Dietary Vitamin D₃ Suppresses Pulmonary Immunopathology Associated with Late-Stage Tuberculosis in C3HeB/FeJ Mice

Tuberculosis (TB) is a significant human disease caused by aerogenic transmission of the intracellular pathogen Mycobacterium tuberculosis, which primarily infects phagocytes in the lung alveoli (1). Recent data from the World Health Organization (WHO) demonstrate there were ∼9.6 million new TB cases in 2014, with 1.5 million individuals dying from TB in the same year (2). Improved socioeconomic conditions, public health practices, and use of standard short-course chemotherapy have reduced the global TB death rate by 45% since 1990 and have helped the WHO achieve its goal of reversing TB incidence by 2015 (2). However, the treatment options available for TB and increasingly common multidrug-resistant TB remain limited and are often complicated by poor patient compliance because of long treatment durations (2). For these reasons, new or adjunctive therapies that shorten treatment or improve treatment outcome are greatly needed and required for the WHO to reach its goal of eliminating TB by 2050 (3).

Host resistance to TB is promoted by a Th1 response that is initiated by lung-resident phagocytes following their internalization of M. tuberculosis bacilli; after these phagocytes migrate and present M. tuberculosis Ag in the draining mediastinal lymph nodes (MLNs), Ag-specific T cells accumulate and the lungs develop granulomas comprising both mononuclear and polymorphonuclear lineages (1). In experimental TB, there is a delay in Ag-specific T cell activation (4), allowing for dissemination of M. tuberculosis to extrapulmonary sites. Phagocyte production of IL-12 is essential for activation of T cells and their subsequent production of IFN-γ and TNF-α (5, 6), which are the two major Th1 cytokines that promote macrophage (MØ) killing of intracellular M. tuberculosis (1). The necessity of IFN-γ during infection is highlighted by the observation that T cells incapable of IFN-γ production are unable to control M. tuberculosis burdens (5, 7). Paradoxically, although Th1 responses are required to control M. tuberculosis, pulmonary inflammation comes at the expense of immunopathology that is potentially fatal to the host (8, 9). In animal models, this is evidenced by the postinfection mortality of mice deficient in IL-27R (10) or PD1 (11, 12), both of which function to suppress Th1 responses. In humans, the danger of immunopathology is evidenced by the morbidity and mortality associated with a failure to resolve lung immunopathology (13). Given the negative consequences of too little or too much inflammation, adjunctive TB therapies must create and/or support an immune environment that can effectively contain the bacteria without damaging the lung or other organs. Although adjunct immunosuppressive therapies can enhance antibiotic efficacy (14), others can reduce immunopathology and markedly improve animals’ survival even in the absence of effecting M. tuberculosis burden (15).

There is currently widespread interest in using vitamin D₃ (VitD₃) as an adjunctive therapy for TB (16). VitD₃ is a fat-soluble secosteroid that can be obtained through the diet or synthesized in the skin. VitD₃ enters the circulation and is hydroxylated in the liver to form 25-OH hydroxyvitamin D₃ (25(OH)D) (17). 25(OH)D can either freely diffuse into target cells or bind the VitD binding protein and enter cells via receptor-mediated endocytosis (18). 25(OH)D must then undergo a second hydroxylation by 25-hydroxyvitamin D 1α hydroxylase (CYP27b1) to form 1,25(OH)₂ dihydroxyvitamin D (1,25(OH)₂D), the biologically active form referred to as VitD. VitD binds the vitamin D receptor (VDR) in the cytosol and translocates into the nucleus where it binds VitD response elements on the DNA and regulates transcription of multiple genes in both hematopoietic and non-hematopoietic cell types (17). After it was discovered that both sun exposure and animal fat consumption (e.g., butter and cod liver oil) are sources of VitD₃, it was widely reasoned that VitD₃ contributed to the therapeutic efficacy of these early TB treatments (19). VitD’s direct bactericidal effect on M. tuberculosis has been studied in vitro (20, 21), as have VitD’s indirect bactericidal effects on M. tuberculosis–infected human monocyte/MØ cultures (22–24). These in vitro studies have led to the following prevailing model (25): Following TLR2/1 recognition of M. tuberculosis, MØs increase their expression of VDR and CYP27B1 (22); any 25(OH)D that enters the MØ is then converted to VitD, binds the VDR, translocates to the nucleus, and through synergy with IFN-γ increases production of cathelicidin—an antimicrobial peptide that is directly cytotoxic to M. tuberculosis (26). However, despite the number of in vitro studies that point to VitD₃ and its metabolites as having both direct and indirect bactericidal activities, several large clinical trials have demonstrated that VitD₃’s effect on sputum conversion is largely nil (27–30). These negative data raise the question of what if any role does VitD₃ have during TB in vivo.

