NOS2-derived nitric oxide regulates the size, quantity and quality of granuloma formation in Mycobacterium avium-infected mice without affecting bacterial loads


Dr Ehlers Division of Molecular Infection Biology, Research Center Borstel, Parkallee 22, D-23845 Borstel, Germany.


Granuloma formation in response to mycobacterial infections is associated with increased expression of inducible nitric oxide synthase (NOS2) within granuloma macrophages and increased levels of nitrate/nitrite in the sera of infected mice. Continuous treatment with 5 mm or 10 mm l-N6-(1-imino-ethyl)-lysine (L-NIL), a selective NOS2-inhibitor, in acidified drinking water for up to 7 weeks consistently reduced infection-induced nitrate/nitrite to background levels in mycobacteria-infected BALB/c mice. Oral treatment with 5 mm L-NIL initiated at the time of infection significantly exacerbated growth of Mycobacterium tuberculosis, but had no effect on Mycobacterium avium colony-forming unit development in the liver, spleen and lungs of intravenously infected mice. In order to examine the role of nitric oxide in mycobacteria-induced granulomatous inflammation in the absence of any effect on the bacterial load, M. avium-infected mice were treated with 5 mm L-NIL from day 1 through 38 and the development of granulomatous lesions in the liver was assessed by histology, immunohistology and reverse-transcription–polymerase chain reaction (RT-PCR). Computer- and video-assisted morphometry performed at 4 and 7 weeks post-infection showed that treatment with L-NIL led to markedly increased number, cellularity and size of granulomatous lesions in infected mice regardless of the virulence of the M. avium isolate used for infection. Immunohistology of the liver revealed that in mice treated with L-NIL, the numbers of CD3+ T cells, CD21/35+ B cells, CD11b+ macrophages and RB6-8C5+ granulocytes associated with granulomatous lesions was increased. RT-PCR of the liver showed that in L-NIL-treated mice infected with M. avium, mRNA levels of tumour necrosis factor, interleukin-12p40, interferon-γ, interleukin-10 and interferon-γ-inducible protein-10 (IP-10) were up-regulated, while mRNA levels of interleukin-4, monocyte chemotactic protein-1 (MCP-1) and MCP-5 were similar to those in untreated control infected mice. When M. avium-infected mice were treated with 5 mm L-NIL between the 5th and 12th weeks of infection, similar changes in granuloma number and size were found in the absence of any effect on the bacterial load. These findings demonstrate that nitric oxide regulates the number, size and cellular composition of M. avium-induced granulomas independently of antibacterial effects by modulating the cytokine profile within infected tissues.


Nitric oxide (NO) is a freely diffusible radical gas that is enzymatically generated by oxidative deamination of l-arginine.1 Constitutively expressed isoforms of nitric oxide synthases (NOS) are found in the endothelium and the brain, while an inducible isoform (NOS2) is only transiently expressed in macrophages following cytokine stimulation.1[2]–3 NO has been shown to have potent antimicrobial and antiparasitic effects. Among others, the clearance of Leishmania major and Mycobacterium tuberculosis from infected murine tissues depends on the generation of NO via the up-regulation of NOS2 in mouse macrophages.2, 4[5][6][7]–8

In addition, NO is clearly involved in inflammatory processes, although the role for NOS2-derived NO is less clear than for NOS1- and NOS3-derived NO. Depending on the experimental system used, both proinflammatory and anti-inflammatory effects were described.9[10][11][12][13][14]–15 Thus, NOS2-inhibitors NG-monomethyl-L-arginine (L-NMMA) and LG-nitro-L-arginine methyl ester (L-NAME) significantly attenuated formyl-methionyl-leucyl-phenylanine (fMLP)-induced human peripheral blood monocyte chemotaxis in vitro.16In vivo, clinical symptoms of experimental allergic encephalomyelitis were prevented by inhibition of NOS2-induction,9, 10, 12 and NOS2-inhibitors reduced monocyte/macrophage infiltrations in response to intratracheal instillation of zymosan or silica.17 On the other hand, NOS2-deficient mice have enhanced leucocyte–endothelium interactions in endotoxaemia, suggesting that induction of NOS2 is a negative regulator of leucocyte recruitment.18, 19 NO was also shown to influence cytokine secretion by T cells and may modulate inflammation by inducing apoptosis in regulatory cells.13, 20[21][22]–23

The hallmark of mycobacterial infections is the formation of organized mononuclear cell infiltrations called granulomas. NOS2 expression was previously demonstrated in mycobacteria-or particle-induced granulomas,7, 24, 25 but it is not known whether the generation of NO from within these lesions is involved in regulating the recruitment of new monocytes into the lesion. In a model that uses a non-replicating secondary antigenic stimulus (purified protein derivative; PPD) to induce T-cell-mediated short-term granuloma formation, reduction of NO levels was found to increase monocyte and granulocyte recruitment to the lung.26

