Deficiency in Galectin-3, -8, and -9 impairs immunity to chronic Mycobacterium tuberculosis infection but not acute infection with multiple intracellular pathogens

Macrophages employ an array of pattern recognition receptors to detect and eliminate intracellular pathogens that access the cytosol. The cytosolic carbohydrate sensors Galectin-3, -8, and -9 (Gal-3, Gal-8, and Gal-9) recognize damaged pathogen-containing phagosomes, and Gal-3 and Gal-8 are reported to restrict bacterial growth via autophagy in cultured cells. However, the contribution of these galectins to host resistance during bacterial infection in vivo remains unclear. We found that Gal-9 binds directly to Mycobacterium tuberculosis (Mtb) and Salmonella enterica serovar Typhimurium (Stm) and localizes to Mtb in macrophages. To determine the combined contribution of membrane damage-sensing galectins to immunity, we generated Gal-3, -8, and -9 triple knockout (TKO) mice. Mtb infection of primary macrophages from TKO mice resulted in defective autophagic flux but normal bacterial replication. Surprisingly, these mice had no discernable defect in resistance to acute infection with Mtb, Stm or Listeria monocytogenes, and had only modest impairments in bacterial growth restriction and CD4 T cell activation during chronic Mtb infection. Collectively, these findings indicate that while Gal-3, -8, and -9 respond to an array of intracellular pathogens, together these membrane damage-sensing galectins play a limited role in host resistance to bacterial infection.


Introduction
Infectious diseases caused by intracellular bacterial pathogens are among the leading causes of mortality worldwide [1]. Bacterial replication within host cells is critical for pathogenesis yet also exposes bacteria to a myriad of innate immune sensing mechanisms [2]. Detection of pathogen-associated molecular patterns (PAMPs) by pattern recognition receptors (PRRs) activates multiple cell-intrinsic antimicrobial pathways, including autophagy and inflammasome assembly [3,4], and proinflammatory signaling pathways required for coordinating protective adaptive immune responses [5,6]. However, the relative importance and precise function of distinct pathogen-sensing pathways during in vivo infection remains poorly understood.
While it is clear that Gal-3, -8, and -9 are able to sense and respond to damaged phagosomes, the contribution that these membrane damage-sensing galectins make to antibacterial immunity is less certain. Most prior studies that have explored the ability of galectins to restrict the growth of intracellular bacteria have used immortalized cell lines. In an early seminal study, it was shown that depletion of Gal-8 in HeLa cells led to increased Stm replication [9]. Subsequent studies have also shown that depletion of Gal-3 or knockout of Gal-8 in the macrophage-like RAW 264.7 cell line led to increased Mtb replication [15,20]. However, only limited studies of galectin-deficient mice challenged with bacterial pathogens have been carried out. Mice deficient in either Gal-3 or Gal-8 subjected to a high-dose Mtb challenge succumbed slightly faster to Mtb infection [15,16], although bacterial burden was not reported in either study. Mice deficient in Gal-9 challenged with high-dose intranasal Mtb showed a mild, twofold increase in bacterial burden without any change in host survival reported [27]. Similarly, mice inoculated intravenously with Gal-9 knockout macrophages infected ex vivo with Mtb also showed a mild, two-fold increase in bacterial burden in comparison to wild-type (WT) control [13]. The potential for redundancy between these galectins, which all respond to endomembrane damage, could explain the seemingly mild phenotypes of individual galectin knockouts. Thus, the role of membrane damage-sensing galectins in host resistance to bacterial infection remains unclear.
In the present study, we identified Gal-9 in a mass spectrometry-based search for host proteins that bind to the surface of Mtb. We found that Gal-9 bound Mtb in a carbohydratedependent manner and, similar to Gal-3 and Gal-8, localized to bacteria in response to phagosome damage during macrophage infection. To further characterize the collective role of membrane damage-sensing galectins in vivo, we generated Gal-3, -8, and -9 triple knockout (TKO) mice. Primary bone marrow-derived macrophages (BMMs) from TKO mice exhibited reduced autophagic flux yet maintained normal bacterial growth restriction during infection with multiple intracellular pathogens. Furthermore, TKO mice exhibited no defects in resistance to acute bacterial infection with Mtb, Lm, or Stm, and showed only modest susceptibility to chronic Mtb infection. These results suggest that membrane damage-sensing galectins play a limited role in antibacterial immunity in vivo and may fulfill roles other than mediating direct antibacterial resistance.

Gal-9 binds to Mtb
To identify host proteins that bind to the Mtb surface, we performed pull-down experiments with formaldehyde-fixed Mtb incubated with lysates from differentiated THP-1 macrophages ( Fig 1A). After washing the bacteria, the bound proteins were eluted with 8M urea and analyzed by liquid-chromatography mass spectrometry (LC-MS/MS) (Fig 1A), which identified 300 human proteins (Tab A in S1 Table). We ranked the identified proteins using the Contaminant Repository for Affinity Purification (Tab B in S1 Table) [35], which identifies common contaminants in published pull-down experiments. We then focused on lipid-and carbohydrate-binding proteins, as these molecules are likely to recognize cell wall components during infection, and noted the presence of Gal-1 and Gal-9. Gal-1 is often an LC-MS/MS contaminant, whereas Gal-9 is predicted to be specific (Tab B in S1 Table) [35]. Since prior studies

