Graphpad Prism was used to analyze data and obtain p-values. All experiments contained biological triplicates and were done at least two independent times.

The procedures of sample collection and analyses were approved by HSC-SPH-12-0037.

Samples from healthy donors and TB patients were obtained following approved IRB procedures. Within TB patient and healthy controls, the samples were randomized for age, sex, and other demographics for unbiased analysis ( Supplementary Table 1 ). From 10 ml of blood, 1 ml was subjected to RBC lysis, centrifuged at 400g for 5 min. PBS washed cells were stained with 100 µl of viability dye for 5 min. Cells were then washed with PBS + 1% FBS and blocked with 50 µl of PBS + 1% FBS containing human Fc block for 10 min, followed by 50–100 µl of antibodies ( Supplementary Table 1 ) against various immune cells using our CyTOF (cytometry by time of flight or mass cytometry; HMRI) core. Cells were mixed and incubated for 30 min, and washed four times with PBS + 1% FBS followed by overnight fixation. From the remaining 9 ml of blood, PBMCs were also isolated using Ficoll®Paque Plus (GE Healthcare #17-1440-02) and were stained with antibodies as above and used for CyTOF analysis.

MDSCs were isolated as described previously (17). Briefly, PBMCs from the buffy coat of healthy donors were isolated using Ficoll®Paque Plus (GE Healthcare #17-1440-02), which resulted in a yield of about 200 million PBMCs. PBMCs were then subjected to CD14+ cell fractionation using an EasySep Human CD14+ positive selection kit (Stem cells; #17858) as per the manufacturer’s instructions. PBMCs from one buffy coat unit yielded about 80 million CD14+ cells. The CD14+ cells were cultured for 6 days in IMDM medium supplemented with 10% fetal bovine serum in the presence of GM-CSF (10 ng/ml) in combination with IL-6 (10 ng/ml) to induce their proliferation into Monocytic-MDSCs. M-MDSCs were then purified from CD14+ cells using the EasySep Human CD33 positive selection kit (Stem cells; #17876) as per the manufacturer’s instructions. The purity of M-MDSCs (CD15-CD14+CD33+HLADR-) was found to be >90% by flow cytometry ( Supplementary Figure 1 ). The total yield for M-MDSCs was in the range of 16–20 million cells per donor from one unit of buffy coat. To test whether MDSCs naturally occurring in blood show bactericidal function after mab activation, for some experiments ( Figure 4 ), CD33+ MDSCs were purified using FACsorting of PBMCs from healthy donors.

Flow cytometric analysis was done to assess the purity of M-MDSCs (henceforth referred to as MDSCs). Cells were stained with Fluorochrome-conjugated antibodies, FITC anti-human CD14 (Biolegend, #367115), PE anti-human CD15 (Biolegend, #301905), PE-Cy7 anti-human HLADR (Biolegend, #307615), APC anti-human CD33 (Biolegend, #366605), and Fixable Viability Stain 510 (BD Biosciences, #564406). Cells were analyzed on the Fortessa flow cytometer 202 (Beckton Dickinson), and the data were processed using FlowJo v10 software (Tree Star, Inc.).

MDSCs were plated at 10⁶ per well in 24-well adherent plates. Mtb H37Rv (ATCC 27294) was grown in Middlebrook 7H9 broth supplemented with 0.05% (v/v) Tween 80 and Middlebrook AODC enrichment (Difco, Becton Dickinson) to mid-log phase (OD 600 nm 0.6± 0.8). Early log phase cultures of Mtb suspensions were sonicated at 4 W for 60 s using a sonicator to prepare a uniform single-cell suspension and then MDSCs were infected with Mtb at a MOI of 1. After 4 h of infection, cells were washed and added with LILRB-2 mab (mab-B2) or its IgG control antibody for 18 h at 1 µg/ml per 10*6 MDSCs for 7 days. Mtb counts were expressed as log10 CFU counts averaged for triplicate wells of MDSCs per group or combination and experiments were repeated two times. p-value for differences in CFUs was determined using 1-way ANOVA using Tukey’s posttest.

