Robust lysosomal rewiring in Mtb infected macrophages mediated by Mtb lipids restricts the intracellular bacterial survival

Intracellular pathogens commonly manipulate the host lysosomal system for their survival, however whether this affects the organization and functioning of the lysosomal system itself is not known. Here, we show using in vitro and in vivo infections that the lysosomal content and activity is globally elevated in M. tuberculosis infected macrophages. The enhanced lysosomal state is sustained over time and defines an adaptive homeostasis of the infected cell. Lysosomal alterations are caused by mycobacterial surface components, notably the cell wall lipid SL-1, which functions through the mTORC1-TFEB axis. Mtb mutant defective for SL-1 levels shows reduced lysosomal content and activity compared to wild type. Importantly, this phenotype is conserved during in vivo infection. The alteration in lysosomal phenotype in mutant Mtb lead to decreased lysosomal delivery of Mtb, and importantly, increased survival of intracellular Mtb. These results define the global alterations in the host lysosomal system as a crucial distinguishing feature of Mtb infected macrophages that is host protective and contribute to the containment of the pathogen.

Intracellular pathogens commonly manipulate the host lysosomal system for their survival, 18 however whether this affects the organization and functioning of the lysosomal system itself 19 is not known. Here, we show using in vitro and in vivo infections that the lysosomal content 20 and activity is globally elevated in M. tuberculosis infected macrophages. The enhanced 21 lysosomal state is sustained over time and defines an adaptive homeostasis of the infected 22 cell. Lysosomal alterations are caused by mycobacterial surface components, notably the cell 23 wall lipid SL-1, which functions through the mTORC1-TFEB axis. Mtb mutant defective for 24 SL-1 levels shows reduced lysosomal content and activity compared to wild type. 25 Importantly, this phenotype is conserved during in vivo infection. The alteration in 26 lysosomal phenotype in mutant Mtb lead to decreased lysosomal delivery of Mtb, and 27 importantly, increased survival of intracellular Mtb. These results define the global 28 alterations in the host lysosomal system as a crucial distinguishing feature of Mtb infected 29 macrophages that is host protective and contribute to the containment of the pathogen. Whether or not M. tuberculosis or its components impacts these processes is not known. 11 In this study, we focus on the global alterations in the macrophage lysosomal system and 12 show that it is significantly increased in M. tuberculosis infected macrophages compared to 13 non-infected cells. This increase is robust and defines an altered homeostatic state in the 14 infected cells. Modulations in the lysosomal system are mediated by diverse mycobacterial 15 surface components, such as the Sulfolipid SL-1 and PIM6. Purified SL-1 induces lysosomal 16 biogenesis in an mTORC1-TFEB dependent manner, while an M. tuberculosis mutant strain 17 lacking SL-1 shows correspondingly reduced altered lysosomal homeostasis both in vitro 18 and in vivo. The attenuated lysosomal rewiring in SL-1 mutant results in reduced trafficking 19 to lysosomes and an enhanced intracellular survival of the mutant bacteria. 20

Material and methods 21
Mycobacterial strains and growth conditions 22 Mycobacterium tuberculosis (H37Rv) expressing GFP was provided by Dr. Amit Singh (Indian 23 Institute of Science, Bangalore). Mycobacterium bovis BCG expressing GFP was a kind gift 24 from Jean Pieters (University of Basel  Each dataset has multiple features relating to cells, bacteria and  2  lysosomes, as well as their associations with each other. Out of these features, 16 lysosomal  3  parameters and 11 cellular parameters were chosen for further analysis. The data was split  4 into training and test set (7:3) and the model was trained using logistic regression with an 5 L1 penalty. Logistic regression uses a logistic function to model one or more independent 6 variables in order to predict a categorical variable (Pedregosa et al., 2011). 