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity have shown that Gal-1 does not localize to bacterial phagosomes or appear to play a role in pathogen sensing [9,20], we focused on Gal-9 ( Fig 1B). To biochemically validate the LC-MS/ MS results, we expressed FLAG-tagged mouse Gal-9 in RAW 264.7 mouse macrophages and performed similar pull-down experiments with formaldehyde-fixed Mtb as described above, followed by anti-FLAG immunoblotting. Consistent with our findings with human Gal-9 from THP-1 macrophages, mouse Gal-9 also bound to the surface of Mtb (Fig 1C). This interaction was dependent upon carbohydrate binding, as addition of lactose to the in vitro binding reaction blocked Gal-9 binding (Fig 1C). We used pull-down assays to further explore whether Gal-9 binds to other phylogenetically diverse intracellular pathogens and found that Gal-9 bound to Stm and did not bind to Lm or the fungal pathogen Cryptococcus neoformans (Cn) under the tested conditions ( Fig 1C).
To determine whether Gal-9 localizes to Mtb in infected macrophages, we analyzed Mtbinfected Gal-9-FLAG cells using immunogold electron microscopy with anti-FLAG antibody (S1 Fig). Gal-9-FLAG was found primarily on extra-phagosomal membranes throughout the macrophage cytoplasm, with roughly 20% of puncta localized to phagosomal membranes and 5% localizing to bacteria (S1 Fig). Taken together, these results demonstrate that Gal-9 can bind directly to Mtb via carbohydrate-binding and that Gal-9 localizes to Mtb in an infected macrophage.

Gal-9 recruitment to Mtb is ESX-1-dependent
To determine whether ESX-1-mediated membrane damage is required for the recruitment of endogenous Gal-9, we infected BMMs with fluorescent Mtb and used automated confocal microscopy to quantify Gal-9 recruitment to bacteria. Mtb-Gal-9 colocalization was evident as early as 3 hours post-infection and reached a maximum of 15% at 8 hours post-infection before returning to 10-12% at later times (Fig 1D and 1E). We observed minimal colocalization between Gal-9 and the ESX-1 secretion mutant ΔeccC, indicating that Gal-9 recruitment requires ESX-1-mediated membrane damage and cytosolic access. As a comparison, we also measured endogenous Gal-3 recruitment to Mtb and found that Gal-3 similarly colocalized with approximately 8-11% of WT Mtb but showed no recruitment to ΔeccC Mtb (Fig 1D and 1E). Thus, endogenous Gal-9 is recruited to Mtb in an ESX-1-dependent manner soon after infection and maintains steady localization with a subpopulation of bacteria throughout infection.

Generation and validation of Gal-3, -8, and -9 triple knockout (TKO) mice
While many cell-intrinsic roles have been described for Gal-3, -8, and -9, their contributions to host immunity are incompletely understood. Prior studies have shown that Gal-3, -8, and -9 each make partial contributions to Mtb immunity in animal models [15,16,27]. Because the mild phenotypes of individual galectin knockouts could be due to functional redundancy between damage-responsive galectins, we sought to test the combined role of all three membrane damage-sensing galectins in antibacterial immunity in vivo by generating Gal-3, -8, and -9 triple knockout (TKO) mice. Using CRISPR/Cas9 genome editing, we generated two independent TKO mouse strains (strain 1 and strain 2), both containing frameshift mutations in each of the three galectins that result in premature stop codons. Western blotting of BMMs from both TKO strains confirmed that these mutations led to loss of protein expression (Figs  2A and S2A). Gal-3, -8, and -9 have documented roles in the response to lysosomal damage, which entail activation of lysophagy and ESCRT-mediated repair [15][16][17][18]. To determine whether TKO cells have defective lysophagy induction, we treated WT and TKO BMMs with the lysosomal damaging agent Leu-Leu-OMe (LLOMe) and quantified LC3 puncta. There was a modest yet statistically significant reduction in LC3 puncta per cell in TKO BMMs treated

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity with LLOMe (S2B and S2C Fig), indicating an impairment in damage-induced LC3 puncta formation and lysophagy in galectin-deficient cells. Thus, we have established a genetic system to evaluate the collective roles of membrane damage-sensing galectins in vivo.

Deficiency in Gal-3, -8, and -9 leads to reduced autophagic flux and accumulation of autophagy targeting intermediates
Given the characterized roles of Gal-3 and Gal-8 in antibacterial autophagy induction [9,15,20], we tested whether TKO macrophages were impaired for Mtb-targeted autophagy, which is characterized by recruitment of polyubiquitin chains and ubiquitin-binding autophagy receptors to cytosolic Mtb [31,32]. We used automated confocal microscopy to quantify the recruitment of polyubiquitin, polyubiquitin linkage subtypes K48 and K63, and p62 to Mtb in both WT and TKO BMMs. We hypothesized that decreased Mtb-ubiquitin colocalization would indicate an impairment of autophagy initiation, whereas an increase would indicate either a block in Mtb delivery to the lysosome or a defect in phagosome repair, which could expose bacteria to increased autophagy targeting. TKO macrophages exhibited a 40-70% increase in Mtb colocalization with all autophagy markers tested (Figs 2B, 2C, S3A, and S3B). The increase in Mtb-ubiquitin colocalization persisted from 3 to 24 hours post-infection ( Fig 2D). Both mock-and Mtb-infected WT and TKO cells had equivalent numbers of ubiquitin puncta (S3E Fig), suggesting that the increase in Mtb-ubiquitin colocalization in TKO cells is not due to accumulation of ubiquitin structures at baseline. As expected, the increase in ubiquitin-labeled bacteria was dependent upon ESX-1-mediated phagosome permeabilization (Fig 2B and 2C). We also measured Mtb colocalization with Tax1bp1 and Optineurin, which are autophagy receptors implicated in Mtb-targeted autophagy [20,22,31]. In contrast to p62, there were no differences in Tax1bp1 and Optineurin recruitment to Mtb between WT and TKO macrophages (S3C and S3D Fig), suggesting that galectins function downstream of p62 recruitment but do not affect targeting of Tax1bp1 or Optineurin.
To assess if autophagosome maturation was impaired, we measured Mtb delivery to the lysosome by infecting cells with Mtb and staining with Lysotracker. TKO macrophages showed a~15% decrease in Mtb-Lysotracker colocalization relative to WT (Fig 2E and 2F). This loss in colocalization was not due to reduced lysotracker staining from defects in lysosomal homeostasis or repair in TKO cells, as both mock-or Mtb-infected WT or TKO macrophages had similar numbers of Lysotracker puncta (S3F Fig), indicating that TKO cells maintain lysosomal integrity at baseline and during infection. In contrast to WT Mtb, infection with ΔeccC Mtb resulted in similar levels of Mtb-Lysotracker colocalization in WT and TKO cells ( Fig 2E  and 2F), indicating that galectin-dependent lysosomal trafficking requires ESX-1 and cytosolic access. We initially hypothesized that the reduction in Mtb lysosomal delivery in TKO cells might be due to defects in LC3+ phagophore recruitment to bacteria. However, LC3 immunostaining revealed no differences in steady-state Mtb-LC3 colocalization between WT and TKO BMMs during infection (Figs 2G, 2H, and S3G), while BMMs depleted of the core autophagy gene Atg5 expectedly showed marked reduction in Mtb-LC3 colocalization relative to control. As expected, ESX1 mutants elicited minimal colocalization with LC3 ( Fig 2G and 2H) [31]. Although we detected no difference in Mtb-LC3 colocalization, the reduction in Mtb lysosomal delivery in TKO cells suggested that autophagic flux was disrupted. To test this, we inhibited autophagosome degradation in the lysosome by treating infected cells with bafilomycin A1 and measured accumulation of LC3+ Mtb autophagosomes. TKO BMMs had a 50% reduction in accumulated LC3+ Mtb after treatment with bafilomycin A1 (Fig 2I), indicating that galectin deficiency leads to reduced autophagic flux of LC3+ Mtb. Taken together, these data suggest that while Gal-3, -8, and -9 are dispensable for ubiquitin and p62 recruitment, they function to promote autophagic flux and efficient delivery of cytosolic Mtb to the lysosome.