The goal of this study was to determine what effect dietary VitD₃ has on the outcome of experimental TB. This determination is important given the current interest in using VitD₃ as an adjunct therapy for TB, despite its effect on TB never having been tested in an in vivo model. Specifically, we used the C3HeB/FeJ model of TB to compare several key experimental TB readouts in the presence or absence of dietary VitD₃. Among mouse models of TB, the C3HeB/FeJ strain is considered by many to be the best at recapitulating the pathology observed in human TB (31–35). Given the number of in vitro studies demonstrating that VitD has both direct and indirect bactericidal effects, we began our study with the hypothesis that dietary VitD₃ would protect the host by reducing M. tuberculosis burden; we were surprised to observe this is not the case, and instead that VitD₃ protects the host by limiting TB-associated immunopathology. Collectively, our results demonstrate that dietary VitD₃ functions through hematopoietic lineages to suppress the pulmonary immunopathology associated with late-stage TB.

C3HeB/FeJ and B6.129S4-Vdr^tm1Mbd/J (i.e., vdr^+/− (36)) breeder pairs were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred at the Medical College of Wisconsin. All mice were treated according to National Institutes of Health and Medical College of Wisconsin Institute Animal Care and Use Committee guidelines.

vdr^+/+ and vdr^−/− progeny were used to generate radiation bone marrow chimeras according to our reported protocols, the percent reconstitution being ≥94% (5, 6). Mice were allowed 10 wk to reconstitute prior to their use.

Beginning 1 mo prior to infection, C3HeB/FeJ mice were fed specially formulated diets from Harlan Laboratories (Madison, WI) that contained either 0 IU VitD₃/g (VitD^LO, Harlan number TD.89123), 2 IU VitD₃/g (VitD^Std, Harlan number TD.1108), or 20 IU VitD₃/g (VitD^HI, Harlan number TD.110799). These diets were chosen because their formulations were identical with the exception of VitD₃ and were similar to the diets used in other studies of dietary VitD₃ (37–39). For experiments involving vdr^−/− and vdr^+/+ mice and radiation bone marrow chimeras: beginning at weaning and continuing throughout each experiment, vdr^−/− and vdr^+/+ progeny were kept on a VitD^HI rescue diet containing 20 IU VitD₃/g and excess lactose (20% lactose, Harlan number TD.140326). Excess lactose is necessary to prevent the bone and immune abnormalities that would otherwise develop in vdr^−/− mice (40–42).

Mice were aerosol infected with virulent M. tuberculosis H37Rv according to our reported protocols (5, 6). For bacterial load determinations, lungs and half a spleen were removed and individually homogenized in normal sterile saline; serial dilutions of each homogenate were then plated on 7H11, and colonies were counted after ≥2-wk incubation at 37°C and 5% CO₂. Lungs from control mice were plated on day 1 postinfection to confirm the delivery of 46–108 bacteria/lung.

According to the methods of Chackerian et al. (43), half of each spleen from M. tuberculosis–infected VitD^HI and VitD^LO mice were collected at the indicated time points, pushed through a 70-μm nylon screen, and treated with sterile RBC lysis buffer. A total of 2 × 10⁶ cells/spleen in 2 ml complete RPMI 1640 medium (RPMI 1640 medium + 10% FBS) were then stimulated with 1 μg/ml Con A and live M. tuberculosis (H37Rv) at multiplicity of infection (MOI) = 0, 0.5, or 1 for 48 h. Supernatants were then collected, filtered through a 0.22-μm filter, and stored at −80°C until being assayed for murine IFN-γ and IL12/23p40 by sandwich ELISA (BD Biosciences).

Lungs from M. tuberculosis–infected mice were formalin-fixed and paraffin-embedded and used for H&E and Acid Fast staining (Children’s Hospital of Wisconsin Histology Core). Images were taken with a Labophot-2 upright microscope (Nikon, Tokyo, Japan) using a Retiga 2000R camera (QImaging, Surrey, BC, Canada); NIS Elements software (Nikon) was used for morphometric analysis (Supplemental Fig. 1).

Paraffin-embedded lungs were sectioned, deparaffinized, hydrated in deionized water, and treated with Target Retrieval Solution (DakoCytomation, Carpinteria, CA) at either pH 9 (anti-CD3, anti-Foxp3, and anti-Ly6C/Ly6G) or pH 6 (anti-F4/80). All slides were then stained using a Dako Autostainer Plus (DakoCytomation) with the standard labeled streptavidin–biotin protocol. Lung sections were stained at room temperature for 60 min with polyclonal rabbit anti-CD3 (1:100; DakoCytomation), anti-F4/80 (clone A3-1; 1:250; AbD Serotec), anti-Foxp3 (clone FJK-16s; 1:25; eBioscience), or anti-Ly6C/Ly6G (clone NIMP-R14; 1:250). Negative controls were performed with omission of the primary Ab. Slides were then developed with tertiary streptavidin–HRP (DakoCytomation) and counterstained with hematoxylin.