Granuloma formation in response to a slowly replicating stimulus in vivo is, of necessity, a more chronic and less synchronized process. Moreover, assessment of the role of NO in granuloma formation during actual infection with mycobacteria is complicated by the fact that some mycobacterial species, such as M. tuberculosis, are exquisitely sensitive to the direct or indirect bactericidal effects of NO. Thus, M. tuberculosis infection is dramatically exacerbated in NOS2-knockout mice.7

M. avium is an opportunistic pathogen mainly afflicting patients with end-stage acquired immune deficiency syndrome (AIDS).27 Most strains of M. avium are resistant to the bacteriostatic effects of NO, at least in vitro.28 In a murine model of infection, M. avium was shown to cause chronically persisting or progressive pathology depending on the strain chosen for infection.29, 30 We therefore selected M. avium infection as a model to investigate the effects of selectively inhibiting NOS2-activity on chronic granuloma development, because bacterial load would probably be unaffected. Our data show that NOS2-derived NO is involved in down-regulating granuloma formation by altering the cellular composition and the cytokine levels at the site of infection in the absence of any discernible effect on mycobacterial load.



Specific pathogen-free BALB/c mice (8–12 weeks old) were purchased from Charles River Wiga (Sulzfeld, Germany) and were sex-matched for any given experiment. Mice were housed under barrier conditions in the animal facilities at the Borstel Research Centre, or, for infection experiments involving M. tuberculosis, in the BL3 animal facilities at Colorado State University, Fort Collins, CO. All experiments were approved by the ethics committee instituted by the Ministry of Nature, Environment and Forestation (Kiel, Germany).


Mycobacterium avium TMC724 (originally obtained from Dr F. Collins, Trudeau Institute, Saranac Lake, NY), M. avium SE01 (an isolate from the blood culture of an AIDS patient) and M. tuberculosis Erdman were passaged in susceptible mice twice and cultured in Middlebrook 7H9 (Difco, Detroit, MI) medium supplemented with OADC (oleic acid, albumin, dextrose, catalase; Becton Dickinson, Heidelberg, Germany) to a mid-logarithmic phase. Aliquots were frozen at −70° until needed. An inoculum of bacteria was prepared by thawing an aliquot and diluting it in phosphate-buffered saline (PBS). Mice were infected intravenously via a lateral tail vein with 105 colony-forming units (CFU) of the indicated mycobacterial strain in 0·2 ml PBS. Mice were anaesthetized and killed at the indicated time-points to follow the course of infection. Organs were removed aseptically and homogenized in 10 ml distilled water to determine bacterial loads by plating serial tenfold dilutions of whole organ homogenates on nutrient Middlebrook 7H10 agar (Difco) supplemented with OADC. Bacterial colony numbers (CFU) were determined after 14–21 days of incubation at 37° in humidified air. The natural course of infection and the kinetics of granuloma formation in mice infected with these strains has been described previously.29

Treatment with L-N6-(1-imino-ethyl)-lysine

The l-N6-(1-imino-ethyl)-lysine (L-NIL) was purchased from Alexis (Läufelfingen, Switzerland) and dissolved in water (adjusted to pH 2·7 in order to prevent microbial growth). Fresh acidified drinking water with or without the indicated concentration of L-NIL was provided to the mice every other day. Except for the experiment detailed in Fig. 7, treatment was started at the time of infection and continued for the indicated number of days. Water and food consumption was similar in control and treatment groups. A toxicity study previously demonstrated that L-NIL at the concentrations used here does not influence water or food intake or impair weight gain in naive mice.31

Figure 7.

Bacterial loads and granuloma size and number in BALB/c mice infected with M. avium SE01 and treated with L-NIL in the chronic phase of infection. BALB/c mice were infected with M. avium SE01 and treated from day 35 to day 84 with 5 mm L-NIL (open columns) or left untreated (filled columns). Bacterial CFU (a), granuloma number (b) and granuloma area (c) were determined on liver tissue obtained at day 84 post-infection. Data represent the means±SD from five mice per group. *P<0·005.


One cranial and one caudal liver lobe per mouse was fixed in 4% formalin-PBS, set in paraffin blocks, sectioned (2–3 μm), and stained using haematoxylin and eosin (HE). Granuloma numbers were determined by counting focal mononuclear infiltrations in five non-sequential sections derived from both cranial and caudal liver lobes per animal (four to five mice per group) in a superimposed 0·25 cm2 grid. For the purpose of quantification, a granuloma was defined as the focal accumulation of more than nine mononuclear cells. Data represent the means of 20 determinations±SD.