Gal-3, -8, and -9 are not required for control of intracellular bacterial pathogens in primary macrophages
We next tested whether membrane damage sensing galectins were important for cell-intrinsic restriction of bacterial growth. Our initial finding that human Gal-9 could bind directly to Mtb motivated us to evaluate its role in bacterial control during infection of human cells. To this end, we generated two different Gal-9 shRNA-knockdown THP-1 cells lines, infected them, and measured bacterial growth of a bioluminescent Mtb reporter strain (Mtb-LUX) [36]. Gal-9 knockdown cells exhibited normal bacterial control during infection (S4A Fig). We then took a genetic approach and used CRISPR/Cas9 to knockout Gal-9 in the RAW 264.7 mouse macrophage cell line. Interestingly, Gal-9 -/cells were more permissive for Mtb replication, whereas knockout of Gal-3 yielded no differences (S4B Fig). Previous reports have also demonstrated antimicrobial roles for certain galectins in immortalized cell lines [9,15,20], but galectins have never been evaluated for restriction of Mtb growth in primary cells. Therefore, we further evaluated Gal-9 by measuring Mtb growth in Gal-9 -/-BMMs but observed no defects in bacterial control (S4C Fig). Similarly, we found that Gal-8, which was previously reported to promote bacterial control in immortalized cell lines [20], was not required for Mtb growth restriction in primary macrophages (S4D Fig). Taken together, these data indicate that although Gal-8 or Gal-9 can promote bacterial growth restriction in certain cell lines, they are dispensable for restriction of Mtb growth in primary macrophages.
We speculated that galectin single knockout BMMs exhibit normal bacterial control due to functional redundancy between membrane damage-associated galectins. To test this, we evaluated whether TKO BMMs were more permissive for Mtb replication. We found that despite defects in autophagic flux in TKO macrophages, bacterial growth kinetics were nearly identical between WT and TKO macrophages (Fig 3A and 3B), a finding we confirmed in separate experiments by directly enumerating Mtb colony-forming units (CFUs) (Fig 3C). In addition, TKO macrophages exhibited normal interferon-gamma (IFN-γ) dependent restriction of Mtb ( Fig 3A). Thus, despite previously described roles for Gal-3, -8, and -9 in autophagy and restriction of bacterial growth in immortalized cell lines [9,15,20], we found no defects in cell-intrinsic Mtb restriction in the absence of all three membrane damage-sensing galectins.
The inability of Gal-3, -8, and -9 to restrict Mtb growth led us to examine whether other intracellular bacterial pathogens are restricted by the membrane damage-sensing galectins. We performed infections with Stm, Lm, or Legionella pneumophila (Lp), and found that TKO macrophages also exhibited normal restriction of bacterial growth with these pathogens (Fig 3D-3F). TKO macrophages additionally showed no defects in IFN-γ-dependent restriction of Stm ( Fig 3D). Furthermore, negligible differences were found between WT and TKO macrophages infected with ΔflaA Lp (Fig 3E), which does not activate the NAIP5-NLRC4 inflammasome [37], or an autophagy-sensitive Lm strain (ΔactA PlcA H86A PlcB H69G , Fig 3F) [38], indicating that Gal-3, -8, and -9 are dispensable for inflammasome-independent immunity to Lp and Lm-targeted autophagy. Thus, across an array of conditions in both resting and activated primary macrophages, we found that Gal-3, -8, and -9 are not required for restricting the growth of several intracellular bacterial pathogens.