MDSCs were subjected to siRNA knockdown for autophagy using manufacturer’s instructions (18). The kits for human siRNAs (mixture of duplexes) were purchased from Origene technologies (Beclin-1 siRNA including scrambled controls: SR322490; iNOS siRNA: SR303202). MDSCs were treated with siRNA and the scrambled control, washed, rested, and added with Mtb (H37Rv) for 4 h of infection (MOI of 1). Cells were washed and added with mab-LILRB2/3 (1 µg/ml each) or isotype (1 µg/ml) and incubated as required. Mab was combined with inhibitors (1 µg/ml each) as needed: SHP1 inhibitor (IACS-13909; #HY-137092/CS-0136474) and SHP2 inhibitor (TPI-1 #HY-100463/CS-6450) were purchased from MCE Med Chem Express. At time points, cells were lysed in 0.05% SDS-PBS buffer and 10-fold dilutions were plated on 7H11 agar plates for CFU counts, which were read after 21 days of incubation. To assess the effects of NO or ROS on Mtb growth in MDSCs, replicate cultures of mab-treated or untreated MDSCs were incubated with diamino-fluorescein diacetate (DAN) fluorescent probe (1 µM) followed by fluorometry using Ascent Fluoroscan and relative light units (RLUs) plotted for triplicate wells per treatment group (19). Whole-cell ROS was similarly detected using fluorometry and dihydro-dichloro-fluorescein diacetate (1 µM) using live cells as described (20).

Naive MDSCs were infected using gfp-expressing Mtb (MOI of 1) for 4 h, washed and treated using either mab-B2 or isotype, and incubated at 18 h. Post infection, the cells were washed, plated onto coverslips, fixed in 3.7% paraformaldehyde, and permeabilized using Tween 80-digitonin-BSA buffer as described before (18). A specific mab to LC3 (Cell Signaling #3868) was used to stain gfpMtb phagosomes followed by Alexa fluor 555 conjugates. Confocal images were acquired using an N90 Nikon microscope equipped with a Metaview RT deconvolution software. Colocalization was determined by scoring gfpMtb phagosomes of 25 MDSCs in triplicate chambers per treatment group (18, 21).

Total RNA was extracted from control or treated MDSCs using the RNAeasy mini kit (Qiagen, Germany). RNA concentration and purity ratios (OD260/280, OD260/230) were measured using the NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, USA). cDNA synthesis was performed on a CFX96 Real-Time PCR System (Biorad, USA) using the 2X OneStep qRT-PCR Mastermix Kit (Applied Biosystems, USA) according to the manufacturer’s instruction. Quantitative PCR (qPCR) was performed using SYBR green probe and gene specific primers [CD80-5’->3’: CTCTTGGTGCTGGCTGGTCTTT Forward primer; 5’->3’: GCCAGTAGATGCGAGTTTGTGC Reverse primer; CD206-5’->3’: AGCCAACACCAGCTCCTCAAGA Forward primer; 5’->3’: CAAAACGCTCGCGCATTGTCCA Reverse primer; GBP2-5’->3’: AGCCAACACCAGCTCCTCAAGA Forward primer; 5’->3’: CAAAACGCTCGCGCATTGTCCA Reverse primer; IRF3-5’->3’: TCGTGATGGTCAAGGTTGT Forward primer; 5’->3’: AGGTCCACAGTATTCTCCAG Reverse primer; IRF5-5’->3’: ATGCTGCCTCTGACCGA Forward primer; 5’->3’: GCCGAAGAGTTCCACCTG Reverse primer; IRF7-5’->3’: GAGCCCTTACCTCCCCTGTTAT Forward primer; 5’->3’: CCACTGCAGCCCCTCATAG Reverse primer]. Threshold cycle numbers were transformed to ΔΔCt values, and the results were expressed relative to the reference gene, β-actin, and GAPDH. Gene expression data were performed using GraphPad Prism ver. 6.0 suite (GraphPad Software). Student’s t-test was used for means comparison between both uninfected and Mtb-infected MDSCs. Significance was set at the 0.05 level.