At least 100 randomly selected granulomas per group – from at least five non-sequential HE-stained sections, taken from two different liver lobes of four to five mice per group – were video-scanned at ×40 magnification (Panasonic KR 222 camera; Main Actor v1.6 software from ELSA, Aachen, Germany) and the granuloma area was analyzed using Optimas 6.2. Bioscan software (Optimas corporation, Bothell, WA) after calibration. An image of 0·506 mm2 was represented by 704×576 pixels.


For formalin-fixed tissues, sections were deparaffinated and placed in 10 mm sodium citrate buffer (pH 6) followed by pressure-cooking for exactly 1 min.24 After blocking for 20 min in 1% H2O2 solution, slides were incubated with appropriately diluted polyclonal rabbit anti-mouse-NOS2 (Genzyme-Virotech, Rüsselsheim, Germany) in tris-buffered saline (TBS)/10% fetal calf serum (FCS) for 30 min in a humid chamber. As bridging antibody, appropriately diluted goat-anti-rabbit-IgG-peroxidase (Dianova, Hamburg, Germany) and as tertiary antibody, diluted rabbit-anti-goat-IgG-peroxidase (Dianova) was used in sequential incubations of 30 min each. For the detection of CD3+ cells, a rat anti-mammalian CD3 monoclonal antibody (mAb; clone CD3-12; Biotrend, Cologne, Germany) was used as a primary antibody, diluted rabbit-anti-rat-IgG (Dianova) as a secondary, and goat-anti-rabbit-IgG peroxidase as a tertiary antibody.For the detection of proliferating cells, a rabbit-anti-mouse-Ki-67 antiserum was used.32 Development was performed with 3-3′-diaminobenzidine (DAB; Sigma, Deisenhofen, Germany) and urea superoxide (Sigma), and haemalum was used to counterstain the slides. For the detection of CD4+ cells, CD8+ cells, B cells and granulocytes, frozen tissue sections were prepared using a cryostat (Frigocut E 2800; Leica, Bensheim, Germany). 4-μm sections were air-dried and fixed in acetone before storage at −70°. After acetone–chloroform treatment, sections weere blocked with 0·3% superoxide and incubated with mAb KT174 (anti-CD4), KT15 (anti-CD8α), 5C6 (anti-CD11b), all purchased from Biosource International, Camarillo, CA, 7G8 (anti-CD21/35), purchased from Pharmingen, San Diego, CA, or RB6-8C5 (anti-neutrophil), kindly provided by Dr Rui Appelberg, Porto, Portugal. Incubations with secondary antibodies and subsequent development steps were performed as described for paraffin-embedded tissues.

Reverse-transcription–polymerase chain reaction (RT-PCR)

A detailed protocol of the procedure employed for semi-quantitative RT-PCR was published elsewhere.24 Briefly, weighed liver samples (≈150 mg each) were homogenized in 5 ml 4 m guanidinium-isothiocyanate buffer, diluted to obtain equalized amounts of mg liver/ml buffer, and after acid phenol extraction of total RNA, cDNA was obtained using Moloney murine leukaemia virus (MMLV) -RT (Gibco-BRL, Eggenstein, Germany) and oligo-dT(12–18 mer) (Sigma) as a primer. After amplification (denaturation at 94° for 30 seconds, annealing at 60° for 30 seconds, and extension at 72° for 30 seconds) and electrophoresis on a 2% agarose gel, amplicons were blotted onto a nylon membrane (Hybond N+, Amersham-Pharmacia, Freiburg, Germany), and hybridization was performed at 42° with specific internal oligo probes followed by two washes at room temperature and 45° under increasingly stringent salt conditions. Chemiluminescent labelling and detection of hybridized oligonucleotides were performed using the Amersham ECL-kit and autoradiography for ≈90 min on Hyperfilm-ECL (Amersham-Pharmacia). The following specific cytokine primers and probes and PCR cycle numbers were used: β2-microglobulin (14 cycles): sense 5′-TGACCGGCTTGTATGCTATC-3′, antisense 5′-CAGTGTGAGCCAGGATATAG-3′, probe 5′-GAAGCCGAACATACTGAACTGCTAC-3′; interleukin-10 (IL-10; 26 cycles): sense 5′-CGGGAAGACAATAACTG-3′, antisense 5′-CATTTCCGATAAGGCTTGG-3′, probe 5′-GGACTGGCTTCAGCCAGGTGAAGAC-3′; IL-12p40 (26 cycles): sense 5′-CGTGCTCATGGCTGGTGCAAAG-3′, antisense 5′-CTTCATCTGCAAGTTCTTGGGC-3′, probe 5′-TCTGTCTGCAGAGAAGGTCACA-3′; IL-4 (28 cycles): sense 5′-GAATGTACCAGGAGCCATATC-3′, antisense 5′-CTCAGTACTACGAGTAATCCA-3′, probe 5′-AGGGCTTCCAAGGTGCTTCGCA-3′; interferon-γ (IFN-γ; 24 cycles): sense 5′-AACGCTACACACTGCATCTTGG-3′, antisense 5′-GACTTCAAAGAGTCTGAGG-3′, probe 5′-GGAGGAACTGGCAAAAGGA-3′; tumour necrosis factor (TNF; 22 cycles): sense 5′-GATCTCAAAGACAACCAACTAGTG-3′, antisense 5′CTCCAGCTGGAAGACTCCTCCCAG-3′, probe 5′-CCCGACTACGTGCTCCTCACC-3′; monocyte chemotactic protein-1 (MCP-1; 20 cycles): sense 5′-AGAGAGCCAGACGGAGGAAG-3′, antisense 5′-GTCACACTGGTCACTCCTAC-3′, probe 5′-CCAGATGCAGTTAACGCCC-3′; MCP-5 (24 cycles): sense 5′-GATTTCCACACTTCTATGCC-3′, antisense 5′-CAAGGATGAAGGTTTGAGAC-3′, probe 5′-GCTACAGGAGAATCACAAGC-3′; interferon-γ-inducible protein- 10 (IP-10; 22 cycles): sense 5′-GTGCTGCCGTCATTTTCTGC-3′, antisense 5′-CTTAGATTCCGGATTCAGAC-3′, probe 5′-GAGATCATTGCCACGATGAA-3′.