Immunohistochemistry (IHC)-stained slides were scanned on a high-resolution, whole slide scanner (NanoZoomer HT 2.0, Hamamatsu, Japan) at original magnification ×40. For image data quantification, images were imported into Visiopharm software (Visiopharm, Hørsholm, Denmark), and original magnification ×20 region-of-interest (ROI) images were extracted to estimate detectable Ab (DAB)–positive areas. DAB-negative areas were also measured to include in the total tissue area calculations. We used four ROIs per tissue section, and morphometric analysis was performed. All images were processed with these preset thresholds and linear Bayesian classification to generate a processed image (Supplemental Fig. 2). Total DAB-positive area per ROI (and total tissue area of the ROI) was measured in microns.

MLN cells were prepared and stained for surface and intracellular markers per our previous protocols (5), using Abs specific for CD4 (clone L3T4), CD8 (clone 53-6.7), IFN-γ (clone XMG1.2), and TNF-α (clone MP6-XT22) (BD Biosciences).

Lung RNA was isolated from snap frozen tissue using the IBI Total RNA extraction kit (IBI Scientific, Peosta, IA). RNA was reverse transcribed using Applied Biosciences reagents (Life Technologies, Grand Island, NY), and the resulting cDNA was amplified using Bullseye Evagreen Master Mix (MidSci) and Bio-Rad iQ5 detection system. The primer sequences used for cDNA amplification of ifng, cyp27b1, cramp, vdr, and gapdh are listed in Supplemental Table I. Cycle threshold values were determined using the Bio-Rad iQ5 bundled software, and all samples were normalized to gapdh. cDNA from specific time points was additionally used in the RT² Profiler PCR Array for Mouse Innate and Adaptive Immune Responses (Qiagen, Valencia, CA). Data were analyzed using the online RT² Profiler PCR Array data analysis software (Qiagen). All samples were normalized to at least one housekeeping gene between plates.

For group analysis, two-way ANOVA with Bonferroni correction was used with p < 0.05 considered significant. For all other experiments, unpaired Student t test was used to determine significance (p < 0.05). All results are expressed as mean ± SD.

VitD₃ and VitD₃ metabolites are capable of killing M. tuberculosis in vitro via both direct and indirect mechanisms (20, 21, 25). To determine whether dietary VitD₃ affects M. tuberculosis burden in vivo, adult C3HeB/FeJ mice were placed on either a VitD₃-replete (VitD^HI mice) or VitD₃-deficient (VitD^LO mice) diet for 30 d, aerosol-infected with 46–108 CFU of virulent M. tuberculosis (H37Rv), and kept on their respective diets for up to 110 d postinfection (Fig. 1A). We then measured and compared organ CFU burden in VitD^LO and VitD^HI mice over a broad period of time postinfection. Standard CFU changes in C3HeB/FeJ mice were established by comparing VitD^LO and VitD^HI values to those of C3Heb/FeJ mice kept on standard mouse chow (VitD^Std mice; Fig. 1B–D). Throughout the entire time course, plasma 25(OH)D levels were significantly higher in VitD^HI mice relative to VitD^LO and VitD^Std mice; no differences in circulating Ca²⁺ and PO₄³⁻ levels were observed between VitD^LO, VitD^Std and VitD^HI mice (data not shown).

We observed that in both VitD^HI and VitD^LO groups, lung M. tuberculosis burdens increased logarithmically during the first 25 d postinfection (Fig 1B). Between days 25 and 78 postinfection, M. tuberculosis burden continued to increase in both VitD^HI and VitD^LO groups albeit more slowly than days 1–25. At day 92 postinfection, we observed a modest increase in M. tuberculosis burden in VitD^HI mice relative to VitD^LO mice; however, this difference was directionally reversed by the last time point (day 110). Two-way ANOVA comparison of both VitD^LO and VitD^HI curves demonstrated that although the effect of time on bacterial burden was significant (p < 0.0001), the effect of diet was not (p = 0.1444). Similarly, in the spleen—an organ to which M. tuberculosis disseminates 11–14 d postinfection in mice (43) —the effect of time was significant (p < 0.0001), whereas no significant difference existed between the M. tuberculosis burdens of VitD^HI and VitD^LO groups (p = 0.6109) (Fig. 1C). On the basis of the growth characteristics of M. tuberculosis in VitD^LO and VitD^HI lungs, we termed postinfection days 1–40 as representing early-stage TB, and the period thereafter as representing late-stage TB; these terms are consistent with the terminology of others to describe the same periods (1, 4, 35). The abundance of bacteria in VitD^LO and VitD^HI mice during both early and late-stage TB was also apparent by acid fast staining of lungs from these periods (Fig. 1D–G). Collectively, these data demonstrate that dietary VitD₃ does not significantly affect bacterial burden during either early or late stages of experimental TB.