TKO mice exhibit partial impairment in host resistance to chronic Mtb infection
To assess the role of Gal-3, -8, and -9 during Mtb in vivo infection, TKO and WT mice were infected with aerosolized Mtb (Figs 4, S5, and S6). During the initial 21 days of infection, when innate immunity predominates, there were no differences in bacterial burden in TKO mice compared to WT (Fig 4A and 4B). During chronic infection, at 9 weeks post-infection, there

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity was a slight 2-3-fold increase in bacterial burden in the lungs of TKO mice in two independent experiments (S5 Fig). Although this effect did not meet a p = 0.05 threshold for statistical significance for each experiment, pooling the CFU-fold change data from both experiments showed a statistically significant effect (Fig 4A). TKO spleens also exhibited a trend towards higher bacterial burden at this time point (Fig 4B). In addition, TKO mice succumbed slightly faster to infection than WT mice (Fig 4C). To evaluate whether these changes were dependent on inoculum size, we infected mice with a higher bacterial load (~450 CFUs) and observed a similar trend toward slightly increased bacterial burdens in the lungs and spleens of TKO mice (S6 Fig). Taken together, these data show that mice deficient for Gal-3, -8, and -9 have no discernable defect in immunity to Mtb during the early acute phase of infection and have only modest defects in host resistance to chronic Mtb infection.

Deficiency in Gal-3, -8, and -9 leads to a specific defect in splenic Ag85Bspecific T cell responses
Because TKO mice exhibited resistance defects exclusively during chronic infection, when T cell-mediated immunity promotes bacterial control, we assessed whether Gal-3, -8, and -9 could play a role in shaping the adaptive immune response. To this end, we performed IFN-γ enzyme-linked immunospot (ELISpot) assays to enumerate Mtb-specific T cells in both WT and TKO mice infected with Mtb. Splenocytes were isolated 21 days post-infection when bacterial CFU in WT and TKO animals were equivalent, and ELISpot was performed with the immunodominant I-A b epitopes from Mtb antigens ESAT-6 and Ag85B (S7 Fig). Consistent with a contribution to adaptive immunity, we detected a reduction in IFN-γ-secreting Ag85Bspecific T cells in TKO spleens harvested three weeks post-infection (S7A Fig). This difference seemed selective, as there were no differences in the number of IFN-γ-secreting ESAT-6-specific T cells (S7A Fig). We also measured T cell responses in the lung at this time point but found no differences between WT and TKO mice for either antigen (S7B Fig), possibly reflecting the robust enrichment of highly activated T cells in this organ [39,40]. Finally, we measured splenic T cell responses in a high-dose infection and, in contrast to our low-dose infection results, observed no differences in Ag85B-specific T cell responses (S7C Fig), suggesting that Mtb infection can promote robust T cell activation in the absence of Gal-3, -8, and -9 if sufficient antigen is present. These findings suggest that membrane damage-sensing galectins promote full activation of CD4 T cells in some physiologic contexts, but not universally.

TKO mice exhibit normal resistance to infection with Lm or Stm
Given the surprisingly mild immune defects in response to Mtb, we broadened our analysis and challenged WT and TKO mice with Lm or Stm. Using a Lm intravenous infection model, we detected similar bacterial burdens in spleen and liver from WT and TKO mice at 48 hours post-infection (Fig 5A and 5B). Likewise, bacterial burdens were unaltered in TKO mice infected with Stm via the intraperitoneal route (Fig 5C and 5D). Thus, despite their well-established role in membrane damage-sensing and our data demonstrating direct pathogen recognition, simultaneous loss of all three phagosome-localized galectins has no discernable impact on the ability of the host to restrict several phylogenetically-diverse bacterial pathogens during the acute phase of infection and results in only modest impairments to host immunity during chronic Mtb infection.

Discussion
Given the previous work that demonstrated roles for galectins in antibacterial autophagy and bacterial control [9,15,20], we were surprised that mice lacking Gal-3, -8, and -9 had no

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity resistance defects during the acute stage of infection with Stm, Lm, or Mtb. The absence of a significant immune deficit in TKO mice also stands in contrast to the clear resistance defects seen in autophagy-deficient animals during Mtb infection [30]. There are several possible explanations for the seemingly minimal contribution of Gal-3, -8, and -9 to antibacterial immunity during acute infection. First, redundancy in galectin function (e.g. different galectins or other pathways) may be compensating for Gal-3, -8, and -9 deficiency, although no other galectins are currently known to be targeted to damaged membranes. Alternatively, as several intracellular pathogens actively inhibit autophagy using specific virulence factors [41][42][43], these pathogens may inactivate galectins or downstream autophagy steps as a virulence strategy. Finally, it is possible that galectins play roles in cell types other than macrophages or in different physiologic contexts, such as mucosal immunity in the gut, where epithelial cells rather than professional phagocytes are the initial line of defense against an array of pathogens. Autophagy plays a protective role in the gut epithelium [44], and certain innate immune mechanisms are only evident during infection of the gut epithelium, exemplified by NOD2mediated restriction of Lm growth during oral but not intravenous infection [45]. Thus, galectins may still have an important antibacterial role in other physiologic contexts beyond those we have tested here.
We found that TKO mice exhibited modest defects in bacterial control and survival during the chronic phase of Mtb infection and these findings are supported by previous Mtb studies using galectin single knockout mice and autophagy-related mouse models [15,16,27,33]. In previous survival experiments, Gal-3 -/and Gal-8 -/mice exhibited mild defects in susceptibility to chronic Mtb infection following high dose aerosol challenge (200-3000 CFUs) [15,16]. Similarly, Gal-9 -/mice infected intranasally with an extremely high dose of Mtb (2 x 10 6 CFUs) had a modest two-fold increase in lung bacterial burden [27]. While there are many differences between these studies and ours (e.g. animal housing conditions, animal microbiota, Mtb strain and dose, and/or infection route), these data are in general agreement with our finding that membrane damage-associated galectins play a limited role in vivo during Mtb infection. The increased susceptibility of TKO mice during the chronic phase is also similar to mice that lack the autophagy-related ubiquitin ligase Smurf1, which also have higher Mtb burdens exclusively during the chronic phase and exhibit a very similar survival phenotype [33]. Whether the susceptibility of TKO mice is due to defects in direct antimicrobial killing or