7 Applying an L1-regularization penalty using the parameter c forces the weights of many of 8 the features to go to zero. The best regularization parameter was identified by 20-fold cross-9 validation on the training set. Since all the features were selected for the best regularization 10 parameter ( = 1), we further reduced c. Classification metrics, accuracy, precision, recall 11 and F1 score were calculated for infected and non-infected cells for a range of regularization Col inhalation exposure chamber (at the indicated CFU). Mice were sacrificed post-infection 3 at the indicated timepoints, and infected lungs were dissected out, minced and placed in 4 Miltenyi GentleMACS C-tubes containing 2ml dissociation buffer (RPMI media with 5 0.2mg/ml Liberase (Sigma Aldrich 5466202001) and 0.5mg/ml DNAase (Sigma Aldrich 6 11284932)) and subjected to the inbuilt lung dissociation protocol 1 of Miltenyi 7 GentleMACS, followed by incubation at 37°C and 5% CO2 for 30 min with low agitation (50 8 rpm) and a second 20-second dissociation with lung dissociation protocol 2 (Miltenyi 9 GentleMACS). The suspension was passed through 70-micron cell strainer and then pelleted 10 at 1200 rpm for 5 min. Pellet was re-suspended in 1ml erythrocyte lysis buffer (155 mM 11 NH4Cl, 12 mM NaHCO3 and 0.1 mM EDTA) for 1 min and immediately added to 10 ml RPMI 12 media. Cells were centrifuged again, re-suspended in RPMI media with 10% fetal bovine 13 serum, and plated for 2 hours in RPMI media with 10% fetal bovine serum for macrophage 14 selection based on adherence. Typically, 50-60% cells were infected under these experimental conditions; hence, statistics 31 could be obtained from a reliably large number of both infected and uninfected cells from 32 the same population at a single cell resolution. First, we compared the total number and 33 total intensity of lysosomes between individual infected and uninfected cells. The results 34 show that infected cells on an average have more lysotracker positive vesicles and 35 integrated intensity than uninfected cells (Fig 1A). Similar results were obtained in THP-1 36 monocyte derived macrophages infected with M. tuberculosis H37Rv expressing GFP (Fig  37  1B) and stained with lysotracker red. Lysotracker red stains acidic vesicles but is not 38 specific for lysosomes. In order to further confirm the global alterations in lysosomes upon 39 M. tuberculosis infection, we repeated these experiments with two lysosome activity probes, 40 Magic Red Cathepsin B (MRC) and DQ-BSA, which are cell-permeable fluorogenic dyes that 41 fluoresce when exposed to the hydrolytic lysosomal proteases. The results (Fig 1C, D) show 42 that the number and total fluorescence of both MRC and DQ-BSA positive vesicles are higher 43 in M. tuberculosis infected cells compared to non-infected cells suggesting that the enhanced 44 lysosomes in infected cells are functional in terms of their proteolytic activity. In order to 45 independently verify this result, we immunostained M. bovis BCG infected THP1 1 macrophages with antibodies against two commonly used lysosomal markers, Lamp1 (Fig  2  S1A, C) and Lamp2 (Fig S1B, D), and assessed the lysosomal content by imaging assay. In 3 both the cases, infected cells showed higher lysosomal content than uninfected cells. 4 Moreover, the elevated lysosomal content in mycobacteria infected cells was observed at 5 both 2 hours post infection (hpi) and 48 hpi (Fig 1B-D, S1A-D). Independently, Lamp1 levels 6 were also measured at 2, 24 and 48hrs in M. bovis BCG infected THP1 macrophages by flow 7 cytometry. In all the timepoints measured, infected cells showed higher lysosomal content 8 than uninfected cells ( Fig S1E). Together, these results show that the enhanced lysosomal 9 content and activity are sustained over time in Mtb infected macrophages. In cultured 10 macrophages in vitro, it is well established that majority of the pathogenic mycobacteria are 11 not delivered to lysosomes but remain in an arrested phagosome. We tested the co-12 localisation of Mtb with the different lysosomal probes quantitatively ( Fig S1F). Analysis of 13 the lysosomal delivery of more than 10,000 intracellular Mtb using multiple lysosomal 14 probes shows that inline with earlier observations, majority of Mtb did not co-localise with 15 lysosomes (Fig 1E-G). 16

17