Dietary VitD₃ has been repeatedly demonstrated to suppress organ-specific inflammation in autoimmune disease models (44). To determine whether dietary VitD₃ limits proinflammatory gene expression in M. tuberculosis–infected organs, lungs from M. tuberculosis–infected VitD^LO and VitD^HI groups were collected on postinfection days 35 and 92, used for mRNA extraction and cDNA synthesis, and analyzed via real-time PCR array for the expression of select genes of interest. Day 35 was chosen to represent early-stage TB because, by this time, the adaptive immune response against M. tuberculosis has been established and limits M. tuberculosis growth (1); day 92 was chosen to represent late-stage TB. The results of this analysis are shown in Fig. 2A and demonstrate that, on postinfection day 35, the expression of multiple proinflammatory genes was increased in VitD^HI lungs relative to VitD^LO lungs. However, by postinfection day 92, this difference in expression had inversed relative to that observed on postinfection day 35; namely, the expression pattern of many of the same genes was decreased in VitD^HI lungs relative to VitD^LO lungs. We confirmed this dynamic expression pattern for select genes of interest (e.g., ifng) by separate PCR (data not shown). In addition to the genes assessed via PCR array, we also compared VitD^LO and VitD^HI lungs’ expression of cramp (the mouse homolog of cathelicidin) (Fig. 2B), vdr (Fig. 2C), and cyp27b1 (Fig. 2D). Cathelicidin is an effector of VitD₃-elicted M. tuberculosis growth restriction in vitro (24); VDR and CYP27b1 are essential for VitD3 signaling and metabolism, respectively. Our data demonstrate that cramp transcription was increased in both VitD^HI and VitD^LO groups following M. tuberculosis infection (Fig. 2B); the expression levels of vdr and cyp27b1 varied with the time point examined but did not significantly differ between VitD^HI and VitD^LO groups at each timepoint (Fig. 2C, 2D). Therefore, any phenotypic differences between VitD^HI and VitD^LO mice are likely not due to differences in VitD metabolism or antimicrobial peptide production.

To measure and compare the recall response in the spleens of M. tuberculosis–infected VitD^HI and VitD^LO mice, splenocytes were collected at four time points (postinfection days 1, 35, 62, 98), normalized for cell number, and cultured in the presence of a polyclonal mitogen (Con A) and three different M. tuberculosis multiplicities of infection (MOI 0, 0.5, and 1). Two days later, culture supernatants were collected and used for ELISA measurements of both IL12/IL23p40 and IFN-γ; MØs and dendritic cells are the principal sources of IL12/IL23p40 during experimental tuberculosis (6), whereas αβ T cells, NK cells, and neutrophils are the sources of IFN-γ (5). The results of these recall experiments are shown in Fig. 2E and 2F and demonstrate that VitD^LO splenocytes secreted significantly higher levels of IL12/IL23p40 and IFN-γ on postinfection day 62 relative to VitD^HI mice. This difference was observed at each MOI. Collectively, our lung transcription studies (Fig. 2A) and splenocyte response studies (Fig. 2E, 2F) demonstrate that dietary VitD₃ suppresses proinflammatory gene expression during late-stage TB.