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity aberrant inflammatory responses is currently unknown [33,46,47]. However, defects in host resistance during the chronic phase in both galectin-deficient mice and Smurf1 -/mice may suggest that full and sustained activation of the autophagy pathway is important for long-term containment of Mtb.
Previous reports demonstrated that Gal-3 and Gal-8 promote restriction of bacterial growth in immortalized cell lines [9,15,20]. Similarly, we found that loss of Gal-9 in RAW 264.7 cells also resulted in increased Mtb replication. In contrast, when studying primary immune cells, we found that individual loss of Gal-8 or Gal-9 in BMMs resulted in no discernable change in intracellular Mtb replication. Furthermore, TKO BMMs lacking all membrane damage-sensing galectins exhibited normal restriction of Mtb, as well as Stm, Lm, and Lp. Taken together, this suggests that while membrane damage-sensing galectins have the potential to restrict bacterial growth in cell lines, there may be key differences between these cells and primary macrophages that account for the differential requirements for galectin activity to restrict bacterial growth. Indeed, previous reports have identified significant differences in the phagosome proteome, proteolytic capacity, and kinetics of lysosomal trafficking and proinflammatory signaling between primary BMMs and immortalized RAW 264.7 macrophages that may impact the role galectins play in antibacterial immunity [48,49].
Our results showed that TKO macrophages exhibit impaired autophagic flux of LC3+ Mtb and slightly reduced Mtb lysosomal delivery. TKO cells also showed increased levels of ubiqui-tin+ and p62+ Mtb during infection, suggesting that Gal-3, -8, and -9 function downstream of ubiquitin and p62. However, Mtb association with the autophagy receptors Tax1bp1 and Optineurin were unchanged in TKO cells. Although a mechanistic understanding awaits further investigation, these observations suggest that there are different subsets of targeted phagosomes and that galectins play a role in promoting maturation of certain phagosomes. NDP52 also promotes autophagy via Gal-8 binding in human cells [9], however murine Ndp52 lacks the C-terminal ubiquitin-and Gal-8-binding regions and therefore is likely not at play in mouse cells [20,50]. Of note, we do not see complete loss of autophagy in TKO cells, likely because numerous autophagy-promoting pathways are activated during Mtb infection [3,[31][32][33]. Indeed, robust ubiquitin recruitment to Mtb occurs in TKO macrophages, suggesting that ubiquitin-mediated autophagy pathways, promoted via Parkin, Smurf1, Trim16, or cGAS/ Sting [15,[31][32][33], are largely intact in TKO macrophages and may partially compensate for defects in galectin-driven autophagy.
Our finding that galectin deficiency leads to slight impairments in both autophagy and antigen-specific CD4 T cell activation raises the possibility that these phenotypes could be linked and suggests how an autophagy phenotype might manifest during later stages of infection. Autophagy mediates the capture and degradation of cytosolic pathogens and antigens, after which the peptide constituents are trafficked to MHC-class-II-containing late endosomes and then loaded onto MHC class II complexes for presentation to CD4 T cells [51][52][53][54]. During Mtb infection, inhibition of autophagy was shown to impair Ag85B-specific, but not ESAT-6-specific CD4 T cell activation [55], possibly suggesting that some microbial antigens, but not others, require autophagy for efficient antigen presentation. Similarly, the CD4 T cell activation defect we observed in TKO mice was specific to Ag85B, while ESAT-6-specific CD4 T cell responses were unchanged. How some antigens, but not others, might access MHC class II compartments via autophagy remains unknown. It is possible that highly expressed and secreted antigens, like ESAT-6 [56], do not require autophagy for antigen presentation and instead access the antigen presentation pathway via phagocytosis and endolysosomal trafficking. In contrast, Ag85B, a cell wall antigen that is downregulated at three weeks post-infection and exhibits low abundance during chronic infection [57,58], might rely on autophagy for efficient antigen presentation.

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity Although we and Wu et al. found that Gal-9 can bind directly to the surface of Mtb [27], the ligands recognized by Gal-9 during infection remain unclear. The mycobacterial cell wall, from inside to out, is comprised of an inner membrane, a layer of peptidoglycan covalently linked to a subsequent layer of arabinogalactan (AG), an extremely hydrophobic and relatively-impermeant outer membrane (OM) consisting mainly of long-chain mycolic acids, and a loosely associated polysaccharide "capsule" at the very surface [59,60]. Wu et al. reported that Gal-9 recognized Mtb AG [27], though this layer is internal to the OM and unlikely to be accessible to Gal-9 in healthy, dividing bacteria [60]. A possible solution to this conundrum is that permeabilization of the outermost capsule and mycolic acid layer in vivo (or with detergents routinely used in mycobacterial growth medium) [59], might expose AG to Gal-9. Moreover, it is possible that Gal-9 primarily recognizes dead or dying bacteria that have already been damaged during infection in order to clear bacterial corpses (and their PRR-activating ligands) from the cytosol via autophagy. Gal-9 localizes to damaged endosomes and lysosomes independently of bacterial infection [9,18], indicating that Gal-9 also recognizes an endogenous carbohydrate ligand. In line with this, our immuno-EM studies showed that most Gal-9 colocalizes with host membranes, while a small fraction colocalizes with Mtb in infected macrophages. Recognition of host glycoproteins on damaged membranes or release of AG or other Gal-9 bacterial ligands could localize Gal-9 to extra-phagosomal compartments [61]. Lastly, our observation that Gal-9 also binds Stm but not Lm suggests Gal-9 recognition of bacterial constituents is not limited to AG and is selective for a subset of bacteria. Indeed, other galectin-bacteria interactions have been identified, including Gal-3/Gal-9-Leishmania major, Gal-4/Gal-8-Escherichia coli, and Gal-3-Neisseria meningitidis [24][25][26]. This suggests a scenario where galectins have divergent binding preferences, allowing the recognition of a broader array of microbes.
In conclusion, our data suggest that although Gal-3, -8, and -9 contribute to autophagy in macrophages, they play a limited role in the immune response to systemic bacterial infection. These membrane damage-sensing galectins had no discernable impact on acute infection with Mtb, Stm, or Lm in vivo, and had a modest effect on host resistance to chronic Mtb infection. These data suggest that the role of Gal-3, -8, and -9 in antibacterial immunity is not generalized, but rather is limited to specific physiologic contexts in vivo.

