpL7) and the mucosal commensal-pathogens Staphylococcus aureus and Pseudomonas aeruginosa (Figures 1F–1H). In addition to mononuclear phagocytes, S. aureus and P. aeruginosa elicited the early recruitment of neutrophils, which were distinguished from monocytes and macrophages using the transgenic lyz::EGFP zebrafish, through TLR-Myd88 signaling (Figures 1G and 1H). In all cases, resident macrophage recruitment was independent of TLR-Myd88 signaling, as they were still responding toward infection in Myd88-deficient fish (Figures 1F–1H). Thus, tissue-resident macrophages appear to be default first-responders to invading bacteria, even those that elicit a robust protective neutrophilic response, with their recruitment to bacteria being independent of the TLR-Myd88 pathway.

8We ruled out the possibility that mechanosensing of a foreign body at the infection site was driving resident macrophage recruitment by showing that neither resident macrophages nor monocytes were recruited to sterile beads (Figure 1I). To examine whether resident macrophage recruitment is mediated by bacterial signals, we assayed recruitment of resident macrophages to supernatants of cultures of Mm, S. aureus, and P. aeruginosa; supernatants from these bacterial cultures recruited resident macrophages (and in the case of the latter two, neutrophils) but not monocytes (Figures 1J–1L). Thus, tissue-resident macrophages are recruited in response to a secreted factor(s) produced by both Gram+ and Gram− bacteria as well as mycobacteria.

9For mycobacterial infection, our findings that resident macrophages are rapidly recruited through a PGL- and Ccl2-independent pathway followed by PGL- and Ccl2-dependent monocyte recruitment, led us to ask whether monocyte recruitment was dependent on resident macrophage recruitment. We first used zebrafish larvae depleted of myeloid cells (by morpholino-mediated inhibition of myeloid transcription factor pu.1 expression; ) and evaluated ccl2 expression following intravenous infection with PGL-competent Mm. Myeloid-deficient fish were unable to induce ccl2 consistent with myeloid cells being responsible for Ccl2 production in response to mycobacterial infection (Figure 2A). Next, to specifically determine whether resident macrophages could induce ccl2, we infected bacteria into the HBV and used in situ hybridization analysis with an antisense ccl2 RNA probe. At 1 hr post infection, when the recruited phagocytes comprise almost entirely resident macrophages (Figure 1E), ccl2-positive phagocytes were present, but only following wild-type Mm infection and not PGL-deficient Mm infection (Figures 2B–2D). Together, these data showed that resident macrophages, like peripheral monocytes, induce ccl2 in response to mycobacteria. This induction is PGL-dependent in both cases, suggesting the presence of a common activation program in both cell types.

10Next, to directly test whether resident macrophages are required for monocyte recruitment, we used zebrafish mutants in which colonization of the brain by resident macrophages is delayed due to a genetic mutation in colony-stimulating factor receptor 1 (CSF1R). Therefore, at the time of our recruitment assay (2 days post fertilization), csf1r−/− fish have normal numbers of circulating monocytes but very few resident macrophages (Figure 2E). Following wild-type Mm infection into the HBV of csf1r−/− fish, resident macrophage recruitment was decreased and delayed, consistent with the lack of available cells in the brain (Figure 2F). Importantly, monocyte recruitment was also markedly decreased, consistent with our hypothesis that resident macrophages mediate monocyte recruitment (Figure 2F). In conjunction with our earlier finding that mycobacterial PGL was also required for monocyte recruitment (Figure 1B), these findings supported a model where resident macrophages, recruited in response to generic bacterial signals, engulf the mycobacteria. Mycobacterial PGL then induces them to express Ccl2 that mediates monocyte recruitment. Because PGL is heat-stable, this model would predict that heat-killed PGL-expressing Mm would both induce Ccl2 and recruit monocytes. It did neither, suggesting that live PGL-expressing mycobacteria are required to recruit monocytes through Ccl2 induction in resident macrophages (Figures 2G and 2H). Notably, heat-killed bacteria did recruit resident macrophages (Figure 2H), consistent with the secreted factor responsible for resident macrophage recruitment being heat-stable.

11Our finding that peripheral monocytes were dependent on signals from resident macrophages to participate in mycobacterial infection was surprising, and we wondered whether this requirement was unique to PGL-expressing mycobacteria. To test this, we used PDIM-deficient Mm, which recruits monocytes through TLR-Myd88 signaling, not Ccl2. Csf1r−/− zebrafish recruited monocytes normally to PDIM-deficient bacteria (Figure 2I). Moreover, in contrast to wild-type mycobacteria, heat-killed PDIM-deficient Mm recruited monocytes (Figure 2J). These results suggested a passive detection of the surface-exposed TLR ligands of this mutant bacterium in contrast to an active recruitment process mediated through live PGL-expressing bacteria. A head-to-head comparison of the recruitment kinetics of wild-type and PDIM-deficient strains revealed earlier monocyte recruitment to PDIM-deficient bacteria (Figure 2K), consistent with their recruitment to this strain being independent of resident macrophages. In sum, resident macrophages specifically promote Ccl2-dependent monocyte recruitment in response to virulent mycobacteria, and this is dependent on mycobacterial PGL.