PCR conditions were first optimized to ensure that the PCR was in the exponential phase. Serial twofold dilutions of a positive control cDNA were included in each reaction and subsequent hybridization to ascertain that the PCR had not reached a plateau. This titration is included in Fig. 6(a) and was used as a calibrated, semi-quantitative scale for comparison of amplicon intensities from different experimental samples. As a control for calibrating an equivalent amount of input cDNA, amplification for β2-microglobulin was performed, and all samples were equalized (if necessary) for β2-microglobulin cDNA content prior to analysis of cytokine cDNA. Quantitation was performed after scanning Hyperfilm-radiographs with a Studiostar-scanner (Agfa, Köln, Germany), and the number of black pixels over background film was measured using Photoshop software (Adobe, Edinburgh, UK) in a preset frame of 3750 total pixels. Statistical analysis was carried out on pixel values obtained from four mice per experimental group. Fold-increase over background was calculated from the titration of the standard cDNA, assigning the titre with background pixel levels an arbitrary value of 1. Fold differences between experimental groups were calculated by comparing fold increases from L-NIL-treated versus untreated mice infected with the same strain.

Figure 6.

Semi-quantitative RT-PCR determination of mRNA expression in the livers of M. avium-infected BALB/c mice following treatment with L-NIL. Mice infected with M. avium TMC724 were killed on day 28 post-infection, mice infected with M. avium SE01 were killed on day 38 post-infection. (a) Liver samples from four mice per group were individually processed for RT-PCR and amplified product was hybridized with a specific internal probe and exposed to X-ray film. Uninfected control mice were analysed in parallel, and a reference cDNA in twofold dilution steps was included in each run for calibration. (b) Evaluation of pixel values and calculated x-fold increases over pixel background values from original data shown in (a). Significant differences in pixel values from L-NIL treated (open columns) and untreated infected mice (filled columns) are indicated. Pixel values from uninfected control mice (hatched columns) are shown for comparison.

Nitrite/nitrate measurements

Plasma was obtained after centrifugation of heparinized blood drawn from the inferior vena cava of anaesthetized mice and stored at −70° until further use. As an indicator of NO production, the concentration of the final reaction products nitrate and nitrite were determined in the plasma by a colorimetric method (Boehringer, Mannheim, Germany), appropriately scaled down to microtitre plate format to accommodate 50 μl samples. For use in this assay, plasma samples were diluted twofold, added to a microconcentrator (cut-off 10 000 MW; Amicon, Beverly, MA) and centrifuged at 3300 g for 15 min followed by 6000 g for 30 min in an Eppendorf microcentrifuge.


Quantifiable data are expressed as the means of individual determinations±SD. Statistical analysis was performed using Student’s t-test for unpaired data or anova for multiple comparisons.


NOS2 expression and NO levels following treatment with L-NIL in mycobacteria-infected mice

Immunohistological staining in mice infected with M. tuberculosis or M. avium demonstrated that NOS2 protein was always present in granulomatous lesions of the liver (data not shown). NOS2 expression was confined to the macrophages located in the centre of granulomas.