The Th1 response is required to limit TB and manifests in M. tuberculosis–infected mouse lungs by the accumulation and organization of CD3⁺ lymphocytes and innate myeloid-derived lineages (histiocytes) into granulomatous regions. To determine whether dietary VitD₃ limits the manifestation of lung Th1 immunity, we visually and quantitatively compared the contribution of CD3⁺ lymphocytes to VitD^HI and VitD^LO lung granulomatous regions during early- and late-stage TB (Fig. 3). Specifically, we used morphometric analysis of sections stained with H&E and CD3-specific mAb (IHC) to determine the area of each granulomatous region and the extent to which lymphocytes contributed to each granulomatous region (Supplemental Fig. 1); four sections per lung were analyzed in this manner, with each section containing two to three distinct granulomatous regions. H&E stains of representative granulomatous regions from VitD^LO and VitD^HI mice are shown in Fig. 3A and 3B, respectively; shown in each row are successively higher magnifications (original magnifications ×4, ×10, ×20, and ×100) demonstrating that hematoxylin-rich areas primarily comprise cells with lymphocyte morphology (small cells with a large nucleus:cytoplasm ratio). IHC on VitD^LO lungs (Fig. 3C) and VitD^HI lungs (Fig. 3D) demonstrated these lymphocytes to be highly reactive with anti-CD3. The average area of VitD^LO and VitD^HI granulomatous regions were similar during late-stage TB (Fig. 3G). However, a notable difference between VitD^LO and VitD^HI mice was that the extent to which lymphocytes contributed to each granuloma area was significantly lower in VitD^HI mice relative to VitD^LO mice during late-stage TB; this difference was both visually apparent (compare the sizes of the hematoxylin staining regions in Fig. 3A to those of Fig. 3B) and quantifiable (Fig. 3H). Regarding the extent to which these lymphocytes were CD3⁺, digital IHC quantification (Supplemental Fig. 2) demonstrated that during early-stage TB, VitD^HI mice had a greater frequency of CD3⁺ lymphocytes in the lung; this difference did not exist during late-stage TB. These CD3⁺ cells most likely represent effector lineages because Foxp3 IHC (Fig. 3E, 3F) and digital Foxp3 IHC analysis (Fig. 3J) demonstrate low frequencies of Foxp3⁺ cells in the granulomatous regions of both VitD^LO and VitD^HI lungs. Collectively, these data demonstrate that dietary VitD₃ attenuates CD3⁺ lymphocytes’ contribution to M. tuberculosis granulomatous regions.

To characterize the T cells that develop in VitD^HI and VitD^LO mice following M. tuberculosis infection, MLNs of M. tuberculosis–infected VitD^LO and VitD^HI mice were removed on days 25, 35, and 92 postinfection and used for flow cytometric analysis of CD4, CD8, IFN-γ, and TNF-α staining. For reference purposes, MLN preparations from M. tuberculosis–infected VitD^STD mice were analyzed in an identical manner. The results of this analysis are shown in Fig. 4 and demonstrate that, on day 25 postinfection, MLN preparations of both groups comprised roughly equal frequencies of CD4⁺ and CD8⁺ cells (Fig. 4A, 4E, 4F). By postinfection days 35 and 92, the frequency of CD4⁺ cells had grown more prominent in both groups, albeit significantly higher in VitD^LO mice compared with VitD^HI mice (Fig. 4B, 4C, 4E); MLN CD8⁺ frequencies were also higher in VitD^LO mice on postinfection day 92 (Fig. 4C, 4F). Regarding MLN cells’ capacity to produce cytokines, VitD^HICD4⁺ cells displayed a greater capacity to produce IFN-γ on days 25 and 35 (Fig. 4G); by day 92, the frequency of CD4⁺IFNγ⁺ cells in VitD^HI and VitD^LO mice were equal (Fig. 4G). At all time points, the frequency of CD4⁺TNFα⁺, CD8⁺IFNγ⁺, and CD8⁺TNFα⁺ cells were similar between VitD^HI and VitD^LO groups (Fig. 4H–J). Collectively, these data demonstrate that dietary VitD₃ limits the accumulation of IFN-γ– and TNF-α–producing CD4⁺ (i.e., Th1) cells during experimental TB.

That the granulomatous regions of VitD^HI mice contained fewer CD3⁺ lymphocytes but were nevertheless equal in size to the granulomatous regions of VitD^LO mice raised the question of whether histiocyte and/or polymorphonuclear lineages were overrepresented in VitD^HI granulomatous regions. To address this question, lung sections serial to those shown in Fig. 4C and 4D were stained with Abs specific to F4/80 and Ly6C/Ly6G. In M. tuberculosis–infected C3HeB/FeJ mice, F4/80 is a marker of both MØs and myeloid-derived suppressor cells, whereas Ly6C/Ly6G is a marker of both neutrophils and myeloid-derived suppressor cells (45). Representative F4/80 staining from both days 35 and 92 are shown in Fig. 5A–D, along with quantitation of the same images in Fig. 5E; representative Ly6C/Ly6G staining from both days 35 and 92 are shown in Fig. 5F–I, along with quantitation of the same images in Fig. 5J. Visual and quantitative analysis was limited to granulomatous regions; uninvolved lung regions were excluded from our analysis. The data demonstrate on postinfection day 35 that the representation of both F4/80⁺ and Ly6C/Ly6G⁺ histiocytes was higher in VitD^HI granulomas relative to VitD^LO granulomas (Fig. 5E, 5J). The elevated representation of F4/80⁺ histiocytes in VitD^HI granulomas continued to postinfection day 92 (Fig. 5E). Collectively, these data demonstrate that dietary VitD₃ accelerating the accumulation of F4/80⁺ and Ly6C/Ly6G⁺ lineages during experimental TB.