12Taken together, our findings suggest that heat-stable bacterial PAMPs of PDIM-deficient Mm trigger a program of microbicidal monocyte recruitment that is not dependent on resident macrophages. In contrast, when bacterial PAMPs are masked by PDIM, PGL-mediated recruitment of permissive monocytes is absolutely dependent on both resident macrophages and live bacteria, suggesting an active bacterial manipulation of these default first-responders.

13How might PGL induce Ccl2 in resident macrophages? Because PGL operated in the context of live bacteria, we wondered whether a cytosolic sensing pathway was involved. Activation of the cytosolic signaling pathway STING can induce CCL2, so we tested whether Sting was the intermediary in PGL-mediated Ccl2 induction. Sting depletion using a splice-blocking morpholino resulted in a lack of ccl2 induction in response to wild-type Mm in both peripheral monocytes (Figure 3A) and resident macrophages (Figures 3B and 3C). Consistent with the inability to induce ccl2 in resident macrophages, Sting-deficient animals had reduced monocyte recruitment to Mm (Figure 3D). The initial recruitment of resident macrophages in these animals was intact, consistent with the prior finding that it was PGL-independent. Importantly, Sting-deficient animals recruited monocytes normally to PDIM-deficient Mm confirming that their inability to elicit monocytes was specifically in the context of Ccl2-mediated and not Myd88-dependent monocyte recruitment (Figure 3E). Finally, our model would predict that like Ccr2 deficiency, Sting deficiency should compromise the ability of wild-type bacteria to establish infection. Mycobacterial infectivity can be stringently tested by infecting animals with very low inocula that resemble human infection; in the zebrafish we have developed an infectivity assay which determines how many animals remain infected 4–5 days after infection with 1–3 mycobacteria. Using this infectivity assay, we found that wild-type Mm had reduced infectivity in Sting-deficient animals (Figure 3F), similar to PGL-deficient bacteria in wild-type animals and wild-type bacteria in Ccr2-deficient animals.

14STING can induce CCL2 either through type I interferons (IFNs), or independently of them. We evaluated expression of the zebrafish type I IFNs, ifnΦ1-3, that are induced during viral infection of larvae and adults, promote an antiviral gene program, and are protective against viral infection. They were not induced appreciably at 3 hpi with wild-type Mm, and the minimal induction observed was not PGL-dependent (Figure 3G). As expected, ccl2 was robustly induced in a PGL-dependent fashion (Figure 3G). This lack of dependence of type I IFNs on STING activation was distinct from the two previously reported pathways by which mycobacteria activate STING either through bacterial c-di-AMP or bacterial nucleic acid. The latter of these requires the bacterial ESX-1 secretion system to permeabilize the bacterial phagosome in order to induce type I IFN that activates STING. Having ruled out the involvement of type I IFNs, we used functional studies to further rule out that STING activation of our pathway was ESX-1-dependent. If STING activation of CCL2 is reliant on ESX-1 induction of type I IFNs, then monocyte recruitment should be ESX-1-dependent. We found that it was not. ESX-1 mutant bacteria recruited both resident macrophages and monocytes normally to the initially infecting bacteria (Figure 3I). Consistent with this finding, ESX-1-deficient Mm established infection at wild-type levels (Figure 3J). Our prior work has found that ESX-1 partners with host MMP9 to accelerate macrophage recruitment to the forming granuloma. These new findings showed that initial macrophage recruitment occurs through a distinct mechanism—PGL-dependent activation of STING that directly induces CCL2. It is not surprising that this process is ESX-1 independent because of the timing of ccl2 induction (prior to 3 hr post infection) versus ESX-1-induced phagosome permeabilization which takes ∼24 hr. Whether PGL is directly sensed by STING or works through an intermediary remains to be determined. It also remains to be determined how PGL or its intermediary contacts the cytosolic signaling pathway. One possibility is through mycobacterial vesicles that can be secreted out of the phagosomes of infected macrophages. Formation of these vesicles requires bacterial viability but not ESX-1, both consistent with our findings.