Lysosomal features alone can predict the infection status of a cell 18 Given the reproducible alterations in lysosomes upon mycobacterial infection, we tested if 19 an infected cell can be predicted solely based on the lysosomal features, in the absence of 20 any information from bacteria. We used multiple features of the lysosomes, which report 21 diverse aspects of lysosomal biology such as the intensity, size, elongation and distribution 22 within the cell for this purpose. We used two separate datasets of human primary monocyte 23 derived macrophages infected with M. bovis BCG-GFP (Exp1 and Exp2 features selected are data independent. Further, analyses using an independent algorithm 38 (random forest) reiterated the importance of these features as they are once again ranked in 39 the top 11 and contribute >70% during classification. We obtained similar results in another 40 dataset describing THP-1 monocyte derived macrophages infected with M. bovis BCG-GFP. 41 The accurate prediction of an infected cell solely based on lysosomal parameters in the 42 absence of any information from the bacterial channel, and the remarkable consistency 43 across different experimental datasets and infection conditions shows the robustness of the 44 alterations in lysosomes upon mycobacterial infection. 45 46 1 Next, we tested if similar lysosomal rewiring is observed during in vivo infection. We 2 infected BALB/c mice with M. tuberculosis expressing GFP using aerosol infection. After four 3 weeks, we prepared single cell suspensions from infected lungs and isolated macrophages. 4 The identity of these cells were tested using F4/80 and CD11b, two markers frequently used 5 to characterize murine macrophages (Zhang et al., 2008), and were found to be over 90% 6 positive ( Fig S2A-D). We stained these cells with lysotracker red, or immunostained for 7 antibodies against Lamp1 and Lamp2 followed by assessment of the total lysosomal content 8 between infected and uninfected cells. The results, compiled from four individual mice ( Fig  9 2A-C), show increased lysosomes specifically in infected cells. Similarly, single cell 10 suspensions from infected mice pulsed with functional lysosomal probes MRC and DQ-BSA 11 showed higher number and total cellular fluorescence of lysosomes in infected cells 12 compared to non-infected (Fig 2D, E). Similar results were obtained with C57BL/6J mice 13 showing that these alterations are robust and strain independent. 14 15 While these results suggest that lysosomes are rewired in vivo during Mtb infection, there 16 are two potential confounding factors for this interpretation. First, the time point used for 17 these infections (4 or 6 weeks) could result in immune activation, which could influence our 18 results. Second, we used high aerosol inocula (~5000 CFU). Although, both the high inocula 19 and longer infection time was necessary to obtain sufficient number of infected cells from 20 mice for robust statistical analysis, they could cause artefacts. In order to test if these factors 21 are significantly influencing the results, we first infected THP-1 monocyte derived 22 macrophages with Mtb-GFP and treated with 25 ng/ml IFN-g for 48 hpi followed by staining 23 with lysotracker red. Quantification of total cellular lysotracker content reveals that, while 24 as expected, there is an increase in net lysosomal content upon IFN g treatment, Mtb infected 25 cells showed a further increase (Fig S2E, F). These results suggest that the lysosomal 26 rewiring during Mtb infection is autonomous of immune activation status. As expected, the 27 co-localisation of Mtb with lysotracker red was also higher in IFN g treated condition (Fig S2  28 G). Next, we infected BALB/c mice with low aerosol inocula (~150 cfu) for shorter time 29 point. We isolated infected macrophages from mice lungs ~2 weeks post infection and 30 stained with lysotracker red or MRC. Data, pooled from multiple infected mice show ( macrophages. The result (Fig 3D) shows that the distribution of lysosomal integral intensity 24 of E. coli infected macrophages, despite a relative increase compared to uninfected cells 25 immediately after infection, remained within the bounds of naive macrophages (Fig 3D), 26 suggesting that the lysosomal response observed during Mtb infections is distinct. 27 Moreover, the classifier previously trained to predict the infection status of a cell solely 28 based on its lysosomal features failed to predict E. coli infected cells ( Fig 3E) identical concentration resulted in elevated lysosomal levels ( Fig 4A). Two of the lipids, SL-1 44 and PIM6, showed strong response, we validated them in independent assay at lower doses 45 (Fig 4B, C). We further validated this by adding increasing amounts of SL-1 to THP-1 46 monocyte derived macrophages, which resulted in increasing levels of lysotracker red 1 fluorescence (Fig S4 A).