A negative consequence of inflammation in the M. tuberculosis–infected lung is immunopathology (8, 9, 46, 47). Given our observation that dietary VitD₃ limits the inflammatory response in M. tuberculosis-infected lung and spleen, we wished to determine whether VitD^HI mice were protected from M. tuberculosis-associated lung pathology. In C3HeB/FeJ mice and other TB-susceptible mouse strains, the progression of lung pathology associated with the appearance of acid fast bacilli (AFB) in acellular areas of lung granulomatous regions (34). The acellular areas of granulomatous regions are a consequence of necrosis and are enriched in fragmented nuclei and membrane cholesterol (48). On day 35 postinfection, AFB were readily apparent in the granulomatous regions of VitD^LO mice (Fig. 1C) and VitD^HI mice (Fig. 1E); in neither group were AFB found outside of the granuloma at this time point. By day 92 postinfection, AFB were found to associate with foamy MØs in both VitD^LO mice (Fig. 1D) and VitD^HI mice (Fig. 1F). However, primarily in VitD^LO granulomas (postinfection day 92), we also observed acellular areas with abundant AFB (Fig. 6A, 6B). These acellular areas were characterized by the appearance of lipid droplets in which AFB could occasionally be found within (Fig. 6A) and cholesterol clefts to which AFB were consistently in proximity (Fig. 6B). Quantitative analysis confirmed VitD^LO lung sections contained significantly more cholesterol clefts than VitD^HI lungs (Fig. 6C). Additional examination of VitD^LO granulomas demonstrated that fragmented nuclei were also present in these acellular regions, as evidenced by hematoxylin-staining puncta (Fig. 6D); acid-fast staining demonstrated AFB were often in close proximity to these fragmented nuclei (Fig. 6E). Finally, when cholesterol cleft-containing regions of the lung were stained with either F4/80 or Ly6C/Ly6G, we observed that cholesterol clefts were consistently surrounded by Ly6C/Ly6G-staining tissue (Fig. 6F–H); only occasionally did F4/80 staining associate with cholesterol clefts (data not shown). Collectively, these data demonstrate that dietary VitD₃ suppresses pulmonary immunopathology during late-stage TB.

There are several distinct lineages within an M. tuberculosis–infected animal that are responsive to VitD: host hematopoietic cells, host nonhematopoietic cells, the microbiome, and M. tuberculosis itself. To determine whether host cells must be VitD responsive to limit TB-associated immunopathology, we aerosol-infected vdr^−/− mice and compared the lung immunopathology that subsequently developed to that observed in C57BL/6 controls. vdr encodes the vitamin D receptor that is necessary for VitD responsiveness (36); when placed on a rescue diet containing excess lactose, vdr^−/− mice are nearly indistinguishable from vdr^+/+ mice in terms of hematopoietic and nonhematopoietic development (40–42). Analogous to our experiments with C3HeB/FeJ mice, vdr^+/+ and vdr^−/− mice were placed on a rescue diet containing high levels of VitD for 1 mo prior to M. tuberculosis infection and were kept on the same diet throughout the experiment time course; at indicated times, the lungs were removed and used for an analysis of M. tuberculosis burden and lung immunopathology.

Consistent with our studies of VitD^LO and VitD^HI C3HeB/FeJ mice, neither lung (Fig. 7A) nor spleen (Fig. 7B) M. tuberculosis burdens differed between vdr^+/+ and vdr^−/− mice. However, upon visual examination of M. tuberculosis–infected vdr^+/+ and vdr^−/− lungs, we noted a greater representation of lymphocytes in vdr^−/− granulomas and that the histiocyte-rich regions of vdr^−/− granulomas also showed evidence of necrosis (Fig. 7C, 7D). Specifically, picnotic and karyorrhectic nuclei were abundant in histiocyte-rich regions (compare Fig. 7Ci to Fig. 7Di), and parenchyma destruction was also occasionally observed (Fig. 7Dii). To quantify these histological differences and to determine whether these features were due to vdr deficiency in hematopoietic or nonhematopoietic lineages, we generated and infected radiation bone marrow chimeras in which vdr deficiency was restricted to either the hematopoietic compartment (i.e., donor vdr^−/−→ vdr^+/+ hosts) or nonhematopoietic compartment (i.e., donor vdr^+/+→vdr^−/− hosts); vdr^+/+→vdr^+/+ and vdr^−/−→vdr^−/− radiation bone marrow chimeras were also generated and used as controls. As shown in Fig. 7E–H, despite each group having equivalent M. tuberculosis burdens (Fig. 7E, 7F), vdr^−/−→vdr^+/+ and vdr^−/−→vdr^−/− lungs displayed higher levels of immunopathology as measured by the increased representation of lymphocytes (Fig. 7G) and picnotic or karyorrhectic nuclei (Fig. 7H) within their granulomatous regions. These characteristics did not appear in uninfected vdr^−/−→ vdr^+/+ controls (data not shown). Collectively, these data demonstrate that hematopoietic responses to VitD are required to limit TB immunopathology.