15Human TB is thought to result from infection with only 1–3 bacteria. In the zebrafish, 1–3 Mm are sufficient to establish infection in the majority of zebrafish larvae provided that bacterial PGL and host Sting and Ccl2-Ccr2 are present; without these factors, infectivity is reduced (Figure 3F). Therefore, it was important to examine myeloid cell recruitment in response to these low inocula where the role of PGL and CCR2 is most relevant. To enable a detailed temporal analysis of the HBV by time-lapse confocal microscopy, we used mpeg::yfp or mpeg::tdtomato transgenic zebrafish with fluorescent myeloid cells, again using Hoechst dye to distinguish monocytes from resident macrophages. Imaging each animal every 10 min from 1–11 hr post infection, we quantified the total number of resident macrophages and monocytes occupying the HBV at each time point. We observed that resident macrophages arrived early whereas monocytes were rarely seen during this period (Figure 4A and Table S1). In contrast, even with these low inocula, both cell types were recruited early to PDIM-deficient mutants (Figure 4A). Accordingly, when we analyzed the phagocytosis event for each bacterium, we found that wild-type bacteria were phagocytosed only by resident macrophages whereas PDIM-deficient bacteria were phagocytosed by both resident macrophages and monocytes (Figure 4B).

16Previously, we had shown that the increased infectivity of PGL-competent bacteria is abrogated by Ccr2 deficiency. Now we had found that both PGL-competent and PGL-deficient bacteria are initially in resident macrophages that are recruited in a Ccr2-independent manner, with the critical difference between the two strains being whether there is subsequent recruitment of Ccr2-dependent monocytes or not. Taken together, the two findings suggested that these monocytes were responsible for the increased infectivity of PGL-competent bacteria. This could be because the monocytes comprised a more permissive niche into which the bacteria were transferring, or because their presence was modulating the microbicidal capacity of the originally-infected resident macrophages.

17In order to determine whether bacteria were being transferred to new cells, we had to image infection for the first several days. Continuous imaging of the infection site in the same animal for several days is precluded by photobleaching. So we devised a strategy where we divided the infected larvae into 14 groups, and imaged each group for one of consecutive 6 or 12 hr periods that together spanned 4.5 days of infection (Table S2). For wild-type bacteria, transfer events were observed starting at 54 hr and peaking in the 66–72 hr window (Figure 4C and 4E and Movie S1). These transfers were accomplished as follows (Movie S1): the infected resident macrophage was approached by an uninfected peripheral monocyte. The cells then converged for a period of time before separating again, with the bacteria now being associated with the peripheral monocyte. Transfer events were not observed for PGL-deficient infection in the 66–72 hr window. (Figure 4E and Table S2). Thus, PGL-deficient bacteria largely remained within resident macrophages longer than wild-type bacteria. Furthermore, we documented clearance events of PGL-deficient Mm by the initially infected macrophage (Figure 4D and Movie S2). In contrast, clearance events were not observed during wild-type Mm infection.

18To rigorously examine the kinetics of clearance in relation to the bacterial transfer events we had observed, we monitored ∼30 animals for bacterial clearance by imaging them once every 24 hr. Because wild-type bacteria only transfer into permissive monocytes starting at 54 hr, the differential clearance of wild-type and PGL-deficient bacteria should become apparent only after this time-point. This was the case (Figure 4F). In contrast, PDIM-deficient bacteria started to be cleared within 24 hr (Figure 4F) consistent with their recruiting microbicidal monocytes within 2 hr and being phagocytosed by them within 12 hr (Figures 4A and 4B).

19Imaging of these early mycobacterium-beneficial transfer events revealed they were distinct in their cellular morphology from subsequent intercellular bacterial transfer observed in the forming granuloma, which is dependent on the apoptotic death of the infected macrophage, the bacterial contents of which are engulfed by newly arriving macrophages. In contrast the PGL-dependent transfer event was characterized by movement of the “donor” infected resident macrophage until the time that it converged with the “recipient” peripheral monocyte (Movie S1). Because the ESX-1 locus promotes apoptosis of infected macrophages, our finding that ESX-1-deficient Mm were not compromised during early infectivity (Figure 3H), suggested that efferocytosis is not mediating this transfer event. To confirm this, we used the pan-caspase inhibitor QVD-OPH that reduces apoptotic cells substantially (7.2-fold) in the context of Mm infection of the zebrafish. QVD-OPH treatment did not reduce the early infectivity of Mm (Figure 4G), further suggesting that this transfer event is not dependent on efferocytosis. Rather, transfer was occurring between living cells, similar to the findings that, following intimate contact between macrophages in culture, intracellular Gram-negative pathogens can transfer between the two cells in a process known as trogocytosis.

