Presence of Infected Gr-1intCD11bhiCD11cint Monocytic Myeloid Derived Suppressor Cells Subverts T Cell Response and Is Associated With Impaired Dendritic Cell Function in Mycobacterium avium-Infected Mice

Myeloid-derived suppressor cells (MDSC) are immature myeloid cells with immunomodulatory function. To study the mechanism by which MDSC affect antimicrobial immunity, we infected mice with two M. avium strains of differential virulence, highly virulent Mycobacterium avium subsp. avium strain 25291 (MAA) and low virulent Mycobacterium avium subsp. hominissuis strain 104 (MAH). Intraperitoneal infection with MAA, but not MAH, caused severe disease and massive splenic infiltration of monocytic MDSC (M-MDSC; Gr-1intCD11bhiCD11cint) expressing inducible NO synthase (Nos2) and bearing high numbers of mycobacteria. Depletion experiments demonstrated that M-MDSC were essential for disease progression. NO production by M-MDSC influenced antigen-uptake and processing by dendritic cells and proliferation of CD4+ T cells. M-MDSC were also induced in MAA-infected mice lacking Nos2. In these mice CD4+ T cell expansion and control of infection were restored. However, T cell inhibition was only partially relieved and arginase (Arg) 1-expressing M-MDSC were accumulated. Likewise, inhibition of Arg1 also partially rescued T cell proliferation. Thus, mycobacterial virulence results in the induction of M-MDSC that block the T cell response in a Nos2- and Arg1-dependent manner.


INTRODUCTION
Non-tuberculous mycobacteria (NTM, also known as ubiquitous or environmental mycobacteria or mycobacteria other than tuberculosis) can lead to serious infections in humans and animals. In recent years, the prevalence and incidence of human NTM infections have increased at an alarming rate, mostly due to the rising number of immunocompromised patients (1)(2)(3)(4)(5).
Among the numerous NTM species, Mycobacterium (M.) avium is a paradigm for a pathogenic NTM. It is most frequent cause of infections (1). M. avium, which belongs to the M. avium intracellulare complex, comprises three major subspecies, M. avium subsp. avium (MAA), M. avium subsp. hominissuis (MAH) and M. avium subsp. paratuberculosis (MAP) (6). The M. avium subspecies differ strongly in their host range, virulence and tissue tropism (6,7). MAA causes tuberculosis in birds and is a potential zoonotic and opportunistic pathogen in humans (7). MAP is the well-known causative agent of Johne's disease, a chronic fatal enteritis of ruminants (8). MAH can cause systemic disease in immunocompromised as well as localized disease in immunocompetent humans (9,10). All M. avium subspecies are known to elicit chronic infections and granuloma formation in inbred mouse models (11). However, degree and outcome of such infections vary between subspecies and individual bacterial strains (11)(12)(13).
It is now well established that pathogenic mycobacteria not only reside in macrophages, but also in other phagocytes including myeloid derived suppressor cells (MDSC). MDSC represent a heterogeneous population of immature myeloid cells. They are broadly characterized by co-expressing the myeloid lineage differentiation antigen Gr-1 (also known as Ly6C/G) and CD11b (also known as α M -integrin). MDSC can be further subdivided into polymorphonuclear MDSC (PMN-MDSC; CD11b + Ly6G + Ly6C − ) and monocytic MDSC (M-MDSC; CD11b + Ly6G − Ly6C hi ) (14). M-MDSC usually lack surface markers of inflammatory monocytes such as CD11c and MHC class II (15,16). In mice, normal bone marrow contains 20-30% of cells with MDSC phenotype. In contrast, only a low number is found in naive spleen (2-4%) and they are absent from lymph nodes (17). The number of MDSC can dramatically expand during pathologic conditions, like cancer or infection (17).
MDSC use a plethora of effector molecules to regulate innate and adaptive immune responses. Most commonly employed are inducible nitric oxide (NO) synthase (Nos2) and arginase 1 (Arg1) (14,17). NO might be the major effector molecule to attenuate immune responses during inflammation and cancer (18,19). On the other hand, local production of microbicidal NO at the site of infection was also shown to control various viral, bacterial and parasitic pathogens (20)(21)(22).
Early studies already reported the generation of T cellsuppressing macrophages during the course of tuberculosis in humans. Similarly, suppressor macrophages were found in mice experimentally infected with M. tuberculosis (MTB) or NTM species (23)(24)(25)(26). Today, these cells are defined as MDSC and have been described in MTB as well as other bacterial infections (27,28). Despite several previous reports, the role and phenotype of MDSC in mycobacterial infection is still not completely understood and controversies exist (29).
In the present study we dissected the role of M-MDSC in M. avium-infected C57BL/6 mice that are usually able to control mycobacterial infection. Large numbers of NOproducing M-MDSC were induced which were found to be responsible for disease aggravation. These M-MDSC used NO and Arg1 to subvert the immune response in MAA-infected mice by directly inhibiting T cell proliferation and indirectly by a novel mechanism, i.e., by impeding antigen-processing and presentation by conventional dendritic cells. Together, our data for the first time clearly demonstrate that the induction of Nos2-and Arg1-expressing M-MDSC provides an immune escape mechanism that supports infection and pathology by virulent M. avium subspecies.

Mice
Female C57BL/6J mice were purchased from Janvier (La Genestsaint-Isle, France) and maintained under specific pathogenic free conditions (SPF) at the animal facility of the Helmholtz Centre for Infection Research (HZI), Braunschweig, Germany. Thy1.1 OVA albumin transgenic II (OT-II) C57B6L/J mice were bred at HZI. Nos2 −/− C57BL6/J mice were obtained from the breeding facility of the Universitätsklinikum Erlangen. Mice were infected at the age of 7-12 weeks. In all experiments mice were placed on ad-libitum feeding and unlimited access to water. Mice were maintained for maximum of 5 weeks after infection. All animals experiments were conducted in accordance with the German law for animal protection (Tierschutzgesetz). Approval of study was granted from research ethics committee of the local authority LAVES in Lower Saxony (permission No. 3392 42502-04-13/1192).