Despite there having been longstanding interest in VitD₃ as an adjunctive therapy for TB (49), there have until now been no in vivo studies of VitD₃’s effect on the pulmonary immune response to M. tuberculosis. This represents a significant gap in the TB immunology research continuum. To fill this gap, we used C3HeB/FeJ mice–fed diets either deficient or replete with VitD₃ and aerosol infected them with virulent M. tuberculosis. At time points representing early- and late-stage TB, both groups were examined for changes in VitD status and metabolism as well as standard experimental TB readouts. Our results demonstrate that dietary VitD₃ modulates the pulmonary immune response by altering the accumulation dynamics of CD3⁺ lymphocytes and F4/80⁺ and Ly6C/Ly6G⁺ histiocytes in granulomatous regions; the overall consequence of these changes is a microenvironment that suppresses immunopathology while still permitting M. tuberculosis control. Collectively, our data support a model wherein VitD is acting in vivo during TB not as a bactericide but rather as an immunosuppressant that limits pulmonary inflammation to levels that are sufficient to control M. tuberculosis but limited in the ability to cause pathology.

VitD₃ is well recognized in many areas of immunology research as having immunosuppressive activity (44). However, in the field of TB immunology, the prevailing model of how VitD affects TB has until this point centered around VitD’s direct and indirect mycobactericidal activities (25). VitD’s direct mycobactericidal activity in M. tuberculosis broth culture has been demonstrated repeatedly (20, 21). The indirect mycobactericidal activity of VitD has also been repeatedly demonstrated using cocultures of primary human MØs and virulent M. tuberculosis, a system in which VitD synergizes with IFN-γ to increase production of cathelicidin (23, 24)—an antimicrobial peptide that is cytotoxic in M. tuberculosis broth cultures (26). At the patient level, there is an inverse correlation between circulating 25(OH)D concentrations and risk of both active and latent TB (27, 50, 51). However, clinical trials testing VitD₃’s use as an adjunctive therapy for TB have failed to produce striking changes in sputum conversion rate, despite involving hundreds of individuals and independent statistical analyses (27–30). We propose that rather than serving to enhance killing of M. tuberculosis, VitD instead protects the host by limiting M. tuberculosis–elicited inflammation. Put in terms of the disease tolerance framework (9), VitD reduces the fitness cost of M. tuberculosis infection by promoting tolerance (i.e., reducing the negative impact of infection) instead of promoting resistance (i.e., reducing pathogen burden once an infection is established).

The appropriateness of mice as a model organism for studies of TB pathology has been a subject of deliberation for several years (52). One reason for this is that the granulomas that develop in common mouse strains (e.g., C57BL/6 and BALB/c) do not, as a general rule, display caseous necrosis (34)—a defining feature of granulomas in untreated human TB (53). Although there are exceptions to this general rule (47), the infrequency of caseous necrosis in common strains makes it difficult to systematically determine the biology underlying formation of necrotic lesions. For this reason, the development of the C3HeB/FeJ model by Kramnik et al. (54) was a major step forward for the field because mice permit investigations that otherwise cannot be performed in other animal models for either technical or economic reasons. Relative to C57BL/6 mice, M. tuberculosis–infected C3HeB/FeJ mice have been reported to die more quickly, have accelerated M. tuberculosis growth in the lung and spleen, and develop necrotic caseous lung lesions (54). Contrasting with these previous studies, none of the C3HeB/FeJ mice in our study developed the large caseous lesions, high M. tuberculosis burdens (∼10^8–9 CFU), and mortality reported to develop in C3HeB/FeJ mice by ∼30 d postinfection (54). Instead, we found the M. tuberculosis growth characteristics in C3HeB/FeJ mice to resemble those historically observed in C57BL/6 mice (52). Nevertheless, the immunopathology we observed in C3HeB/FeJ mice was more severe than any we have ever seen in C57BL/6 mice (5, 6). Specifically, we observed necrotic areas in H37Rv-infected VitD^LO mice during late-stage TB that contained abundant AFB. The reason we did not observe the caseous lesions reported by others is unknown at this time but likely relates to the use of different M. tuberculosis strains (Erdman versus H37Rv) or passaging protocols (55). It is also possible that more dramatic differences in either M. tuberculosis burden, immunopathology, or mRNA expression readouts occur between VitD^LO and VitD^HI mice at other time points, including the period between postinfection days 35 and 92.