Next, we checked if the increase is specific for lysotracker red staining, or if lysosomal 4 activity is increased as well. Hence, we pulsed SL-1 treated THP-1 monocyte derived 5 macrophages with lysosomal activity probes DQ-BSA and MRC (Fig 4D, E) and obtained 6 similar results showing that total cellular lysosomal content and activity increases upon SL-7 1 treatment. The increase in lysosomal content upon SL-1 treatment was further confirmed 8 by immunoblotting lysates of SL-1 treated THP1 cells for lysosomal marker Lamp1 (Fig 4F). 9 Importantly, RAW macrophages, as well as non-macrophage cells like HeLa cells treated 10 with SL-1 showed similar phenotypes (Fig S4B, C), showing that the increased lysosomal 11 phenotype mediated by SL-1 is not cell-type specific and suggesting that SL-1 could 12 influence a molecular pathway broadly conserved in different cell types. To assess if SL-1 13 effect is specific for lysosomes or if it influences the upstream endocytic pathway, we pulsed 14 SL-1 treated cells with two different endocytic cargo, fluorescently tagged transferrin or 15 dextran. The results (Fig S4D, E) show that SL-1 does not affect endocytic uptake suggesting 16 that its effect is specifically modulating lysosomes. 17 18 Next, we aimed to gain insights into the molecular mechanism by which SL-1 influences 19 lysosome biogenesis. positive control in these assays. The results (Fig 5D, S5A) show a significant nuclear 39 translocation of TFEB upon SL-1 treatment. Finally, to confirm the involvement of TFEB in 40 SL-1 mediated increase in lysosomes, we silenced TFEB expression in THP-1 macrophages 41 with esiRNA for TFEB. Silencing was confirmed by western blotting for TFEB ( Fig S5B). We 42 treated TFEB and universal negative control (UNC) silenced cells with SL-1, and quantified 43 the change in lysosomal number between the different conditions ( Fig 5E). The result shows 44 a significant reduction in the number of lysosomes upon TFEB silencing in SL-1 treated cells. 45 Similar results were obtained with the positive control Torin1 (Fig 5E). These results 1 confirm that SL-1 acts through the mTORC1-TFEB axis to induce lysosomal biogenesis. 2 In the assays described above, we have treated cells with purified SL-1. The presentation of 3 lipids to the host cells, and consequently its response, can be different when added 4 externally in a purified format, or presented in the context of Mtb bacteria. Hence, we tested 5 the relevance of SL-1 mediated alteration in lysosomal homeostasis in the context of Mtb 6 infection. WhiB3 is a mycobacterial protein that controls the flux of lipid precursors through 7 the biosynthesis of lipids such as SL-1. MtbDWhiB3 mutants show significantly reduced 8 levels of SL-1 both in vitro culture and within macrophages (Singh et al., 2009). If SL-1 9 presentation from Mtb is relevant for lysosomal alterations, we expected cells infected with 10 MtbDwhiB3 to show reduced lysosomes relative to cells infected with wild type (wt) Mtb. In 11 order to test this, we infected THP-1 cells with wild type Mtb H37Rv and MtbDwhiB3 and 12 assessed the total lysosomal content of infected macrophages by staining for lysotracker 13 red, DQ-BSA and MRC. The results (Fig S6A-C) show that indeed cells infected with 14 MtbDwhiB3 have reduced lysosomal levels compared to wild type Mtb infected cells. 15 Importantly, chemical complementation of MtbDwhiB3 with purified SL-1 rescued the 16 lysosomal phenotype (Fig S6A-C). showed a significantly reduced delivery to lysosomes, with only ~ 20% of the mutant Mtb 35 delivered to lysosomes (Fig 6B, D) relative rescue in the inhibition of mTORC1 in the absence of SL-1 (Fig 6E). Next, we tested 43 the nuclear translocation of TFEB upon Mtb infection by infecting RAW macrophages 44 transfected with TFEB-GFP. The results show that similar to Torin1 treatment, wt Mtb 45 infection results in nuclear translocation of TFEB (Fig 6F). Importantly, MtbDpks2 infected 1 cells show a partial rescue in nuclear translocation compared to wt infected cells (Fig 6F). show that the number of bacteria per infected cell (Fig 6G) is significantly higher in 11 MtbDpks2 compared to wt Mtb. Finally, to confirm this phenotype, we lysed infected cells 12 and plated on 7H11 agar medium