Growth of Mycobacterial Strains and Infection
Bacterial strains for infection were grown in Middlebrook 7H9 broth. Mice were infected intraperitoneally (i.p.) by injection of 200 µl (∼10 8 CFU) of the bacterial suspension. Bacterial strains (M. avium ATCC 25291 and M. avium 104) were grown in Middlebrook 7H9 broth (BD) supplemented with 0.5% glycerol and 10% OADC. To attain early logarithmic phase of bacterial growth, broth was inoculated with mycobacteria to an optical density 600 (OD 600nm ) of 0.2 and cultures were grown at 37 • C under stirring (130 rpm) until a final OD of ∼1. The bacterial culture was washed 3 times with Dulbecco's phosphate buffered saline (DPBS). To avoid bacterial clumping, the suspension was intensively vortexed including 3 mm glass beads. Bacterial suspension was adjusted to OD 600 of 5 in DPBS for infection. In addition, the same OD of heat inactivated bacteria (85 • C for 15 min) was applied in some experiments.

Flow Cytometry and Cells Sorting
Spleen cell suspensions were prepared by gently flushing the organs with Iscove's complete medium (IMDM, 10% heat inactivated fetal calf serum, penicillin 100 unit/ml, streptomycin 100 µg/ml, 2 mM L-glutamine, 50 µM 2-mercaptoethanol). Then cells were filtered through 70 µm and finally through 50 µm diameter cell strainers. Red blood cells were removed by erythrocyte lysis buffer (14.2 mM sodium hydrogen carbonate (NaHCO 3 ), 155 mM ammonium chloride (NH 4 Cl), 0.1 mM EDTA, at final pH of 7.3). Cells were stained in FACS buffer (PBS containing 2 mM EDTA, 2% FBS). Anti-CD16/32 (clone 2.4G2, FCR block) was applied before staining with specific antibodies. Dead cells were excluded with DAPI staining. Separate staining was done for each fluorochrome conjugated antibody to determine positive and negative cell populations. All antibodies used, with their respective clones are indicated in Table S1. Data were acquired on LSR II analyzer (BD, NJ, USA). Data analysis was done using FACSDiva software (BD) or FlowJo (TreeStar). Cell sorting was done on BD FACSAria-II. Re-analysis of sorted cells was done for purity check. The gating strategy is provided in Figure S5.

Intracellular Staining
For intracellular staining, 10 7 spleen cells were incubated in 200 µl IMDM containing 5 µg brefeldin A (BioLegend) for 2 h at 37 • C. Cells were stained for surface markers using standard staining protocol. Then cells were fixed in fix/perm buffer (BD), stained for intracellular markers in perm/wash buffer (BD) and analyzed using flow cytometry.

Ex vivo T Cell Proliferation Assay
Untouched CD4 + T cells were purified from infected and PBS control mice spleen. Isolated cells were stimulated (2 × 10 5 cells/well) with either plate bound anti-CD3 (5 µg/ml, clone 45-2c11) or in combination with soluble anti-CD28 (500 ng/ml, clone 37.51). Four days after stimulation, percentage proliferation of cells was analyzed by CFSE dilution.

Nos2 Dependent ex vivo T Cell Inhibition Assay
Infection induced spleen M-MDSC were sorted and co-cultured with naïve, CFSE labeled CD4 + T cells in the presence of immobilized anti-CD3 antibody (5 µg/ml). Different ratios of T cells and M-MDSC were tested in the presence or absence of Nos2 inhibitor, L-NIL (Cayman chemicals) at a final concentration of 40 µM. T cells proliferation was measured after 4 days in culture.

In vivo T Cell Proliferation
Naïve CFSE labeled Thy1.1 OT-II CD4 + T cells (2 × 10 6 ) were injected intravenously via lateral tail vein. After 24 h, 200 µg ovalbumin protein was injected i.p. Three days after immunization, mice were sacrificed and spleen was collected. In vivo proliferation of Th1.1 expressing CD4 + T cells was monitored by CFSE dilution using flow cytometry. The gating strategy is provided in Figure S5B.

Antigen Uptake and Processing Assay
A standard antigen uptake (OVA-cy5 endocytosis) and processing (DQ-OVA degradation) protocol was followed (30). Briefly, cDC were incubated with 100 µg/ml OVA-cy5 or 62.5 µg/ml DQ-OVA for 90 min and analyzed using flow cytometry. To determine the rate and compartment in which degradation of DQ-OVA takes place in cDC, spleen cDC were sorted and seeded on poly-L-lysine (Life science, Sigma Aldrich) coated coverslip at a density of 2 × 10 5 cells per well overnight. After changing to new complete medium, cells were pulsed with 62.5 µg/ml DQ-OVA and incubated for 45 min at 37 • C. After 3x washings, cells were further incubated for 2 h at 37 • C. Then cells were washed and fixed in 20% eBioscience TM fixation buffer. Fixed cells were stained with lysosome associated membrane protein 1 (LAMP-1) antibodies (BioLegend). Finally, images were taken by confocal fluorescent microscopy.

Depletion of Gr-1 Expressing Cells
Eleven days after infection, 250 µg of anti-Gr-1 antibody (clone RB6-8C5) was given intraperitoneally at day 11, 14 and 17 post infection (p.i.) and mice were sacrificed on day 20. Control mice were injected with similar concentration of rat IgG (Dianova). Spleen cells were prepared for flow cytometry and livers were collected for histology and plating.

Splenocytes Nitrite Assay
Spleen single cell suspensions were prepared following the above standard protocol. A total of 10 7 cells were seeded in 2 ml complete IMDM in 6-well cell culture plates and incubated for 48 h. Cell culture supernatants were collected and nitrite content analyzed using the Griess reagent system (Promega).