Pathogen pull-down for mass spectrometry and immunoblotting
For Mtb pull-down mass spectrometry, Mtb was fixed in PBS with 4% PFA, washed three times with PBS, and pelleted. 50 μL of Mtb pellet was incubated for 3 hours with 1 mL of 10 mg/mL differentiated THP-1 lysate, washed four times with PBS, and eluted in 8M urea with Rapigest (Waters) before LC-MS sample preparation. For Mtb, Lm, Stm, and Cn pull-down for Gal-9-FLAG immunoblotting, pathogens were fixed, washed, and pelleted as described above.

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity 1 mL of 10 mg/mL RAW 264.7 Gal-9-FLAG lysate was incubated, washed, and eluted as described above prior to immunoblotting.

Mass spectrometry
Samples were denatured and reduced in 2 M urea, 10 mM NH 4 HCO 3 , 2 mM DTT for 30 minutes at 60˚C, then alkylated with 2 mM iodoacetamide for 45 minutes at room temperature. Trypsin (Promega) was added at a 1:100 enzyme:substrate ratio and digested overnight at 37˚C. Following digestion, samples were concentrated using C18 ZipTips (Millipore) according to the manufacturer's specifications. Desalted samples were evaporated to dryness and resuspended in 0.1% formic acid for mass spectrometry analysis.
Peptides were resuspended in 0.1% formic acid and 3% directly injected on a 75 μm ID column (New Objective) packed with 25 cm of Reprosil C18 3 μm, 120 Å particles (Dr. Maisch). Peptides were eluted in positive ion mode into an Orbitrap Elite mass spectrometer (Thermo Fisher) via a Nanospray Flex Ion Source (Thermo Fisher). Elution of peptides was achieved by an acetonitrile gradient delivered at a flow rate of 400 nL/min by an Easy1000 nLC system (Thermo Fisher). All mobile phases contained 0.1% formic acid as buffer A. The total gradient time was 70 minutes, during which mobile phase buffer B (0.1% formic acid in 90% acetonitrile) was ramped from 5% to 30% B over 55 minutes, followed by a ramp to 100% buffer B over a subsequent 5 minutes, and held at 100% for 10 minutes to wash the column. MS data was collected over the first 60 minutes of the gradient. The ion transfer tube was set to 180˚C and the spray voltage was 1500V. MS1 scan was collected in the ion trap in centroid mode with 120K resolution over a scan range of 200-2000 m/z, an S-Lens RF of 68%, a 50 ms maximum injection time, and a AGC target of 3e4. Species with a charge state z = 1 or unassigned charge states were excluded from selection for MS/MS. Resonance excitation MS/MS was performed on the 20 most abundance precursors with a minimum signal intensity of 200, an isolation width of 2 m/z, a 0.25 activation Q, a 10 ms activation time, a 100 ms maximum injection time, and an AGC target of 1e3. Dynamic exclusion was employed to exclude a list of 50 previously selected precursors within 180 seconds and using a +/-1.5 Da window.
The results raw data was matched to protein sequences as previously described [62].

Bacterial and fungal cultures
Mycobacterium tuberculosis. All Mtb experiments were performed with Mtb Erdman strain or strains derived from Mtb Erdman strain. Low passage frozen stocks of Mtb were grown to log phase in 7H9 (BD) liquid media supplemented with 10% Middlebrook OADC (Sigma), 0.05% Tween-80 and 0.5% glycerol in roller bottles at 37˚C. Mtb-LUX expressing luciferase from the luxCDABE operon has been described previously and was cultured as described above [36]. WT or ΔeccC Mtb expressing eGFP under control of the MOP promoter was a gift from Dr. Sarah Stanley's laboratory [63], and was cultured as described above. For BMM infections, log phase bacteria were washed twice with PBS and sonicated three times at 90% amplitude for 30 seconds each then diluted into BMM media (DMEM with 10% FBS, 2 mM L-glutamine, 10% MCSF, 11 mg/mL sodium pyruvate) prior to infection. For THP-1 and RAW 264.7 infections, 10 mL of the Mtb culture was washed twice in PBS with 10% heat-inactivated horse serum. After washing, Mtb was resuspended in 5 mL PBS with 10% heat-inactivated horse serum and sonicated twice at 50% amplitude for 15 seconds. Mtb was diluted into DMEM with 10% heat-inactivated horse serum, and this inoculum was added to THP-1 cells prior to spinfection.
Salmonella enterica serovar Typhimurium. For pathogen pull-down experiments, Stm strain SL1344 was cultured overnight at 37˚C with shaking in Luria-Bertani (LB) broth and backdiluted 1:100 1 day prior to infection, then allowed to recover to mid-log. BMM infections were performed with Stm strain IR715. For mouse infections, Stm strain IR715 was cultured in LB broth overnight at 37˚C with shaking.
Listeria monocytogenes. All Lm strains used in this study were derived from strain 10403S (streptomycin-resistant). ΔactA PlcA H86A PlcB H69G has been previously described [38]. For macrophage infections, Lm was grown to stationary phase overnight, slanted at 30˚C in filtersterilized BHI media with 200 μg/mL streptomycin then diluted into PBS for OD600 measurement and inoculum preparation.
Legionella pneumophila. All Lp strains used are JR32 and express the Photorhabdus luminescens luxCDABE operon. LuxCDABE-expressing WT or isogenic mutant for flaA have been previously described [64]. Bacteria were cultured in Charcoal-Yeast Extract Agar (CYE, 10 g/L 4-morpholinepropanesulfonic acid [MOPS], 10 g/L Yeast extract, 15 g/L technical agar, 2 g/L activated charcoal, supplemented with 0.4 g/L L-cysteine and 0.135 g/L Fe(NO 3 ) 3 ) at 35-37˚C for 4 days from frozen stocks. Single colonies were grown on fresh plates for another 2 days. For macrophage infections, bacteria were grown on solid plates and resuspended in autoclaved water and diluted in RPMI.