20Together, these findings are consistent with the model that PGL-competent bacteria transfer into the permissive monocytes they recruit. Conversely, our finding that PGL-deficient mycobacteria have a more prolonged sojourn in resident macrophages in which they are cleared, suggests that resident macrophages are more microbicidal than Ccl2-recruited monocytes.

21Our findings linking increased time in the resident macrophage to increased bacterial killing suggested that resident macrophages are more microbicidal than Ccr2-recruited monocytes. To address this question, we took advantage of our finding that following infection with 1–3 bacteria, only resident macrophages harbor PGL-deficient bacteria for at least the first 4.5 days (Figure 4E). We found that in Pu.1 morphant animals lacking myeloid cells and therefore the resident macrophage niche they occupied at this stage, PGL-deficient bacteria were able to establish infection at wild-type levels (Figure 5A). These data further suggested that resident macrophages are microbicidal to PGL-deficient bacteria. We found that PGL-deficient infection resulted in more inducible nitric oxide synthase (iNOS)-positive cells than wild-type infection (Figure 5B). This was similar to the case of the PDIM-deficient mutant whose TLR-recruited monocytes express more iNOS than Ccl2-elicited monocytes (Figure 5B). However, since PGL-deficient Mm recruits only resident macrophages, the increased iNOS production must be coming from the resident macrophages—i.e., resident macrophages, like TLR-recruited monocytes, also produce more iNOS than Ccl2-elicited permissive monocytes following infection. If this were the case then following delivery directly to monocytes via caudal vein infection (Figure 1A), PGL-deficient bacteria should result in the same low number of infected iNOS-positive cells as wild-type bacteria, and they did (Figure 5B). PDIM-deficient infection induced more iNOS in the caudal vein also (Figure 5B), suggesting that myeloid cells responding to PDIM-deficient bacteria are more activated regardless of location. Finally, we showed that the increased iNOS expression in the resident macrophages contributed to their increased microbicidal activity, as it does for TLR-recruited monocytes—treatment of animals with the nitric oxide scavenger CPTIO increased the infectivity of PGL-deficient bacteria delivered into the HBV (Figure 5C). Together these results suggested that the reduced infectivity of PGL-deficient bacteria is due to their prolonged sojourn in resident macrophages. If so, then the infectivity of PGL-deficient bacteria should be restored when delivered directly to monocytes by intravenous infection. It was (Figure 5D), and this result further showed that mycobacterial PGL does not protect mycobacteria from the microbicidal activity of resident macrophages but rather promotes their escape into the more permissive monocytes. Both Ccr2-deficiency and Sting-deficiency, which produced the expected decrease in infectivity of wild-type Mm upon hindbrain ventricle infection, failed to do so when the bacteria were delivered directly to monocytes through caudal vein infection (Figures 5E and 5F). Together, these findings highlighted the role of STING and CCL2 as early host susceptibility factors that work by enabling recruitment of peripheral monocytes to sites of infection.

22Finally, we asked whether the 54–90 hr sojourn in resident macrophages was at all detrimental to wild-type bacteria. The infectivity assay we had used so far only assessed whether the animals had cleared the bacteria or not, and not the extent of bacterial growth in the animals that did not clear them. We now tested this following infection of animals with a single bacterium. We found twice as much bacterial growth in the caudal vein compared to the HBV (Figure 5G). Together these results show that resident macrophages are more microbicidal than the permissive monocytes to which the wild-type bacteria eventually gain access. Moreover, the resident macrophage plays a growth-restrictive role even to wild-type PGL-expressing bacteria during the truncated time period that they remain in it.

23Our prior work had shown that pathogenic mycobacteria establish infection by recruiting and infecting permissive monocytes while having specialized strategies to avoid recruiting microbicidal cells, neutrophils, and TLR-stimulated monocytes. The latter strategy requires that mycobacteria initiate infection in the lower lung, so as to avoid the TLR-stimulated microbicidal monocytes by the mucosal flora of the upper airway. The present work had now identified the resident macrophage as another default rapid first-responder microbicidal cell that mycobacteria cannot avoid even in the lower airways. It must therefore co-opt them into their escape strategy by inducing them to secrete CCL2. In terms of human relevance of our zebrafish findings, our findings that resident macrophages are more microbicidal than peripheral monocytes already had support from human studies: human alveolar macrophages have substantial mycobactericidal activity ex vivo, in contrast to peripheral blood monocytes which not only fail to kill mycobacteria but are growth-permissive. Moreover, consistent with our findings, the microbicidal activity of human alveolar macrophages is at least in part mediated by nitric oxide.