In order to verify that treatment with the selective NOS2 inhibitor L-NIL effectively reduced NO levels to those detectable in uninfected mice, mice were treated with either 5 mm or 10 mm L-NIL in acidified drinking water starting on the day of infection and for up to 38 days post-infection. Mice were intravenously infected with 105 CFU of either a virulent strain of M. avium (TMC724), an M. avium strain of intermediate virulence (SE01), or M. tuberculosis Erdman. Figure 1 depicts the reduction of nitrate/nitrite levels in the sera of mice at the time when peak levels were reached in control infected mice, i.e. after 2 weeks of treatment (M. tuberculosis Erdman), 4 weeks of treatment (M. avium TMC 724) or 7 weeks of treatment (M. avium SE01) with 5 mm L-NIL showing that treated and infected mice had levels comparable to uninfected controls. Treatment with 10 mm L-NIL did not further augment this effect (data not shown). Background NO levels were higher than in previously published experiments because no attempt was made to reduce l-arginine intake in the mouse diet. This was done in order not to affect the constitutive levels of NO production that were probably involved in regulating normal vascular tonus.

Figure 1.

Nitrate/nitrite levels in mycobacteria-infected BALB/c mice following treatment with L-NIL. (a) BALB/c mice were infected with M. tuberculosis Erdman and treated with 5 mm L-NIL in acidified drinking water (+L-NIL) or were left untreated (control). Nitrite/nitrate levels were determined in mouse sera obtained 15 days post-infection. (b) BALB/c mice were infected with M. avium TMC724 and treated with 5 mm L-NIL (+L-NIL) or were left untreated (control). Nitrite/nitrate levels were determined in mouse sera obtained 28 days post-infection. (c) BALB/c mice were infected with M. avium SE01 and treated with 5 mm L-NIL (+L-NIL) or were left untreated (control). Nitrite/nitrate levels were determined in mouse sera obtained 38 days post-infection. The data represent the means±SD of four or five mice per group. The mean levels of nitrite/nitrate in the sera of uninfected mice are shown for comparison (n=6). *P<0·05, **P<0·01 (when compared to untreated controls and not statistically different from uninfected controls)

Course of M. tuberculosis and M. avium-infection in L-NIL-treated and control mice

Since treatment with 5 mm L-NIL effectively reduced NO levels, we next determined the effect on the CFU development in infected organs. As expected, M. tuberculosis infection was greatly exacerbated in the liver, spleen and lungs with an increase in CFU counts of up to 2 log10 units when compared to untreated control-infected mice at day 30 of infection, while M. avium CFU numbers (determined at day 28 post-infection for TMC724 or at day 38 post-infection for SE01) were not higher in treated than in untreated controls regardless of the virulence of the M. avium isolate used for infection (Fig. 2).

Figure 2.

Mycobacterial CFU counts in the spleens and livers of infected BALB/c mice following treatment with L-NIL. (a) BALB/c mice were infected with M. tuberculosis Erdman and treated with 5 mm L-NIL in acidified drinking water (open columns) or left untreated (filled columns). Bacterial organ loads were determined on day 30 post-infection. (b) BALB/c mice were infected with M. avium TMC724 and treated with 5 mm L-NIL (open columns) or left untreated (filled columns). Bacterial organ loads were determined on day 28 post-infection. (c) BALB/c mice were infected with M. avium SE01 and treated with 5 mm L-NIL (open columns) or left untreated (filled columns). Bacterial organ loads were determined on day 38 post-infection. The data represent the means±SD of four or five mice per group. *P<0·001 when compared to untreated controls.

NOS2 expression in the granulomatous lesions of infected mice was identical whether treated with L-NIL or not in all groups of mice (see Fig. 5). Nor did L-NIL affect mRNA expression for NOS2, as evidenced by semi-quantitative RT-PCR in the livers of infected animals (data not shown).

Figure 5.

Immunohistological evaluation of the cellular composition of liver granulomas in M. avium-infected BALB/c mice following treatment with L-NIL. BALB/c mice were infected with M. avium TMC724 and treated with 5 mm L-NIL (b,d,f,h,j,l) or left untreated (a,c,e,g,i,k) and killed on day 28 post-infection. Liver sections were stained with mAb specific for CD3 (a,b), CD21/35 (c,d), Ki-67 (e,f), polymorphonuclear leucocytes (g,h), CD11b (i,j), and NOS2 (k,l). Immunoperoxidase stains, ×128. Identical results were obtained following infection with M. avium SE01.