Macrophage infections
Mycobacterium tuberculosis. To ensure bacterial replication or autophagy targeting was not impacted by differences in macrophage concentration, WT BMMs were plated in 96-well plates at the target cell density, while TKO BMMs were plated on a cell density gradient in order to ensure equivalent plating numbers of each genotype. After plating, cells were allowed to adhere and equilibrate for 2 to 3 days prior to infection. On day 0 of infection, one plate was fixed in PBS with 4% PFA, washed, DAPI stained, and nuclei were counted at 5X magnification on an Opera Phenix microscope, thus allowing determination of cell density-matched wells. During analysis, wells of different genotypes were only compared if cell counts were matched within <5% of each other. For microscopy, CFU, and Mtb-LUX experiments, cells were infected by removing media from wells, overlaying cell monolayers with bacterial suspensions, and spinfecting for 10 minutes at 300 rcf.
For Mtb-LUX infections, cells were plated at 4 x 10 4 cells per well (target density) in Nunc 96-well white walled TC treated plates (Thermo Fisher). For IFN-γ treatment, cells were incubated with 1.5 ng/mL IFN-γ the day before infection. BMMs were infected at a MOI of 1 or 5. Luminescence of Mtb-LUX-infected BMMs in cell density-matched wells was measured as previously described [62].
For all Mtb microscopy experiments, WT Mtb-GFP or ΔeccC Mtb-GFP were used. BMMs were plated in either 24-well glass-bottom plates (Cellvis) at 2 x 10 5 cells per well, or 96-well PerkinElmer Cell Carrier Ultra plates at 2.5 x 10 4 cells per well (target density). All infections for microscopy used an MOI of 2. At indicated time points, microscopy plates were washed in warm PBS, fixed in PBS with 4% PFA, washed three times with PBS, and stored in PBS at 4˚C before immunostaining. For lysotracker experiments, cells were stained with 50 nM Lysotracker Red DND-99 in BMM media for 3 hours prior to fixation and imaged immediately after washing out the fixative.

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity For Mtb CFU experiments, BMMs were plated in 24-well TC treated plates at 2 x 10 5 cells per well, infected with WT Mtb Erdman strain using an MOI of 2, and Mtb CFU enumeration was performed as previously described [63].
For Mtb infection of THP-1 cells, THP-1 cells were differentiated for 3 days in DMEM with 2 mM L-glutamine, 1 mM sodium pyruvate, and 10% heat-inactivated fetal calf serum with 10 ng/ mL PMA and 0.1 ng/mL Vitamin D. On day 0 of differentiation, THP-1 cells were plated at 4 x 10 4 cells per well in a 96-well plate. The media was exchanged with fresh differentiation media daily. Bacterial suspension was prepared as described above and THP-1 cells were spinfected for 5 minutes. Cells were incubated at 37˚C for 30 minutes. THP-1 cells were washed 3 times with PBS with 1% heat-inactivated horse serum. Differentiation media lacking phenol red with PMA and Vitamin D was replaced after the washes. Luminescence of Mtb-LUX-infected THP-1 cells was measured as previously described [62]. Half of the media was discarded and replaced with fresh differentiation media with PMA and Vitamin D on days 1, 3, and 5 post-infection.
Salmonella enterica serovar Typhimurium. 1 x 10 5 BMMs were plated on 24-well plates 24 hours prior to infection. For IFN-γ treatment, cells were incubated with 0.01 μg/mL IFN-γ the night before infection. Bacteria grown to mid-log phase as described above were pelleted and resuspended in BMM media. Suspension was incubated at 37˚C for 30 minutes for opsonization. Bacteria were then pelleted and resuspended to 1 x 10 9 CFU/mL prior to spinfection with BMMs for 5 minutes at 250 rcf followed by 37˚C incubation for 30 minutes for phagocytosis. Cells were washed with PBS and incubated with BMM media with 100 μg/mL gentamicin at 37˚C incubation for 30 minutes, then washed with PBS and incubated with BMM media with 5 μg/mL gentamicin. For CFU enumeration, cells were washed with PBS and lysed in 500 uL water for 10 minutes prior to lysate serial dilution and plating on LB agar plates.
Listeria monocytogenes. A total of 3 x 10 6 BMMs were plated on 14 coverslips in 60 mm non-TC treated dishes and infected the next day at an MOI of 0.25. Lm CFU enumeration was performed as previously described [65].
Legionella pneumophila. Bacterial replication was measured using a luminescence-based replication assay as previously described [66]. Briefly, 10 5 WT or TKO BMMs were plated on 96-well plates and incubated overnight. Medium was replaced with bacterial suspension in BMM media at an MOI of 0.01. At indicated time points, luminescence emission was measured at 470 nm with a Spectra-L plate reader (Molecular Devices).

Immunofluorescence
Following fixation, washing, and storage in PBS at 4˚C, cells were blocked and permeabilized in PBS with 5% FBS or 2% BSA and 0.1% Triton X-100 (blocking/permeabilization buffer) for 30 minutes at room temperature. For LC3 immunostaining, cells were incubated with ice cold methanol for 10 minutes on ice, then washed three times with PBS prior to the blocking and permeabilization step above. Cells were then stained in blocking/permeabilization buffer overnight at 4˚C with the following primary antibodies at the indicated dilutions: anti-Gal-9 (Abcam ab275877, primary 1:100, secondary 1:500), anti-Gal-3 (SCBT sc-32790, primary 1:50,

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity 30-60 minutes using anti-mouse or anti-rabbit Alexa Fluor 647 or 488 (Invitrogen) at the above indicated dilutions. Cells were washed three times with PBS for 5 minutes then imaged.