24Our model would further predict that human alveolar macrophages would rapidly produce CCL2 upon mycobacterial infection in a PGL-dependent fashion. To test this prediction, we performed a pilot experiment with human alveolar macrophages obtained by bronchoalveolar lavage. We infected them with either PGL-expressing or PGL-deficient Mm. CCL2 was induced in a PGL-dependent fashion at 60 min post-infection (Figure 6A and 6B, Donor 1). We then recruited 12 additional donors and infected their alveolar macrophages with PGL-expressing or PGL-deficient mycobacteria as well as with LPS (100 ng/ml), a known CCL2 inducer. LPS induced CCL2 (> 1.2 fold over uninfected) in 5 of 12 donors suggesting that the remaining were not capable of inducing CCL2 rapidly in response to a known inducer (Table S3). The LPS-nonresponding macrophages also did not induce CCL2 upon mycobacterial infection (Table S3). This nonresponsiveness is consistent with significant donor variation in human alveolar macrophage cytokine secretion after mycobacterial infection. Of the LPS-responding macrophages, four of five induced CCL2 upon mycobacterial infection and this response was PGL-dependent (Figures 6A and 6B, and Table S3). In order to see whether CCL2 induction occurred even earlier than 60 min, we had collected supernatants at 30 min. Only those donor alveolar macrophages that induced CCL2 in response to LPS and mycobacterial infection at the 60 min time point, did so at the 30 min time point (Figure 6C and Table S3). Again, CCL2 induction was PGL-dependent (Figures 6C and 6D). These experiments suggest that the rapid induction of CCL2 in human alveolar macrophages in response to mycobacterial infection is PGL-dependent.

25By tracking the dynamics and kinetics of the earliest myeloid cell responses in the first hours of mycobacterial infection, we found that tissue-resident macrophages are the first cells to come in contact with any infecting bacteria in response to a ubiquitous heat-stable secreted bacterial signal. Arriving to virulent mycobacteria, resident macrophages were rapidly infected and could subsequently eradicate infection. In turn, mycobacterium’s counterstrategy to circumvent this first-line host defense that it cannot evade was to engineer its escape from these cells.

26The PGL-STING-mediated pathway of ccl2 induction was shared by both resident macrophages and monocytes. It is interesting then that mycobacteria deploy PGL-CCL2-mediated recruitment only initially. The involvement of CCL2 in subsequent monocyte recruitment to the forming granuloma is less clear. CCR2-deficient animals are not compromised for granuloma formation (C.J.C. and L.R., unpublished data). Rather monocyte recruitment to the granulomas is dependent on another bacterial virulence determinant, the ESX-1 locus that induces monocyte recruitment through induction of the host matrix metalloproteinase 9. Intercellular bacterial transfer in the granuloma requires the apoptotic death of a highly-infected macrophage that is then engulfed by multiple new recruits so as to expand the bacterial niche. Therefore, this mechanism of granuloma expansion depends upon bacteria being in a growth-permissive cell and might be less effective when the bacteria are still within the more microbicidal-resident macrophage. In contrast, this work shows that PGL-induced CCL2 occurred even under the bacteriostatic or bactericidal conditions imposed by the resident macrophage, allowing even the few remaining bacteria to escape into permissive cells. On the other hand, it remains unclear why the PGL-CCL2 pathway becomes less relevant during granuloma formation. Perhaps the kinetics of ESX-1-MMP9 recruitment are faster, benefitting the bacterium by allowing for its greater intercellular expansion and spread at this stage.

27We have recently shown that Mycobacterium leprae’s PGL-1, differing from Mm’s and Mtb’s PGL in the carbohydrate domain, is required for monocyte-mediated demyelination at a later step of the infection. However, M. leprae also mediates recruitment of monocytes through CCL2-CCR2 signaling, suggesting that its specialized PGL-1 still retains the basal function of eliciting permissive monocytes to promote its infectivity at the first steps of infection. It is noteworthy that both PGL-mediated functions—establishment of infection and demyelination—are through manipulation of host myeloid cells (, and this work).