MDSCs were collected and centrifuged for 5 min at 400g and lysed by using RIPA buffer in the presence of protease inhibitors. For the analysis of proteins, the quantitative Wes capillary immunoassay was used, in which the lysates were separated and detected using Wes separation capillary cartridge 12–230 kDa along with the Wes Anti-Rabbit Detection Module (Simple Western system and Compass Software, Protein Simple). ( Supplementary Figure 2 ) In brief, glass microcapillaries were loaded with stacking and separation matrices followed by sample loading. During capillary electrophoresis, proteins were separated by size and then immobilized to the capillary wall. Samples were loaded at 1 mg/ml dilution and the primary rabbit antibody was used at a dilution of 1:50 and β-actin (Rabbit monoclonal, Sigma-Aldrich # SAB5600204) was used at 1:50 dilution. The antibodies for iNOS (Abclonal, #A0312) and STAT1 (#14994T), p-STAT1 (#9167S), ERK1/2 (#4695T), p-ERK1/2 (#4370 T), p38 MAPK (# 8690T), p-p38 MAPK (#4511T), NF-kB (#8242T), and p-NF-kB (#3033T) were purchased from Cell Signaling Technology. Data were analyzed with the Compass software (version 2.6.7). The area under the curve (AUC), which represents the signal intensity of the chemiluminescent reaction, was analyzed for all the antibodies and β-actin. Values given for protein expression were normalized to β-actin. Quantitation of protein levels (area under each peak; arbitrary units [A.U.]) were performed using the Compass software (version 2.6.7).

For the analysis of LC3B protein, a traditional Western blot was used as the protein was of small molecular weight. Protein concentrations were determined using the Bradford Assay Reagent (Thermo Scientific: cat. no. 23238) and 25 µg of total protein was loaded per well of 4%–20% Mini-Protean TGX gels (Bio-Rad #4561095) and transferred to nitrocellulose membranes (Trans-Blot Turbo Transfer pack, Bio-Rad #1704158) using the Bio-Rad trans-blot turbo transfer system. Antibody against LC3B was used at a dilution of 1:500 (LC3B Antibody; Novus Biologicals: #NB100-2220SS). Secondary antibody was added (1:1,000 Anti-rabbit IgG, HRP-linked Antibody, Cell Signaling #7074s), and the membranes were then developed using an ECL kit. Densitometry analysis was done using ImageJ software and LC3 band density was normalized relative to GAPDH/β-actin (Anti-GAPDH antibody, Rabbit monoclonal, Sigma-Aldrich #SAB5600208) or β-actin (Anti-β Actin antibody, Rabbit monoclonal, Sigma-Aldrich #SAB5600204).

During natural infection, Mtb are inhaled and phagocytized by alveolar MФs. They infiltrate into the lung parenchyma and secrete chemokines to recruit interstitial and blood-deirved monocyte-MФs, neutrophils, dendritic cells, and various subsets of T cells to form an organzied structure called granuloma (22). TB granulomas contain a central core of Mtb-infected macrophages surrounded by macrophages, T cells, DCs, and PMNs. Interestingly, MDSCs also occur in and around tuberculosis granulomas and in the peripheral blood of TB patients (10). To determine their frequency in circulation, both PBMCs purified on Ficoll and whole blood RBC lysis preparations, from tuberculosis patients (n = 5) and heathy donors (n = 5), were analyzed using CyTOF. Figure 1 illustrates that neutrophilic/granulocytic-type N-MDSCs were higher among PBMCs of TB patients and there was a marked reduction in the activation of CD8+T cells compared to CD4 T cells. Although M-MDSCs were lesser in PBMCs, herein, we pursued their analysis of because of their potential role during TB. CyTOF analysis was done to confirm the presence of N-MDSCs and M-MDSCs in samples of TB patients and compared with donors; Supplementary Figure 3 shows the t-SNE plots of immune cell composition and Supplementary Table 1 indicates the clinical status of TB patients and controls.

Using CD14 and CD33 bead fractionation, M-MDSCs were purified from PBMCs of healthy donors. They were infected for 4 h with Mtb, washed, and then treated with either mab-B2 or its isotype followed by incubation for Mtb growth assay. Figure 2A indicates that the Mtb CFU counts progressively declined over 3 days, whereas their numbers remained stable in untreated or IgG isotype-treated MDSCs. Because MDSCs were infected for 4 h, we sought to determine if there are differences in the uptake of Mtb after a 4-h infection. Microscopic counting of Mtb within MDSCs using acid fast mycobacterial stains did not show a significant difference in the uptake by naive MDSCs (day 0 CFU; Figure 2A ). Thus, we demonstrate that LILRB2 mab-mediated antagonism rapidly enhances the killing of Mtb, increasing to a 0.8-log10 decline in CFU by 3 days.