Quantification and morphometric analyses of granulomatous lesions in L-NIL-treated and control M. avium-infected mice

When granulomatous lesions in mycobacteria-infected mice were compared at the same time bacterial CFU were determined, it was evident that in mice treated with L-NIL, granulomas were much larger and more numerous (Fig. 3). This was due to both increased cellularity in the peripheral cuff of granulomas and the appearance of a more differentiated macrophage phenotype with larger cytoplasm in L-NIL-treated mice. Since in the case of M. tuberculosis infection, this effect could be accounted for by the increase in bacterial CFU which probably accelerated the inflammatory response, quantitative assessments of granuloma number and size were subsequently carried out only for M. avium-infected mice. Quantification in non-sequential liver sections showed that there were about two- to threefold more granulomas in L-NIL-treated mice than in control infected mice, regardless of the M. avium strain used for infection (Fig. 4; P<0·01). Video- and computer-assisted morphometry showed that the average size of granulomas was also enlarged approximately twofold in L-NIL-treated mice, averaging 4440 and 6018 μm2 for untreated SE01- or TMC724-infected mice, respectively, and 10 480 and 7680 μm2 for L-NIL-treated SE01- or TMC724-infected mice (Fig. 4; P<0·01)). When L-NIL treatment was performed in SCID mice infected with M. avium, there was no difference either in the size or in the number of granulomas induced in response to M. avium infection (data not shown).

Figure 3.

Microscopic morphology of granulomas in the livers of mycobacteria-infected BALB/c mice following treatment with L-NIL. (a,b) BALB/c mice were infected with M. tuberculosis Erdman and treated with 5 mm L-NIL (b) or left untreated (a). Liver sections were prepared from mice killed on day 30 post-infection. HE×128. (c,d) BALB/c mice were infected with M. avium TMC724 and treated with 5 mm L-NIL (d) or left untreated (a). Liver sections were prepared from mice killed on day 28 post-infection. HE 128. (e,f) BALB/c mice were infected with M. avium SE01 and treated with 5 mm L-NIL (f) or left untreated (e). Liver sections were prepared from mice killed on day 38 post-infection. HE×128.

Figure 4.

Granuloma number and size in the livers of M. avium-infected BALB/c mice following treatment with L-NIL. (a,c) BALB/c mice were infected with M. avium TMC724 and treated with 5 mm L-NIL (open columns) or left untreated (filled columns) and killed on day 28 post-infection. (b,d) BALB/c mice were infected with M. avium SE01 and treated with 5 mm L-NIL (open columns) or left untreated (filled columns) and killed on day 38 post-infection. (a,b) The number of granulomas per 0·25 cm2 were determined in HE-stained sections from different areas of the liver (five sections per mouse, four mice per group); the data represent the means of 20 determinations±SD. (c,d) The granuloma area was determined by video-assisted morphometry on different, non-sequential HE-stained sections from different areas of the liver of four mice per group; the data represent the means of 100 determinations±SD. *P<0·005.

Immunohistological analyses of granulomatous lesions in L-NIL-treated and control M. avium-infected mice

In order to determine whether a particular cellular component in the enlarged granulomas made up for the increased cellularity within the peripheral cuff in L-NIL-treated mice, immunohistological staining was performed employing mAb with specificity for granulocytes, macrophages, B and T cells. All types of cells investigated were found to be increased in numbers in granulomas of L-NIL-treated mice, regardless of the M. avium strain used for infection (Fig. 5a–l). CD3+ T cells were increased within the lymphocytic cuff surrounding epithelioid macrophages, and this increase was accounted for by both CD4+ and CD8+ T cells (data not shown). Staining for the proliferation-associated antigen Ki-6732 was not quantitatively different between groups, particularly when the increased granuloma size was taken into account (Fig. 5f). This showed that cells in L-NIL-treated mice were not actively dividing within the lesion to a greater extent than in untreated mice. Some cells which occasionally appeared in small clusters in the periphery of the granuloma defied definition by the staining mAb used in this study.

Semi-quantitative RT-PCR in the livers of L-NIL-treated and control M. avium-infected mice

In order to elucidate the mechanism by which nitric oxide might mediate its regulatory effect on the cellular composition in granulomatous lesions, semi-quantitative RT-PCR was performed on liver samples taken at the same time that bacterial loads and granuloma morphology were assessed (Fig. 6a). It was evident that while mRNA levels for IL-4, MCP-1 and MCP-5 were similar in treated and untreated mice, mRNA for IL-12p40, IFN-γ, TNF, IL-10 and IP-10 were enhanced at 7 weeks post-infection with SE01, and to a lesser extent at 4 weeks post-infection with TMC724 (Fig. 6b). Specifically, mRNA expression for IL-12p40 was increased two- to threefold in mice treated with L-NIL when compared to untreated infected mice, regardless of whether infection was performed with TMC724 or with SE01. Similarly, IP-10 mRNA levels were approximately two- to threefold higher in L-NIL-treated mice following infection with either strain. However, IFN-γ mRNA expression appeared similar in the livers of TMC724-infected mice treated with L-NIL when compared to untreated control infected mice, but was increased fourfold in mice infected with SE01and treated with L-NIL. Likewise, TNF mRNA expression was similar in mice infected with TMC724 and treated with L-NIL when compared to untreated infected mice, but was increased 10-fold in SE01-infected mice treated with L-NIL. IL-10 mRNA levels were found to be approximately twofold higher in SE01-infected mice treated with L-NIL when compared to untreated infected mice infected with that strain. Thus, differential cytokine gene expression during treatment with L-NIL differed quantitatively depending on the strain of M. avium used for infection, presumably because of the different kinetics of the inflammatory response to infection with either strain.