28Our findings highlight not only both phylogenetic and ontogenic conservation of resident macrophage function but also suggest that different tissue resident macrophages—even the most specialized brain-resident macrophages—all retain their primal function as sentries against invading pathogens. The finding that resident macrophages can make short shrift of mycobacteria, notoriously pernicious pathogens, is particularly noteworthy given their key role in tissue homeostasis. It is curious that CCL2-elicited monocytes provided a safe-haven to mycobacteria as CCR2+ monocytes are broadly microbicidal against bacterial, fungal protozoan, and viral pathogens. Indeed, these cells, also called inflammatory monocytes, are implicated in the pathogenesis of multiple inflammatory diseases affecting the brain, gut, and vascular system. On the other hand, CCR2+ myeloid cells have been implicated in promoting an immunosuppressive tumor environment. Our data identified a permissive role for these cells in the context of an important intracellular infection. Consistent with our findings, CCL2-recruited monocytes have been previously shown to be more permissive to Mtb growth in the lungs of mice, and mice overexpressing CCL2 were found to be more susceptible to challenge with Mtb. Their reduced microbicidal capacity in response to mycobacterial infection might simply reflect the masking of activating TLR ligands by mycobacteria, though it is notable that even in the absence of TLR-mediated activation, resident macrophages are more microbicidal to mycobacteria than monocytes. Of course TB is a complex infection and it is possible that as infection progresses, these same inflammatory monocytes could take on a host-beneficial role in delivering mycobacterial antigens to pulmonary lymph nodes to eventually lead to antigen-specific T cell responses. However, even this role might have complex consequences—while T cell responses are clearly protective for individuals, they might also be paradoxically benefitting bacteria by promoting transmission to new individuals. Overall, our findings add to the discussion of the plasticity and context-dependent function of myeloid cells, for which there is increasing appreciation particularly with the advent of in vivo studies suggesting that myeloid cell functions defy rigid classifications.

29Finally, we note that while evolutionary ancestors of Mtb e.g., Mm and Mycobacterium cannetti uniformly express PGL, the prevalence of PGL-expression in modern-day Mtb strains is not clear. This work emphasizes the need to assess the prevalence of PGL-positive strains, and to thoroughly examine TB transmission epidemiology in regions where PGL-expressing strains abound, while devising therapeutic strategies to block PGL to prevent TB infection and transmission.

30Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Lalita Ramakrishnan (lr404@cam.ac.uk).

31Wild-type AB (University of Washington), csf1raj4blue homozygous mutant (csf1r−/−) zebrafish, Tg(mpeg1:YFP)w200, and Tg(mpeg1:Brainbow)w201 (expressing tdTomato), and the Tg(lyz:EFGP)nz117 lines were maintained in buffered reverse osmotic water systems. Fish were fed twice daily a combination of dry feed and brine shrimp and were exposed to a 14 hr light, 10 hr dark cycle to maintain proper circadian conditions. Larvae (of undetermined sex given the early developmental stages used) were infected at 48 hr post-fertilization (hpf) via caudal vein (CV) or hindbrain ventricle (HBV) injection using single-cell suspensions of known titer. Number of animals to be used for each experiment was guided by pilot experiments or by past results with other bacterial mutants and/or zebrafish. On average 35 to 40 larvae per experimental condition were required to reach statistical significance and each experiment was repeated at least twice. Larvae were randomly allotted to the different experimental conditions. All experiments where csf1r−/− zebrafish were used, csf1r−/− were either in-crossed or outcrossed to wild-type ABs to generate csf1r+/− which are phenotypically wild-type. The zebrafish husbandry briefly described above and all experiments performed on them were in compliance with guidelines from the UK Home Office (Cambridge experiments) and in compliance with the U.S. National Institutes of Health guidelines and approved by the University of Washington Institutional Animal Care and Use Committee (Seattle experiments) and the Stanford Institutional Animal Care and Use Committee (Stanford experiments).

32Human alveolar macrophages (AMs) were retrieved at bronchoscopy as approved by the Research Ethics Committee of St. James’s Hospital (Reference number 2008/17/17), and previously reported. Briefly all donors were patients undergoing clinically indicated bronchoscopy and written informed consent for retrieving additional bronchial washings for research was obtained prior to the procedure. Thirteen donors were recruited to this study, of which 8 were male and 5 were female. The mean age of donors was 56yrs ± 3.4yr, with a range 32-70yrs. Bronchial washing fluid was filtered through a 100 μm nylon strainer (BD Falcon, BD Bioscience, Belgium) and centrifuged at 390 g for 10min. Alveolar macrophages were resuspended in RPMI 1640 culture media supplemented with 10% fetal bovine serum (FBS, GIBCO), 2.5ug/ml fungizone and 50 μg/ml cefotaxime. AMs were seeded at a density of 5 × 104 cells/well in 96-well plates (Corning Costar, Nijmegen, Netherlands). AMs were purified by plastic adherence, non-adherent cells were removed by washing after 24hrs.

33Mm strain M (ATCC BAA-535) ΔmmpL7, Δpks15, and Δesx-1 mutants expressing either TdTomato or Wasabi under the control of the msp12 promoter were grown under hygromycin (Mediatech) selection in 7H9 Middlebrook’s medium (Difco) supplemented with oleic acid, albumin, dextrose, and Tween-80 (Sigma). To prepare heat-killed Mm, bacteria were incubated at 80°C for 20 min. To prepare bacterial supernatants, bacteria were grown to an OD600 of 0.6, pelleted and the supernatant was then filtered twice through a 0.2μm filter. The P. aeruginosa PAO1 fluorescent strain has been described. The S. aureus Newman strain expressing pOS1-SdrC-mCherry #391 was a gift from Dr. Juliane Bubeck Wardenburg.