In vitro NO Susceptibility Testing
For testing nitric oxide susceptibility bacteria were harvested at early logarithmic growth and diluted in PBS − to an OD 600nm of 0.01. Following 10 µl (∼10 4 CFU) were incubated in PBS − containing two different concentrations of nitric oxide donor Snitrosoglutathione (GSNO) (Sigma) for 4 h at 37 • C. Plating was done on Middlebrook 7H10 agar to compare the CFU with and without treatment. The relative survival was calculated referring the number of CFU from bacteria incubated in the presence of the NO donor to that of the input CFU.

Quantification of Intracellular Bacteria by PCR
Whole cell (eukaryotic and bacterial) DNA was extracted from sorted and 3% paraformaldehyde fixed cells. Briefly, zirconium beads were added and cells were disrupted using a tissue homogenizer. The homogenate was sonicated using Branson sonifier 450. Supernatant was collected after centrifugation. After adding an equal volume of TE buffer, RNA was removed by adding 10 µl RNase A (Roche) followed by a 1 h incubation at 37 • C. To reverse the cross link, 15 µl of 4M NaCl was added and incubated for 5 h at 65 • C. Finally, DNA was extracted using standard phenol chloroform extraction method. Bacterial DNA PCR was performed using the IS901 specific primers and normalized against eukaryotic Cxcl2 (Mip2a) promoter. IS901: for_GTGATCAAGCACCTTCGGAA, rev_GCTGCGAGTTGCTTGATGAG; Mip2a: for_GAAGGG CAGGGCAGTAGAAT, rev_ ATGGCGCTAGGCTGAAGTG.

ELISA
Blood was collected via cardiac puncture in 500 Serum-Gel tubes (Sarstedt). Serum was separated by centrifugation at 10,000 rpm for 4 min at room temperature and kept at −80 • C until analysis. Serum concentration of IL-6 and CCL5 were determined by using PEPROTECH kit following the manufacturer's kit protocol. In addition serum concentration of cytokines and chemokines were quantified by LUMINEX based mouse cytokine 23-plex assay following manufacturer's instruction (Bio-Rad, USA).

Histopathology
Upon isolation, specimens were fixed with 4% (v/v) formalin and embedded in paraffin. Approximately 3 µm thick sections were cut and stained with hematoxylin/eosin according to standard laboratory procedures. Immuno-histo-chemical staining was performed to detect cleaved caspase-3 (Asp175), using the 3,3 ′diaminobenzidine, Zytomed Systems DAB530 as chromogen. Hematoxylin was used for counterstaining. Immunofluorescence staining was performed using following primary antibodies: polyclonal rabbit anti-mouse Nos2 (eBioscience), polyclonal goat anti-human Arg1 (Santa Cruz Biotechnology), anti-mouse MAC-2 (Biozol diagnostics) and self-produced anti-heparin binding hemagglutinin (HBHA). Alexa Fluor anti-rabbit/goat 488 and Alexa Fluor anti-rabbit 594 were used as secondary antibodies. Sections were analyzed by light blinded to the experimental groups.

Ziehl-Neelsen (ZN) Staining
Sorted spleen cells were fixed in 3% paraformaldehyde on ice for 10 min. Fixed cells were suspended in FACS buffer. Fifty micro liter of this cell suspension was spun on the glass slide using a cytospin centrifuge. ZN staining was done following a standard protocol.

Immunoblot Analysis
Splenic myeloid cells were sorted from naïve and infected mice into CD11b + CD11c int , CD11b + CD11c − cell populations according to CD11b and CD11c expression. Cell lysates were prepared in 40 mM Tris buffer (pH 8.0) including 1mM PMSF and a protease inhibitor mix (cOmplete mini EDTA-free, Roche). After sonication, cell lysates were separated on 10% SDS-PAGE and transferred to PVDF membrane (0.45 µm, Millipore Immobilon-P, 1 h, constant current of 1 A) using tank block technique. Nonspecific binding sites were blocked by blocking buffer (5% dry milk, 0.1% Tween 20 in PBS) for 1 h. Target proteins were stained by primary antibodies (goat anti-Arg1, clone V-20, Santa Cruz Biotechnology; mouse anti-GRB2, clone 81/GRB2, BD Biosciences) and HRP conjugated secondary antibodies (donkey anti-goat, goat anti-mouse, both Dianova). Reaction signals were detected by ECL based chemiluminescence. Images were processed with help of the Adobe Photoshop CS5software (Adobe Systems, San José, CA, USA).

Statistical Analysis
Data were entered into graph pad prism version 5 software. Mean ± standard error of the mean (Mean ± SEM) was used for data description. Statistical test between two groups was determined using Student's t-test. Difference between more than two groups was determined either with one way or two way analysis of variance (ANOVA) using Dunnett's multiple comparisons against the PBS control group. One way ANOVA Bonferroni's multiple comparison tests has been used in some figures. CFU counts were compared by Mann Whitney U test. Cut off p <0.05 was considered as statistically significant difference ( * p < 0.05, * * p ≤ 0.01, * * * p < 0.001).