Automated confocal microscopy and image analysis
All microscopy experiments were performed with an Opera Phenix High-Content Screening System confocal microscope (PerkinElmer). Microscopy plates were imaged at 10X magnification for nuclei counting and cell density matching, 40X or 63X magnification for colocalization or puncta quantification, or 63X magnification for representative images. Automated analysis for colocalization, nuclei counting, and puncta quantification was performed with PerkinElmer Harmony software package. For colocalization analysis, a maximum intensity projection was generated from multiple z-stacks and autophagy markers and bacteria were identified using a spot finder algorithm. Autophagy marker spots were then filtered based on size and only spots >25-40 pixels in area were selected for analysis to limit nonspecific colocalization of small puncta with bacteria. Bacteria with at least one pixel of overlap with size filtered autophagy marker spots were counted as a colocalization event, and % colocalization was defined as the total number of colocalization events divided by the total number of bacteria.

Electron-microscopy
RAW 264.7 cells expressing mouse Gal-9-3xFLAG were infected with Mtb at an MOI of 1 for 6 hours. Cells were then detached using 1 mM EDTA in PBS and fixed for 10 minutes at room temperature using freshly prepared periodate-lysine-paraformaldehyde stain (0.2 M lysine-HCl pH 7.4, 4% paraformaldehyde, 0.1 M sodium m-periodate). Following fixation, cells were pelleted and resuspended in PBS without washing and stored at 4˚until embedding was performed. Samples were embedded in 10% gelatin and infiltrated overnight with 2.3 M sucrose/ 20% polyvinyl pyrrolidone in PIPES/MgCl 2 at 4˚C. Samples were trimmed, frozen in liquid nitrogen, and sectioned with a Leica Ultracut UCT7 cryo-ultramicrotome (Leica Microsystems). Ultrathin sections of 50 nm were blocked with 5% FBS/5% NGS for 30 minutes and subsequently incubated with mouse anti-FLAG (Sigma F3165) for 1 hour at room temperature. Following washes in block buffer, sections were incubated with goat anti-mouse IgG + IgM 18 nm colloidal gold conjugated secondary antibody (Jackson ImmunoResearch Laboratories) for 1 hour. Sections were stained with 0.3% uranyl acetate/2% methyl cellulose and viewed on a JEOL 1200 EX transmission electron microscope (JEOL USA) equipped with an AMT 8 megapixel digital camera and AMT Image Capture Engine V602 software (Advanced Microscopy Techniques). All labeling experiments were conducted in parallel with controls omitting the primary antibody.

Mycobacterium tuberculosis aerosol infection
To prepare infection inoculum, Mtb Erdman strain frozen stock was diluted in sterile water and sonicated at 90% amplitude three times for 30 seconds. 8 to 12-week-old male and female mice were infected with approximately~75 CFUs (conventional dose) or >400 CFUs (high dose) of prepared inoculum using a Glas-Col inhalation exposure system. At indicated time points, lungs and spleens were isolated, homogenized, and plated on 7H10 plates with Polymyxin B (200,000 U/L), Carbenicillin (50 mg/L), Trimethoprim lactate (20 mg/L) and Amphotericin B (5 mg/L). Plates were incubated at 37˚C for 25-28 days before counting colonies. For survival experiment, mouse weight was measured weekly and mice were sacrificed after losing 15% of maximum body weight.

Salmonella enterica serovar Typhimurium intraperitoneal infection
Stm strain IR715 was cultured as described above. Mice received 100 μl of sterile PBS or 1000 CFUs diluted in 100 μl PBS by intraperitoneal injection. Mouse weights were monitored daily during infection. Systemic bacterial levels were characterized at day 1, 2, or 3 post-infection by harvesting liver and spleen and enumerating CFU. Liver and spleen were collected in PBS, weighed, and homogenized. Homogenates were serially diluted and plated on LB agar plates. CFU per gram tissue was calculated after overnight growth at 37˚C.

Listeria monocytogenes intravenous infection
Lm was grown at 37˚C with shaking at 200 rpm to mid-log phase. Bacteria were washed in PBS and resuspended at 5 x 10 5 colony-forming units (CFU) per 1 mL of sterile PBS. Mice were then injected with 1 x 10 5 CFU via the tail vein. 48 hours post-infection, livers and spleens were collected, homogenized, and plated on LB streptomycin 200 μg/mL to determine the number of CFU per organ.

Interferon-γ enzyme-linked immunospot (ELISpot)
Spleens were isolated and pressed through a 70 μm cell strainer to generate a single cell suspension. Lung cells were isolated and digested in RPMI with Liberase TM (70 μg/ml) and DNase I (30 μg/ml) in gentleMACS C-tubes then placed at 37˚C for 30 minutes, followed by gentle-MACS tissue homogenization for single cell suspension. Cells were pelleted and RBCs were lysed using ACK lysing buffer. Cells were washed, pelleted, resuspended in X-VIVO TM 15 (Lonza) and counted on a hemocytometer or using fluorescent count beads on an SH800 Sony Sorter. Cells were plated, stimulated overnight with indicated peptides (2 μg/ml) or Concanavalin A (ConA, 2.5 μg/ml), developed, and analyzed as previously described [68]. ConA-stimulated cells were used as a positive control and to determine total cell counts. IFN-γ spot-

PLOS PATHOGENS
The role of Galectin-3, -8, and -9 in antibacterial immunity forming cell (SFC) counts from peptide-stimulated wells were normalized to SFC counts from ConA-stimulated wells to control for differences in cell density between samples.

Statistical analysis
Analysis of statistical significance was performed with Prism 9 (GraphPad).