34Sterile red-fluorescent 1μm beads (Thermo-Fisher Scientific F8821) were diluted ten fold with sterile PBS resulting in 3.64 × 103 beads/nL. Approximately 5 nL of the bead mixture was injected into the hindbrain ventricle of 2 dpf larvae for a total of 1.8 x104 beads per larva.

35To detect iNOS in infected larvae, larvae were euthanized by tricaine overdose, fixed overnight at 4°C in 4% paraformaldehyde (Sigma), permeabilized for 30 min with proteinase K (Thermofisher) at 10μg/mL in PBST (PBS + 0.1% Tween20 (Sigma)), then stained overnight at 4°C in iNOS antibody (see Key Resources Table) diluted 1:200, as described. After washing in PBST, secondary antibodies conjugated to Alexa Fluors (Molecular Probes) were added at 1:500 and incubated overnight at 4°C.

36CPTIO or QVD-OPH (Sigma) was used at a final concentration of 50 mM and 50 μM, respectively, in 0.5% dimethylsulphoxide in fish water. Control fish were incubated in 0.5% dimethylsulphoxide only. Fish were incubated immediately following infection and fresh inhibitor was added every 24 hr until experiment end point.

37Larvae were embedded in 1.5% agarose (low melting point). A series of z stack images with a 2 μm step size was generated through the infected HBV, using the galvo scanner (laser scanner) of the Nikon A1 confocal microscope with a 20x Plan Apo 0.75 NA objective. Bacterial burdens were determined by using the 3D surface-rendering feature of Imaris (Bitplane Scientific Software).

38Macrophage recruitment assays were performed as previously described, 2 dpf zebrafish were injected in the HBV with the bacterial strain or reagent and dose reported in the figure legends. At the specified time post injection, the number of myeloid cells in the HBV was quantified using differential interference contrast microscopy as described below. For assays distinguishing resident macrophages from monocytes, 200 μg/ml Hoechst 33342 (ThermoFisher) was injected via the caudal vein as previously described 2 hr prior to infection into the HBV. Differential interference contrast and fluorescent imaging using Nikon’s Eclipse E600 was done every ∼30 min to identify resident macrophages (Hoechst negative) and monocytes (Hoechst positive). Objectives used in this assay included 20x Plan Fluor 0.5 NA and 40x Plan Fluor 0.75 NA.

39The Sting morpholino 5′TGGAATGGGATCAATCTTACCAGCA3′ (see Key Resources Table) was designed to block the exon 2 intron 2 border. The following primer pair 5′CTGCTGGACTGGGTTTTCTTACTC3′ and 5′TGGGTGATCTTGTAGACGCTGTTA3′ was used to assess morpholino efficiency. Sting morpholino injection led to nonsense mediated decay of mRNA transcripts out to 5dpf. The Sting morpholino and the Ccr2, Pu.1, and Myd88 morpholinos (see Key Resources Table) previously described were injected into the 1-4 cell stage of the developing embryo.

40Total RNA was isolated from pools of 20-40 larvae as previously described and described herein, using TRIzol Reagent (Life Technologies), followed by chloroform precipitation. Isolated RNA was used to synthesize cDNA with Superscript III reverse transcriptase and oligo DT primers (ThermoFisher Scientific). Quantification of ccl2, ifnΦ1, ifnΦ2, and ifnΦ3 RNA levels were determined using SYBR green PCR Master Mix (Applied Biosystems) on an ABI Prism 7300 Real-Time PCR System (Applied Biosystems) using the following primer pairs; ccl2: 5′GTCTGGTGCTCTTCGCTTTC3′ and 5′TGCAGAGAAGATGCGTCGTA3′, ifnΦ1: 5′TTAATACACGCAAAGATGAGAACTC3′ and 5′GCCAAGCCATTCGCAAGTAG3′, ifnΦ2: 5′CCTCTTTGCCAACGACAGTT3′ and 5′CGGTTCCTTGAGCTCTCATC3′, ifnΦ3: 5′GAGGATCAGGTTACTGGTGT3′ and 5′GTTCATGATGCATGTGCTGTA3′. Average values of technical triplicates of each biological replicate were plotted. Data were normalized to β-actin for ΔΔCt analysis using the following primer pair for β-actin: 5′AGAGGGAAATCGTGCGTGAC3′ and 5′CAATAGTGATGACCTGGCCGT3′.