Chronic MAA Infection Induces Accumulation of Mycobacteria Harboring Histiocytic Cells in Murine Spleen
It is now clear that pathogenic mycobacteria are capable to infect diverse subsets of myeloid cells. To gain more insight into the role of such myeloid cells during progressive M. avium infection, we infected mice intraperitoneally with highly virulent M. avium subsp. avium (strain ATCC 25291), in the following abbreviated as MAA, or with the genetically closely related, but less virulent M. avium subspecies hominissuis strain 104 (MAH). MAA-infected mice lost body weight upon infection and were not able to regain it during the course of observation. In contrast, MAH-infected mice quickly recovered after an initial weight loss ( Figure 1A). We also found high numbers of bacteria (determined as colony-forming units [CFU]) in the liver of MAA-infected mice at 5 weeks p.i., whereas the bacterial burden was roughly 1,000-fold lower in MAH-infected mice ( Figure 1B).
In accordance with the differential weight loss induced by the two bacterial strains, significant differences were seen in the blood levels of proinflammatory cytokines. High concentrations of IL-6 and CCL5, both of which are involved in regulating immune cell migration, were detected by ELISA in MAA-infected mice, but not in mice infected with MAH ( Figure 1C). Similarly, multiplex analysis revealed differential expression of IL-1β, IL-1α, IL-5, TNF, and CCL3, whereas most other cytokines and chemokines including IFN-γ were comparably upregulated by the two mycobacterial strains ( Figure S1A).
Defined granulomas were formed in the spleens of MAHinfected mice. In contrast, diffuse inflammation dominated by histiocytes and loss of defined lymphoid follicles was seen in the spleens of MAA-infected mice ( Figure 1D). In addition, classical granuloma harboring mononuclear cells and peripheral lymphocytes were observed in the liver of MAH-infected mice, whereas granuloma with higher numbers of mononuclear cells and low numbers of lymphocytes were found in the livers of MAA-infected mice ( Figure 1D).
To study the diffuse histiocytic granulomatous inflammation in more detail, hematoxylin/eosin (HE) and Ziehl-Neelsen (ZN) stainings of the spleens of MAA-infected mice were analyzed at different time points (Figure 1E). At 1 week of infection, there were little changes of splenic structures; lymphoid follicles were clearly visible. However, after 3 weeks the general architecture of the spleen changed completely. Lymphoid follicles disappeared and the progressive granulomatous inflammation with increasing numbers of myeloid cells became apparent. Furthermore, numerous acid-fast bacteria became detectable by ZN staining (Figure 1E). In fact, a 10 5 -fold increase in bacterial CFU numbers between week 1 and week 3 p.i. was observed ( Figure S1B).