Effects of L-NIL treatment in mice chronically infected with M. avium

The experiments documented above show the effect of L-NIL treatment during the induction phase of granuloma formation. In order to assess the role of NOS2-derived NO during the phase of granuloma maintenance, mice infected with either TMC724 or SE01 were treated with 5 mm L-NIL between days 35 and 84 of infection. There was no effect of treatment on CFU counts in the liver, spleen and lungs of TMC724- (data not shown) or SE01-infected mice (Fig. 7a). Again, granuloma number and size were increased approximately twofold in SE01-infected mice after treatment (Fig. 7b,c). Granulomas also appeared more numerous and larger in TMC724-infected mice. However, since beyond 5 weeks of infection granulomas coalesced in TMC724-infected mice, it was impossible to quantify the exact granuloma number and size in these mice. Immunohistology revealed that the cellular changes observed during the early phase of infection were also apparent when L-NIL treatment was carried out in the chronic phase of infection (data not shown). Thus, regardless of whether treatment was initiated early or late during infection, granuloma size, number and cellular composition were consistently altered although bacterial loads in the organs of infected mice remained unaffected.


The model of chronic M. avium infection was used to assess the effects of manipulating NOS2 enzyme activity on granuloma formation induced by mycobacterial infection. In the absence of any discernible effect on M. avium replication, NOS2 activity was found to regulate granuloma development, in that treatment with L-NIL, a selective and potent NOS2 inhibitor, significantly increased the number, size and cellularity of granulomas in M. avium-infected mice, both during the early phase of granuloma induction and during the chronic phase of granuloma maintenance. Reduced NO levels brought about by treatment with L-NIL were associated with an increased number of T cells, B cells, granulocytes and macrophages within the granulomatous lesions. Similarly, L-NIL-treated mice infected with M. avium had increased mRNA levels for TNF, IFN-γ, IL-10 and IP-10 within infected livers.

In accordance with previous studies using M. tuberculosis-infected mice, we confirmed that NOS2-inhibitors dramatically exacerbate infection, leading to toxic organ loads and widespread tissue necrosis.4 In view of this effect on CFU counts, the M. tuberculosis model provided no unambiguous access to studying the effect of NO on granuloma formation. In contrast, since NOS2-inhibitors failed to show any untoward effect on M. avium growth, regardless of the virulence of the isolate used for infection, experimental M. avium infection afforded a unique opportunity to study NO regulation of chronic inflammatory processes.

Our data corroborate and extend those of Hogaboam et al. who used a model of acute T-cell-mediated granuloma formation induced by non-persisting antigens (purified protein derivative of M. tuberculosis (PPD)) and schistosomal egg antigen (SEA) in primed mice.26 Those investigators also found that selective NOS2-inhibitors led to increased granuloma size and number in their model and attributed the increase in cellularity to enhanced recruitment of neutrophils and eosinophils.26 The PPD-model is a short-lived model of acute myelomonocytic cell recruitment, and granulomas wane ≈4–6 days after induction. Thus, the differences in both cellular composition and chemokine mRNA expression evident between those authors’ studies and our own probably reflect the entirely different time frame examined here. We studied long-term granuloma development in chronically M. avium-infected mice, and treatment with L-NIL was carried out for 4–7 weeks during the phase of granuloma initiation or for 7 weeks during a later phase of granuloma maintenance. We cannot exclude a more pronounced effect on early neutrophil and eosinophil recruitment from our own data; however, we did not document a specific influx of polymorphonuclear leucocytes per granulomatous lesion over other cell types in L-NIL-treated mice when comparing them to untreated infected controls. In view of the fact that L-NIL-treated mice had more granulomas, it is however, obvious that overall granulocyte numbers within the livers of L-NIL-treated mice were higher than in untreated infected mice, in a similar way as other cell types (T cells, B cells and macrophages) were found to be increased in numbers.