412 dpf larvae were infected via the hindbrain ventricle with an average of 0.8 bacteria per injection as previously described. Fish harboring 1-3 bacteria for some experiments or 1 bacterium for others were identified at 5 hr post infection by confocal microscopy. These infected fish were then evaluated at 5 dpi, or every 24 hr following infection, and were scored as infected or uninfected, based on the presence or absence of fluorescent bacteria.

42In situ hybridization was performed as previously described and described herein: Zebrafish ccl2 (ENSDARG00000041835) was cloned from adult pooled cDNA constructed from isolating RNA from homogenized adult tissues using Trizol (ThermoFisher), chloroform extraction and purification using RNeasy mini kit (QIAGEN). Superscript III reverse transcriptase (ThermoFisher) was used to make cDNA and the following primer pair 5′GTCAGCTAGGATCCATGAGGCCGTCCTGCATCC3′ and 5′GTCAGCTATCTAGATTAGGCGCTGTCACCAGAG3′ was used to clone zebrafish ccl2. ccl2 cDNA was then cloned into the pCS2+ plasmid (A gift from Marc Kirschner, Addgene plasmid #17095), the plasmid was then linearized with the restriction factor HindIII (Thermofisher) and in vitro antisense RNA was synthesized with the T7 Megascript kit (Thermofisher) using DIG RNA labeling mix (Sigma) to make the antisense RNA in situ probe. Mm infected fish were then overdosed in tricaine and fixed overnight in 4% paraformaldehyde and then dehydrated by storage at −20C overnight in methanol. Fish were then rehydrated in PBS with 0.1% Tween 20 (PBST) and digested in 10μg/ml Proteinase K (Thermofisher) for 30min at room temperature. Fish were then refixed in 4% paraformaldehyde, washed in PBST and then hybridized with the antisense probe at 65C for 3hours. Fish were washed in PBST and then incubated with blocking reagent (PBST, 5% sheep serum (Sigma) and 2 mg/ml BSA (Sigma)) for 2hrs at room temperature. Fish were then incubated with anti-DIG-AP antibody (Sigma) at 1:5000 in blocking reagent overnight at 4C. Fish were then washed with PBST and developed with BM-purple (Sigma). Fish were then stored in glycerol and imaged.

43On the day of infection Mm wild-type and Δpks15 growing in Middlebrook 7H9 medium were centrifuged at 2900 g for 10min and resuspended in RPMI 1640 containing 10% FCS. Clumps were disrupted by passing the bacilli through a 25-gauge needle 6-8 times and the sample was centrifuged at 100 (x)g for 3 min to remove any remaining clumps. To assess the adequacy of dispersion and to determine the MOI, macrophages were infected with varying amounts of resuspended Mm wild-type and PGL-deficient for 2hrs. Extracellular bacteria were washed off, and cells were fixed with 2% paraformaldehyde for 10mins. Macrophage nuclei were counterstained with 10 μg/ml of Hoechst 33258 (Sigma). The percentage of infected cells and the number of bacilli per cell were determined by fluorescent microscopy (Olympus IX51, Olympus Europa GmbH, Germany) for each donor, as previously described. Based on this result alveolar macrophages were infected at an estimated MOI of 1-10 bacilli. At 1hr post-infection supernatants were harvested for CCL2 (MCP-1) assay.

44Human MCP-1 chemokine kit (Meso Scale Discovery®, Maryland, USA) was used as per manufacturers’ instructions, briefly samples, standards and controls were added at 25 μL per well. Detection antibody was added at 25 μL per well, 150 μL of the MSD Read Buffer was added to each well and the MSD plates were analyzed on the MSD Sector Imager 2400 plate reader. The raw data was measured as electrochemiluminescence signal (light) detected by photodetectors and analyzed using the Discovery Workbench 3.0 software (MSD). A 4-parameter logistic fit curve was generated for CCL2 (MCP1) using the standards and the concentration of each sample calculated.

45The following statistical analyses were performed using Prism 5.01 (GraphPad): One-way ANOVA with Bonferroni’s post-test, Fisher’s exact test, Student’s unpaired t test, and one sample t test. Error bars represent standard error of mean. Post-test P values are as follows: ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. The statistical tests used for each figure can be found in the corresponding figure legend. Where the n value is given and not represented graphically in the figure, n represents the number of zebrafish used for each experimental group.

46C.J.C. and L.R. conceived, designed and analyzed the zebrafish experiments and C.J.C. performed them. C.J.C, S.M.O., M.P.O., J.K., and L.R. designed and analyzed the human experiments, and S.M.O performed them. C.J.C. and L.R. wrote the paper with input from S.M.O., M.P.O., and J.K.