immediately after infection, or at 48 hpi, and counted the 13 number of colonies obtained. The results (Fig 6H) show similar CFU counts immediately 14 after infection, indicating that the uptake is not altered. Importantly, at 48 hpi, significantly 15 higher number of colonies were seen in mutant bacteria infected cells (Fig 6H) confirming 16 the higher intracellular survival of MtbDpks2 compared to wild type Mtb. 17 Next, we assessed the role of SL-1 in modulating lysosomal response in vivo. We infected 18 C57BL/6NJ mice with wt and MtbDpks2 and assessed lysosomal content in macrophages 19 obtained from single cell suspensions from infected lungs using lysotracker red and MRC 20 staining. The results shows a decreased total lysosomal content in MtbDpks2 infected cells 21 compared to wt Mtb infected cells (Fig 7A, B) demonstrating that indeed SL-1 is involved in 22 lysosomal biogenesis also during in vivo infections. Despite this difference, both wt and 23 MtbDpks2 infected cells showed higher lysosomal content compared to their respective 24 uninfected controls based on lysotracker red and MRC staining (Fig S7A-D), showing that 25 the redundancy in the system is also conserved in vivo. Importantly, under in vivo infection 26 conditions as well, MtbDpks2 showed reduced localization with lysosomal probes (Fig 7C,  27 D). 28

29
Our where they continue to survive, albeit at a reduced rate ( suggested that mycobacteria specific factors modulate lysosomal but not endosomal 39 parameters. In this study, we identify few mycobacterial surface components that increase 40 the macrophage lysosomes, even in the absence of infection, in a cell autonomous way. 41 Of the lipids tested, SL-1 showed a prominent effect on host lysosomes. Although considered 42 non-essential for mycobacterial growth in culture, SL-1 is an abundant cell wall lipid, 43 contributing up to 1-2% of the dry cell wall weight (Goren, 1970 suggested that SL-1 inhibits phagosome maturation in murine peritoneal macrophages 31 (Goren et al., 1976). These differences could be attributed to the different assay systems 32 employed, or to the intrinsic differences between human and mouse macrophages. differential sub-cellular localization to the distinct inflammatory responses elicited by these 2 phylogenetically distant strains will be interesting to explore. 3 Presence of lipids like SL-1 on the surface could provide Mtb with a means to regulate or fine 4 tune its own survival by modulating lysosomes and their trafficking. Generation of reliable 5 probes to accurately quantify individual lipid species such as SL-1 on the bacteria during 6 infection could play a key role in exploring this idea and enable accurate assessment of 7 variations within and across different mycobacterial strains and infection contexts. The 8 discovery that structurally unrelated lipids independently exhibit the same phenotype of 9 enhancing lysosomal biogenesis shows the redundancy in the system.  respectively, of these macrophages. Data are pooled from three mice. Results are 23 representative of two independent infections with at least three mice each. Statistical 24 significance was assessed using Mann-Whitney test, and *** denotes p-value less than 0.001. 25 Scale bar is 10 μm. For panels A to E and G to I, data are represented as box plots, with 26 median highlighted by red line. Individual data points corresponding to single cells are 27 overlaid on the boxplots.    was assessed using Mann-Whitney test, ** denotes p-value of less than 0.01 and *** denotes 20 p-value of less than 0.  Approximately 100 cells were analyzed per category in each experiment and significance 8 assessed by Mann-Whitney test. Data presented is representative of two independent 9 experiments. Scale bar is 10 μm. (B, C) Immunoblots and quantification of phosphorylated 10 and total levels of indicated proteins in THP-1 monocyte-derived macrophage lysates 11 treated with DMSO (control) or SL-1 (B) or PIM6 (C). Torin1 (1μM) was used as a positive 12 control. Bar graphs show average of at least three biological replicates and error bars 13 represent standard deviation. Change in phosphorylation status of indicated protein (S6 14 Kinase) is assessed by normalizing phosphorylated protein to the respective total protein.