Chronic MAA Infection Induces Accumulation of Disease-Aggravating Gr-1 int CD11b hi CD11c int Cells in the Spleen
To investigate the mechanism underlying the different host reactions in MAA-versus MAH-infected mice, we analyzed the monocytic cell populations in the spleens of infected mice in more detail. MAA and MAH infection induced accumulation of Gr-1 hi as well as Gr-1 int cells. Especially in MAA-infected mice, the numbers of Gr-1 int CD11b hi cells were strongly increased compared to Gr-1 hi CD11b hi cells (Figure 2A, Figures S2A,B). Furthermore an accumulation of CD11b hi CD11c int cells was detected in spleens of mice infected by MAA but not in MAH-infected mice or mice treated with heat inactivated MAA (Figures 2B,C, Figure S2C). These cells were Gr-1 int (Figure 3A).
The massive influx of myeloid cells was associated with changes in the lymphoid compartment (Figure S1C). At 5 weeks after MAA infection there was a striking reduction of lymphoid cells. The numbers of splenic B cells, CD4 + and CD8 + T cells were 3-to 4-fold lower when compared to spleens from uninfected mice. Reduction of CD4 + T cells was also observed in MAH-infected mice but to a considerably lower degree ( Figure S1D).
We tested whether increased apoptosis accounted for the lymphoid ablation. However, enhanced levels of activated caspase 3 indicative for apoptosis were not detected by immunohistochemistry. Instead, apoptotic events in the spleens of infected mice appeared to decrease with time of infection ( Figure S1E).
To determine whether Gr-1 + splenic myeloid cells contributed to pathology, we depleted Gr-1-expressing cells in MAA-and MAH-infected mice. Accordingly, anti-Gr-1 depleting antibody was administered intraperitoneally on day 11, 14, and 17 and mice were sacrificed on day 20. As seen by flow cytometry, anti-Gr-1 treatment reduced the number of Gr-1 int CD11b hi CD11c int cells in spleens of MAA-infected mice to the level observed in MAH-infected mice, while the numbers of other Gr-1 int CD11b hi cells were slightly increased in depleted MAA-infected mice and of Gr-1 hi CD11b hi cells were not influenced. In contrast, Gr-1 depletion did not affect the myeloid cell pool in MAH-infected mice (Figures 2D-F,  Figures S2D-G). This is consistent with previous reports where anti-Gr-1 treatment results in increased granulopoiesis after initial depletion (32,33) At necropsy, the splenomegaly induced by MAA was reduced by the anti-Gr-1 treatment. This was not observed in MAH-infected mice treated in the same manner ( Figure 2G). Plating of liver homogenates from MAA-or MAH-infected Gr-1-depleted mice revealed that the anti-Gr-1 treatment reduced the bacterial burden only in MAA-infected mice. Correspondingly, histo-morphometry of HE stained liver sections revealed reduced granuloma sizes and liver pathology in MAA-infected, but not in MAH-infected mice after anti-Gr-1 treatment (Figures 2H-J, Figure S2H). Together, these results clearly demonstrate a strong impact of Gr-1 int CD11b hi CD11c int cells for disease aggravation in MAA-infected mice.
Gr-1 int CD11b hi CD11c int Ly6c hi Cells Are Monocytic Myeloid Derived Suppressor Cells and Heavily Infected With MAA Next, we further characterized the Gr-1 int CD11b hi cells mediating disease aggravation. To this end, we analyzed CD11b hi CD11c neg (P3) and CD11b hi CD11c int (P4) cells for their Gr-1 and Ly6C expression. The CD11b hi CD11c neg cells could be separated into Gr-1 hi CD11b hi CD11c neg Ly6C int (P6) and Gr-1 int CD11b hi CD11c neg Ly6C hi (P7) cells ( Figure 3A). In addition, nearly all CD11b hi CD11c int (P4) were found to be monocytic Gr-1 int CD11b hi CD11c int Ly6C hi cells (P5). Cytospins of these cells were ZN stained to analyze mycobacterial load. As shown in Figure 3B, MAA was almost exclusively found in the monocytic Gr-1 int CD11b hi CD11c int Ly6C hi cell population (P5), whereas the other two monocytic populations (P6 and P7) were largely ZN-negative.   To corroborate the findings of the ZN staining, DNA was extracted from the sorted cells and bacterial content was evaluated by qPCR using primers for the mycobacteria-specific insertion sequence 901 (IS901). Signals were normalized to the number of host cells by qPCR using primers for the promoter of macrophage inflammatory protein 2 (mip2). As expected, this analysis confirmed the almost exclusive presence of MAA DNA in the monocytic Gr-1 int CD11b hi CD11c int Ly6C hi cell population (P5; Figure 3C).
Next we extracted RNA from sorted splenic CD11b hi CD11c int and CD11b hi CD11c neg cells (Figure 3 P4 and P3, respectively) and mRNA expression of selected genes was determined by qRT-PCR. Infected CD11b hi CD11c int cells expressed comparatively little Il1b (Figure 4A). In contrast, mRNA expression of other pro-inflammatory cytokines like Ifng, Tnf as well as Il6 was considerably higher than in CD11b hi CD11c neg cells from infected and control animals. Likewise, mRNA expression of the anti-inflammatory cytokine Il10 as well as of Arg1 and Nos2 were markedly higher in CD11b hi CD11c int cells as compared to uninfected control mice or CD11b hi CD11c neg cells from MAAinfected animals (Figure 4A).
The mRNA expression profile of infected cells was confirmed by immuno-histochemistry. Tissue sections of spleens of MAAinfected mice were stained with antibodies against the common monocyte/macrophage marker Mac2 (galectin 3) (34) and the mycobacterial hemagglutinin binding protein HBHA (35). As shown in Figure 4, MAA was associated with Mac2 expressing cells. In addition, mycobacteria-positive areas could be stained with antibodies against Nos2 or Arg1, indicating that myeloid cells in such areas expressed Nos2 and some of them also Arg-1 (Figures 4B-D). Furthermore, strong intracellular Nos2 and TNF expression was observed for CD11b hi CD11c int cells by flow cytometry (Figure 4E). Finally, medium conditioned by splenocytes from MAA-infected mice contained large amounts of nitrite as stabile metabolite of NO as detected by Griess reaction (Figure 4F). Overall, the immature phenotype of the MAA-infected monocytic Gr-1 int CD11b hi CD11c int Ly6C hi cell population as well as the expression of TNF, Nos2, and Arg1 strongly suggested that these cells represent a monocytic myeloid-derived suppressor cell (M-MDSC) population.
The residence of MAA in these NO-producing cells and the proliferation in an NO/nitrite-containing milieu of the spleen suggested resistance of the MAA strain against NO. Indeed, treatment of MAA and MAH with the NO donor GSNO (4 and 8 mM) for 4 h did not influence viability of MAA although it considerably affected MAH ( Figure 4G).
NO Produced by Gr-1 int CD11b hi CD11c int M-MDSC Is Responsible for CD4 + T Cell Ablation and Disease Control NO can exert a direct pro-apoptotic and/or anti-inflammatory effect in T cells (19,36). Even though apoptotic events were low at the tissue level, the question arose whether ablation of T and B cells in the spleen of MAA-infected mice is due to production of NO by Gr-1 int CD11b hi CD11c int M-MDSC. Therefore, we infected Nos2 −/− mice with MAA. After 5 weeks, spleens of MAA-infected wild-type (wt) and Nos2 −/− mice showed a massive influx of histiocytic cells. In contrast to wt mice, MAAinfected Nos2 −/− mice exhibited mostly granuloma harboring more outer rim lymphocytes ( Figure 5A).
Interestingly, spleens of MAA-infected Nos2 −/− mice exhibited similar frequency of M-MDSC-like Gr-1 int CD11b hi CD11c int cells as MAA-infected wt mice ( Figure 5B). However, in contrast to wt Gr-1 int CD11b hi CD11c int cells, Nos2 −/− Gr-1 int CD11b hi CD11c int cells harbored lower amounts of bacteria ( Figure 5C). Ablation of B cells and CD8 + T cells remained unaltered in Nos2 −/− mice, whereas the CD4 + T cell compartment was fully restored ( Figure 5D). Thus, in chronically MAA-infected mice NO produced by Gr-1 int CD11b hi CD11c int M-MDSC appears to exacerbate the infection and affects specifically CD4 + T cells.