Earlier, we have shown that mab-B2 blockade of LILRB2 switches the MDSC phenotype into an M1-MФ phenotype (16). To determine whether mab-B2 induces a similar switch, Mtb-infected MDSCs were collected on days 1 and 3 post mab-B2 treatment, and tested for the mRNA expression of CD80 and CD206, which are markers of M1-MФs and M2-MФs, respectively (23). In addition, we evaluated mRNA for Guanylate binding protein 2 (GBP2), which is an IFN-γ-inducible marker associated with the control of mycobacterial infection in mice (24–26). Mtb-infected MDSCs treated with mab-B2 showed an upregulation of CD80 and a downregulation of CD206 consistent with an M1-MФ phenotype ( Figure 2B ). GBP2, a biomarker for M1-MФs, was maintained between day 1 and day 3 after LILRB2 blockade.

Since MФs can kill Mtb using NO, we sought to determine whether mab-B2 induced iNOS, which, in turn, can generate NO to boost the bactericial funciton of MDSCs. iNOS synthesis in MDSCs was therefore blocked using siRNA vs. iNOS followed by Mtb infection and incubation with mab-B2 or IgG isotype. Treatment with mab-B2 reduced the Mtb viable counts compared to IgG-treated MDSCs ( Figure 2C ). Moroever, mab-B2-induced intracellular killing of Mtb was abrogated when iNOS was inhibited ( Figures 2C, D ). Replicate cultures of similarly treated MDSCs showed that mab-B2-enhanced NO was confirmed by the presence of nitrite in the medium detected using a diaminonapthalene (DAN) fluorescent indicator and fluorometry (19) ( Figure 2D ). Western blot analysis of iNOS confirmed that the barely detetable iNOS protein in Mtb-infected MDSCs was enhanced after mab-B2 treatment ( Figures 2E, F ). Specific siRNA knockdown of iNOS but not scramled siRNA reduced the levels of iNOS protein ( Figure 2F ). We note here that LILRB2 signals through the src-homolgy containing phosphatases (SHP)-1 and 2 (14). Because SHP-1 mediates negative reguation of iNOS (27), we suggest that mab-B2-LILRB2 blockade suppressed SHP-1 to increase NO, essentially switching MDSCs to an M1-MФ-like phenotype and controlling Mtb growth.

Since MФs can also induce the killing of Mtb using ROS, we sought to determine whether blockdade of LILRB2 induces the ROS; ROS was marginally increased in Mtb-infected MDSC after mab-B2 treatment ( Figure 2G ). Together, these data confirm that mab-B2 induces Mtb-infected MDSCs to switch into an M1-MФ like phenotype, increasing NO and ROS and controlling the growth of Mtb.

Besides oxidants like NO and ROS, MФs can also kill Mtb by sorting Mtb phagosomes to lysosomes through either Rab GTPase and SNARE-dependent phagosome–lysosome (PL) fusion or through autophago-lysosome fusion regulated by ATGs (28, 29). To determine whether autophagy plays a role after LILRB2 antagonism, MDSCs were subjected to siRNA knockdown of beclin1 (ATG6) and scrambled siRNA controls. Following Mtb infection, MDSCs were treated with mab-B2 or IgG isotype and incubation for Mtb growth assay. Unexpectedly, siRNA-mediated autophagy blockade of MDSCs did not reverse the Mb growth in mab-B2-treated cells ( Figure 3A ). This was interesting because mab-B2 induced an enrichment of the autophagy marker protein, microtubule-associated light chain-3 (LC3-II) ( Figures 3B, C ). During autophagolysosome fusion, the lipidated form of LC3-II is enriched on the phagosomes of Mtb. Fluorescent immunostaining using antibodies to LC3 also confirmed that many gfpMtb phagosomes were enriched for LC3 labeling ( Figures 3D, E ). Therefore, a lack of autophagy-mediated killing of Mtb in MDSCs was in contrast with LC3 enrichment of their phagosomes. However, we note that Mtb phagosomes can transiently label with LC3 even without causing the autophagic flux, which delivers them to lysosomes (30). Alternatively, Mtb may evade mab-B2-induced autophagy through secretion of sapM phosphatase, which cleaves phosphatidyl-3 inositol phosphate (PI3P) that is associated with phagosomes preventing autophagolysosome fusion. In this direction, we reported that a ΔfbpAΔsapM mutant of Mtb is lysosome fusion competent (31, 32).