In experimental M. avium infection, granuloma formation is accelerated by specific CD4+ T cells and depends to a large extent on TNF and IFN-γ levels reached within the infected tissues.29, 33, 34 Our data show that there was an increase in CD3+ T cells (accounted for by both CD4+ and CD8+ T cells) in the livers of L-NIL-treated infected mice, which was due to an increase in both the average T-cell number per lesion and the average number of granulomas in L-NIL-treated mice. It is likely that the increased number of T cells was directly reponsible for the increased mRNA levels for TNF and IFN-γ (the latter further up-regulating IP-10 expression) observed following L-NIL treatment, explaining accelerated recruitment of inflammatory cells and increased granuloma size. This is corroborated by our experiments using SCID mice, which lack T and B cells and in which L-NIL treatment did not alter the kinetics and morphology of granuloma formation.

The documented increases in mRNA expression for IL-12p40 in L-NIL-treated mice may also be attributed to enhanced T-cell stimulation, as IL-12 can be up-regulated by IFN-γ.35 The observed increased expression of IL-10 mRNA could also be a result of enhanced monocyte recruitment. The fact that the differences in cytokine expression were observed only to a limited extent after infection with TMC724 show that additional factors, such as the virulence of the infecting strain, are involved in the magnitude and kinetics of cytokine gene expression. As mRNA expression was determined at only one time-point, it is possible that a more detailed kinetic study may have revealed a more pronounced differential cytokine profile in L-NIL-treated mice also with the virulent strain. We cannot decide at present whether increases in mRNA expression are the cause or the consequence of an increase in lesion cellularity, or whether cytokine gene expression is directly modulated by NO levels. The studies by Hogaboam et al. documented an increase in mRNA for MCP-1, C10 and macrophage inflammatory protein-1α (MIP-1α) following pharmacological NOS2 inhibition.26 While we did not specifically examine the latter two chemokines, we could not detect differences in mRNA expression for MCP-1 and MCP-5 in L-NIL-treated mice infected with either strain of M. avium. Again, this discrepancy may reflect differences in the time frame studied or the antigenic stimulus used in these independent investigations.

NO has been shown to exert a negative feedback regulatory effect on T cells either by inhibiting cytokine secretion13, 23, 36 or by inducing apoptosis.20, 21 We can exlude the possibility that NO inhibits cellular proliferation within granulomas, as no evidence was found pointing to differential proliferation in treated versus non-treated mice based on staining for Ki-67 antigen.32

Our findings are consistent with the interpretation that NO has a detrimental effect on T-cell viability and/or responsiveness during the natural course of infection, and that treatment with L-NIL alleviates this ‘immunosuppressive’ effect. Interestingly, CFU counts in the spleens of M. avium-infected mice treated with L-NIL were always lower than in untreated control infected mice, although the absolute differences were only small (0·5 log10 CFU/spleen) and reached significance only in one of three experiments (P<0·05). We speculate that the documented increase in CD3+ cells and IFN-γ within the lesion may be responsible for this enhanced antibacterial effect, and it is possible that further prolongation of L-NIL treatment might lead to a further reduction in the bacterial load. Recent experiments in NOS2-knockout mice infected with M. avium indeed confirm this hypothesis in that bacterial CFU counts at 4 months post-infection were significantly lower in the spleens and lungs of these mice than in littermate immunocompetent controls.37

There are a number of conflicting reports in in vitro and in vivo systems demonstrating virtually opposite effects of NO modulation on recruitment of cells of the myelomonocytic lineage.2, 9, 11, 14[15]–16, 18, 19, 38 These discrepancies are partially due to the use of less specific inhibitors of NOS-enzymes and partially due to different read-out systems, e.g. acute versus chronic inflammatory events. Our study in a chronic infection model using a slowly replicating mycobacterium with low intrinsic toxicity clearly shows that endogenously produced NO down-regulates the inflammatory response because it is involved in the regulation of the cellular composition and the expression of inflammatory mediators within the tissue lesion. One implication of our findings is that endogenous NO contributes to the reduction of immunopathology during the natural course of M. avium infection. This is in stark contrast to an earlier report in which NOS2-knockout mice apparently showed decreased granuloma formation in response to M. avium infection.39 That study, however, compared the histopathology of M. avium-infected C57BL/6 mice (carrying the bcg susceptible allele) with that of NOS2-knock-out mice on a 129Sv/J background (carrying the bcg-resistant allele). Bacterial containment and granuloma formation in response to M. avium infection, however, are known to depend dramatically on the genetic background of mice and in particular on the allelic form of bcg.40 Thus, the histomorphological comparisons in the quoted work39 may have missed the striking differences observed when mice of the same background are studied. Indeed, more recent studies by another group of investigators in which NOS2-knockout mice were infected with M. avium corroborate our own findings in that NOS2-knockout mice also developed larger and more numerous granulomas in their livers and had significantly higher numbers of T cells in their spleens and IFN-γ protein in their sera when directly compared to immunocompetent mice of the same genetic background.37