15
Actin/GAPDH was used as the loading control. Significance is assessed using unpaired-one 16 tailed Student's t-test with unequal variance, * represents p-value less than 0.05 and ** less 17 than 0.01. (D) RAW macrophages were transfected with TFEB-GFP for 24 hours and treated 18 with 25 μg/ml SL-1, or negative and positive controls, DMSO and Torin1 (250nM) 19 respectively. Representative images and quantification of nuclear to cytoplasmic ratio of 20 TFEB-GFP between the different conditions are shown. Results are representative of atleast 21 three independent experiments. (E) Differentiated THP1 macrophages were transfected 22 with either control siRNA (Universal negative control 1-UNC1) or TFEB siRNA for 48hrs 23 followed by treatment with SL-1 (25 μg/ml for 24hrs) or Torin1 (1 μM for 4hrs) and were 24 pulsed with lysotracker red and imaged. Representative images and quantification of 25 control, SL-1 and torin1 treatment in UNC1 or TFEB siRNA transfected macrophages are 26 shown. Results are representative of two biological experiments. Statistical significance for 27 A, D and E was assessed using Mann-Whitney test, and *** denotes p-value of less than 1 0.001. Scale bar is 10 μm. For A, D, E, data are represented as box plot. Individual datapoints 2 overlaid on the box plot in D and E represent single cells. lysosomal probe. Bacteria overlapping by more than 50% with the lysosomal compartment 12 were considered co-localised. More than 1000 phagosomes were analyzed in each 13 experiment for colocalization analysis. nuclear to cytoplasmic ratio of TFEB-GFP was compared between unexposed, torin1 treated, 22 wt and pks2 mutant infected cells. Torin1 treatment (250nM for 4hrs) was used as positive 23 control. TFEB-GFP channel images are shown in Fire LUT for better visualization of the 24 fluorescence intensities. Data points are pooled from two independent biological 25 experiments. (G) THP1 monocyte-derived macrophages were infected with Mtb wt or 26 MtbDpks2 M. tuberculosis-GFP for 48hrs, fixed and imaged. Images and boxplot show the 1 number of bacteria per cell for the two conditions. Statistical significance for boxplots in 2 figure A, B, C, D, F and G was assessed using Mann-Whitney test, * denotes p-value of less 3 than 0.05, ** denotes p-value of less than 0.01 and *** denotes p-value of less than 0.001. 4 (H) CFUs of Mtb wt or MtbDpks2 infected THP1 monocyte-derived macrophages 5 immediately after infection and 48 hours post infection. Results are the average and 6 standard error of data compiled from three biological experiments, each containing four 7 technical replicates. For E and H, significance is assessed using unpaired-one tailed Student's 8 t-test with unequal variance, ** denotes p-value less than 0.01, ns denotes non-significant, * 9 denotes p-value less than 0.05. For A, C, F, G, scale bar is 10μm, data are represented as box 10 plots, with individual data points corresponding to single cells overlaid. infected cells complemented with 5μg/ml purified SL-1 for 48hrs. Results are representative 2 of three biological experiments. Statistical significance was assessed using Mann-Whitney 3 test, * denotes p-value less than 0.05, ** denotes p-value less than 0.01 and *** denotes p-4 value less than 0.001. Scale bar is 10 μm. probes were compared between Mtb wt or MtbDpks2 CDC1551 infected cells. Lysotracker 7 red (C) and MRC (D) index represent the intensity of the respective probe in Mtb wt or 8 MtbDpks2 CDC1551 containing phagosomes. Results are compiled from four wild type Mtb 9 and three MtbDpks2 infected mice. Statistical significance was assessed using Mann-Whitney 10 test, ** denotes p-value of less than 0.01 and *** denotes p-value of less than 0.001. Scale bar 11 is 10 μm.