Because LILRB2 blockade led to increased killing of Mtb by MDSCs ( Figure 2A ), we sought to determine whether it can also “pre-condition” MDSCs to prevent a subsequent Mtb infection and growth. First, CD33+ MDSCs were purified using FACsort from donor PBMCs and treated with mab-B2 and LILRB3-specific mab-B3, followed 1 day later by infection with Mtb and growth assay over 7 days. Figure 4A indicates that mab-B2 and mab-B3 (antagonistic for LILRB3) treatment signficantly reduced the uptake of Mtb by MDSCs on day 1. However, both mabs still activated MDSCs to kill Mtb over 7 days. Next, CD14+CD33+ MDSCs bead purified from PBMCs cultured in GM-CSF/IL-6 (used for Figures 1 – 4 ) were treated using mab-B2/3 followed by Mtb infection and CFU assay. Similar to MDSCs directly purified from blood ( Figure 4A ), culture-grown MDSCs also enhanced the killing Mtb upon LILRB2/3 blockade despite the difference in uptake of Mtb on day 1 ( Figure 4B ).

Inflammatory phenotypes of M1-MФs are associated with NF-κB/STAT1 activation (33). Because LILRB2 antagonism led to an M1-type switch in MDSCs ( Figure 2B ), we evaluated downstream signaling mechanisms after mab treatment in Mtb-infected MDSCs. LILRB2 mab-treated MDSCs showed increased ERK1/2, p38 MAPK, and NF-κB/STAT1 phosphorylation in Mtb-infected MDSCs after Mtb infection compared to IgG controls ( Figure 5A ). We reported earlier that LILRB receptors constitutively recruit and activate SHP-1 (SH2 domain-containing protein tyrosine phosphatase-2) in macrophages (16). Consistent with this observation, LILRB2-associated SHP-1 phosphorylation was reduced in mab-B2 treated Mtb-infected MDSCs compared to IgG controls ( Figure 5B ). These data suggest that LILRB2 antagonism disrupts SHP-1 and increases downstream inflammatory signaling cascades through phosphorylated ERK1/2, p38 MAPK pathways, and NF-κB and STAT1 transcription factors.

Because LILRB2 antagonism in Mtb-infected MDSCs reduced SHP-1 phosphorylation ( Figure 5B ), we pharmacologically blocked SHP1/2 to boost mab-B2-mediated antagonism. To demonstrate synergy, Mtb-infected MDSCs were plated for CFUs by 18 h post treatment. Remarkably, mab-B2 in combination with TPI-1 and IACS, which are inhibitors of SHP-1 and SHP-2, respectively, showed a rapid and synergistic reduction of Mtb burden in MDSCs. Synergy suggests that LILRB2 may not solely go through the SHP1/2 pathway; alternatively, other LILRB members may signal through SHP1/2. MDSCs treated with mab-B2 showed 0.6 log10 reduction of Mtb CFU compared to IgG control by day 3, whereas we found ~1.8 log reduction of Mtb CFUs in the mab-B2–SHP1/2 inhibitor combination ( Figure 6A ). This observation correlated with an increased phosphorylation of ERK1/2 and p38 MAPK activation compared to MDSC treated with IgG and SHP-1 and SHP-2 inhibitors ( Figure 6B ). Thus, LILRB2 antagonism through mab in combination with the disruption of SHP1/2 enhances ERK1/2, p38 MAPK, and the robust killing of intracellular Mtb.

A downstream mechanism of SHP2 is its inhibition of transcriptional activity of IRF-1 and IRF-8 at the IFN-β promoter. Whereas IRF1, IRF5, and IRF8 promote M1-type polarization of macrophages, IRF3 and IRF4 promote M2 differentiation (34). LILRB2 treatment with mab-B2 upregulated IRF3 and IRF5 compared to IgG controls in Mtb-infected MDSCs ( Figure 7A ), confirming a switch of MDSCs into the M1 phenotype. To determine whether IRF changes are associated with cytokine secretion, MDSCs were treated with mab-B2, SHP1/2 inhibitors (TPI-1 and IACS), or their combination followed by cytokine assay. Treatment with mab-B2 alone was enough to increase pro-inflammatory TNF-α, whereas a combination of mab-B2 blockade and SHP1/2 inhibitor led to elevated TNF-α, IL-6, and IL-1β ( Figure 7B ). In contrast, mab-B2 or SHP1/2 inhibitors decreased the levels of anti-inflammatory IL-10. We recall here that both IL-1β and TNF-α are macrophage-activating cytokines (35, 36).