Regarding the mechanism by which VitD functions through hematopoietic cells to suppress TB, there are two possibilities that are the subject of future investigations in our laboratory. First, given transgenic Ipr1’s ability to suppress pathology in C3HeB/FeJ mice (56), it is possible that VitD increases the bioactivity of Ipr1 in M. tuberculosis–infected MØs. In terms of basic Ipr1 biology, little is known about Ipr1 other than it is an IFN-inducible nuclear protein that switches MØ programming from necrosis to apoptosis. Second, despite Foxp3 staining being similar in VitD^LO and VitD^HI granulomatous regions, it remains possible that dietary VitD₃ elicits regulatory T cells (T_REGs) that are more capable of suppressing lung immunopathology. T_REGs develop early following mouse M. tuberculosis infection and delay the onset of Th1 immunity (4). Although T_REGs’ role in limiting TB immunopathology is suggested by comparing lung T_REG numbers in TB-resistant and TB-susceptible mouse strains (57), this role has not yet been empirically tested.

Finally, we believe it is important to relate the findings of the current study to that of the human clinical trial data reported by Martineau and colleagues (27, 58). Martineau et al. (27) demonstrated that although administering adjunct VitD₃ increases serum 25(OH)D levels in patients receiving treatment for pulmonary TB, it does not significantly affect time to sputum culture conversion. When their results were stratified based on participants TaqI genotype (a silent polymorphism in VDR exon 9), a modest reduction in time to sputum culture conversion was observed in patients with the tt genotype relative to those with either the TT or Tt genotype. Individuals with the tt genotype make up <9% of populations most effected by TB (59). Among the same study participants, Coussens et al. (58) observed a more profound effect of adjunct VitD₃: namely, VitD₃ supplementation reduced markers of systemic inflammation in humans during TB treatment. Specifically, PBMCs from VitD₃-treated TB patients produced less IL-1R1A, IL-6, IL-12, and TNF-α following ex vivo stimulation with M. tuberculosis CFP-10 (relative to PBMCs from placebo-treated TB patients). Adjunctive VitD₃ also lowered the monocyte:lymphocyte balance in the circulation of TB patients; a high monocyte:lymphocyte ratio associates with increased disease hazards in some TB patient subpopulations (60–62). Also, and contrary to what Coussens et al. (58) expected based on previous literature, adjunct VitD₃ suppressed circulating cathelicidin levels in TB patients. Our M. tuberculosis–infected VitD^HI mice phenocopy the VitD₃-treated TB patients of Coussens et al. in the following ways: VitD reduced pulmonary expression of IL-1r1, IL-6, and TNF-α (Fig. 2A); VitD reduced splenocyte IL12/23p40 secretion following ex vivo stimulation (Fig. 2E); VitD reduced the frequency of lymphocytes in the lung granuloma and draining lymph node (Figs. 3H, 4D, 4E); and VitD affected neither M. tuberculosis burden nor cathelicidin-related antimicrobial peptide expression.

After considering the clinical trial data of Coussens et al. (58) alongside the findings of the current study, we propose that in most individuals VitD does not function to enhance M. tuberculosis killing in the manner advocated by Modlin and Bloom (25). Instead, VitD protects the host by reducing M. tuberculosis–elicited inflammation to a level that is still sufficient to restrict M. tuberculosis growth but unable to cause pathology. Although the model of VitD being bactericidal is firmly rooted in reviews of TB literature, the alternative model of VitD being a suppressor of TB immunopathology is already being promoted (63). Our study supports this alternative view of the VitD–TB interaction. Our study also illustrates that a small animal model can be used to advance the understanding of an effect that has been demonstrated in a human clinical trial (that vitamin D supplementation modulates inflammatory responses in TB). This is important given recent concerns that small-animal TB models misled development of MVA85A, a new tuberculosis vaccine that was ultimately nonefficacious in human infants (64). As future clinical trials are designed to investigate VitD₃ as an adjunct therapy for TB—as well as TB associated immune reconstitution inflammatory syndrome—it behooves us to move beyond traditional readouts of success (i.e., reduced time to sputum conversion) to also consider reductions in lung pathology and individual morbidity as successful outcomes (63).

We thank Drs. Paula North and Suresh Kumar (Children’s Hospital of Wisconsin Histology Core) for assistance in histological analysis; and Dr. Aniko Szabo (Medical College of Wisconsin Biostatistics Consulting Service) for assistance with statistical analysis.

The authors have no financial conflicts of interest.