NO Produced by Gr1 int CD11b hi CD11c int M-MDSC Affects CD4 + T Cell Responses and cDC Function ex vivo
NO derived from M-MDSC or other myeloid cells is capable to directly suppress T cells activation (19,36). To see whether this phenomenon is also seen in MAA-or MAH-infected mice, we isolated splenic CD4 + T cells and stimulated them with anti-CD3 or anti-CD3 plus anti-CD28. As shown in Figure 6A, CD4 + T cells from MAA-infected mice exhibited significantly reduced proliferation compared to uninfected control mice or CD4 + T cells from MAH-infected mice. Next, we tested whether the suppressive activity of splenic Gr-1 int CD11b hi CD11c int M-MDSC from MAA-infected mice is dependent on Nos2-derived NO. Isolated control (naïve) CD4 + T cells were stimulated with anti-CD3 in the presence or absence of splenic Gr-1 int CD11b hi CD11c int cells from MAA-infected mice. Strong proliferation was observed in the absence of Gr-1 int CD11b hi CD11c int M-MDSC ( Figure 6B). In contrast, when Gr-1 int CD11b hi CD11c int M-MDSC from MAA-infected mice were present, proliferation was significantly lower. Inhibition correlated with the ratio of M-MDSC to T cells ( Figure 6C). Importantly, the inhibitory activity of Gr-1 int CD11b hi CD11c int M-MDSC was completely abolished, when the Nos2 inhibitor L-N 6 -(1-iminoethyl) lysine di-hydrochloride (L-NIL) was added to the co-cultures (Figures 6B,C). These data suggest that Gr-1 int CD11b hi CD11c int M-MDSC from MAA-infected mice exhibit Nos2/NO-dependent inhibitory activity toward T cells ex vivo.
While NO can directly inhibit CD4 + T cell proliferation, NO is also known to modulate the function of antigenpresenting cells (20,21,37). Thus, we analyzed the dominant cell populations expressing CD11c, i.e. conventional dendritic cells (cDC; CD11c high CD11b +/− ) as well as Gr-1 int CD11b hi CD11c int M-MDSC for surface markers involved in the activation or inhibition of T cells. MHC class II and costimulatory CD86 were found on cDC and Gr-1 int CD11b hi CD11c int M-MDSC from PBS-treated control mice and were upregulated upon MAA infection ( Figure 6D). Interestingly, expression of programmed death ligand 1 (PD-L1) was significantly higher on Gr-1 int CD11b hi CD11c int M-MDSC of MAA-infected mice as compared to cDC, which would be compatible with direct involvement of Gr-1 int CD11b hi CD11c int M-MDSC in negative T cell regulation.
For functional testing, we first investigated T cell stimulatory capacity of cDC and Gr-1 int CD11b hi CD11c int M-MDSC. Isolated cDC (>95% purity) from MAA-infected or control mice were sensitized with ovalbumin (OVA) and co-incubated with OVA specific OTII CD4 + T cells. Strong proliferation was observed with cDC from uninfected, PBS-treated mice, whereas cDC from MAA-infected mice were strikingly impaired in their T cell stimulatory capacity (Figure 6E, top). This could indicate a defect in the antigen-processing ability as OVA protein requires processing. Therefore, the experiment was repeated by using an OVA-derived peptide as antigen which does not require processing. Strong OTII T cell proliferation was elicited by both cDC populations under these conditions (Figure 6E,  bottom). Apparently, antigen-presentation and T cell stimulation is undisturbed in cDC from MAA-infected mice, but antigenprocessing might be altered. This was in agreement with the sustained expression of MHC-II and the CD86 co-stimulatory molecules. In contrast, Gr-1 int CD11b hi CD11c int M-MDSC had considerably lower capacity to directly stimulate OT-II T cells with either protein or peptide (Figure 6E).
It is possible that splenic cDC from MAA-infected mice could be directly infected by mycobacteria. This might affect their antigen-processing compartments as has already been suggested for infected macrophages (38). However, cytospins of sorted cDC from MAA-infected mice revealed that hardly any cDC were positive for mycobacteria ( Figure S3A).
We then studied whether impaired antigen-processing is a general feature of mycobacterial infection or specific for cDC from MAA-infected mice. Splenic cDC were sorted from MAHor MAA-infected mice and assessed for their capacity to process and present antigen. As shown in (Figure 6F, Figure S3B), impairment of cDC sensitized with OVA protein was specific for MAA-infected mice, while cDC from MAH-infected mice were fully capable to induce CD4 + T cell proliferation. The T cell stimulatory capacity was retained when OVA peptide was used as seen before (Figures 6E,F, bottom). Inhibition of antigenprocessing required viable mycobacteria since administration of heat-killed MAA did not result in suppression of T cell stimulation ( Figure S3C).
The impaired antigen-processing capacity of cDC from MAA-infected mice might be due to NO produced by the M-MDSC. Therefore, splenic cDC were sorted from MAAinfected wt or Nos2 −/− mice and assessed for their capacity to process and present antigen. As shown in Figure 6G, cDC from infected Nos2 −/− mice induced significantly higher OTII T-cell proliferation than cDC from infected wt mice.
Next, we analyzed whether the antigen uptake or processing was affected in cDC from MAA-infected wt mice. As shown in Figures 6H,I, OVA-Cy5 uptake of cDC from infected wt mice was severely impaired, while it was improved in cDC from infected Nos2 −/− mice. Surprisingly, cDC from all groups of mice had comparable levels of DQ-OVA signals, which is an indicator for intact antigen-processing. To localize the trafficking of DQ-OVA in the cellular compartments, cDC from wt and Nos2 −/− mice were stained with LAMP-1, an indicator of late endosome. As shown in Figure S4, co-localization of DQ-OVA and LAMP-1 was observed in both groups. Since cDC from MAA-infected wt mice apparently took up antigen less efficiently, there must be a compensatory enhancement of antigen processing in such cells.
Overall, these data provide clear evidence that in the spleens of MAA-infected mice NO derived from M-MDSC is also responsible for a defect in the T cell stimulatory capacity of cDC. Nevertheless we cannot exclude the possibility that other NO producing cells might be involved.

Occurrence of Gr-1 int CD11b hi CD11c int M-MDSC Correlates With T Cell Inhibition in MAA-Infected Mice
The above data clearly demonstrated that splenic Gr-1 int CD11b hi CD11c int M-MDSC as well as cDC from MAA-infected mice when analyzed ex vivo hindered CD4 + T cell proliferation. These effects were Nos2/NO dependent. As the in vivo situation is significantly more complex, we also analyzed T cell proliferation in MAA-infected mice. To this end, mice were infected with MAA. Then, OVA-specific CD4 + OT-II T cells labeled with CFSE were adoptively transferred at 1, 3, 4, and 5 weeks p.i. according to the schedule displayed in Figure 7A. After 24 h, antigen-specific in vivo proliferation was induced by i. p. administration of OVA. Three days later, proliferation of CD4 + OT-II T cells was evaluated. Consistent with the histology shown before (Figure 1D), in vivo CD4 + T cell proliferation was not affected during the initial phase of infection. However, after 3 weeks an inhibitory effect on proliferation was first observed. By 5 weeks of infection, T cell proliferation was strongly reduced. This inhibition was only seen when viable bacteria were applied ( Figure 7B). Thus, T cell-inhibitory activity positively correlated with severity of inflammation and with the appearance of Gr-1 int CD11b hi CD11c int M-MDSC in the spleen of MAA infected mice (Figure 1).
To investigate, whether in vivo inhibition of T cell proliferation was dependent on NO as observed ex vivo, we tested OTII CD4 + proliferation using MAA-infected Nos2 −/− mice. Similar to the ex vivo results in Figure 6, CD4 + OT-II T cell proliferation was significantly improved in Nos2 −/− mice infected with MAA. However, proliferation did not reach the levels observed in uninfected mice (Figure 7C). This suggested that even though NO production by Gr-1 int CD11b hi CD11c int M-MDSC was abolished in Nos2 −/− mice, alternative T cell-inhibitory mechanisms were still active. As the population of Gr-1 int CD11b hi CD11c int M-MDSC was numerically unaltered in MAA-infected Nos2 −/− mice, we assumed that M-MDSC inhibited T cell proliferation under in vivo conditions via compensatory mechanisms such as the expression of Arg1 which we had already seen in M-MDSC from MAA-infected mice ( Figure 4A). Indeed, higher Arg1 protein expression was observed in splenic sections of MAAinfected mice in the absence of Nos2 (Figure 7D). This was confirmed at cellular level. Gr-1 int CD11b hi CD11c int M-MDSC sorted from MAA-infected mice lacking Nos2 expressed roughly 100-fold higher mRNA levels of Arg1 than Gr-1 int CD11b hi CD11c int M-MDSC from infected wt mice.
Expression of mRNA encoding Tnf was reduced and Il10 was nearly unaffected (Figure 7E). Upregulation of Arg1 protein in Gr-1 int CD11b hi CD11c int M-MDSC in absence of Nos2 was confirmed by immunoblot ( Figure 7F). Thus, in Nos2 −/− mice the lack of Nos2 i.e. NO appeared to be functionally compensated by Arg1.
Arg1 expressed by M-MDSC is known to cause T cell inhibition by depletion of the semi-essential amino acid Larginine (19,39). Therefore, to evaluate the effect of Arg1 expression, we treated MAA-infected wt mice or Nos2 −/− mice with the arginase inhibitor N ω -hydroxy-nor-L-arginine (nor-NOHA) 2 days before and after OTII cell transfer and immunization for consecutive 6 days. As shown in Figure 7G, this treatment partially restored T cell proliferation in wt mice indicating that in vivo Gr-1 int CD11b hi CD11c int M-MDSC also impede T cell activation via Arg1 expression. Unexpectedly the treatment did not alter T cell activation in Nos2 −/− mice ( Figure 7G) suggesting a markedly higher Arg1 activity in such mice.

DISCUSSION
In the present study, we unraveled the functional role of a M-MDSC population that is particularly induced after infection of mice with a highly virulent strain of M. avium subsp. avium. This cell population exhibited a hitherto unknown CD11c int phenotype and considerably expanded during the progression of the disease. M-MDSC acted as target cells for mycobacterial growth, created an inflammatory milieu prone to cause T cell suppression and aggravated the infection by the expression of Nos2 and Arg1.
The interaction of CD4 + T cells with antigen-presenting cells plays a critical role in controlling mycobacterial disease (40,41). Studies on bacterial survival and CD4 + T cell responses after infection of mice with M. avium revealed that the result of infections and immunopathology strongly depended on the subspecies and strain of M. avium investigated and the route of infection (12,(42)(43)(44). Therefore, to dissect the relevance of myeloid cells during chronic M. avium infection of mice, we used the MAA strain 25291, commonly used as a highly virulent model organism for disseminated mycobacterial disease (45), and the MAH strain 104, which shows intermediate virulence (11,46). We selected the intraperitoneal route of infection as it results in fast systemic dissemination and reproducible chronic infection with granulomatous inflammation in deep organs.
Five weeks after inoculation, mice exhibited systemic mycobacterial infection regardless of the strain used. As expected, more severe disease and higher bacterial burden were observed in MAA-infected mice. However, granulomatous inflammation in liver and spleen was induced by both strains as described before (11,12,47). Importantly, the quality of the granulomatous inflammation in both liver and spleen considerably differed between mice infected with MAA or MAH. The granulomatous inflammation in MAA-infected mice was particularly characterized by the appearance of increasing numbers of mononuclear cells. The high production of IL-6 and CCL5 in MAA-infected mice is most likely responsible for this accumulation. Both cytokines have been shown to induce proliferation of MDSC in tumors (48,49).
Phenotypic characterization of the splenic monocytic population in MAA-and MAH-infected mice revealed that the overall number of Gr-1 and CD11b expressing cells increased. However, accumulation of Gr-1 int cells in MAA-infected mice was significantly higher. Cells expressing Gr-1 and CD11b markers have been described as MDSC (50). Such cells are classified as polymorphonuclear or PMN-MDSC, when expressing CD11b + Ly6G + Ly6C lo , and as M-MDSC, when expressing CD11b + Ly6G − Ly6C hi (51). M-MDSC usually lack surface markers of monocytes such as CD11c and MHC class II (15). In contrast, Gr-1 int cells in MAA-infected mice exhibited a Gr-1 int CD11b hi CD11c int phenotype, expressed high levels of Ly6C as well as MHCII and CD86. Nevertheless, these cells were unable to present antigen to CD4 + T cells but fulfilled criteria of M-MDSC, both functionally and phenotypically (14). They expressed PD-L1, IL-10, Arg1 and Nos2, produced NO and inhibited T cell activation.
The immune-regulatory activity of M-MDSC on T cells largely depends on the metabolic consumption of arginine by the activity of the inducible enzymes Nos2 or Arg1 (14,17). M-MDSC thereby create an arginine starved milieu which prevents T cell growth (52). In addition, NO suppresses T cell function through various mechanisms, including the alteration of signaling cascades in T cells, the inhibition of MHC class II expression, and the induction T cell apoptosis (17,19,20). Indeed, immunohistochemistry revealed that in MAA-infected mice splenic M-MDSC harboring mycobacteria expressed high levels of Nos2 and moderate levels of Arg1 protein. Furthermore, we were able to show that the ablation of CD4 + T cells observed in the spleens of MAA-infected mice was restored in Nos2 −/− mice. This pointed toward an Nos2/NO-mediated immunosuppressive activity of the Gr-1 int CD11b hi CD11c int M-MDSC. In agreement, our ex vivo analyses of Gr-1 int CD11b hi CD11c int M-MDSC from MAA-infected mice showed that these cells were able to inhibit T cell proliferation in a Nos2/NO-dependent manner.
During mycobacterial infection, DC execute two key functions: participating in antigen presentation and in granuloma formation (53). In the present study we showed that the suppressive activity of Gr-1 int CD11b hi CD11c int M-MDSC was not only limited to T cells but also affected the splenic DC population. Absence of T cell stimulatory capacity seems to be specific for DC from MAA-infected mice and partially restored in Nos2 −/− mice. Obviously, NO from M-MDSC influenced the up-take and/or processing of protein antigen, but not its presentation, since the presentation of peptide remained intact. Our findings are reminiscent to the findings of Zietara et al. in RAG deficient mice (30). They are a novel example to an emerging number of reports demonstrating a direct or indirect influence of MDSC on other immune cells including DC. For instance, MDSC have been reported to impair DC functions in mouse tumor models and thereby aggravated tumor-induced immune suppression and tumor growth (54)(55)(56). However, our data, for the first time demonstrate a critical role of NO produced by M-MDSC in these processes.
In addition to DC, CD4 + T cells of MAA-infected mice were impaired as well. Splenic T cells purified from MAAinfected mice were not responsive to CD3/CD28 stimulation. They could not be stimulated ex vivo. Ex vivo data suggested that the lack of T cell response was due to NO production by Gr-1 int CD11b hi CD11c int M-MDSC. To confirm these findings in the more complex in vivo situation, we took advantage of adoptive transfer of ovalbumin-specific OT II cells. These experiments clearly demonstrated that during late phases of MAA infection antigen-specific proliferation of CD4 + T cells was inhibited. The extent of inhibition of OT II cell proliferation correlated with the extent of pathology in the spleens and the number of mycobacteria-containing cells. The inhibition of T cells in the spleen appeared to be of general nature and not restricted to MAA-specific CD4 + T cells.
Interestingly, Gr-1 int CD11b hi CD11c int M-MDSC were even induced in the absence of Nos2/NO as shown using Nos2 −/− mice. This is in line with studies on MTB infected Nos2 −/− mice (29). However, it contrasts with studies on tumors where Nos2/NO were shown to be essential for the induction of MDSC (57).
The deletion of Nos2 had a positive effect on the overall number of CD4 + T cells emphasizing the detrimental effect of Nos2/NO on the proliferation of these cells. However, OT II T cell proliferation was only partially restored in Nos2 −/− mice, indicating that Gr-1 int CD11b hi CD11c int M-MDSC are able to express an alternative mechanism of suppression. We showed that in the absence of Nos2/NO, Arg1 influenced OT II cell proliferation in MAA-infected mice. Indeed, Arg1 was highly expressed under these conditions. It is known that the products of Arg1 and Nos2 reciprocally regulate each other's enzymatic activity (58). Thus, polyamines produced by the Arg1/ornithine decarboxylase pathway are capable to inhibit Nos2 activity, whereas N-hydroxy-L-arginine (LOHA), the intermediate of the Nos2 reaction, functions as an inhibitor of arginase (58). Furthermore, inhibitors of Nos2 transcription have been shown to upregulate Arg1 expression in murine macrophages (59). In our study, we observed that in the absence of Nos2 the expression of TNF mRNA was reduced. As TNF is a potent negative regulator of the expression of Arg1 (22), the deficiency of TNF might have contributed to the enhanced expression of Arg1 seen in Nos2 −/− mice.
In line with previous studies (60), our data confirm that the absence of Nos2 in MAA-infections improved bacterial clearance. This is in remarkable contrast to MTB where infected mice lacking Nos2 suffer from severe TB disease (61)(62)(63) confirming striking differences in the pathogenicity of virulent mycobacteria (60). The higher resistance of Nos2 −/− mice might result from (a) the normalized number and function of CD4 + T cells or (b) the absence of NO-mediated detrimental effects on other immune cells such as DC. Nevertheless, formation of classical granuloma during MAA-infection might be sufficient to reduce bacterial spreading even in the presence of T cell suppressive M-MDSC activity.
In conclusion, the results of this study demonstrated the critical role of M-MDSC in mycobacterial infection. By ex vivo and in vivo analyses we showed that the ability of NTM to cause accumulation of M-MDSC with a unique Gr-1 int CD11b hi CD11c int phenotype is an important virulence trait. The monocytic nature of such M-MDSC rendered them susceptible to mycobacterial replication. Their immature phenotype and their ambiguous response to infection considerably influenced the outcome of the local innate and adaptive immune response, thereby facilitating progression of mycobacterial disease. In addition, our study provided evidence that M-MDSC utilize distinct mechanisms like expression of Nos2 and Arg1 to suppress T cells during mycobacterial infection, either directly or indirectly via a Nos2/NO-dependent impairment of the function of DC.

AUTHOR CONTRIBUTIONS
KA, SW, and RG designed experiments. KA, AN, NJ, NR, UH, VP, and CF performed the experiments. KA, AB, US, CB, DB, SW, and RG analyzed the data. CB provided methodological advice and contributed to the writing of the manuscript. KA, SW, and RG wrote the paper.     Table S1 | List of antibodies for flow cytometry used in this study.